The tidal disruption event AT2017eqx: spectroscopic evolution from hydrogen rich to poor suggests an atmosphere and outflow
M. Nicholl, P. K. Blanchard, E. Berger, S. Gomez, R. Margutti, K. D. Alexander, J. Guillochon, J. Leja, R. Chornock, B. Snios, K. Auchettl, A. G. Bruce, P. Challis, D. J. D'Orazio, M. R. Drout, T. Eftekhari, R. J. Foley, O. Graur, C. D. Kilpatrick, A. Lawrence, A. L. Piro, C. Rojas-Bravo, N. P. Ross, P. Short, S. J. Smartt, K. W. Smith, B. Stalder
MMNRAS , 1–17 (2019) Preprint 5 August 2019 Compiled using MNRAS L A TEX style file v3.0
The tidal disruption event AT2017eqx: spectroscopicevolution from hydrogen rich to poor suggests anatmosphere and outflow
M. Nicholl, , (cid:63) P. K. Blanchard, E. Berger, S. Gomez, R. Margutti, K. D. Alexander, , J. Guillochon, J. Leja, , R. Chornock, B. Snios, K. Auchettl, , A. G. Bruce, P. Challis, D. J. D’Orazio, M. R. Drout, , T. Eftekhari, R. J. Foley, O. Graur, , , C. D. Kilpatrick, A. Lawrence, A. L. Piro, C. Rojas-Bravo, N. P. Ross, P. Short, S. J. Smartt, K. W. Smith, B. Stalder Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, EH9 3HJ, UK Birmingham Institute for Gravitational Wave Astronomy and School of Physics and Astronomy, University of Birmingham,Birmingham B15 2TT, UK Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge,Massachusetts, 02138, USA Center for Interdisciplinary Exploration and Research in Astrophysics (CIERA) and Department of Physics and Astronomy,Northwestern University, Evanston, IL 60208, USA Einstein Fellow NSF Astronomy and Astrophysics Postdoctoral Fellow Astrophysical Institute, Department of Physics and Astronomy, 251B Clippinger Lab, Ohio University, Athens, OH 45701, USA Center for Cosmology and Astro-Particle Physics and Department of Physics, The Ohio State University, 191 West WoodruffAvenue, Columbus,OH 43210, USA DARK, Niels Bohr Institute, University of Copenhagen, Lyngbyvej 2, 2100 Copenhagen, Denmark The Observatories of the Carnegie Institution for Science, 813 Santa Barbara St., Pasadena, CA 91101, USA Department of Astronomy and Astrophysics, University of Toronto, 50 St. George St., Toronto, Ontario, M5S 3H4 Canada Department of Astronomy and Astrophysics, University of California, Santa Cruz, CA 95064, USA Department of Astrophysics, American Museum of Natural History, New York, NY 10024, USA Astrophysics Research Centre, School of Mathematics and Physics, Queens University Belfast, Belfast BT7 1NN, UK Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, HI 96822, USA
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
We present and analyse a new tidal disruption event (TDE), AT2017eqx at redshift z = . , discovered by Pan-STARRS and ATLAS. The position of the transientis consistent with the nucleus of its host galaxy; the spectrum shows a persistentblackbody temperature T (cid:38) , K with broad H I and He II emission; and it peaksat a blackbody luminosity of L ≈ erg s − . The lines are initially centered at zerovelocity, but by 100 days the H I lines disappear while the He II develops a blueshiftof (cid:38) , km s − . Both the early- and late-time morphologies have been seen in otherTDEs, but the complete transition between them is unprecedented. The evolutioncan be explained by combining an extended atmosphere, undergoing slow contraction,with a wind in the polar direction becoming visible at late times. Our observationsconfirm that a lack of hydrogen a TDE spectrum does not indicate a stripped star,while the proposed model implies that much of the diversity in TDEs may be dueto the observer viewing angle. Modelling the light curve suggests AT2017eqx resultedfrom the complete disruption of a solar-mass star by a black hole of ∼ . M (cid:12) . Thehost is another Balmer-strong absorption galaxy, though fainter and less centrallyconcentrated than most TDE hosts. Radio limits rule out a relativistic jet, while X-ray limits at 500 days are among the deepest for a TDE at this phase. Key words: accretion, accretion disks – galaxies: nuclei – black hole physics (cid:63)
E-mail: [email protected] © a r X i v : . [ a s t r o - ph . H E ] A ug M. Nicholl et al.
A tidal disruption event (TDE) occurs when an unfortunatestar passes so close to a supermassive black hole (SMBH)that the tidal force of the SMBH exceeds the self-gravityof the star (Hills 1975). If this takes place outside of theSchwarzschild radius, the result is a luminous flare with L bol ∼ − erg s − , powered either by accretion onto theSMBH (Kochanek 1994; Jiang et al. 2016b; Lodato 2012;Piran et al. 2015). Observationally, these are differenti-ated from more common transients like supernovae by theirhigher blackbody temperatures ( T ∼ , − , K) andcoincidence with the centres of galaxies.Although TDEs were initially expected to peak at X-ray wavelengths if they are powered by accretion (Ko-mossa 2002), TDE candidates have now been discoveredin the rest-frame UV and optical by surveys such as SDSS(van Velzen et al. 2011), Pan-STARRS (Gezari et al. 2012;Chornock et al. 2014; Blanchard et al. 2017; Holoien et al.2018b), ASASSN (Holoien et al. 2014, 2016a,b), OGLE(Wyrzykowski et al. 2017), PTF (Arcavi et al. 2014; Hunget al. 2017; Blagorodnova et al. 2017, 2019) and ZTF (vanVelzen et al. 2019). Up to ∼ of these events are in factfaint in X-rays (Auchettl et al. 2017), suggesting either thatsome events are not powered by direct accretion, or thatX-rays can only escape along certain sight-lines (Dai et al.2018).TDEs have been found in galaxies with stellar massesranging from ∼ . − M (cid:12) (van Velzen 2018; Weverset al. 2017, 2019a), corresponding to black hole masses ∼ − M (cid:12) (Mockler et al. 2019), and even greater in thecase of rapidly rotating SMBHs (Leloudas et al. 2016). Theyprovide a novel way to probe the properties of otherwisedormant SMBHs and their environments, especially at thelower end of the mass spectrum. TDEs are more commonin galaxies with a high stellar mass surface density and acentrally-concentrated light profile (Law-Smith et al. 2017;Graur et al. 2018), but there remains an unexplained over-representation of Balmer-strong absorption galaxies (Arcaviet al. 2014; French et al. 2016; French & Zabludoff 2018).Studying TDEs is complicated by the wide diversity intheir observed characteristics, in the optical, X-rays and ra-dio (Auchettl et al. 2017; Zauderer et al. 2011; Levan et al.2011; van Velzen et al. 2016; Alexander et al. 2016). Theiroptical spectra can exhibit lines of H I, He II, or both (Ar-cavi et al. 2014), but usually retain their spectroscopic signa-tures over time. A few instead show broad absorption lines(Chornock et al. 2014; Leloudas et al. 2016). More recently,metal lines have been detected in some TDEs (Blagorodnovaet al. 2019; Brown et al. 2018; Leloudas et al. 2019; Weverset al. 2019b), while yet others show spectra dominated bypre-existing broad-line and narrow-line regions from an ac-tive galactic nucleus (Blanchard et al. 2017; Kankare et al.2017). Determining the nature of line formation in TDEs iskey to understanding the physical processes in these events(e.g. Guillochon et al. 2014; Roth & Kasen 2018).In this paper, we present a new TDE, AT2017eqx, whichwe observed to undergo a radical evolution in its spectro-scopic properties over time. The spectrum initially showedprominent Balmer and He II emission lines centred at closeto zero velocity. Later spectra showed no evidence for hy-drogen emission, while the He II feature became blueshifted by (cid:38) km s − . Both of these morphologies have beenobserved before in other TDEs, but to our knowledge sucha transition within the same event has not. Understandingthis evolution can shed new light on the geometry and line-formation in these events.Our study is organised as follows. We describe the dis-covery of AT2017eqx in section 2 and detail our multiwave-length follow-up observations in section 3. We analyze thelight curve in section 4 to derive bolometric properties andinfer physical parameters. We present and interpret the sur-prising spectroscopic evolution in section 5. In section 6, westudy the host galaxy in the context of other TDE hosts,before concluding in section 7. AT2017eqx (survey name, PS17dhz) was discovered by thePanSTARRS Survey for Transients (Chambers et al. 2016)on 2017-06-07 UT, at right ascension 22h 26m 48.370s, dec-lination 17 ◦
08’ 52.40”. The source was coincident with thecentre of a galaxy catalogued in the Sloan Digital Sky Surveyas SDSS J222648.38+170852.2, with an apparent magnitudeof g = . mag in Data Release 14 (Abolfathi et al. 2018).We determined the offset of AT2017eqx from the nucleusof this galaxy using a deep g -band image of the transientobtained with LDSS3 on 2017-07-22 and an archival pre-disruption image from PanSTARRS DR1 (Flewelling et al.2016). We geometrically aligned these images, with 30 com-mon stars for reference, using the geomap and geotran tasks in pyraf . After transforming to a common coordinategrid and measuring the centroids of the transient and hostgalaxy, we find a relative offset of . (cid:48)(cid:48) ± . (cid:48)(cid:48) . Thus theorigin of the source is fully consistent with the nucleus ofthe galaxy.We observed AT2017eqx spectroscopically, beginning on28-06-2017, as part of a search for nuclear transients. Thespectrum with the highest signal-to-noise ratio is shown inFigure 1. Our observations show the hallmarks of TDEs: ablue continuum, with a roughly constant blackbody tem-perature of (2–3) × K, and emission lines with widths > km s − . We also detect narrow absorption lines from thehost galaxy, from which we measure the redshift z = . ± . . The host is a Balmer-strong absorption galaxy – arare type of galaxy that is greatly over-represented amongTDE hosts (Arcavi et al. 2014; French et al. 2016; Law-Smithet al. 2017; Graur et al. 2018).The field was frequently monitored by the AsteroidTerrestrial-impact Last Alert System (Tonry et al. 2018).The ATLAS automated search pipeline triggers on σ de-tections (Stalder et al. 2017; Tonry et al. 2018) and nonew source (with multiple σ detections) was found bythe pipeline. We applied forced photometry on template-subtracted images at the transient coordinates, and manu-ally binned the resulting magnitudes to a nightly cadence.This resulted in several 3 σ detections in the ATLAS o -band, including one prior to the PSST discovery. Syntheticphotometry on our earliest spectrum of the transient indi-cates a colour o − i = − . mag at the time of discovery, insome tension with the PanSTARRS and ATLAS photom-etry ( o − i = . mag). While not entirely explained, thismay be related to a broad emission feature at the blue edge MNRAS , 1–17 (2019)
TDE changes its spots F ( a r b it r a r y un it s ) Figure 1.
