Tidal disruption flares are coming

# The first tidal disruption flare in ZTF: from photometric selection to multi-wavelength characterization

###### Abstract

We report the independent discovery of the transient source AT2018zr during commissioning observations of the Zwicky Transient Facility (ZTF), the first tidal disruption event (TDE) found in this survey. The ZTF light curve of the TDE samples the rise-to-peak exceptionally well, with 50 days of and -band detections before the time of maximum light. We also present our multi-wavelength follow-up observations that were triggered by the discovery of this flare: the detection of a weak X-ray source ( erg s) and a stringent upper limit to the radio emission. The X-ray emission shows a thermal spectrum ( eV) and is two orders of magnitude fainter than the contemporaneous optical/UV blackbody luminosity. We use observations of 128 known active galactic nuclei (AGN) to assess the quality of the ZTF astrometry, finding a median host-flare distance of 02 for genuine nuclear flares. Using ZTF observations of variability from known AGN and SNe we show how these sources can be separated from TDEs. A combination of light curve shape, color, and location in the host galaxy can be used to select a clean TDE sample from multi-band optical surveys such as ZTF or LSST.

galaxies: nuclei, accretion, accretion disks, surveys
journal: ApJ{CJK*}

UTF8bkai

\correspondingauthor

Sjoert van Velzen

0000-0002-3859-8074]Sjoert van Velzen

0000-0003-3703-5154]Suvi Gezari

0000-0003-3124-2814]James C. A. Miller-Jones

0000-0002-6485-2259]Nathaniel Roth

0000-0003-1710-9339]Lin Yan

0000-0001-9676-730X]Sara Frederick

0000-0002-3168-0139]Matthew J. Graham

0000-0001-8018-5348]Eric C. Bellm

0000-0002-6540-1484]Thomas Kupfer

0000-0001-5390-8563]Shrinivas R. Kulkarni

0000-0003-2242-0244]Ashish Mahabal

0000-0002-8532-9395]Frank J. Masci

0000-0001-7648-4142]Ben Rusholme

0000-0001-6753-1488]Maayane T. Soumagnac

## 1 Introduction

Stars that pass within the tidal radius of a supermassive black hole are disrupted and a sizable fraction of the resulting stellar debris gets accreted onto the black hole. When this disruption occurs outside the black hole event horizon (Hills, 1975), the result is a luminous flare of thermal emission (Rees, 1988). These stellar tidal disruption flares provide a unique tool to study black hole accretion and jet formation (e.g., Giannios & Metzger, 2011; van Velzen et al., 2011; Tchekhovskoy et al., 2014; Coughlin & Begelman, 2014; Piran et al., 2015; Pasham & van Velzen, 2018).

Optical transient surveys currently dominate the discovery of tidal disruption events (TDEs); about a dozen candidates have been found to date (for a recent compilation see van Velzen, 2018). All-sky X-ray surveys provide a second avenue for discovery (e.g., Saxton et al., 2017). By late 2019, the eROSITA mission (Merloni et al., 2012) should significantly increase the number of TDEs discovered via their X-ray emission (Khabibullin et al., 2014).

The observed blackbody radii of known TDEs (Gezari et al., 2009) suggest the soft X-ray photons of these flares originate from the inner part of a newly-formed accretion disk ( cm), while the optical photons are produced at much larger radii,  cm. Only a handful of optically selected TDEs have received sensitive X-ray follow-up observations within the first few months of discovery. So far, every case has been different; optically selected TDEs can be X-ray faint (Gezari et al., 2012), or have (Holoien et al., 2016b), or even show a decreasing optical to X-ray ratio (Gezari et al., 2017b).

Unification of optical and X-ray properties of TDEs is possible if the optical emission is powered by shocks from intersecting stellar debris streams (Piran et al., 2015) and the X-ray photons are produced when parts of the stream get deflected towards a few gravitational radii and accreted (Shiokawa et al., 2015; Krolik et al., 2016). In this scenario, TDEs with low X-ray luminosities can be explained as inefficiencies in this circularization process. If instead most of the stellar debris is able to rapidly form an accretion disk (Bonnerot et al., 2016; Hayasaki et al., 2016), the X-rays from the disk have to be reprocessed at larger radii (e.g., Loeb & Ulmer, 1997; Bogdanović et al., 2004; Strubbe & Quataert, 2009; Guillochon et al., 2014) to yield the observed optical emission. When the reprocessing layer is optically thick to X-rays, TDE unification is established via orientation (e.g., Metzger & Stone, 2016; Auchettl et al., 2017; Dai et al., 2018).

Discriminating between the different unification proposals might be possible by comparing the X-ray and optical TDE rate (e.g., if the reprocessing picture is correct, the ratio of the X-ray/optical rate yields the covering factor of the region that absorbs the X-ray photons). However this requires a larger body of TDEs since the current rate estimates are dominated by Poisson uncertainties (van Velzen & Farrar, 2014; Holoien et al., 2016b; Hung et al., 2018). Moreover, rates from different X-ray surveys (Donley et al., 2002; Esquej et al., 2007) are discrepant by a factor —although this discrepancy can be fixed by a more uniform analysis of the X-ray TDEs (Saxton et al. 2018 in prep).