Spectrum of AT2017eqx obtained with Magellan andLDSS3. The blue continuum and broad emission lines, leading toclassification as a TDE, can be easily seen. Lower panels showzoom-ins around host galaxy absorption lines from Ca II H&K,H δ , H γ , H β , Mg I 5175, Na I D, H α , Ca II NIR triplet, fromwhich we securely measure a redshift z = . . of the i -band. To compensate, we added a constant shift of0.31 mag to all ATLAS measurements so that the colour isconsistent with the higher signal-to-noise ratio PanSTARRSphotometry. We imaged AT2017eqx in the optical g , r , i , z filters using theInamori Magellan Areal Camera and Spectrograph (IMACS)(Dressler et al. 2011) and the Low Dispersion Survey Spec-trograph 3 (LDSS3) on the 6.5-m Magellan Baade and Claytelescopes at Las Campanas Observatory, and KeplerCamon the 1.2-m telescope at Fred Lawrence Whipple Observa-tory (FLWO). All images were reduced using pyraf to applybias subtraction and flat-fielding. Photometry was measuredusing a custom wrapper for daophot , using stars in the fieldfrom PanSTARRS Data Release 1 (Flewelling et al. 2016) todetermine the point-spread function (PSF) and photomet-ric zeropoint of each image. We downloaded Pan-STARRS1DR1 g , r , i images (Flewelling et al. 2016) as templates forthe field and convolved and subtracted them from the im-ages using the hotpants algorithm (Becker 2015) to isolatethe transient, before measuring its flux with the PSF model.We also observed AT2017eqx in g , r , i with the 1-mSwope Telescope at Las Campanas Observatory. The datawere reduced using the photpipe photometry and differenceimaging pipeline (Rest et al. 2005; Kilpatrick et al. 2018). photpipe is a flexible software package that performs op-timal bias-subtraction and flat-fielding, image stitching, as- A pp a r e n t m a gn it ud e UVW2UVM2UVW1Ug rizo
Figure 2.
Optical and UV light curves of AT2017eqx from PS1,ATLAS, Magellan, FLWO, Swope and
Swift . Host fluxes havebeen removed by subtracting reference images where possible( g , r , i , z and o bands), or otherwise by subtracting fluxes derivedfrom a host SED model (section 6). A third-order polynomial fitto the early ATLAS and PS1 data (dashed line) suggest maximumlight occured on MJD 57921.6. trometry, and photometry using dophot (Schechter et al.1993). We again subtracted PS1 reference images using hot-pants . Final photometry was performed on the differenceimages using dophot .We obtained further images on 2017-08-16 with the low-resolution imaging spectrograph (LRIS) on the Keck-I 10-mtelescope on Mauna Kea, Hawaii. Observations were per-formed in the blue and red channels simultaneously withthe B + R filters and V + I filters and the D560 dichroic, andreduced using photpipe . For our photometric calibration,we used secondary calibrators in each image with magni-tudes derived from SDSS standard stars transformed to theBVRI system (Bilir et al. 2011; Alam et al. 2015). Differenceimaging was performed using PS1 g -band images for the B -and V -band images, r -band images for the R -band images,and i -band for the I -band images.Imaging in the UV was obtained using the UV-OpticalTelscope (UVOT) on board the Neil Gehrels Swift observa-tory. We downloaded the data from the
Swift public archiveand extracted light curves in the
UVW2 , UVM2 , UVW1 and U filters following the procedures outlined by Brown et al.(2009), using a 5”aperture. The magnitudes are calibrated inVegamags in the Swift photometric system (Breeveld et al.2011). No reference images were available in the UV, so weestimated the host galaxy contribution using our best-fitspectral energy distribution (SED) model (section 6). Thegalaxy flux in each UVOT filter was determined by applyingthe s3 synthetic photometry package (Inserra et al. 2018) tothe SED model; this was then subtracted from the UVOTmeasurements of AT2017eqx. Our optical and UV photom-etry is shown in Figure 2 and listed in Table 1.Finally, we checked for variability in public data fromthe Wide-field Infrared Survey Explorer (WISE; Wrightet al. 2010). The latest data release includes detections inthe W1 and W2 bands. These are consistent with the his-torical magnitudes of the host galaxy. If there is significant MNRAS000
UVW2 , UVM2 , UVW1 and U filters following the procedures outlined by Brown et al.(2009), using a 5”aperture. The magnitudes are calibrated inVegamags in the Swift photometric system (Breeveld et al.2011). No reference images were available in the UV, so weestimated the host galaxy contribution using our best-fitspectral energy distribution (SED) model (section 6). Thegalaxy flux in each UVOT filter was determined by applyingthe s3 synthetic photometry package (Inserra et al. 2018) tothe SED model; this was then subtracted from the UVOTmeasurements of AT2017eqx. Our optical and UV photom-etry is shown in Figure 2 and listed in Table 1.Finally, we checked for variability in public data fromthe Wide-field Infrared Survey Explorer (WISE; Wrightet al. 2010). The latest data release includes detections inthe W1 and W2 bands. These are consistent with the his-torical magnitudes of the host galaxy. If there is significant MNRAS000 , 1–17 (2019)
M. Nicholl et al. F ( - e r g s - c m - Å - ) + c on s t a n t
10d +1.728d +1.431d +162d +0.7112d +0.4145dHost -0.6 Rest-frame wavelength (Å)10 λ L λ ( e r g s - c m - ) Figure 3.
Rest-frame spectra of AT2017eqx, labelled by phase from maximum light. The final spectrum contains only host galaxy light,and has been subtracted from the others. Darker lines show data after Savitsky-Golay smoothing. Dashed lines indicate blackbody fitswith temperatures of ≈ , K. Broad emission features from He II λ and H α are also apparent, with the latter disappearing inlater spectra. The inset shows the full X-ray to radio spectral energy distribution from Chandra and the VLA around day 50. dust in the nucleus of this galaxy, future WISE data mayshow an infrared echo over the next few years (Jiang et al.2016a). We obtained six epochs of spectroscopy between 2017-06-28and 2017-11-26 using LDSS3, IMACS, and the BlueChan-nel spectrograph on the 6.5-m MMT telescope (Schmidtet al. 1989). Spectra were reduced in pyraf , including biassubtraction, flat-fielding, wavelength calibration using arclamps, and flux calibration using standard stars observedon the same nights. We obtained one additional spectrum on2018-08-06 using Binospec on MMT. This was reduced usinga dedicated pipeline. The final spectrum shows no evidenceof TDE features, and we therefore consider it a pure hostgalaxy spectrum. All spectra were scaled to contemporane-ous photometry (interpolated were necessary) and telluricfeatures removed using model fits. We corrected for extinc-tion using the Galactic dust maps of Schlafly & Finkbeiner(2011), and assumed negligible extinction in the transienthost galaxy. Host-subtracted spectra are plotted in Figure3; a log of spectra is provided in Table 2.
We observed AT2017eqx using the Karl G. Jansky VeryLarge Array (VLA) in C configuration on 2017-07-14 and2017-08-22 (Program ID: 16B-318; PI: Alexander). Each ob-servation lasted one hour and was split between C and Kbands (central frequencies 6.0 GHz and 21.7 GHz). We re-duced the data using the VLA CASA Calibration Pipeline(CASA version 4.7.2) and imaged the data using standardCASA routines (McMullin et al. 2007). Both observationsused 3C48 as the flux calibrator and J2232+1143 as thephase calibrator.No significant radio emission was detected in eitherepoch. We therefore combined the epochs to derive deeperlimits on the radio flux, at a mean phase of 55 days af-ter discovery. The 6 GHz limit corresponds to ν L ν < . × erg s − at the distance of AT2017eqx. All limits are listedin Table 3. AT2017eqx has one of the deepest radio lim-its among TDEs, particularly at this phase from disruption(Figure 4). Only iPTF16fnl (Blagorodnova et al. 2017) andAT2018zr (van Velzen et al. 2019) have deeper constraints.The non-detection in the radio rules out a powerfuljet similar to those associated with the relativistic TDEsSwift J1644+57 (Zauderer et al. 2011; Bloom et al. 2011;Burrows et al. 2011) and Swift J2058+05 (Cenko et al. MNRAS , 1–17 (2019)
TDE changes its spots Rest-frame days since discovery10 ν L ν ( e r g s - ) Figure 4.
Limit on radio emission at 6 GHz obtained with theVLA. The deep limit rules out a powerful jet in our line of sight, asseen in the brightest radio TDEs. The possibility of a weaker non-relativistic outflow cannot be excluded, but the limit places strictrequirements that the radio luminosity from any an outflow doesnot exceed that seen in nearby events like ASASSN-14li (Alexan-der et al. 2016; van Velzen et al. 2016). Further observations willbe required to rule out an off-axis jet such as that seen in Arp299-B (Mattila et al. 2018). Comparison data are from Alexanderet al. (2017); Eftekhari et al. (2018); Blagorodnova et al. (2017);Zauderer et al. (2011); Berger et al. (2012); Cenko et al. (2012b);Alexander et al. (2016, 2017); Eftekhari et al. (2018); Komossa(2002); Bower et al. (2013); van Velzen et al. (2013); Arcavi et al.(2014); Chornock et al. (2014); Mattila et al. (2018); van Velzenet al. (2019) . -1 Rest-frame days since discovery10 . - k e V L u m i no s it y ( e r g s - ) Figure 5.
Limits on X-ray emission obtained using
Swift and
Chandra . The limits imply an X-ray to optical ratio < − , muchlower than for any TDEs that do have observed X-ray emission.Comparison data are from Auchettl et al. (2017); Gezari et al.(2017). TDEs plotted with large symbols use the same colourscheme as in Figure 4. Initial X-ray data was obtained using the X-ray Telescope(XRT) onboard
Swift . Analysis of 5.7 ks of data collected be-tween 2017-07-11 and 2017-07-24 revealed no significant fluxin the 0.3-10 keV range. We carried out deeper observationsusing the
Chandra
X-ray Observatory with 10 ks integra-tions on 2017-08-16 and 2019-01-04 (Programs 20625, 21437;PI: Nicholl, Berger), again resulting in non-detections. Wealso checked for archival imaging before the optical flare torule out previous AGN activity. This field was observed byXMM-Newton on 2015-05-20 (Observation ID: 076247020).We analysed the image using the online XMM-Newton Sci-ence Archive tools. No source is detected at the position ofthe host galaxy.For all epochs, we assume a power law spectral modelwith Γ = , and a Milky Way hydrogen column density of N H = . × cm − along this line of sight, in order to con-vert count rates to fluxes over the range 0.3-10 keV. We veri-fied that our results are only weakly dependent on our choiceof model, and we derive similar constraints if we instead as-sume a blackbody SED with a temperature of 0.1 keV. Thelimiting count rates and fluxes are given in Table 4.Our limits from Chandra imply an X-ray/optical ratio < − . TDEs with optical and X-ray detections have generallyexhibited X-ray/optical ratios ∼ . For some TDEs, the ratiocan be much greater. These are generally relativistic TDEs,though XMMSL1 J074008.2-853927 (Saxton et al. 2017) ap-pears to be an example of a TDE with a thermal componentthat also exhibited a large X-ray/optical ratio. However,many TDEs have X-ray non-detections that imply a ratio (cid:28) (Auchettl et al. 2017). AT2017eqx falls firmly in thiscategory. Compared to other TDEs which have exhibitedstrong X-rays, e.g. ASASSN-14li (Miller et al. 2015; Brownet al. 2017), ASASSN-15oi (Holoien et al. 2016b; Gezari et al.2017; Holoien et al. 2018a), and Swift J1644+57 (Levan et al.2011, 2016), the X-ray emission from AT2017eqx is at least MNRAS000
X-ray Observatory with 10 ks integra-tions on 2017-08-16 and 2019-01-04 (Programs 20625, 21437;PI: Nicholl, Berger), again resulting in non-detections. Wealso checked for archival imaging before the optical flare torule out previous AGN activity. This field was observed byXMM-Newton on 2015-05-20 (Observation ID: 076247020).We analysed the image using the online XMM-Newton Sci-ence Archive tools. No source is detected at the position ofthe host galaxy.For all epochs, we assume a power law spectral modelwith Γ = , and a Milky Way hydrogen column density of N H = . × cm − along this line of sight, in order to con-vert count rates to fluxes over the range 0.3-10 keV. We veri-fied that our results are only weakly dependent on our choiceof model, and we derive similar constraints if we instead as-sume a blackbody SED with a temperature of 0.1 keV. Thelimiting count rates and fluxes are given in Table 4.Our limits from Chandra imply an X-ray/optical ratio < − . TDEs with optical and X-ray detections have generallyexhibited X-ray/optical ratios ∼ . For some TDEs, the ratiocan be much greater. These are generally relativistic TDEs,though XMMSL1 J074008.2-853927 (Saxton et al. 2017) ap-pears to be an example of a TDE with a thermal componentthat also exhibited a large X-ray/optical ratio. However,many TDEs have X-ray non-detections that imply a ratio (cid:28) (Auchettl et al. 2017). AT2017eqx falls firmly in thiscategory. Compared to other TDEs which have exhibitedstrong X-rays, e.g. ASASSN-14li (Miller et al. 2015; Brownet al. 2017), ASASSN-15oi (Holoien et al. 2016b; Gezari et al.2017; Holoien et al. 2018a), and Swift J1644+57 (Levan et al.2011, 2016), the X-ray emission from AT2017eqx is at least MNRAS000 , 1–17 (2019)
M. Nicholl et al. l og L bo l ( e r g s - ) T BB ( K )
25 0 25 50 75 100 125 150Rest-frame days from maximum light2.55.07.5 R BB ( c m ) Figure 6.
Bolometric light curve of AT2017eqx using blackbodyfits, calculated at all epochs with i or o observations. Emptysymbols indicate early epochs where the temperature was esti-mated assuming zero colour evolution. We also show the deepX-ray limits corresponding to (cid:46) of the optical luminosity.The light curve evolution is similar to other TDEs from the liter-ature (van Velzen et al. 2019; Holoien et al. 2016b; Gezari et al.2012; Chornock et al. 2014; Arcavi et al. 2014; Holoien et al. 2014,2016a; Blagorodnova et al. 2017). The middle and lower panelsshow the best-fit temperature and radius for each epoch. Theemitting material contracts at roughly constant temperature. an order of magnitude less than that observed in those events(Figure 5). We determine the bolometric luminosity of AT2017eqx usingthe following procedure, implemented via superbol (Nicholl2018). We first interpolate or extrapolate all light curves toa common set of observation times, defined by having anobservation in i or o bands, using low (first- or second-) or-der polynomial fits. At each epoch, we integrate the spectralluminosity over all bands, and fit blackbody curves to esti-mate the flux falling outside of the UV-optical wavelengthrange. These fits also allow us to constrain the temperatureand radius of the emitting material.The bolometric light curve and blackbody parametersare shown in Figure 6. The best-fitting blackbody tempera-ture is ≈ , − , K and shows no significant evolutionin time within the errors. Visually there may be a slight rise,though this is sensitive to extrapolating the UV photome-try. The blackbody radius decreases from . × cm to . × cm. AT2017eqx peaks at a luminosity ≈ erg s − (with a solid lower limit of > . erg s − in the observedbands), and emits a total of > . × erg over the duration
50 0 50 100 150Rest-frame days from maximum light121416182022242628 A pp a r e n t m a gn it ud e z-5i-3o-1.5 rg+2U+4.5 W1+6M2+7W2+8 Figure 7.
Fits to the light curve of AT2017eqx using the TDEmodel in mosfit , for the case where we assume disruption occuredwithin 30 days before the first detection. Coloured lines show 100MCMC realisations. of our observations. The luminosity and light curve shape istypical of optical TDEs.
We fit a physical TDE model to the observed UV and op-tical photometry using mosfit : the Modular Open SourceFitter for Transients (Guillochon et al. 2018). This is a semi-analytic code employing a range of modules that can belinked together to produce model light curves of astronomi-cal transients, and determine the best fitting model param-eters through Bayesian analysis. The TDE model and asso-ciated modules in mosfit are described in detail by Mockleret al. (2019). The method is based on an older code, tdefit (Guillochon et al. 2014), and uses a combination of scalingrelations and interpolations between the output of numericalTDE simulations to determine the luminosity.The model has ten free parameters: the masses ofthe star and SMBH; the impact parameter (determiningwhether a disruption is full or partial); the efficiency of con-verting fallback energy into radiation; two parameters con-trolling the relationship between the luminosity and radius;the time of disruption relative to first detection; a viscoustimescale over which the accretion disk forms; the extinction(column density) in the host galaxy; and a white-noise termparameterising any unaccounted-for variance.To sample the parameter space we used the affine-invariant ensemble method (Goodman & Weare 2010;Foreman-Mackey et al. 2013) as implemented in mosfit .We ran the Markov Chain with 100 walkers for 50,000 it-erations, checking for convergence by ensuring that the Po-tential Scale Reduction Factor was < . at the end of therun (Brooks & Gelman 1998). We assume the same priorsas used by Mockler et al. (2019), with the exception of thetime of disruption. Given that we have only one or two datapoints weakly constraining the rise-time of AT2017eqx, werun two fits: one where we allow the disruption to occur upto 168 days before first detection (corresponding to the time MNRAS , 1–17 (2019)
TDE changes its spots log (M BH (M )) = 6.32 +0.11-0.12 . . . . M s t a r ( M ) M star (M ) = 0.87 +0.17-0.31 . . . . s c a l e d β scaled β = 1.00 +0.16-0.23 . . . . . l og () log ( ) = -2.37 +0.48-0.37 . . . . . l og ( R ph ( c m )) log (R ph0 ( cm)) = 0.63 +0.23-0.18 . . . . . l ph l ph = 1.09 +0.17-0.13 l og ( T v i s c ( d a y s )) log (T visc ( days)) = -1.37 +1.19-1.10 . . . . . l og ( n H , h o s t ) log (n H , host ) = 20.71 +0.15-0.41 . . . . log (M BH (M )) . . . . . l og () . . . . M star (M ) . . . . scaled β . . . . . log ( ) . . . . . log (R ph0 ( cm)) . . . . . l ph log (T visc ( days)) . . . . . log (n H , host ) .
78 0 .
72 0 .
66 0 .
60 0 . log ( ) log ( ) = -0.66 +0.05-0.05 Figure 8.
Corner plot showing the posteriors for our TDE model fit. Black corresponds to the solution with a tight prior on the disruptiontime ( < days before detection); blue corresponds to the solution with a broader prior using the last non-detection in ATLAS ( < days). of the last ATLAS non-detection); and one where we re-strict disruption to within 30 days before detection, to forcea better fit to the first ATLAS detection.The model light curves for the latter case are shown inFigure 7, and the posteriors of the parameters for both fitsare shown as a corner plot in Figure 8. These posteriors areoverall similar, with narrower constraints in the case withthe stronger prior on disruption time, though some poste-riors show shifts in the median of up to ∼ σ between thefits. The posteriors point to the full disruption (i.e. scaledimpact parameter ≈ ) of a ∼ . M (cid:12) star. The SMBHmass is more sensitive to the assumed rise time, with log ( M BH / M (cid:12) ) = . ± . in the more general case, fallingto log ( M BH / M (cid:12) ) = . ± . when the rise time is constrainedto 30 days. The TDE peak luminosity ( ≈ erg s − ) cor-responds to 10-40% of the Eddington luminosity for thisSMBH mass range. The model employs a power-law photo-spheric radius, R ∝ L l ; the best fit has l (cid:39) , i.e. the radius MNRAS000
Corner plot showing the posteriors for our TDE model fit. Black corresponds to the solution with a tight prior on the disruptiontime ( < days before detection); blue corresponds to the solution with a broader prior using the last non-detection in ATLAS ( < days). of the last ATLAS non-detection); and one where we re-strict disruption to within 30 days before detection, to forcea better fit to the first ATLAS detection.The model light curves for the latter case are shown inFigure 7, and the posteriors of the parameters for both fitsare shown as a corner plot in Figure 8. These posteriors areoverall similar, with narrower constraints in the case withthe stronger prior on disruption time, though some poste-riors show shifts in the median of up to ∼ σ between thefits. The posteriors point to the full disruption (i.e. scaledimpact parameter ≈ ) of a ∼ . M (cid:12) star. The SMBHmass is more sensitive to the assumed rise time, with log ( M BH / M (cid:12) ) = . ± . in the more general case, fallingto log ( M BH / M (cid:12) ) = . ± . when the rise time is constrainedto 30 days. The TDE peak luminosity ( ≈ erg s − ) cor-responds to 10-40% of the Eddington luminosity for thisSMBH mass range. The model employs a power-law photo-spheric radius, R ∝ L l ; the best fit has l (cid:39) , i.e. the radius MNRAS000 , 1–17 (2019)
M. Nicholl et al. F - C on ti nuu m Fe II?
Figure 9.
Continuum-subtracted, smoothed spectra ofAT2017eqx, with important line features marked. The earlyspectra show H I and He II lines, possibly contaminated byHe I, while later spectra show only a blueshifted feature closeto He II. The peak of the He II line at early times appears tohave an additional contribution from N III, indicating the Bowenflourescence mechanism may be in effect (Blagorodnova et al.2019; Leloudas et al. 2019). We also mark the location of Fe IIlines recently identified in AT2018fyk by Wevers et al. (2019b),however we disfavour this identification for AT2017eqx due tothe lack of both X-rays and an optical plateau as seen in thatevent. of the emitting region is directly proportional to the lumi-nosity.The best-fit viscous time is always much shorter thanthe rise time, indicating that the emission has not beendelayed by a long circularisation process. This requires ei-ther that an accretion disk forms promptly after disruptionor that the optical luminosity instead arises from stream-stream collisions (Mockler et al. 2019). Assuming that theenergy is released close to the innermost stable circular orbitof a . M (cid:12) SMBH, the implied blackbody temperature is ∼ × K, which is in some tension with the temperaturerange ruled out by our X-ray non-detections, (cid:38) × K(Figure 3, inset). Thus in an accretion-powered model, thedisk emission would have to be downgraded to a cooler spec-trum by a reprocessing layer.
As shown in the previous section, the overall blue contin-uum in the spectrum shows little change over time. Toanalyse the evolution in our spectra, we therefore first re-move this continuum using fifth-order polynomial fits (e.g.Hung et al. 2017). We excise from the fit any regions within ± ˚A ( ≈ , km s − ) of the rest wavelengths of H α , H β and He II λ α and He II λ ,with several features matching the locations of the otherBalmer lines. The presence of both He II and H I in the spec-trum would qualify AT2017eqx as a transitional event be-tween so-called ‘H-rich’ and ‘He-rich’ TDEs, following thecontinuum identified by Arcavi et al. (2014). However, laterspectra, > days after maximum, look markedly different.The hydrogen lines have largely disappeared, leaving only astrong emission line around He II λ . AT2017eqx there-fore offers direct evidence that disruption of a hydrogen-richstar (required by the early spectra) can form a spectrumwith no visible hydrogen. We will return to this critical pointin section 5.3.There is some indication of neutral He I lines in thespectrum, which could help to explain asymmetries apparentin the line profiles. A red shoulder in the early-time H α profile could be due to He I λ , while a blue shoulder inHe II λ could be contamination from He I λ . Theearliest spectrum shows a potential He I λ , which mightsupport this interpretation, though He I lines do not appearto be present at later times (see section 5.2).While early studies of TDE optical spectra primarilyidentified emission from hydrogen and helium, several au-thors have recently found evidence for metal lines such asN III and O III, attributed to the Bowen flourescence mecha-nism (Blagorodnova et al. 2019; Leloudas et al. 2019) . Thisoccurs when the → transition in recombining He II emitsa photon that happens to resonate with a far-UV transitionin O III, which in turn produces a photon that resonateswith N III (Bowen 1935). Optical photons are also producedin this cascade, which (unlike the far-UV photons) can es-cape to the observer. We mark the positions of N III lineson Figure 9 (the O III Bowen lines are at λ < ˚A). N III λ δ for a firm identification. How-ever, N III λ (or possibly C III λ ) does appear to bepresent in the early spectra of AT2017eqx, manifesting as abump just bluewards of the centre of the broad He II λ peak. This is similar to the profile seen in TDEs iPTF15af(Blagorodnova et al. 2019) and AT2018dyb/ASASSN-18pg(Leloudas et al. 2019). This supports the claim by Leloudaset al. (2019) that such features may be common in TDEs.The He II line appears shifted to bluer wavelengths inthe later spectra. The shift is too large to be explained byblending with N III λ ; furthermore, the N III λ Lines from other ionisation states of oxygen, nitrogen and car-bon have been found in the UV (Cenko et al. 2016; Brown et al.2018). MNRAS , 1–17 (2019)
TDE changes its spots FitHe II He IH I
Figure 10.