Insight into the emission mechanism of TDEs can also be gained from detailed observations of individual events. A lag of the X-ray emission in a cross correlation analysis (Pasham et al., 2017) and decrease of with time (Gezari et al., 2017a) have both been interpreted as evidence against a reprocessing scenario. However these conclusions are not definitive since the X-ray diversity of TDEs has not yet been mapped out.

Clearly more TDEs with multi-wavelength observations are needed to make progress, which brings us to the topic of this paper: a new TDE detected in commissioning data from the Zwicky Transient Facility (ZTF). As discussed in Hung et al. (2018), ZTF has the potential to significantly increase the TDE detection rate; this first TDE candidate is encouraging evidence that this survey will reach the anticipated performance.

Like the Palomar Transient Factory (PTF; Law et al., 2009; Rau et al., 2009), ZTF uses the Samuel Oschin 48" Schmidt telescope at Palomar Observatory. The biggest improvement over PTF is the 47 deg field-of-view of the ZTF camera (Bellm et al., 2018). The public Northern Sky Survey of ZTF (Graham et al., 2018) began on 2018 March 17, and covers the entire visible sky from Palomar in both the and band every three nights (the Galactic Plane, , is covered with a one night cadence) to a typical depth of 20.5 magnitude. Using ZTF images a stream of alerts (Patterson et al., 2018) containing transients and variable sources is generated by IPAC (Masci et al., 2018). Besides the essential photometric information, this stream contains value-added products such as the quality of the subtraction (Mahabal et al., 2018) and probability that the alert is associated with a star versus a galaxy (Tachibana & Miller, 2018).

This paper is organized as follows. In Section 2 we present our observations of AT2018zr, the first TDE with ZTF observations. This source was also detected in Pan-STARRS and is known as PS18kh (Chambers et al., 2016); both ATLAS (Tonry et al., 2018) and ASAS-SN (Shappee et al., 2014) also obtained detections of AT2018zr, see Holoien et al. (2018). Here we present several new observations of this latest TDE: the ZTF light curve, XMM-Newton X-ray spectra, and VLA/AMI radio observations. The results from our HST campaign of UV spectroscopy and ground-based optical spectroscopic monitoring will be presented separately (Hung et al. in prep.). In Section 3 we compare AT2018zr to previous TDEs, SNe, and AGN. In Section 4 we show the astrometric quality of ZTF data for nuclear transients. In Section 5 we discuss the results.

We adopt a flat cosmology with and . All magnitudes are reported in the AB system (Oke, 1974).

## 2 Observations

### 2.1 Selection of nuclear flares in ZTF data

During the commissioning phase of the ZTF camera and IPAC alerts pipeline we assessed the quality of the alert stream, focusing on nuclear transients. A transient was considered nuclear when it had at least one detection with a distance between the location of the source in the reference frame and the location of the transient that was smaller than 06. We also required a match within 1" of a known Pan-STARRS (Chambers et al., 2016) galaxy, selected using a star-galaxy score (Tachibana & Miller, 2018) of gcore . To remove sources with a very small flux increase, we also required that the point spread function (PSF) magnitude in the difference image (agpsf , see \citealt{tp_Masci:18:ZTFDataSystem) and the PSF magnitude of the source in the reference image (agnr ) obey the relation: \verbagpsf agnr $<$ 1.5ag. To apply these filters and to obtain visual confirmation of the alerts we used the GROWTH Marshal (Kasliwal et al., 2018).

The objective of our commissioning effort was to understand the quality of the astrometry of nuclear transients; these results are presented in Section 4. We also obtained spectroscopic follow-up for a subset of nuclear transients, which led to the identification of the transient AT2018zr as a TDE candidate.

### 2.2 Brief history of AT2018zr

On 2018 March 6, the source ZTF18aabtxvd111We internally nicknamed this source ZTF-NedStark. was identified as a nuclear transient by our alert pipeline. Spectroscopic follow-up observations using SEDM (Blagorodnova et al., 2018) were obtained 3 days later; we measured a blue continuum without significant absorption or emission; a redshift and TDE classification was established using additional spectroscopic observations (Hung et al. 2018, in prep). Upon further investigation we noticed the reference frame was contaminated with light from the transient, which prohibited an earlier detection; after rebuilding the reference images and applying an image subtraction algorithm (Zackay et al., 2016) that is similar to the one used in the IPAC pipeline, we found the first ZTF detection was on 2018 February 7 (Fig. 1).

On March 24, Tucker et al. (2018, ATel 11473) reported spectroscopic follow-up and a TDE candidate classification for this source. This follow-up was based on the PS1 photometric observations reported to the Transient Name Server (TNS) on 2018 March 2.

### 2.3 Optical/UV observations

Observations with the Neil Gehrels Swift Observatory (Swift; Gehrels et al., 2004) started on 2018 March 27. We extracted the UVOT (Roming et al., 2005; Poole et al., 2008) flux with the help of the votsorce task, using an aperture radius of 5 arcsec.

The flux of the host galaxy in the UVOT bands was estimated by fitting a synthetic galaxy spectrum (Conroy et al., 2009; Conroy & Gunn, 2010) to the SDSS model magnitudes (Stoughton et al., 2002; Ahn et al., 2014), see Table 1. To construct the synthetic galaxy spectrum we adopt the default assumptions for the stellar parameters: a Kroupa 2001 initial mass function with stellar masses ; Padova isochrones and MILES spectral library Vazdekis et al. 2010. We assume an exponentially declining star formation rate (, with and as free parameters). We account for Galactic extinction by applying the Cardelli et al. (1989) extinction law with to the model spectrum. We also allow for extinction in the TDE host galaxy by modifying the model spectrum using a Calzetti et al. (2000) extinction law. The best-fit parameters for the formation history are  Gyr and  Gyr; the total stellar mass of the galaxy inferred for this model is .