Fits to the main emission features in the spectra ofAT2017eqx. Lines of the same ion (H I, He II, He I) have the samevelocity centroids and widths as well as fixed luminosity ratios.The fits are to the unsmoothed spectra, but smoothed data arealso shown to guide the eye. Derived parameters for H α and He II λ are shown in Figure 11. of these properties in the case of AT2017eqx, and thereforecannot identify Fe II. We analyse the line profiles quantitatively by means of Gaus-sian fits. There are several important caveats to note: blend-ing between broad overlapping lines complicates their ob-served profiles. The fits are also sensitive to the choice ofcontinuum, which we defined using a fifth-order polynomialfit. Finally, the profile of even an isolated line is not neces-sarily Gaussian – lines may have a large electron-scatteringoptical depth, and outflows can induce asymmetries (Roth& Kasen 2018). Nevertheless, we proceed with Gaussian fitsas a simplified means to determine line centres and widths,following other studies in the literature (Arcavi et al. 2014;Blagorodnova et al. 2017; Hung et al. 2017).The most difficult issue to deal with is the effect ofblending. We account for this by fitting the entirety of eachcontinuum-subtracted spectrum simultaneously (Figure 10)with a model that includes H I ( α, β, γ ), He II (4686 ˚A) andHe I (4471, 5876, 6678 ˚A). We neglect the Bowen fluorescencelines (N III and C III) as these are almost completely degen-erate with He II λ at the observed line widths, and we donot see a strong N III λ line to suggest that the Bowenmechanism dominates line formation in this region, as seenin AT2018dyb (Leloudas et al. 2019). We therefore estimatethat the Bowen lines account for no more than (cid:46) of themeasured flux in this linr blend.For each line we allow the centroid offset(red/blueshift), velocity width and luminosity to vary, L ( e r g s - ) He II 4686H102030 v F W H M ( k m s - )
20 40 60 80 100 120 140Rest-frame days from maximum light10505 v s h i f t ( k m s - ) Figure 11.
Evolution of line properties from the fits in Figure 10.The He II/H α ratios changes from ∼ to > , while He II developsa blueshift of 5000-8000 km s − . We are unable to reliably measurea blueshift for H α in the later epochs due to its low luminosity(empty symbol). but fix lines from the same ion to have the same offset andwidth. To further reduce the number of free parameters, wefix the ratios between lines from a given ion. For the Balmerlines we assume Case B recombination (Osterbrock & Fer-land 2006), which predicts H α /H β = . and H α /H γ = . .Our analysis is not sensitive to the precise ratios here, andwe obtain essentially the same results for any reasonablechoices, though we get poor fits at early times if H α /H β ∼ .For He I, we use the model ratios from (Benjamin et al.1999), which show little sensitivity to temperature ordensity: λ / λ ≈ . , λ / λ ≈ . This leaves atotal of 9 free parameters. We fit to the spectra using theOptimize routine in scipy .We show the fits in Figure 10 and plot the derived lumi-nosities, velocity widths and shifts of H α and He II in Figure11 (He I contributes significantly only in the earliest spec-trum). He II exhibits a fairly flat luminosity with time, whileH α fades by at least a factor 5. We therefore determine aninitial ratio He II / H α ∼ , but a markedly different ratio af-ter 100 days of He II / H α > . We measure an anomalouslylow He II luminosity in the 62 day spectrum, for which wecannot rule out an issue due to the low signal-to-noise ratioof the data.At early times, lines are centred close to their rest wave-lengths, but we find a substantial shift in the He II line atlate times (after H α fades) measuring a maximum blueshiftof ∼ km s − at 112 days. We find a smaller but stillsignificant blueshift of ∼ km s − at 145 days, but thesedata are noisier. Such shifts are too large to be explained bythe un-modelled Bowen lines (N III, C III), which are offsetfrom He II only by 3000 km s − . MNRAS000
Evolution of line properties from the fits in Figure 10.The He II/H α ratios changes from ∼ to > , while He II developsa blueshift of 5000-8000 km s − . We are unable to reliably measurea blueshift for H α in the later epochs due to its low luminosity(empty symbol). but fix lines from the same ion to have the same offset andwidth. To further reduce the number of free parameters, wefix the ratios between lines from a given ion. For the Balmerlines we assume Case B recombination (Osterbrock & Fer-land 2006), which predicts H α /H β = . and H α /H γ = . .Our analysis is not sensitive to the precise ratios here, andwe obtain essentially the same results for any reasonablechoices, though we get poor fits at early times if H α /H β ∼ .For He I, we use the model ratios from (Benjamin et al.1999), which show little sensitivity to temperature ordensity: λ / λ ≈ . , λ / λ ≈ . This leaves atotal of 9 free parameters. We fit to the spectra using theOptimize routine in scipy .We show the fits in Figure 10 and plot the derived lumi-nosities, velocity widths and shifts of H α and He II in Figure11 (He I contributes significantly only in the earliest spec-trum). He II exhibits a fairly flat luminosity with time, whileH α fades by at least a factor 5. We therefore determine aninitial ratio He II / H α ∼ , but a markedly different ratio af-ter 100 days of He II / H α > . We measure an anomalouslylow He II luminosity in the 62 day spectrum, for which wecannot rule out an issue due to the low signal-to-noise ratioof the data.At early times, lines are centred close to their rest wave-lengths, but we find a substantial shift in the He II line atlate times (after H α fades) measuring a maximum blueshiftof ∼ km s − at 112 days. We find a smaller but stillsignificant blueshift of ∼ km s − at 145 days, but thesedata are noisier. Such shifts are too large to be explained bythe un-modelled Bowen lines (N III, C III), which are offsetfrom He II only by 3000 km s − . MNRAS000 , 1–17 (2019) M. Nicholl et al. S ca l e d F - c on ti nuu m H IHe IIN III H ,Rest He IIBlueshifted He II
Figure 12.
Spectroscopic comparison of AT2017eqx with other TDEs. All spectra have had continuum removed to highlight lineemission. Phases are listed relative to maximum light or discovery. The early spectra of AT2017eqx show the same lines as the H/He-strong TDEs ASASSN-14li (Holoien et al. 2016a) and iPTF16axa (Hung et al. 2017), while late spectra match the He-strong TDEs withblueshifted lines PTF09ge (Arcavi et al. 2014) and ASASSN-15oi (Holoien et al. 2016b). This demonstrates that a TDE can changeits apparent spectral type, and that observed lines are more indicative of the physical conditions in the stellar debris rather than itscomposition. ASASSN-14ae (Holoien et al. 2014) at first shows only hydrogen, but as these lines become weaker it also develops a He IIline, suggesting a similar evolution to AT2017eqx.
The spectroscopic evolution of AT2017eqx described in theprevious sections can be summarised in two distinct phases:at early times, it is strong in H I and He II, with lines centredclose to zero velocity. At later times, beyond 60-100 days,the spectrum is dominated by a single broad feature closeto He II but with a blueshift of up to ∼ km s − . Both ofthese morphologies have been observed in previous TDEs,but AT2017eqx is unique among the TDE sample to datein showing both a strongly evolving He II/H α ratio and alate-onset blueshift.Given the surprising spectroscopic evolution inAT2017eqx, we compare to literature TDEs in Figure 12.We include H-strong TDEs PTF09djl (Arcavi et al. 2014)and ASASSN-14ae (Holoien et al. 2014), the H/He-strongASASSN-14li (Holoien et al. 2016a) and iPTF16axa (Hunget al. 2017), and the He-strong PS1-10jh (Gezari et al. 2012),PTF09ge (Arcavi et al. 2014) and ASASSN-15oi (Holoienet al. 2016b); the latter latter two events show a signifi-cant blueshift in their He II line profiles. While AT2017eqxfalls neatly into the H/He-strong group around the time of maximum, the late-time spectrum is a close match to theHe-strong events, including the blueshift in He II.The transition in AT2017eqx from a H-strong to H-poor spectrum confirms theoretical arguments that the dif-ference between H-strong and He-strong TDE spectra is dueto physical conditions rather than the composition of thedisrupted star (Guillochon et al. 2014; Roth et al. 2016).The fact that AT2017eqx changes its spectral morphologyon a relatively short timescale indicates that line formationis likely very sensitive to the precise configuration of thesystem.Early models of TDE spectra suggested that largeHe II/H α ratios could be explained as a consequence of near-complete hydrogen ionization throughout the stellar debris(Guillochon et al. 2014). Such a situation implies that asthe ionizing flux from the inner disk fades, hydrogen lines(and He I 5876) could become apparent later in the TDEevolution – the opposite of what we observe in AT2017eqx.However, Roth et al. (2016) argued that under more realisticTDE conditions, wavelength-dependent optical depth is themost important factor in determining line strengths. Theirradiative transfer calculations showed that – all else being MNRAS , 1–17 (2019)
TDE changes its spots Figure 13.
Schematic of AT2017eqx and unified model for TDE line emission. Left (A): If an accretion disk forms quickly, X-rays heata reprocessing layer powering the optical light. In the polar region, a wind can develop (Dai et al. 2018; Metzger & Stone 2016). Right(B): Alternatively, energy is produced by collisions between debris streams. Bound debris forms an atmosphere (Jiang et al. 2016b), whileunbound material escapes along the stream directions, leading to a qualitatively similar inflow/outflow model (see also Hung et al. 2019).In either scenario, extended atmospheres produce He II λ and Balmer lines, whereas compact ones produce only He II (Guillochonet al. 2014; Roth et al. 2016). Outflows produce blueshifted emission lines (Roth & Kasen 2018), but are visible only for certain viewingangles. If the atmosphere contracts, blueshifted emission can be revealed to observers at larger angles. In case A, viewing angles far fromthe disk plane, which show blueshifted lines, may also reveal X-rays earlier, whereas no such correlation exists in case B. equal – a more compact envelope gives a larger He II/H α ra-tio, because H α is self-absorbed at most radii whereas He IIis thermalised at greater depth and therefore emitted overa greater volume. Thus the transition from a H-strong toHe-strong spectrum can potentially be explained if there isa contraction of the envelope towards the SMBH.A contracting envelope is consistent with the observeddecrease in luminosity at constant temperature (Figure 6).Moreover, our results from modelling the photometry withblackbody fits and mosfit indicated a photosphere thatgrows and shrinks in direct proportion to the luminosity.In the time between maximum light and the disappearanceof the hydrogen lines, the luminosity (and therefore radius)decrease by roughly an order of magnitude, which could ac-count for the change in the line ratios (Roth et al. 2016).Equally important is the blueshift in He II. Blendingwith other lines such as He I or the Bowen lines may makea contribution, but these lines are not sufficiently blue toaccount for the size of the shift. One species that does emitat approximately the right wavelength is Fe II. Wevers et al.(2019b) identified these lines in AT2018fyk, and argued thatthey could also account for the apparently blueshifted He IIline profile in TDEs like ASASSN-15oi. While we cannot A reduction in photospheric radius is also possible even in out-flowing material, if the density and ionization are decreasing asin a supernova, but this is generally accompanied by a decreasingtemperature. Moreover, a lower density should favour H α pro-duction, as photons emitted from greater depths could escapewithout destruction by self-absorption. rule out an Fe II contribution in AT2017eqx, these linesare thought to originate from dense gas close to a newly-formed accretion disk, and so would seem to be inconsistentwith the lack of X-ray emission or other disk signatures inAT2017eqx. If a disk was visible, another way to induce ablueshift is Doppler boosting of the blue (approaching) side.We disfavour this for two reasons: first, if the optical depthis low enough to reveal the disk, we would expect to seeH α , as in other events with disk-like line profiles (Arcaviet al. 2014; Holoien et al. 2018b). Second, the disk shouldbe hotter than the envelope, but we see no clear increase intemperature, nor the onset of X-ray emission.Roth & Kasen (2018) explained the blueshifted profilesin TDEs as evidence for electron-scattered line emission inan outflowing gas. Yet we have previously shown that theincreasing He II/H α ratio indicated a net inflow of material.Therefore to account for the full spectroscopic evolution ofAT2017eqx requires both inflowing and outflowing gas, alongwith an appropriate geometry. We show a schematic of sucha model in Figure 13.In this scenario, the luminosity is generated in a smallregion, either from an accretion disk or at the intersectionpoint between colliding debris streams. Initially, the observersees (reprocessed) emission from a quasi-static atmosphereof bound debris (Jiang et al. 2016b). Its extent, ∼ – cm, is roughly proportional to the luminosity from theTDE engine. Following Roth et al. (2016), the dominantemission lines depend on the extent and optical depth ofthis layer. Lines from this region are broadened by electronscattering, but centered at their rest-frame wavelength. MNRAS000
Schematic of AT2017eqx and unified model for TDE line emission. Left (A): If an accretion disk forms quickly, X-rays heata reprocessing layer powering the optical light. In the polar region, a wind can develop (Dai et al. 2018; Metzger & Stone 2016). Right(B): Alternatively, energy is produced by collisions between debris streams. Bound debris forms an atmosphere (Jiang et al. 2016b), whileunbound material escapes along the stream directions, leading to a qualitatively similar inflow/outflow model (see also Hung et al. 2019).In either scenario, extended atmospheres produce He II λ and Balmer lines, whereas compact ones produce only He II (Guillochonet al. 2014; Roth et al. 2016). Outflows produce blueshifted emission lines (Roth & Kasen 2018), but are visible only for certain viewingangles. If the atmosphere contracts, blueshifted emission can be revealed to observers at larger angles. In case A, viewing angles far fromthe disk plane, which show blueshifted lines, may also reveal X-rays earlier, whereas no such correlation exists in case B. equal – a more compact envelope gives a larger He II/H α ra-tio, because H α is self-absorbed at most radii whereas He IIis thermalised at greater depth and therefore emitted overa greater volume. Thus the transition from a H-strong toHe-strong spectrum can potentially be explained if there isa contraction of the envelope towards the SMBH.A contracting envelope is consistent with the observeddecrease in luminosity at constant temperature (Figure 6).Moreover, our results from modelling the photometry withblackbody fits and mosfit indicated a photosphere thatgrows and shrinks in direct proportion to the luminosity.In the time between maximum light and the disappearanceof the hydrogen lines, the luminosity (and therefore radius)decrease by roughly an order of magnitude, which could ac-count for the change in the line ratios (Roth et al. 2016).Equally important is the blueshift in He II. Blendingwith other lines such as He I or the Bowen lines may makea contribution, but these lines are not sufficiently blue toaccount for the size of the shift. One species that does emitat approximately the right wavelength is Fe II. Wevers et al.(2019b) identified these lines in AT2018fyk, and argued thatthey could also account for the apparently blueshifted He IIline profile in TDEs like ASASSN-15oi. While we cannot A reduction in photospheric radius is also possible even in out-flowing material, if the density and ionization are decreasing asin a supernova, but this is generally accompanied by a decreasingtemperature. Moreover, a lower density should favour H α pro-duction, as photons emitted from greater depths could escapewithout destruction by self-absorption. rule out an Fe II contribution in AT2017eqx, these linesare thought to originate from dense gas close to a newly-formed accretion disk, and so would seem to be inconsistentwith the lack of X-ray emission or other disk signatures inAT2017eqx. If a disk was visible, another way to induce ablueshift is Doppler boosting of the blue (approaching) side.We disfavour this for two reasons: first, if the optical depthis low enough to reveal the disk, we would expect to seeH α , as in other events with disk-like line profiles (Arcaviet al. 2014; Holoien et al. 2018b). Second, the disk shouldbe hotter than the envelope, but we see no clear increase intemperature, nor the onset of X-ray emission.Roth & Kasen (2018) explained the blueshifted profilesin TDEs as evidence for electron-scattered line emission inan outflowing gas. Yet we have previously shown that theincreasing He II/H α ratio indicated a net inflow of material.Therefore to account for the full spectroscopic evolution ofAT2017eqx requires both inflowing and outflowing gas, alongwith an appropriate geometry. We show a schematic of sucha model in Figure 13.In this scenario, the luminosity is generated in a smallregion, either from an accretion disk or at the intersectionpoint between colliding debris streams. Initially, the observersees (reprocessed) emission from a quasi-static atmosphereof bound debris (Jiang et al. 2016b). Its extent, ∼ – cm, is roughly proportional to the luminosity from theTDE engine. Following Roth et al. (2016), the dominantemission lines depend on the extent and optical depth ofthis layer. Lines from this region are broadened by electronscattering, but centered at their rest-frame wavelength. MNRAS000 , 1–17 (2019) M. Nicholl et al.
Above and below the plane of the disk, or parallel tothe streams, an outflow forms – either from a disk wind(Metzger & Stone 2016), or material on unbound trajecto-ries (Jiang et al. 2016b). This produces emission lines witha net velocity shift (Roth & Kasen 2018). The crucial pointis that whether the observer sees outflowing gas depends onwhether their line of sight is obstructed by the envelope.But even for obstructed sight lines, outflows can eventuallybe revealed as they expand or the envelope contracts. Ap-plying this to AT2017eqx, as the atmosphere shrinks andsuppresses the H I emission, we are exposed to more of theoutflowing material, causing the blueshift of He II. Thus thismodel naturally accounts for why an evolving He II/H α ratiois associated with a late-onset blueshift. Our model for AT2017eqx suggests that for TDEs moregenerally, viewing angle may be the primary determinantas to whether we see inflowing or outflowing gas, the ra-dial extent of which is important in setting the line ra-tios. Interpreting the spectroscopic diversity of TDEs as amanifestation of viewing similar sources from different an-gles evokes the unified model for active galactic nuclei (An-tonucci 1993), in which the diverse observational propertiesof AGN depend on our sight-line towards any associated jet,torus or broad/narrow-line regions. The term ‘unified model’has recently been applied to TDEs by Dai et al. (2018),who showed using simulations how the diversity of observedTDE X-ray/optical ratios, temperatures and jets may varyas a function of viewing angle. While our interpretation ofAT2017eqx and other TDEs has been developed indepen-dently, and applies primarily to the spectral lines, here wediscuss how our model complements that work towards aunified model of TDEs.In the accretion-powered paradigm, Dai et al. (2018)find that X-rays only escape for viewing angles close tothe pole. In our model, such sight-lines are associated withblueshifted emission in the early spectra (or possibly diskprofiles for very small angles). If most TDEs are pow-ered by accretion, there should exist a correlation betweenblueshifted line profiles and detectable X-ray emission. Inthe case of stream-stream collisions, material may still ulti-mately accrete onto the SMBH and produce X-rays, but be-cause the optical flare occurs off-center, we would expect nocorrelation between X-rays and blueshifted lines. Correlat-ing the X-ray and optical properties of TDEs can thereforebe used to test our proposed model, and possibly provide away to determine whether most TDEs are powered by ac-cretion or stream-stream collisions (see also Pasham et al.2017).There are several complications to this picture, suchas the non-spherical geometry of real TDEs, and the factthat powerful outflows can also produce X-rays (at least insome relativistic TDEs). Moreover, some TDEs show twolight curve maxima, or a far-UV excess several years afterdisruption, which may indicate both accretion and streamcollisions are at work (Leloudas et al. 2016; Wevers et al.2019b; van Velzen et al. 2018). Searching for this correlationwill therefore require large statistical samples of TDEs, butthese will be provided soon by current and next-generationsurveys.