Besides the Swift/UVOT and ZTF photometry, our light curve also includes P60/SEDM photometric data, host-subtracted using SDSS reference images (Fremling et al., 2016).

We correct the difference magnitude in each band for Galactic extinction,  mag (Schlegel et al., 1998), again assuming a Cardelli et al. (1989) extinction law with . The resulting light curve is shown in Fig. 1 and the photometry is available in Table 4.

By adding a blackbody spectrum to the synthetic host galaxy spectrum we find the best-fit temperature of the flare (Fig. 2). During the first 40 days of Swift monitoring the mean temperature was  K, and appears constant with an rms of only  K. In the last 20 days of monitoring (before the source moved out of the Swift visibility window), the temperature increased by a factor of (see also Holoien et al., 2018).

We searched for time lags between the optical and UV measurements taken with Swift using cross-correlations, following the procedure in Peterson et al. (1998). We first correct the light curves with a simple linear trend using a maximum-likelihood approach, and then use linear interpolation between data points in order to sample both UV and optical light curves on the same grid. We find no significant time lags between any combination of the the UV and optical bands at the 95% confidence level.

To search for outbursts in the years prior to the ZTF detection of AT2018zr, we applied a forced photometry method to the difference images from PTF and iPTF (Masci et al., 2017). We obtained 61 images, clustered at 6, 4, and 1 year before the current peak of the light curve. No prior variability was detected to a typical -band magnitude (5).

Using our ZTF observations of AT2018zr, we measure a mean angular distance between the host galaxy center and the flare of 012, or 162 pc. The rms of the offset, combining 46 offset measurements in Right Ascension and Declination and both - and -band detections, is 025. We thus conclude that the position of the flare is consistent with originating from the center of its host galaxy. Indeed, as we will see in Section 4, flares from AGN—which are expected to originate from the center of their galaxy—have a similar mean host-flare distance.

### 2.4 X-ray spectrum of AT2018zr

X-ray follow-up observations of AT2018zr were obtained using XMM-Newton (program 082204, PI: Gezari). Two epochs of XMM-Newton observations of the source were taken on 2018 April 11, and 2018 May 3 (see Table 2 for details). We reduced the XMM-Newton/pn data using the XMM-Newton Science Analysis System (SAS) and the newest calibration files. We started with the observation data files (ODFs) and followed standard procedures. Events were filtered with the conditions

ATTERN<= 4 and \verb FLAG==0 . We checked for high background flares (of which there were none). The source and background extraction regions are circular regions of radius 35". The response matrices were produced using {\sc rmfgen} and {\sc arfgen} in SAS. Spectral fitting was performed using
{\sc XS
EC v12.10 (Arnaud, 1996) with the c-statistic.

Both spectra are well-described by a single blackbody component ( eV) and Galactic absorption ( cm; Kalberla et al. 2005). No additional absorption (at the redshift of the source) was required. The 0.3-1 keV luminosity of the first epoch is . The temperature of the thermal component remained constant between the two observations, though the flux decreased by a factor of 2. In Table 2 we list the full details of the spectral parameters.

The results from our XMM analysis are in mild conflict with the non-thermal spectrum (photon index ) reported by Holoien et al. (2018) using the Swift/XRT data. The XRT measurements overlap with our XMM-Newton epochs, but have a lower signal-to-noise ratio and less spectral coverage at low energies. Given these limitations, a thermal spectrum is easily mistaken for a steep power-law. Indeed, the XRT flux reported by Holoien et al. (2018) is consistent with the XMM-Newton flux. Given the superior quality of the XMM observations, we conclude that the thermal nature of the X-ray spectrum of AT2018zr is firmly established.

### 2.5 Radio upper limits of AT2018zr

Radio observations of AT2018zr were obtained using the Arcminute Microkelvin Imager Large Array (AMI-LA; Zwart et al. 2008; Hickish et al. 2017) and the Karl G. Jansky Very Large Array (VLA; program 18A-373, PI: van Velzen). AMI observed on 2018 March 28, followed by the VLA on 2018 March 30 and April 28. The AMI data reduction was performed using the custom calibration pipeline reduce_dc (e.g. Perrott et al. 2013) with 3C 286 used as the primary calibrator, and J0745+3142 as the secondary calibrator. The same calibrators were used for both VLA observations. For the VLA data analysis, we made use of the NRAO pipeline products and flagged a few additional spectral channels after manual inspection for radio frequency interference (RFI). The calibrated visibilites were imaged using the Common Astronomy Software Application (CASA; McMullin et al., 2007) task clean, with natural weighting.

The source was not detected in any of our radio observations. Our most sensitive observation was the first VLA epoch, which yields a 3 upper limit to the 10 GHz radio luminosity (defined as ) of  erg s. Full details are listed in Table 3.