The schematic shown in Figure 13 implies three possiblescenarios for a typical TDE spectrum:(i) The TDE is viewed approximately parallel to the disk(A) or orbital plane (B). An observer sees only reprocessedoptical radiation from the envelope, with lines centered closeto zero velocity. The ratio of He II/H α depends on the op-tical depth in this material, which may evolve as the debrisexpands or contracts in response to heating from below.(ii) The TDE is viewed perpendicular to the disk (A) ororbital plane (B). This observer can see outflowing material,and spectral lines will have a net blueshift. If the powersource is accretion onto the SMBH, and the outflow carvesa cavity in the envelope, this could also be observed as anX-ray TDE.(iii) The TDE is viewed at an intermediate angle. Thisobserver will most likely see only the envelope initially, butif this layer contracts, the outflow on the near side can be re-vealed leading to blue-shifted emission lines. X-rays may notbecome visible until much later. This scenario may explainthe spectroscopic evolution seen in AT2017eqx.Here we consider some other TDEs within this context.Spectra have been obtained from the Open TDE Cata-log (Guillochon et al. 2017) and Weizmann Interactive Su-pernova Data Repository (WISeREP) (Yaron & Gal-Yam2012).PS1-10jh, one of the earliest optical TDEs, showed onlyHe II lines, but in this case without a blueshift even at 250days. No X-rays were detected despite deep observations.Our model naturally accounts for this, as even if an accre-tion disk did form promptly, it may never become visibledue to an approximately side-on viewing angle through theatmosphere ( θ ∼ ◦ in Figure 13), in agreement with otherstudies (Dai et al. 2018).ASASSN-14ae (Holoien et al. 2014; Brown et al. 2016)initially shows only hydrogen lines in its spectrum, but itgradually develops a He II line over time, while H α be-comes weaker. By 93 days, He II is stronger than H α .Thus ASASSN-14ae evolves from a H-strong to a mixedH/He spectrum, as shown in Figure 12, analogous to howAT2017eqx evolves from a H/He to a He-strong spectrum.Therefore this is another event consistent with a contractingenvelope. ASASSN-14ae did not exhibit X-ray emission.iPTF16fnl showed H I and He II in its early spectra(Blagorodnova et al. 2017; Brown et al. 2018), centered closeto their rest-frame wavelengths, and the X-ray/optical ratiowas constrained to be < − (consistent with backgroundfluctuations; Auchettl et al. 2017; Blagorodnova et al. 2017).Along with these similarities to AT2017eqx, this event mayalso show a comparable evolution in the He II / H α ratio, in-dicating the atmosphere could be contracting in a similarmanner. However, no blueshifts are observed, suggesting aviewing angle closer to side-on.ASASSN-15oi showed blueshifted He II around maxi-mum light (Figure 12), and X-ray emission that graduallyincreased in time as the optical light faded (Holoien et al.2016b; Gezari et al. 2017; Holoien et al. 2018a). In ourmodel, seeing the blueshifted emission at early times im-plies a viewing angle close to the direction of the outflow, soif a disk had formed promptly we may have expected that MNRAS , 1–17 (2019)
TDE changes its spots this event would be X-ray bright from the beginning, ratherthan slowly increasing over a year. Thus if our proposedgeometry applies to ASASSN-15oi, this could support theinterpretation Gezari et al. (2017): that the X-rays were asign of gradual disk formation, and that the earlier opticalemission was therefore collisional.ASASSN-14li is one of the best-studied H- and He-strong TDEs (Holoien et al. 2016a; Brown et al. 2017). Theoptical lines in this event are consistent with their rest-framewavelengths (though He II may show a slight offset due toblending with N III; Leloudas et al. 2019). This event wasalso detected in X-rays and radio, with X-ray absorptionlines and radio luminosity corresponding to a velocity of upto ∼ . c (van Velzen et al. 2016; Alexander et al. 2016; Karaet al. 2018). However, other X-ray and UV lines are seen atmuch lower velocities, of a few hundred km s − (Miller et al.2015; Cenko et al. 2016), implying material with a rangeof velocities. Based on cross-correlating the temporal evolu-tion of the emission in the X-ray, optical and radio regimes,Pasham et al. (2017) and Pasham & van Velzen (2018) ar-gue that the X-rays are produced by an accretion disk thatmodulates the radio jet, while the optical emission comesfrom a stream intersection region further away. In the con-text of our model, seeing early-time X-ray emission withonly low-velocity spectral lines would appear to be possibleonly in case B, supporting the collsional picture preferredby Pasham & van Velzen (2018).Overall, it appears that the limited sample of op-tical TDEs can be accommodated within a picture likethat presented in Figure 13: PS1-10jh, ASASSN-14ae, andiPTF16fnl are consistent with either accretion or stream col-lisions, but ASASSN-15oi and ASASSN-14li seem to favourthe latter. However, to consistently explain the X-ray andoptical properties together likely requires a combination ofboth processes on different timescales. This may make itdifficult to identify the suggested correlation between X-rayemission and blueshifts until much larger TDE samples areavailable. A surprisingly large fraction of TDEs have been found in aspecific class of quiescent Balmer-strong absorption galaxies(Arcavi et al. 2014; French et al. 2016; Graur et al. 2018;Law-Smith et al. 2017), defined spectroscopically by thepresence of Balmer lines in absorption and a lack of nebularemission lines. This combination signifies that star forma-tion was likely significant up to ∼ Gyr ago but has nowlargely ceased, leaving behind A type stars that dominatethe stellar light (whereas the ionizing O and B stars havealready died).From our host galaxy spectrum, we measure a Lickindex (the equivalent width of H δ using line and contin-uum bandpasses defined by Worthey & Ottaviani 1997)H δ A = . ˚A. We also find that H α is visible only in absorp-tion, with an equivalent width of 3.1 ˚A. This combinationsatisfies the cut used by French et al. (2016), encapsulating75% of TDE hosts known at the time, and only 2% of SDSSgalaxies. Using the largest TDE sample to date, a more re-cent study by Graur et al. (2018) found that the fractionof such host galaxies among optical TDEs is ∼ , which log(Z/Z (cid:1) ) . Observed wavelength ( 𝜇 m) log(Z/Z (cid:1) ) = − +0.48-0.42 log(SFR ) = − +0.70-1.17 𝜈 F 𝜈 ( e r g s - c m - ) -13 𝜒 Photometry, best fit Spectrum (median) Observed z=0.11 best-fit 𝜒 /N phot =1.29 log(M ∗ /M (cid:1) ) =9.36 +0.08-0.10 log(SFR )log(M ∗ /M (cid:1) )t lookback (Gyr)
10 3 1 0 . . . SF R ( M (cid:1) y r - ) .
00 9 .
30 9 .
45 9 . − − − − − . − . − . . Figure 14.
Fit to the archival SED of the AT2017eqx host galaxyusing prospector (Leja et al. 2017). The lower panels show thestar-formation history, and posteriors for stellar mass, present-day SFR, and metallicity. The recent decline in SFR is consistentwith other TDE hosts. is still a significant overabundance. The host of AT2017eqxadds to this over-representation of Balmer-strong absorptiongalaxies in the TDE host population.The galaxy is also detected in a number of sky sur-veys with public catalogs. We retrieved u g riz magnitudesfrom SDSS DR14 (Abolfathi et al. 2018), g riz y magnitudesfrom PanSTARRS DR1 (Flewelling et al. 2016), and the W magnitude from WISE (Wright et al. 2010), and fit the re-sultant host SED using prospector (Leja et al. 2017) toderive physical parameters such as stellar mass and star-formation rate. The code includes the effects of stellar andnebular emission, metallicity, dust reprocessing, and a non-parametric star-formation history, within a nested samplingframework.The galaxy SED and prospector fitting results areshown in Figure 14. A small offset is visible between theSDSS and PS1 photometry, but the model well captures theSED shape. We find a stellar mass log ( M ∗ / M (cid:12) ) = . ± . ,with low star-formation rate and metallicity. The star-formation history is of particular interest: this drops by anorder of magnitude over the past Gyr, consistent with ex-pectations for the galaxy spectral type. Using the scalingrelation between bulge stellar mass and SMBH mass fromKormendy & Ho (2013), and assuming that M ∗ , bulge = M ∗ for an elliptical galaxy, we estimate a black hole mass of log ( M BH / M (cid:12) ) ≈ . , consistent with the results of fitting theTDE light curve.The SMBH mass we infer for AT2017eqx is typical foroptical TDE host galaxies, ∼ M (cid:12) , as measured from stel- MNRAS000
Fit to the archival SED of the AT2017eqx host galaxyusing prospector (Leja et al. 2017). The lower panels show thestar-formation history, and posteriors for stellar mass, present-day SFR, and metallicity. The recent decline in SFR is consistentwith other TDE hosts. is still a significant overabundance. The host of AT2017eqxadds to this over-representation of Balmer-strong absorptiongalaxies in the TDE host population.The galaxy is also detected in a number of sky sur-veys with public catalogs. We retrieved u g riz magnitudesfrom SDSS DR14 (Abolfathi et al. 2018), g riz y magnitudesfrom PanSTARRS DR1 (Flewelling et al. 2016), and the W magnitude from WISE (Wright et al. 2010), and fit the re-sultant host SED using prospector (Leja et al. 2017) toderive physical parameters such as stellar mass and star-formation rate. The code includes the effects of stellar andnebular emission, metallicity, dust reprocessing, and a non-parametric star-formation history, within a nested samplingframework.The galaxy SED and prospector fitting results areshown in Figure 14. A small offset is visible between theSDSS and PS1 photometry, but the model well captures theSED shape. We find a stellar mass log ( M ∗ / M (cid:12) ) = . ± . ,with low star-formation rate and metallicity. The star-formation history is of particular interest: this drops by anorder of magnitude over the past Gyr, consistent with ex-pectations for the galaxy spectral type. Using the scalingrelation between bulge stellar mass and SMBH mass fromKormendy & Ho (2013), and assuming that M ∗ , bulge = M ∗ for an elliptical galaxy, we estimate a black hole mass of log ( M BH / M (cid:12) ) ≈ . , consistent with the results of fitting theTDE light curve.The SMBH mass we infer for AT2017eqx is typical foroptical TDE host galaxies, ∼ M (cid:12) , as measured from stel- MNRAS000 , 1–17 (2019) M. Nicholl et al. lar velocity dispersions by Wevers et al. (2019a). The samestudy found that X-ray selected TDEs have a flatter distri-bution in M BH and typically show a smaller emitting region,consistent with an accretion disk rather than an inflated at-mosphere. This may indicate that the dominant emissionsource can vary depending on the SMBH mass.The host of AT2017eqx has a lower stellar mass andstar-formation rate (SFR) than any of the galaxies studiedby Law-Smith et al. (2017), with the possible exception ofthe host of RBS 1032 (Maksym et al. 2014), whose origin as aTDE has been questioned (Ghosh et al. 2006). Typical TDEhosts have a SFR that is ∼ . dex below the star-formingmain sequence for a given mass. The host of AT2017eqx issimilar, though with a slightly larger offset of ≈ dex. Thisis consistent with the positions of other quiescent, Balmerstrong galaxies (French et al. 2016).TDE host galaxies also tend to have centrally-concentrated mass distributions (Graur et al. 2018; Law-Smith et al. 2017), with an average S´ersic index of . + . − . ,significantly higher than typical galaxies in the same massrange (Law-Smith et al. 2017). We retrieved an r -band im-age of the host of AT2017eqx from PanSTARRS DR1, andfit the light profile using galfit (Peng et al. 2002). We findan excellent fit with a S´ersic index of only 0.7 (Figure 15), in-dicating a less sharply-peaked light distribution than typicalTDE hosts.Graur et al. (2018) showed that TDE hosts generallyhave a stellar mass surface density log ( Σ M ∗ /( M (cid:12) kpc − )) > , which is consistent with typical quiescent galaxies, buthigh for star-forming TDE hosts. This can be interpretedas evidence for a high density of stars around the centralSMBH, which could naturally lead to a higher rate of TDEs(Graur et al. 2018), as earlier proposed by (Stone & Metzger2016; Stone & van Velzen 2016). For consistency with thatstudy, we recalculate the stellar mass of the AT2017eqx hostgalaxy by fitting only the SDSS magnitudes, using lephare (Arnouts et al. 1999; Ilbert et al. 2009). We find log M ∗ = . ± . , consistent with our results from prospector .To convert mass into surface density, we use the half-lightradius from our S´ersic fit: . (cid:48)(cid:48) = . kpc, giving a surfacedensity log ( Σ M ∗ ) = . ± . .We plot this compared to other TDE hosts in Figure 15.AT2017eqx resides in one of the faintest galaxies for knownTDEs, and has an unusually low surface mass density. OnlyPS1-10jh has a host with comparable Σ M ∗ . However, thisgalaxy had a Petrosian half-light radius (from SDSS), ratherthan S´ersic. Taking the Petrosian radius of the AT2017eqxhost from SDSS, and applying the conversion R = R petro − . (cid:48)(cid:48) (Graur et al. 2018), we find an even lower stellar surfacemass density of log ( Σ M ∗ ) = . ± . , significantly offset fromother TDE hosts.The low surface mass density compared to other TDEhosts is consistent with the lower S´ersic index. The sur-face mass density is also low compared to average quies-cent galaxies, but is typical for star-forming galaxies in thevolume-weighted sample from Graur et al. (2018). Takingtheir empirical relation between TDE rate and Σ M ∗ , we findthat (cid:46) of TDEs are expected to occur in a galaxy withthis surface density. We therefore conclude that although thehost of AT2017eqx is somewhat unusual for TDEs, it is notoverwhelmingly improbable to find a TDE in such a galaxy. l og M ( M kp c - ) PS1-10jh Sersic R Petrosian R AllSersic
PS1 r-band GALFIT model Residual
Figure 15.
Stellar mass surface density within the half-light ra-dius ( R ) versus r -band magnitude for the host of AT2017eqxand a TDE host comparison sample (Graur et al. 2018).AT2017eqx resides in a galaxy that is significantly less centrally-concentrated than typical TDE hosts, with only PS1-10jh fallingin a similar part of the plot. The bottom panels show the GALFITmodel used to determine the S´ersic index and half-light radius. We have presented an in-depth study of a new TDE,AT2017eqx, discovered by the PanSTARRS Survey for Tran-sients and ATLAS. We followed up this event with opticalphotometry and spectroscopy, UV photometry from
Swift ,X-ray imaging and
Chandra and radio observations with theVLA. The SED maintains a roughly constant colour temper-ature of (cid:38) , K for at least 150 days.Non-detections in the radio indicate that AT2017eqxdid not launch a relativistic jet, and constrain the luminosityof any slower outflow to at most the level seen in ASASSN-14li. Our X-ray limits are among the deepest for any TDEto date. If an accretion disk formed promptly after the dis-ruption, the system would have to remain optically thick toX-rays for at least 500 days.Modelling of the UV and optical light curve with mos-fit suggests that AT2017eqx resulted from the complete dis-ruption of a solar-mass star by a black hole of (cid:38) . M (cid:12) ,with no significant viscous delay, similar to other UV-opticalTDEs (Mockler et al. 2019). This SMBH mass is consistentwith the observed properties of the host galaxy, with SEDfitting indicating a host stellar mass ≈ . M (cid:12) . The in-ferred star-formation history, and analysis of a galaxy spec-trum, show that this is yet another Balmer-strong absorp-tion galaxy hosting a TDE. However, it is one of the leastmassive TDE hosts to date, and the stellar mass surface den- MNRAS , 1–17 (2019)
TDE changes its spots sity is relatively low compared to typical TDE hosts (Grauret al. 2018).The most important new results come from the spec-troscopic evolution of AT2017eqx. In the first months fol-lowing the light curve maximum, the spectrum shows broademission lines from He II λ and the Balmer series, withsome evidence for N III Bowen flourescence (Blagorodnovaet al. 2019), and He I in the earliest spectrum. The H Iand He II lines initially have widths of (cid:38) , km s − , butare centered close to their rest-frame wavelengths. This re-sembles other TDEs such as ASASSN-14li, iPTF16axa andAT2018dyb. However, between 60-100 days after maximum,the H I lines disappear, while He II develops a blueshift of ∼ -8000 km s − , resembling He-strong TDEs such asPTF09ge and ASASSN-15oi. Such a stark transition has notbeen seen in TDEs before, and it confirms theoretical argu-ments that a lack of H I lines in the spectrum of a TDE doesnot indicate a low abundance of hydrogen in the disruptedstar.We propose that the evolution can be explained if thedebris around AT2017eqx consist of a quasi-spherical enve-lope, and a relatively narrow outflow from a disk wind orunbound debris. Viewing from an intermediate angle (i.e.neither parallel nor perpendicular to the outflow), we ini-tially see emission lines of H I and He II from the envelopeat zero velocity. As this layer contracts, either due to sim-ple fallback or in response to decreasing radiation pressure,the increasing optical depth suppresses hydrogen emission.At the same time, we see more of the outflow, which pro-duces the observed blueshift in the He II line at late times.The picture we have proposed for AT2017eqx is compatiblewith many TDEs from the current available samples, thoughthese are limited in size. It has recently been shown thatmany TDEs exhibit outflow signatures in their UV spectra(Hung et al. 2019). Future work should determine whetherit can also be applied other TDEs with featureless spectra orabsorption lines (Cenko et al. 2012a; Chornock et al. 2014;Leloudas et al. 2016; Blanchard et al. 2017), or with disk-likeline profiles (Holoien et al. 2018b).Furthermore, determining whether our model appliesto both X-ray weak and strong TDEs will crucial to un-derstanding TDE physics and geometry. We suggest thatsearching for a correlation between X-rays and blueshiftedspectral lines in TDEs is a promising avenue to distinguishbetween reprocessed accretion power and stream-stream col-lisions as the dominant power source. However, this pic-ture quickly becomes more complicated if many collision-powered optical TDEs also form X-ray accretion disks onrelatively short timescales. Identifying such a correlation be-tween X-ray and optical properties will therefore require amuch larger statistical sample, but this can soon be pro-vided by the many ongoing and planned sky surveys such asASASSN, ATLAS, PanSTARRS, ZTF and LSST. ACKNOWLEDGEMENTS
We thank an anonymous referee for many helpful com-ments that improved this paper. Thanks to Yuri Beletskyfor IMACS observing, Jabran Zahid for help with the galaxyanalysis, Jorge Anais, Jaime Vargas, Abdo Campillay andNahir Mu˜noz Elgueta for Swope observing, and Dave Coul- ter for writing the Swope scheduling software. We also thankJoel Aycock, Percy Gomez, and Yen-Chen Pan for assistancewith Keck/LRIS observations. M.N. is supported by a RoyalAstronomical Society Research Fellowship. The BergerTime-Domain Group is supported in part by NSF grantAST-1714498 and NASA grant NNX15AE50G. We acknowl-edge Chandra Award 20700239. K.D.A. acknowledges NASAHubble Fellowship grant HST-HF2-51403.001. ATLAS ac-knowledges NASA grants NN12AR55G, 80NSSC18K0284,and 80NSSC18K1575. S.J.S. acknowledges STFC GrantsST/P000312/1 and ST/N002520/1. O.G. and J.L. are sup-ported by NSF Astronomy and Astrophysics Fellowshipsunder awards AST-1602595 and AST-1701487. The UCSCteam is supported in part by NASA grant NNG17PX03C,NSF grant AST-1518052, the Gordon & Betty Moore Foun-dation, the Heising-Simons Foundation, and by a fellowshipfrom the David and Lucile Packard Foundation to R.J.F.Data were obtained via the Swift archive, the SmithsonianAstrophysical Observatory OIR Data Center, the MMT Ob-servatory of the Smithsonian Institution and the Univer-sity of Arizona, and Las Campanas Observatory. NRAO isa facility of the NSF operated by Associated Universities,Inc. ATLAS products are made possible by the Universityof Hawaii, Queen’s University Belfast, the Space TelescopeScience Institute, and the South African Astronomical Ob-servatory. ome of the data presented herein were obtained atthe W. M. Keck Observatory, which is operated as a scien-tific partnership among the California Institute of Technol-ogy, the University of California, and NASA; the observa-tory was made possible by the generous financial support ofthe W. M. Keck Foundation. The authors wish to recognizeand acknowledge the very significant cultural role and rever-ence that the summit of Mauna Kea has always had withinthe indigenous Hawaiian community. We are most fortunateto have the opportunity to conduct observations from thismountain.
REFERENCES
Abolfathi B., et al., 2018, ApJS, 235, 42Alam S., et al., 2015, ApJS, 219, 12Alexander K. D., Berger E., Guillochon J., Zauderer B. A.,Williams P. K. G., 2016, ApJ, 819, L25Alexander K. D., Wieringa M. H., Berger E., Saxton R. D., Ko-mossa S., 2017, ApJ, 837, 153Antonucci R., 1993, ARA&A, 31, 473Arcavi I., et al., 2014, ApJ, 793, 38Arnouts S., Cristiani S., Moscardini L., Matarrese S., Lucchin F.,Fontana A., Giallongo E., 1999, MNRAS, 310, 540Auchettl K., Guillochon J., Ramirez-Ruiz E., 2017, ApJ, 838, 149Becker A., 2015, HOTPANTS: High Order Transform of PSFANd Template Subtraction, Astrophysics Source Code Li-brary (ascl:1504.004)Benjamin R. A., Skillman E. D., Smiths D. P., 1999, ApJ, 514,307Berger E., Zauderer A., Pooley G. G., Soderberg A. M., Sari R.,Brunthaler A., Bietenholz M. F., 2012, ApJ, 748, 36Bilir S., Karaali S., Ak S., daˇgtekin N. D., ¨Onal ¨O., Yaz E.,Co¸skunoˇglu B., Cabrera-Lavers A., 2011, MNRAS, 417, 2230Blagorodnova N., et al., 2017, ApJ, 844, 46Blagorodnova N., et al., 2019, ApJ, 873, 92Blanchard P. K., et al., 2017, ApJ, 843, 106Bloom J. S., et al., 2011, Science, 333, 203MNRAS000
Abolfathi B., et al., 2018, ApJS, 235, 42Alam S., et al., 2015, ApJS, 219, 12Alexander K. D., Berger E., Guillochon J., Zauderer B. A.,Williams P. K. G., 2016, ApJ, 819, L25Alexander K. D., Wieringa M. H., Berger E., Saxton R. D., Ko-mossa S., 2017, ApJ, 837, 153Antonucci R., 1993, ARA&A, 31, 473Arcavi I., et al., 2014, ApJ, 793, 38Arnouts S., Cristiani S., Moscardini L., Matarrese S., Lucchin F.,Fontana A., Giallongo E., 1999, MNRAS, 310, 540Auchettl K., Guillochon J., Ramirez-Ruiz E., 2017, ApJ, 838, 149Becker A., 2015, HOTPANTS: High Order Transform of PSFANd Template Subtraction, Astrophysics Source Code Li-brary (ascl:1504.004)Benjamin R. A., Skillman E. D., Smiths D. P., 1999, ApJ, 514,307Berger E., Zauderer A., Pooley G. G., Soderberg A. M., Sari R.,Brunthaler A., Bietenholz M. F., 2012, ApJ, 748, 36Bilir S., Karaali S., Ak S., daˇgtekin N. D., ¨Onal ¨O., Yaz E.,Co¸skunoˇglu B., Cabrera-Lavers A., 2011, MNRAS, 417, 2230Blagorodnova N., et al., 2017, ApJ, 844, 46Blagorodnova N., et al., 2019, ApJ, 873, 92Blanchard P. K., et al., 2017, ApJ, 843, 106Bloom J. S., et al., 2011, Science, 333, 203MNRAS000 , 1–17 (2019) M. Nicholl et al.