## 3 Comparison to known TDEs, supernovae, and AGN flares

To date, only four optical TDEs have a resolved rise to peak: PS1-10jh (Gezari et al., 2012), PS1-11af (Chornock et al., 2014), PTF-09ge222PTF-09djl and PTF-09axc (Arcavi et al., 2014) also have pre-peak detections, but no post-peak detections.(Arcavi et al., 2014) and iPTF-16fnl (Blagorodnova et al., 2017). The earliest ZTF detection of AT2018zr is 50 days before the peak of the light curve. Measurements of the rise time to peak are important since this parameter is expected to scale with the mass of the black hole that disrupted the star (Rees, 1988; Lodato et al., 2009). In Fig. 4 we show the light curves of the five TDEs with pre-peak detections. We compare both the rest-frame -band luminosity and the blackbody luminosity. The -correction and bolometric correction were estimated using the mean blackbody temperature of the post-peak observations (van Velzen et al., 2018), except for AT2018zr we used the temperature estimated from the nearest Swift/UVOT observation.

Our sample of nuclear flares from ZTF data is large enough to provide a meaningful comparison of the photometric properties of TDEs, SNe, and AGN flares. Selecting alerts that were discovered between May 1 and August 8, we obtain 840 sources. Of these, 331 can be classified as AGN using the Million quasar catalog (Flesch, 2015, v5.2.). Our sample contains 81 spectroscopically confirmed SNe (of which 62 SNe Type Ia) and 3 cataclysmic variables (CVs). The spectroscopic observations for SN/CV classifications were obtained by the ZTF collaboration, with SEDM serving as the main instrument; SNe typing was established using SNID (Blondin & Tonry, 2007).

To be able to compare TDEs, SNe, CVs, and AGN flares, we use a single light curve model to describe the observations of these transients. We compute the fading timescale of the light curve with respect to the rise time to the peak of the light curve using an exponential decay, while rise to peak is modelled using a Gaussian function:

 F(t)=Fpeak×{e−(t−tpeak)2/2σ2)t≤tpeake−(t−tpeak)/τt>tpeak (1)

This simplistic light curve model is a compromise between using specific models for each type of object (e.g., SN Ia templates) and using a completely non-parametric approach (e.g., interpolating the light curve to measure the FWHM). Using all available ZTF data (i.e., including upper limits prior to the first detection of the alert), we fit both the -band and -band simultaneously. To model the SED, we use a constant color for the observation before the peak of the light curve, and a linear change of color with time for the post-peak observations:

 (g−r)(t)={⟨(g−r)⟩t≤tpeak⟨(g−r)⟩+a(t−tpeak)t>tpeak (2)

To summarize, our light curve model has six free parameters: rise timescale (), fade timescale (), the time of peak (), flux at peak (), mean pre-peak color (), and rate of color change (, with units time).

For AT2018zr, we have no ZTF photometry post peak (because the flux of the transient is contained in the reference image of the public survey, which is not available for reprocessing until the first ZTF data release; Graham et al. 2018) and we instead use the Swift/UVOT photometry. For the other four TDEs with a resolved peak of the light curve, we include the published photometry333Obtained using the Open TDE Catalog, http://TDE.space. up to 100 days post-peak (when a longer temporal baseline is used, an exponential decay no longer provides a good description of the TDE light curve).

In Fig. 5 we show the result of applying our light curve model to AT2018zr, other known TDEs, as well as the AGN, SNe and CVs in our sample of nuclear flares. We discuss these results in Section 5.4.

## 4 Host-flare astrometry in ZTF data

In the previous section we found that our sample of nuclear flares contains about 10% spectroscopically confirmed SNe. As explained in Section 2.1, the sample of nuclear flares was constructed from alerts with at least one detection with a host-flare distance smaller than 06. However we expect that the mean host-flare distance can be measured with a precision that is better than 06, thus facilitating a better separation of nuclear flares (AGN/TDEs) and SNe.

To understand how the measurement of the offset scales with the signal-to-noise ratio of the detection, we collected ZTF measurements for a sample of known AGN. To obtain a good measurement of the mean and rms of the offset, we required at least 7 detections and a median host-flare distance , leaving 128 AGN. Under the assumption that the variability of these sources originates from the photometric center of their host galaxy, the observed rms of the offset yields the uncertainty, .

In Fig. 6 we show binned by the PSF magnitude of the flare in the difference image (). We find the following dependence between these two parameters

 σoffset=0.24+0.04(mdiff−20) arcsec. (3)

We can use this relation to compute the inverse-variance weighted mean of the offset. In Fig. 7 we show the weighted mean host-flare distance for the SNe, AGN, and unclassified flares in our nuclear flare sample (again using only sources with at least 7 detections). Multiple observations of the same flare lead to an increased accuracy on the mean host-flare distance, yielding a typical uncertainty of 02 of nuclear flares with a few tens of detections (cf. the peak of the AGN distribution in Fig. 7).

The measurements of the host-flare offset are not independent because each measurement depends on the same reference frame to yield the position of the host. To show the astrometric accuracy without the contribution of the reference frame, we also report the rms of the offset with respect to the median position of the flare in the difference image (Fig. 6).

## 5 Discussion

### 5.1 Origin of the thermal emission mechanism

AT2018zr is the fifth optical/UV-selected TDE with an X-ray detection, and only the third source that was detected within a few months of discovery. Its optical to X-ray flux ratio is similar to ASASSN-15oi (Fig. 8), yet in contrast to that TDE, increased between the two epochs of XMM-Newton observations. A longer temporal baseline is needed to confirm whether this high flux ratio persists or decreases to , as observed for most other optical TDEs with X-ray detections (Fig. 8).