Bowen I. S., 1935, ApJ, 81, 1Bower G. C., Metzger B. D., Cenko S. B., Silverman J. M., BloomJ. S., 2013, ApJ, 763, 84Breeveld A. A., Landsman W., Holland S. T., Roming P., KuinN. P. M., Page M. J., 2011, in McEnery J. E., Racusin J. L.,Gehrels N., eds, American Institute of Physics Conference Se-ries Vol. 1358, American Institute of Physics Conference Se-ries. pp 373–376 ( arXiv:1102.4717 ), doi:10.1063/1.3621807Brooks S. P., Gelman A., 1998, Journal of computational andgraphical statistics, 7, 434Brown P. J., et al., 2009, AJ, 137, 4517Brown J. S., Shappee B. J., Holoien T. W. S., Stanek K. Z.,Kochanek C. S., Prieto J. L., 2016, MNRAS, 462, 3993Brown J. S., Holoien T. W. S., Auchettl K., Stanek K. Z.,Kochanek C. S., Shappee B. J., Prieto J. L., Grupe D., 2017,MNRAS, 466, 4904Brown J. S., et al., 2018, MNRAS, 473, 1130Burrows D. N., et al., 2011, Nature, 476, 421Cenko S. B., et al., 2012a, MNRAS, 420, 2684Cenko S. B., et al., 2012b, ApJ, 753, 77Cenko S. B., et al., 2016, ApJ, 818, L32Chambers K. C., et al., 2016, arXiv e-prints, p. arXiv:1612.05560Chornock R., et al., 2014, ApJ, 780, 44Dai L., McKinney J. C., Roth N., Ramirez-Ruiz E., Miller M. C.,2018, ApJ, 859, L20Dressler A., et al., 2011, PASP, 123, 288Eftekhari T., Berger E., Zauderer B. A., Margutti R., AlexanderK. D., 2018, ApJ, 854, 86Flewelling H. A., et al., 2016, arXiv e-prints, 1612.05243Foreman-Mackey D., Hogg D. W., Lang D., Goodman J., 2013,PASP, 125, 306French K. D., Zabludoff A. I., 2018, ApJ, 868, 99French K. D., Arcavi I., Zabludoff A., 2016, ApJ, 818, L21Gezari S., et al., 2012, Nature, 485, 217Gezari S., Cenko S. B., Arcavi I., 2017, ApJ, 851, L47Ghosh K. K., Suleymanov V., Bikmaev I., Shimansky S.,Sakhibullin N., 2006, MNRAS, 371, 1587Goodman J., Weare J., 2010, Communications in applied mathe-matics and computational science, 5, 65Graur O., French K. D., Zahid H. J., Guillochon J., Mandel K. S.,Auchettl K., Zabludoff A. I., 2018, ApJ, 853, 39Guillochon J., Manukian H., Ramirez-Ruiz E., 2014, ApJ, 783, 23Guillochon J., Parrent J., Kelley L. Z., Margutti R., 2017, ApJ,835, 64Guillochon J., Nicholl M., Villar V. A., Mockler B., Narayan G.,Mandel K. S., Berger E., Williams P. K. G., 2018, ApJS, 236,6Hills J. G., 1975, Nature, 254, 295Holoien T. W.-S., et al., 2014, MNRAS, 445, 3263Holoien T. W.-S., et al., 2016a, MNRAS, 455, 2918Holoien T. W.-S., et al., 2016b, MNRAS, 463, 3813Holoien T. W. S., Brown J. S., Auchettl K., Kochanek C. S.,Prieto J. L., Shappee B. J., Van Saders J., 2018a, MNRAS,480, 5689Holoien T. W.-S., et al., 2018b, arXiv e-prints, 1808.02890Hung T., et al., 2017, ApJ, 842, 29Hung T., et al., 2019, arXiv e-prints, p. arXiv:1903.05637Ilbert O., et al., 2009, ApJ, 690, 1236Inserra C., et al., 2018, MNRAS, 475, 1046Jiang N., Dou L., Wang T., Yang C., Lyu J., Zhou H., 2016a,ApJ, 828, L14Jiang Y.-F., Guillochon J., Loeb A., 2016b, ApJ, 830, 125Kankare E., et al., 2017, Nature Astronomy, 1, 865Kara E., Dai L., Reynolds C. S., Kallman T., 2018, MNRAS, 474,3593Kilpatrick C. D., et al., 2018, MNRAS, 473, 4805Kochanek C. S., 1994, ApJ, 422, 508Komossa S., 2002, in Schielicke R. E., ed., Reviews in Modern Astronomy Vol. 15, Reviews in Modern Astronomy. p. 27( arXiv:astro-ph/0209007 )Kormendy J., Ho L. C., 2013, ARA&A, 51, 511Law-Smith J., Ramirez-Ruiz E., Ellison S. L., Foley R. J., 2017,ApJ, 850, 22Leja J., Johnson B. D., Conroy C., van Dokkum P. G., Byler N.,2017, ApJ, 837, 170Leloudas G., et al., 2016, Nature Astronomy, 1, 0002Leloudas G., et al., 2019, arXiv e-prints, 1903.03120Levan A. J., et al., 2011, Science, 333, 199Levan A. J., et al., 2016, ApJ, 819, 51Lodato G., 2012, in European Physical JournalWeb of Conferences. p. 01001 ( arXiv:1211.6109 ),doi:10.1051/epjconf/20123901001Maksym W. P., Lin D., Irwin J. A., 2014, ApJ, 792, L29Mattila S., et al., 2018, Science, 361, 482McMullin J. P., Waters B., Schiebel D., Young W., Golap K.,2007, in Shaw R. A., Hill F., Bell D. J., eds, AstronomicalSociety of the Pacific Conference Series Vol. 376, AstronomicalData Analysis Software and Systems XVI. p. 127Metzger B. D., Stone N. C., 2016, MNRAS, pp 948–966Miller J. M., et al., 2015, Nature, 526, 542Mimica P., Giannios D., Metzger B. D., Aloy M. A., 2015, MN-RAS, 450, 2824Mockler B., Guillochon J., Ramirez-Ruiz E., 2019, ApJ, 872, 151Nicholl M., 2018, RNAAS, 2, 230Osterbrock D. E., Ferland G. J., 2006, Astrophysics of gaseousnebulae and active galactic nucleiPasham D. R., van Velzen S., 2018, ApJ, 856, 1Pasham D. R., Cenko S. B., Sadowski A., Guillochon J., StoneN. C., van Velzen S., Cannizzo J. K., 2017, ApJ, 837, L30Peng C. Y., Ho L. C., Impey C. D., Rix H.-W., 2002, AJ, 124,266Piran T., Svirski G., Krolik J., Cheng R. M., Shiokawa H., 2015,ApJ, 806, 164Rest A., et al., 2005, ApJ, 634, 1103Roth N., Kasen D., 2018, ApJ, 855, 54Roth N., Kasen D., Guillochon J., Ramirez-Ruiz E., 2016, ApJ,827, 3Saxton R. D., Read A. M., Komossa S., Lira P., Alexander K. D.,Wieringa M. H., 2017, A&A, 598, A29Schechter P. L., Mateo M., Saha A., 1993, PASP, 105, 1342Schlafly E. F., Finkbeiner D. P., 2011, ApJ, 737, 103Schmidt G. D., Weymann R. J., Foltz C. B., 1989, PASP, 101,713Stalder B., et al., 2017, ApJ, 850, 149Stone N. C., Metzger B. D., 2016, MNRAS, 455, 859Stone N. C., van Velzen S., 2016, ApJ, 825, L14Tonry J. L., et al., 2018, PASP, 130, 064505Wevers T., van Velzen S., Jonker P. G., Stone N. C., Hung T.,Onori F., Gezari S., Blagorodnova N., 2017, MNRAS, 471,1694Wevers T., Stone N. C., van Velzen S., Jonker P. G., HungT., Auchettl K., Gezari S., Onori F., 2019a, arXiv e-prints,1902.04077Wevers T., et al., 2019b, arXiv e-prints, 1903.12203Worthey G., Ottaviani D. L., 1997, ApJS, 111, 377Wright E. L., et al., 2010, AJ, 140, 1868Wyrzykowski (cid:32)L., et al., 2017, MNRAS, 465, L114Yaron O., Gal-Yam A., 2012, PASP, 124, 668Zauderer B. A., et al., 2011, Nature, 476, 425van Velzen S., 2018, ApJ, 852, 72van Velzen S., et al., 2011, ApJ, 741, 73van Velzen S., Frail D. A., K¨ording E., Falcke H., 2013, A&A,552, A5van Velzen S., et al., 2016, Science, 351, 62van Velzen S., Stone N. C., Metzger B. D., Gezari S., BrownT. M., Fruchter A. S., 2018, arXiv e-prints,MNRAS , 1–17 (2019)
TDE changes its spots van Velzen S., et al., 2019, ApJ, 872, 198This paper has been typeset from a TEX/L A TEX file prepared bythe author.MNRAS000
TDE changes its spots van Velzen S., et al., 2019, ApJ, 872, 198This paper has been typeset from a TEX/L A TEX file prepared bythe author.MNRAS000 , 1–17 (2019) M. Nicholl et al.
Table 1.
Log of photometric observationsMJD Phase (d) Magnitude Error Band Telescope MJD Phase (d) Magnitude Error Band Telescope57905.6 -14.5 20.14 0.18 o ATLAS 57978 50.9 20.03 0.34 U Swift/UVOT57911.6 -9.1 19.7 0.04 i PS1 57978 50.9 19.39 0.24 W1 Swift/UVOT57926.6 4.5 19.57 0.14 o ATLAS 57978 50.9 19.28 0.18 M2 Swift/UVOT57932.6 9.9 19.99 0.12 o ATLAS 57978 50.9 18.83 0.11 W2 Swift/UVOT57936.6 13.5 20.03 0.14 o ATLAS 57979.2 52 21.1 0.13 g Swope57938.5 15.3 20.09 0.03 i PS1 57979.2 52 21.16 0.14 r Swope57945 21.1 19.05 0.2 U Swift/UVOT 57979.2 52 21.25 0.13 i Swope57945 21.1 18.82 0.18 W1 Swift/UVOT 57981.4 53.9 20.71 0.21 B Keck/LRIS57945 21.1 18.45 0.13 M2 Swift/UVOT 57981.5 53.9 20.76 0.21 V Keck/LRIS57945 21.1 18.28 0.09 W2 Swift/UVOT 57981.4 53.9 20.82 0.19 R Keck/LRIS57951.3 26.8 20.47 0.05 g Magellan/IMACS 57981.5 53.9 21.01 0.18 I Keck/LRIS57951.3 26.8 20.96 0.04 r Magellan/IMACS 57997.4 68.4 21.54 0.13 g FLWO48/Kepcam57951.3 26.8 20.93 0.05 i Magellan/IMACS 57997.4 68.4 21.89 0.24 r FLWO48/Kepcam57951.3 26.8 21.31 0.05 z Magellan/IMACS 57997.4 68.4 21.31 0.2 i FLWO48/Kepcam57952.6 27.9 20.49 0.23 o ATLAS 58002 72.5 19.66 0.49 W1 Swift/UVOT57953 28.3 19.55 0.27 U Swift/UVOT 58002 72.5 19.01 0.28 M2 Swift/UVOT57953 28.3 19.06 0.2 W1 Swift/UVOT 58002 72.5 19.49 0.28 W2 Swift/UVOT57953 28.3 18.75 0.15 M2 Swift/UVOT 58006 76.1 19.56 0.29 M2 Swift/UVOT57953 28.3 18.33 0.09 W2 Swift/UVOT 58006 76.1 19.32 0.22 W2 Swift/UVOT57956.3 31.3 20.59 0.04 g Magellan/LDSS 58010.2 79.9 21.63 0.09 g FLWO48/Kepcam57956.3 31.3 20.88 0.08 r Magellan/LDSS 58012 81.5 19.62 0.24 M2 Swift/UVOT57956.3 31.3 21.12 0.11 i Magellan/LDSS 58012 81.5 19.57 0.21 W2 Swift/UVOT57956.3 31.3 20.92 0.35 z Magellan/LDSS 58012.2 81.7 21.63 0.26 g Swope57956.5 31.5 20.21 0.21 o ATLAS 58012.2 81.7 > > > > > > > + . mag to match the o − i colours inferred from spectroscopy Table 2.
Log of spectroscopic observationsMJD Phase (d) Telescope Instrument Grating Exposure (s) Airmass57933.4 10 MMT Blue Channel 300GPM 900 1.157953.2 28 Magellan IMACS G300-17.5 1500 2.157956.3 31 Magellan LDSS3c VPH-all 3 × × × Table 3.
Log of VLA radio observations. Upper limits correspond to 3 σ .MJD Phase (d) Frequency (GHz) Image RMS ( µ Jy) Flux ( µ Jy)57948.5 25 6.0 9.0 < < < < Combined 43 6.0 6.0 < Combined 43 21.7 17.0 < MNRAS , 1–17 (2019)
TDE changes its spots Table 4.
Log of X-ray observations. Upper limits correspond to 3 σ .MJD Phase Telescope Instrument Duration Count rate 0.3-10 keV flux(d) (ks) (s − ) (erg s − cm − )57162.5 XMM MOS-1 14.1 < . × < . × − < . × < . × − < . × < . × − < . × < . × − MNRAS000