The similarity of to ASASSN-15oi may lead one to consider that the low X-ray luminosity of AT2018zr is explained by a delay in the formation of the accretion disk (Gezari et al., 2017a). If instead disk formation is efficient and the optical emission can be explained by reprocessing (e.g. by bound-free absorption) of emission from the inner disk, the low X-ray luminosity can be explained by obscuration of this disk. If the observed X-ray plus optical emission is similar to the bolometric disk luminosity, the observed value of implies our line of sight to the disk lets through at most 1% of the X-ray photons.

The blackbody radius corresponding to the single temperature model of the X-ray spectrum is  cm for the first epoch of X-ray observations. This corresponds to the Schwarzschild radius () of a black hole with a mass of  . Such a low-mass black hole is not expected given the properties of the host galaxy of AT2018zr. If half of the stellar mass of the galaxy is in the bulge, the predicted black hole mass (Gültekin et al., 2009) is   (consistent with the black hole mass estimate of Holoien et al. 2018). If the observed X-ray photons originated from the inner part of an accretion disk, the intrinsic X-ray luminosity must be times higher to match the blackbody radius to the expected size of the inner disk. Part of this tension can be alleviated if the observed X-ray temperature is higher than the true temperature due to obscuration or a contribution of inverse Compton emission to the 0.3-1 keV spectrum. For example, if we adopt the upper limit to the intrinsic absorption from the X-ray spectral fit of , then the inferred blackbody temperature decreases by a factor of 2 and the unabsorbed X-ray luminosity increases by a factor of 1800, yielding a blackbody radius that is a factor 170 larger and of the same of order as the expected size of the inner disk.

Even after accounting for the potential effect of absorption on the X-ray spectrum, the X-ray blackbody radius from our XMM-Newton observations is two orders of magnitude smaller than the inner radius of of the elliptical disk model proposed by Holoien et al. (2018). While this extended elliptical disk could explain the properties of the optical emission lines, our X-ray observations suggest a small, compact accretion disk is present as well.

### 5.2 Interpretation of radio non-detection

The upper limit to the radio luminosity of AT2018zr is one order of magnitude lower than the observed radio emission of the TDE ASASSN-14li (van Velzen et al., 2016; Alexander et al., 2016; Bright et al., 2018). Currently, only the TDE iPTF-16fnl (Blagorodnova et al., 2017) has received radio follow-up observations close to the peak of the flare with a similar sensitivity. This source was also not detected at radio frequencies. However, this flare was exceptional, being the faintest and fastest fading TDE to date (see Fig. 4). The optical properties of AT2018zr, on the other hand, are similar to the mean properties of the current TDE sample (see Figs. 4 & 5 and Hung et al. 2017).

Our radio non-detection rules out the hypothesis that TDEs with a typical optical luminosity and fade timescale produce radio emission similar to ASASSN-14li (Fig. 9). However the X-ray luminosity of AT2018zr is two orders of magnitude lower than ASASSN-14li (Holoien et al., 2016b). If the radio luminosity scales linearly with the power of the accretion disk, as observed for ensembles of radio-loud quasars (Rawlings & Saunders, 1991; Falcke et al., 1999; van Velzen et al., 2015) and for ASASSN-14li (Pasham & van Velzen, 2018), the expected radio flux of AT2018zr would be too faint to be detectable.

Free-free absorption is unlikely to affect the 10 GHz flux of AT2018zr since it would require an unrealistically high electron density. For an electron temperature of  K, we require an emission measure (EM) of at least  cm pc (e.g., a mean electron density of  cm within one parsec) to yield a significant optical depth () for free-free absorption at 10 GHz (e.g., Condon, 1992). For higher electron temperatures, as expected for galaxy centers (Lazio et al., 1999), the lower limit on the EM would increase even further.

### 5.3 Rise timescale and black hole mass

While our current sample of TDEs with a resolved rise to peak is still small, we can start to search for the anticipated correlation between black hole mass and rise timescale (Rees, 1988; Lodato et al., 2009). To estimate the black hole mass we use the velocity dispersion measurements from Wevers et al. (2017) and Wevers et al. (2018, in prep) and the Gültekin et al. (2009) - relation. For the host galaxy of AT2018zr, a velocity dispersion measurement is not yet available and we adopt the black hole mass from the bulge mass (, see Section 5.1). To provide a more uniform comparison we also consider the relation between rise time and total stellar mass. The results are shown in Fig. 10.

Our measurement of the rise time uses a Gaussian function (Eq.1), which has no defined start time. To be able to compare our measurement to the predicted fallback timescale () we assume the disruption happened at (i.e., when the flux in the model light curve is just 1% of the flux at peak). In Fig. 10 we show the predicted rise time from the theoretical fallback time of Stone et al. (2013), for the disruption of a star with a mass of 1  and an impact parameter of unity,

 tfb=3.5×106(M∙106 M⊙)1/2 s. (4)

We find no correlation between the rise time and black hole mass or total galaxy mass. This could be considered surprising given that a correlation between the fade timescale and the black hole mass has been reported (Blagorodnova et al., 2017; Wevers et al., 2017). However, our results also show that the rise and fade timescales themselves appear to be uncorrelated (Fig. 5). It could be possible that the post-peak light curve provides a better tracer for the fallback rate—and thus a better mass estimate (Mockler et al., 2018)—compared to the rise to peak.

### 5.4 Photometric selection of TDEs in optical surveys

Using only 3 months of ZTF data, we confirm the conclusions from earlier TDE population studies (Gezari et al., 2009; van Velzen et al., 2011; Holoien et al., 2016a; Hung et al., 2017), showing that TDEs are a class of flares with a shared set of photometric properties.

Our work is the first to quantify the distribution of rise and fade timescales of TDEs. We find that most TDEs have both a longer rise time and fade time compared to SNe (Fig. 5, top panel). The TDE iPTF-16fnl is an interesting exception, displaying a rise timescale at the edge of the SN Ia distribution and a fade timescale that is faster compared to most SNe.

Using only pre-peak observations, effective removal of SNe Ia is possible by restricting to flares with a rise time () that is longer than  days. While AGN flares can have a wide range of rise/fade timescales, very few rise and fade within a few months. Only 5% of AGN in our sample of nuclear flares have rise and fade timescales that fall within the range spanned by known TDEs.

The largest contrast between TDEs and SNe is found when we consider the mean color and color change (Fig. 5). We see that TDEs cluster in the region of blue and constant colors (van Velzen et al., 2011). Near their peak, some SNe can be as blue as TDEs, yet these SNe also cool very fast.

We can conclude that a clean sample of TDEs can be selected from using a combination of photometric properties: the rise/fade timescale (Fig. 5, top panel), the flare’s color and its evolution (Fig. 5, bottom panel), and the location of the flare in the host galaxy (Fig. 7). While each of these metrics has exceptions (e.g., TDE from faint galaxies have large astrometric uncertainties on the host-flare offset, some TDE rise and fade rapidly), photometric selection will be unavoidable in the era of the Large Synoptic Survey Telescope (LSST). The TDEs detected by LSST will be too faint and too numerous ( per year; van Velzen et al., 2011) to use spectroscopic follow-up observations for classification.

Acknowledgments — Based on observations obtained with the Samuel Oschin Telescope 48-inch and the 60-inch Telescope at the Palomar Observatory as part of the Zwicky Transient Facility project, a scientific collaboration among the California Institute of Technology, the Oskar Klein Centre, the Weizmann Institute of Science, the University of Maryland, the University of Washington, Deutsches Elektronen-Synchrotron, the University of Wisconsin-Milwaukee, and the TANGO Program of the University System of Taiwan. Further support is provided by the U.S. National Science Foundation under Grant No. AST-1440341. We thank the National Radio Astronomy Observatory (NRAO) staff for the rapid scheduling of the VLA observations. NRAO is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. We thank the staff of the Mullard Radio Astronomy Observatory for their assistance in the operation of AMI. We acknowledge the use of public data from the Swift data archive. This research made use of Astropy, a community-developed core Python package for Astronomy (The Astropy Collaboration et al., 2018). S. Gezari is supported in part by NSF CAREER grant 1454816 and NSF AAG grant 1616566. M. M. Kasliwal acknowledges support by the GROWTH (Global Relay of Observatories Watching Transients Happen) project funded by the National Science Foundation PIRE (Partnership in International Research and Education) program under Grant No 1545949. N.R. acknowledges the support of a Joint Space-Science Institute prize postdoctoral fellowship. J.C.A.M.-J. is supported by an Australian Research Council Future Fellowship (FT140101082). This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 759194 - USNAC).

## References

• Ahn et al. (2014) Ahn, C. P., Alexandroff, R., Allende Prieto, C., et al. 2014, ApJS, 211, 17
• Alexander et al. (2016) Alexander, K. D., Berger, E., Guillochon, J., Zauderer, B. A., & Williams, P. K. G. 2016, ApJ, 819, L25
• Arcavi et al. (2014) Arcavi, I., Gal-Yam, A., Sullivan, M., et al. 2014, ApJ, 793, 38
• Arnaud (1996) Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series, Vol. 101, Astronomical Data Analysis Software and Systems V, ed. G. H. Jacoby & J. Barnes, 17
• Auchettl et al. (2017) Auchettl, K., Guillochon, J., & Ramirez-Ruiz, E. 2017, ApJ, 838, 149
• Bellm et al. (2018) Bellm, E. C., Kulkarni, S. R., Graham, M. J., et al. 2018, submitted to PASP
• Blagorodnova et al. (2017) Blagorodnova, N., Gezari, S., Hung, T., et al. 2017, ApJ, 844, 46
• Blagorodnova et al. (2018) Blagorodnova, N., Neill, J. D., Walters, R., et al. 2018, PASP, 130, 035003
• Blondin & Tonry (2007) Blondin, S., & Tonry, J. L. 2007, ApJ, 666, 1024
• Bogdanović et al. (2004) Bogdanović, T., Eracleous, M., Mahadevan, S., Sigurdsson, S., & Laguna, P. 2004, ApJ, 610, 707
• Bonnerot et al. (2016) Bonnerot, C., Rossi, E. M., Lodato, G., & Price, D. J. 2016, MNRAS, 455, 2253
• Bright et al. (2018) Bright, J. S., Fender, R. P., Motta, S. E., et al. 2018, MNRAS, 475, 4011
• Calzetti et al. (2000) Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682
• Cardelli et al. (1989) Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245
• Chambers et al. (2016) Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, ArXiv e-prints, arXiv:1612.05560
• Chornock et al. (2014) Chornock, R., Berger, E., Gezari, S., et al. 2014, ApJ, 780, 44
• Condon (1992) Condon, J. J. 1992, ARA&A, 30, 575
• Conroy & Gunn (2010) Conroy, C., & Gunn, J. E. 2010, ApJ, 712, 833
• Conroy et al. (2009) Conroy, C., Gunn, J. E., & White, M. 2009, ApJ, 699, 486
• Coughlin & Begelman (2014) Coughlin, E. R., & Begelman, M. C. 2014, ApJ, 781, 82
• Dai et al. (2018) Dai, L., McKinney, J. C., Roth, N., Ramirez-Ruiz, E., & Miller, M. C. 2018, ArXiv e-prints, arXiv:1803.03265
• Donley et al. (2002) Donley, J. L., Brandt, W. N., Eracleous, M., & Boller, T. 2002, AJ, 124, 1308
• Esquej et al. (2007) Esquej, P., Saxton, R. D., Freyberg, M. J., et al. 2007, A&A, 462, L49
• Falcke et al. (1999) Falcke, H., Bower, G. C., Lobanov, A. P., et al. 1999, ApJ, 514, L17
• Flesch (2015) Flesch, E. W. 2015, PASA, 32, e010
• Fremling et al. (2016) Fremling, C., Sollerman, J., Taddia, F., et al. 2016, A&A, 593, A68
• Gehrels et al. (2004) Gehrels, N., Chincarini, G., Giommi, P., et al. 2004, ApJ, 611, 1005
• Gezari et al. (2017a) Gezari, S., Cenko, S. B., & Arcavi, I. 2017a, ApJ, 851, L47
• Gezari et al. (2009) Gezari, S., Heckman, T., Cenko, S. B., et al. 2009, ApJ, 698, 1367
• Gezari et al. (2012) Gezari, S., Chornock, R., Rest, A., et al. 2012, Nature, 485, 217
• Gezari et al. (2017b) Gezari, S., Hung, T., Cenko, S. B., et al. 2017b, ApJ, 835, 144
• Giannios & Metzger (2011) Giannios, D., & Metzger, B. D. 2011, MNRAS, 416, 2102
• Graham et al. (2018) Graham, M., et al. 2018, pASP submitted
• Guillochon et al. (2014) Guillochon, J., Manukian, H., & Ramirez-Ruiz, E. 2014, ApJ, 783, 23
• Gültekin et al. (2009) Gültekin, K., Richstone, D. O., Gebhardt, K., et al. 2009, ApJ, 698, 198
• Hayasaki et al. (2016) Hayasaki, K., Stone, N., & Loeb, A. 2016, MNRAS, 461, 3760
• Hickish et al. (2017) Hickish, J., Razavi-Ghods, N., Perrott, Y. C., et al. 2017, ArXiv e-prints, arXiv:1707.04237
• Hills (1975) Hills, J. G. 1975, Nature, 254, 295
• Holoien et al. (2016a) Holoien, T. W.-S., Kochanek, C. S., Prieto, J. L., et al. 2016a, MNRAS, 463, 3813
• Holoien et al. (2016b) —. 2016b, MNRAS, 455, 2918
• Holoien et al. (2018) Holoien, T. W.-S., Huber, M. E., Shappee, B. J., et al. 2018, ArXiv e-prints, arXiv:1808.02890
• Hung et al. (2017) Hung, T., Gezari, S., Blagorodnova, N., et al. 2017, ApJ, 842, 29
• Hung et al. (2018) Hung, T., Gezari, S., Cenko, S. B., et al. 2018, ApJ, in press, arXiv:1712.04936
• Kalberla et al. (2005) Kalberla, P. M. W., Burton, W. B., Hartmann, D., et al. 2005, A&A, 440, 775
• Kasliwal et al. (2018) Kasliwal, M., et al. 2018, in preparation
• Khabibullin et al. (2014) Khabibullin, I., Sazonov, S., & Sunyaev, R. 2014, MNRAS, 437, 327
• Krolik et al. (2016) Krolik, J., Piran, T., Svirski, G., & Cheng, R. M. 2016, ApJ, 827, 127
• Kroupa (2001) Kroupa, P. 2001, MNRAS, 322, 231
• Law et al. (2009) Law, N. M., Kulkarni, S. R., Dekany, R. G., et al. 2009, PASP, 121, 1395
• Lazio et al. (1999) Lazio, T. J. W., Cordes, J. M., Anantharamaiah, K. R., Goss, W. M., & Kassim, N. E. 1999, in Astronomical Society of the Pacific Conference Series, Vol. 186, The Central Parsecs of the Galaxy, ed. H. Falcke, A. Cotera, W. J. Duschl, F. Melia, & M. J. Rieke, 441
• Lodato et al. (2009) Lodato, G., King, A. R., & Pringle, J. E. 2009, MNRAS, 392, 332
• Loeb & Ulmer (1997) Loeb, A., & Ulmer, A. 1997, ApJ, 489, 573
• Mahabal et al. (2018) Mahabal, A. A., et al. 2018, in preparation
• Masci et al. (2018) Masci, F. J., Laher, R. R., Rusholme, B., & Shupe, D. L. 2018, pASP submitted
• Masci et al. (2017) Masci, F. J., Laher, R. R., Rebbapragada, U. D., et al. 2017, PASP, 129, 014002
• McMullin et al. (2007) McMullin, J. P., Waters, B., Schiebel, D., Young, W., & Golap, K. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 376, Astronomical Data Analysis Software and Systems XVI, ed. R. A. Shaw, F. Hill, & D. J. Bell, 127
• Merloni et al. (2012) Merloni, A., Predehl, P., Becker, W., et al. 2012, ArXiv e-prints, arXiv:1209.3114
• Metzger & Stone (2016) Metzger, B. D., & Stone, N. C. 2016, MNRAS, 461, 948
• Mockler et al. (2018) Mockler, B., Guillochon, J., & Ramirez-Ruiz, E. 2018, ArXiv e-prints, arXiv:1801.08221
• Oke (1974) Oke, J. B. 1974, ApJS, 27, 21
• Pasham et al. (2017) Pasham, D. R., Cenko, S. B., Sadowski, A., et al. 2017, ApJ, 837, L30
• Pasham & van Velzen (2018) Pasham, D. R., & van Velzen, S. 2018, ApJ, 856, 14
• Patterson et al. (2018) Patterson, M. T., et al. 2018, in preparation
• Perrott et al. (2013) Perrott, Y. C., Scaife, A. M. M., Green, D. A., et al. 2013, MNRAS, 429, 3330
• Peterson et al. (1998) Peterson, B. M., Wanders, I., Horne, K., et al. 1998, PASP, 110, 660
• Piran et al. (2015) Piran, T., Svirski, G., Krolik, J., Cheng, R. M., & Shiokawa, H. 2015, ApJ, 806, 164
• Poole et al. (2008) Poole, T. S., Breeveld, A. A., Page, M. J., et al. 2008, MNRAS, 383, 627
• Rau et al. (2009) Rau, A., Kulkarni, S. R., Law, N. M., et al. 2009, PASP, 121, 1334
• Rawlings & Saunders (1991) Rawlings, S., & Saunders, R. 1991, Nature, 349, 138
• Rees (1988) Rees, M. J. 1988, Nature, 333, 523
• Roming et al. (2005) Roming, P. W. A., Kennedy, T. E., Mason, K. O., et al. 2005, Space Sci. Rev., 120, 95
• Saxton et al. (2017) Saxton, R. D., Read, A. M., Komossa, S., et al. 2017, A&A, 598, A29
• Schlegel et al. (1998) Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525
• Shappee et al. (2014) Shappee, B. J., Prieto, J. L., Grupe, D., et al. 2014, ApJ, 788, 48
• Shiokawa et al. (2015) Shiokawa, H., Krolik, J. H., Cheng, R. M., Piran, T., & Noble, S. C. 2015, ApJ, 804, 85
• Stone et al. (2013) Stone, N., Sari, R., & Loeb, A. 2013, MNRAS, 435, 1809
• Stoughton et al. (2002) Stoughton, C., Lupton, R. H., Bernardi, M., et al. 2002, AJ, 123, 485
• Strubbe & Quataert (2009) Strubbe, L. E., & Quataert, E. 2009, MNRAS, 400, 2070
• Tachibana & Miller (2018) Tachibana, Y., & Miller, A. 2018, pASP submitted
• Tchekhovskoy et al. (2014) Tchekhovskoy, A., Metzger, B. D., Giannios, D., & Kelley, L. Z. 2014, MNRAS, 437, 2744
• The Astropy Collaboration et al. (2018) The Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M., et al. 2018, ArXiv e-prints, arXiv:1801.02634
• Tonry et al. (2018) Tonry, J. L., Denneau, L., Heinze, A. N., et al. 2018, PASP, 130, 064505
• Tucker et al. (2018) Tucker, M. A., Huber, M., Shappee, B. J., et al. 2018, The Astronomer’s Telegram, 11473
• van Velzen (2018) van Velzen, S. 2018, ApJ, 852, 72
• van Velzen et al. (2015) van Velzen, S., Falcke, H., & Körding, E. 2015, MNRAS, 446, 2985
• van Velzen & Farrar (2014) van Velzen, S., & Farrar, G. R. 2014, ApJ, 792, 53
• van Velzen et al. (2011) van Velzen, S., Körding, E., & Falcke, H. 2011, MNRAS, 417, L51
• van Velzen et al. (2016) van Velzen, S., Mendez, A. J., Krolik, J. H., & Gorjian, V. 2016, ApJ, 829, 19
• van Velzen et al. (2018) van Velzen, S., Stone, N. C., Metzger, B. D., et al. 2018, ArXiv e-prints, arXiv:1809.00003
• van Velzen et al. (2011) van Velzen, S., Farrar, G. R., Gezari, S., et al. 2011, ApJ, 741, 73
• Vazdekis et al. (2010) Vazdekis, A., Sánchez-Blázquez, P., Falcón-Barroso, J., et al. 2010, MNRAS, 404, 1639
• Wevers et al. (2017) Wevers, T., van Velzen, S., Jonker, P. G., et al. 2017, MNRAS, 471, 1694
• Zackay et al. (2016) Zackay, B., Ofek, E. O., & Gal-Yam, A. 2016, ApJ, 830, 27
• Zwart et al. (2008) Zwart, J. T. L., Barker, R. W., Biddulph, P., et al. 2008, MNRAS, 391, 1545
\startlongtable
You are adding the first comment!
How to quickly get a good reply:
• Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
• Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
• Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters