Discovery and Early Evolution of ASASSN-19bt, the First TDE Detected by TESS

We obtained BVgri magnitudes of the host galaxy from the AAVSO Photometric All-Sky Survey Data Release 10 (APASS; Henden et al. 2015), JHK_S magnitudes from the Two Micron All-Sky Survey (2MASS), and W1 and W2 measurements from the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010) AllWISE data release. 2MASX J07001137−6602251 is at a decl. too far south to be observed by optical surveys such as the Sloan Digital Sky Survey (SDSS) and Pan-STARRS or radio surveys like FIRST and NVSS. It is not detected in archival data from, or was not previously observed by, Spitzer, Herschel, or the Hubble Space Telescope (HST).

Serendipitously, 2MASX J07001137−6602251 is located approximately 85 away from the BL Lac-type object PKS 0700−661, which was observed several times by Swift in 2009, 2010, 2014 and 2018 (target IDs 38456, 41619, and 83377). Due to the 170 × 170 field of view of Swift, 2MASX J07001137−6602251 was captured in several of these archival observations. While it is located at the edge of the field, making it difficult to use these archival images as image subtraction templates, we obtained archival Swift UVOT magnitudes by first summing all of the available data in each of the six UVOT filters using the HEAsoft software task uvotimsum, and then extracting counts from the combined images in a 50 radius aperture using the software task uvotsource, with a sky aperture of ∼400 radius to estimate and subtract the background counts. We converted the archival count rates to magnitudes and fluxes using the most recent UVOT calibration (Poole et al. 2008; Breeveld et al. 2010). The Swift UVOT, APASS, 2MASS, and WISE magnitudes of the host that we used to fit the host galaxy spectral energy distribution (SED) and estimate the host flux (see below) are shown in Table 1.

There were pre-event X-ray observations of 2MASX J07001137−6602251 by the ROSAT All-Sky Survey (Voges et al. 1999) circa 1990 and in the serendipitous Swift X-Ray Telescope (XRT; Burrows et al. 2005) observations. No source was detected by ROSAT to a 3σ upper limit of 4.7 × 10⁻³ counts s⁻¹ in the 0.3–2.0 keV range. Assuming a Γ = 1.75 power-law spectrum typical of an active galactic nucleus (AGN; e.g., Ricci et al. 2017) and an H i column density of 7.1 × 10²⁰ cm⁻², we derive a 3σ absorbed flux upper limit of ∼1.7 × 10⁻¹² erg cm⁻² s⁻¹ over the 0.3–10.0 keV energy range. This corresponds to an absorbed luminosity of 2.7 × 10⁴² erg s⁻¹, implying that the host does not harbor a strong AGN (e.g., Ricci et al. 2017). Interestingly, we detect weak (∼3σ) X-ray emission from the first Swift observation (ID 38450) in 2009, but a follow-up observation (also ID 38450) taken ∼20 days later does not show any significant X-ray emission. Further observations (IDs 41619 and 83377) taken in 2010, 2014 and 2018 show no evidence of X-ray emission.

The detection in the first Swift observation has a count rate of (2.3 ± 1) × 10⁻³ counts/sec in the 0.3–10.0 keV energy range. Assuming the same model we used to derive the ROSAT upper limit, this corresponds to a flux of (9.6 ± 4) × 10⁻¹⁴ erg cm⁻² s⁻¹ and a luminosity of (1.5 ± 0.7) × 10⁴¹ erg s⁻¹ in the 0.3–10.0 keV energy range. This emission could be indicative of a weak AGN, but there are too few counts to determine the origin of the emission.

If we then combine all of the Swift non-detections, we obtain a 3σ upper limit on the count rate of 9.7 × 10⁻⁴, which corresponds to a flux limit of 4.1 × 10⁻¹⁴ erg cm⁻² s⁻¹ and a luminosity limit of 6.3 × 10⁴⁰ erg s⁻¹. For our black hole (BH) mass estimate (see below), the apparent detection corresponds to 2 × 10⁻⁴ of the Eddington luminosity given our estimated BH mass, and the upper limit from the Swift non-detections is two times lower.

We assessed the recent star-formation history by measuring the Hα equivalent width (EW) and Lick Hδ_A index of the 6 dF spectrum as in French et al. (2016), finding that the Hα EW is 17.11 ± 0.76 Å and the Lick Hδ_A index is 6.08 ± 0.92 Å, where we are defining positive values of EW as emission. The left panel of Figure 2 shows the Hα EW compared to the Lick Hδ_A index for 2MASX J07001137−6602251 and several other TDE host galaxies. While its Lick Hδ_A index is similar to those of other TDE hosts, 2MASX J07001137−6602251 exhibits a significantly stronger Hα emission feature. We also measured the N ii6584/Hα and O iii5007/Hβ line ratios in order to analyze the galaxy using the BPT diagram (Baldwin et al. 1981) shown in the right panel of Figure 2. The N ii6584/Hα flux ratio is 0.76 ± 0.03, and the O iii5007/Hβ flux ratio is 12.85 ± 3.36. These line ratios indicate the possible presence of a Seyfert in the host galaxy but could also be generated by shocks (e.g., Rich et al. 2015).

Zoom In Zoom Out Reset image size

As can be seen in the left panel of Figure 2, 2MASX J07001137-6602251 is similar to the "shocked post-starburst" (SPOG) galaxies identified by Alatalo et al. (2016). SPOGs combine strong Balmer absorption with emission line ratios consistent with shocks and inconsistent with star formation. This selection differs from an E+A or K+A selection by allowing for Hα emission, even if much of the emission is unlikely to be from star formation. Most SPOGs have star-formation histories similar to traditionally selected post-starburst galaxies, yet are on average younger and thus may have higher dust obscuration (French et al. 2018). There are several theoretical predictions for a higher TDE rate in young starburst/post-starburst galaxies (Madigan et al. 2018; Stone et al. 2018), which would predict an intrinsically high TDE rate in SPOGs as well, although the observed rate might be lower due to the extra dust.

The right panel of Figure 2 shows a BPT diagram of 2MASX J07001137−6602251 and several other TDE hosts along with a comparison sample from the SDSS main spectroscopic survey (Strauss et al. 2002). The host of ASASSN-19bt lies in the Seyfert or non-star-forming portion of the diagram, along with several other TDE hosts, but it lies outside the region enclosing galaxies with line ratios consistent with shocks. Thus, while it is similar in many ways to SPOGS, it is likely not a member of this population.

We fit the SED of the host galaxy using the Fitting and Assessment of Synthetic Templates (fast; Kriek et al. 2009) code. We used the Swift UVOT UV and U, APASS BVgri, 2MASS JHK_S, and WISE W1 and W2 magnitudes given in Table 1 to constrain the model. We assumed a Cardelli et al. (1989) extinction law with R_V = 3.1, a Galactic extinction of A_V = 0.336 mag (Schlafly & Finkbeiner 2011), a Salpeter initial mass function, an exponentially declining star-formation history, and stellar population models from Bruzual & Charlot (2003) for the fit. The fast fit indicates that 2MASX J07001137−6602251 has a stellar mass of ${M}_{\star }={1.1}_{-0.1}^{+1.3}\,\times {10}^{10}$ M_⊙, an age of ${3.2}_{-0.1}^{+5.8}$ Gyr, and a star-formation rate of ${\rm{SFR}}={1.7}_{-0.1}^{+0.6}\times {10}^{-1}$ M_⊙ yr⁻¹. Using the average stellar-mass-to-bulge-mass ratio from the hosts of ASASSN-14ae, ASASSN-14li, and ASASSN-15oi (Holoien et al. 2014, 2016a, 2016b) to scale the stellar mass of the host, as we have done with previous TDEs (e.g., Holoien et al. 2019b), results in an estimated bulge mass of M_B ≃ 10^9.4 M_⊙. This corresponds to an estimated BH mass of M_BH = 10^6.8 M_⊙ (McConnell & Ma 2013), comparable to the BH mass estimates for other TDE hosts (e.g., Holoien et al. 2014, 2016a, 2016b; Brown et al. 2017; Wevers et al. 2017; Mockler et al. 2019).

ASAS-SN images are processed in real time using a fully automated pipeline incorporating image subtraction that is performed with the ISIS package (Alard & Lupton 1998; Alard 2000). As ASAS-SN monitors the TESS fields with an increased cadence, the field containing ASASSN-19bt was observed on average 1–2 times per night before and after discovery of the transient with cameras in our "Cassius," "Paczynski," and "Payne-Gaposchkin" units, located in Chile and South Africa. To ensure that no transient flux was present in the reference images used for image subtraction, we constructed a reference image for each camera that observed ASASSN-19bt using only data obtained earlier than 2018 December 1, roughly 60 days prior to discovery. We used these references as templates to subtract the background and host emission from all data taken after 2019 January 1, ensuring that we capture the full rise of the TDE in ASAS-SN data.

We performed aperture photometry on each host-subtracted image using the IRAF apphot package and an aperture 3 pixels in radius (roughly equivalent to 210). We calibrated the ASAS-SN g-band magnitudes using several stars near the transient with magnitudes available in APASS. For many pre-discovery epochs, when ASASSN-19bt was very faint or not detected, we combined images taken over several days on the same camera to improve the signal-to-noise of our detections and obtain deeper limits on the TDE emission. All ASAS-SN photometry, including both detections and 3σ upper limits, is presented in Table 2, and we show the ASAS-SN light curve in Figure 3 along with our follow-up photometry and photometry from TESS.

Zoom In Zoom Out Reset image size

While SPOGs typically do have a higher amount of dust obscuration, and the host emission lines indicate that the emission line regions may have a higher amount of obscuration, our host SED fit is sufficiently good to not require additional dust correction. Furthermore, the fact that we clearly detect the TDE in UV filters, and the fact that TDE SED is well fit by a blackbody without any additional dust correction (see Figure 9) implies that any potential host extinction along the line of sight to the TDE must be minimal. For this reason, while we correct all of our photometry for Galactic extinction, we do not apply any host extinction correction to our measurements.

Located in the TESS CVZ near the South Ecliptic Pole, ASASSN-19bt has been observed by TESS almost constantly since science operations commenced in late July of 2018. The location of the TDE fell within a chip gap during Sector 6 observations, but this is the only sector in Cycle 1 for which TESS did not observe it. There are five full sectors of pre-disruption observations and a full orbit's worth of observations obtained in Sector 7 prior to first light from the TDE. After first light, the transient's rise is continuously sampled by TESS throughout the remainder of Sector 7 and all of Sector 8, with the exception of a gap in Sector 8 caused by an instrument anomaly on the spacecraft. The transient's epoch of maximum brightness and initial decline are captured in the Sector 9 observations.

As with the ASAS-SN data, we used the ISIS package (Alard & Lupton 1998; Alard 2000) to perform image subtraction on the TESS full-frame images (FFIs) to produce high-fidelity light curves. Due to the large pixel scale of the TESS observations, we elected not to rotate a single reference image for use across various sectors and instead chose to construct independent reference images for each sector. This was achieved by selecting the first 100 FFIs of good quality obtained during that sector, excluding those with sky background levels or PSF widths above average for the sector. We entirely exclude any FFIs with data quality flags from our analysis. We adopted additional conservative quality cuts, excluding FFIs obtained when the spacecraft's pointing was compromised, when TESS was either impacted by or recovering from an instrument anomaly, or when significant background effects due to scattered light were present in the images.

Because a considerable amount of flux from the TDE is present in the images used to construct the Sector 8 reference, fluxes in the raw difference light curve for this sector are systematically lower than the intrinsic values. We correct for this by applying an offset, which we calculate by matching the first 12 hr of Sector 8 observations (24 epochs) to an extrapolation of the Sector 7 power-law fit shown in Figure 6 (see Section 3.2). We subsequently take a simple linear fit between the last 12 hr of Sector 8 observations and the first 12 hr of Sector 9 observations to calculate the offset for the Sector 9 light curve. The measured fluxes were converted into TESS-band magnitudes using an instrumental zero-point of 20.44 electrons per second in the FFIs, based on the values provided in the TESS Instrument Handbook (Vanderspek et al. 2018). TESS observes in a single broadband filter, spanning roughly 6000–10000 Å with an effective wavelength of ∼7500 Å, and TESS magnitudes are calibrated to the Vega system (Sullivan et al. 2015). The full, host-subtracted light curves for all currently available TESS sectors are shown in Figure 4.

Zoom In Zoom Out Reset image size

For comparison with our other photometry, we binned the TESS data for epochs after MJD = 58,491 in 2 hr bins by taking the median flux in each bin and calculating the TESS magnitude from the median flux. We converted the TESS Vega magnitudes to the AB system using an offset of m_AB−m_Vega = 0.411 to match our other photometry. TESS magnitudes and 3σ limits are presented in Table 2 and shown in Figure 3.

Following our initial classification of ASASSN-19bt as a possible TDE, we requested and were awarded 17 epochs of TOO observations with the Swift UVOT and XRT, with the first epoch of observations obtained within 2 days of discovery.

UVOT observations were obtained in the V (5468 Å), B (4392 Å), U (3465 Å), UVW1 (2600 Å), UVM2 (2246 Å), and UVW2 (1928 Å) filters (Poole et al. 2008) in most epochs, with some epochs having only a subset of filters due to scheduling. Each epoch of UVOT data contains two observations in each filter, and, in order to measure transient fluxes for each epoch, we first combined the two images in every filter using uvotimsum. We then used uvotsource to measure counts from the source in the combined images using a 50 aperture and a background region of ∼400 radius to estimate and subtract the sky background. As with the archival data, we then converted the count rates to magnitudes and fluxes using the most recent UVOT calibration (Poole et al. 2008; Breeveld et al. 2010).

We corrected all of the UVOT photometry for Galactic extinction using a Cardelli et al. (1989) extinction law. To enable the UVOT B and V data to be directly compared with Johnson B and V data obtained from ground-based telescopes, we converted the UVOT B- and V-band data to Johnson B and V magnitudes using publicly available color corrections.²⁴ Finally, we then subtracted the host flux measured from the archival Swift observations (UV filters and U) or taken from the APASS catalog (BV) from each UVOT observation to isolate the transient flux in each epoch. The host-subtracted Swift UVOT photometry is presented in Table 2 and shown in Figure 3. ASASSN-19bt has one of the brightest measured peak UV magnitudes of any TDE to date, and is comparable in brightness to ASASSN-14li (Holoien et al. 2016b) despite being more distant. ASASSN-19bt also has 10 epochs of UV observations obtained prior to peak, making this one of the best-sampled multiwavelength rising light curves obtained to date of a TDE.

In addition to the ASAS-SN, Swift, and TESS observations, we also obtained photometric observations in the BVri filters from Las Cumbres Observatory 1 m telescopes located in Cerro Tololo, Chile, Siding Spring, Australia, and Sutherland, South Africa (Brown et al. 2013). After applying photometric calibrations, we measured aperture magnitudes using the IRAF apphot package, with a 50 aperture region used to extract source counts and a 150–300 annulus used to estimate and subtract background counts. We calibrated the Las Cumbres Observatory magnitudes using several stars in the field with magnitudes available in the APASS DR 10 catalog.

As we did with the UVOT observations, we corrected the ground-based aperture magnitudes for Galactic extinction and subtracted the host flux in each band taken from APASS. We find excellent agreement between the UVOT B and V magnitudes and those measured from the Las Cumbres Observatory telescopes. We present the host-subtracted ground-based photometry in Table 2 and show them in Figure 3.

2.6.1. Swift XRT Observations

2.6.2. XMM-Newton Observations

Since the Swift XRT observations showed evidence of weak X-ray emission as the source rose to peak, we requested two deep (41.4 ks each) XMM-Newton Observatory TOO observations of the source. The first observation was taken on 2019 March 1 (ObsID: 0831791001, PI: Auchettl), approximately 3.5 days before peak (${\mathrm{MJD}}_{{\mathrm{XMM}}_{1}}$ = 58,543.21), while the second observation was taken on 2019 April 15 (ObsID: 0831791101, PI: Auchettl), approximately 42 days after peak (${\mathrm{MJD}}_{{\mathrm{XMM}}_{2}}$ = 58,589). Both the MOS and PN detectors were used for this analysis, and both detectors were operated in full-frame mode using a thin filter. All data reduction and analysis was done using the XMM-Newton science system (SAS) version 15.0.02 with the most up-to-date calibration files.

Due to the fact that XMM-Newton suffers from periods of high background and/or proton flares that may affect the quality of the data, we checked for these periods by generating a count rate histogram of the events that have energies between 10 and 12 keV. We find that our observations are only minimally affected by background flares, giving an effective exposure in the PN and MOS detectors of 32 and 37 ks for the first observation and 40 and 39 ks for the second observation.

For our analysis, we used the standard screening of events, with single to quadruple events (PATTERN ≤ 12) chosen for the MOS detectors. For the PN detector, only single and double events (PATTERN ≤ 4) were selected. We also used the standard screening FLAGS for both the MOS (#XMMEA EM) and PN (#XMMEA EP) detectors. We corrected for vignetting by processing all event files using the task evigweight. We extracted spectra of ASASSN-19bt from both the MOS and PN detectors using the SAS task evselect and the cleaned event files from all detectors. We used the same source region that was used to analyze the Swift observations. To avoid chip gaps, we used a smaller background region centered at (α, δ) = (7:00:38.535, −66:05:43.21) with a radius of 720 for the first observation and (α, δ) = (7:00:06.415, −66:06:31.27) with a radius of 720 for the second observation. To increase the signal-to-noise of the MOS spectra, we combined these spectra together using the SAS task epicspeccombine. To analyze the spectra extracted from our XMM-Newton observations, we used the X-ray spectral fitting package (XSPEC) version 12.10.1f and χ² statistics. Using the FTOOLS command GRPPHA, we grouped both the PN and merged MOS spectra with a minimum of 10 counts per energy bin. These data are further discussed in Section 3.4, and the X-ray luminosity measured from the XMM-Newton observations is given in Table 3.

Table 3.  X-Ray Luminosities (0.3–10.0 keV)

Observation	MJD	Rest-frame Days Relative to Peak	X-Ray Luminosity
Swift 001–005	58521.4	−24.9	<7.68 × 1040
Swift 006–008	58535.7	−10.9	${6.73}_{-3.61}^{+3.57}\times {10}^{40}$
XMM 0831791001	58543.2	−3.6	${4.48}_{-0.78}^{+0.77}\times {10}^{40}$
Swift 009–013	58550.0	3.0	<6.62 × 1040
Swift 014–016	58569.6	22.1	<1.26 × 1041
Swift 017–019	58581.4	33.6	<8.69 × 1040
XMM 0831791101	58589.0	−42.1	${1.24}_{-0.38}^{+0.40}\times {10}^{40}$

Note. X-ray luminosities and 3σ upper limits on the X-ray luminosity from our Swift XRT and XMM-Newton observations. Swift data were binned in time to increase the signal-to-noise of the observations, and the epochs combined for each bin are indicated in Column 1.

Download table as:  ASCIITypeset image

After obtaining our first classification spectrum of ASASSN-19bt, we began a program of spectroscopic follow-up to complement our photometric data set. Our follow-up spectra were obtained with LDSS-3 on the 6.5 m Magellan Clay telescope, the Inamori-Magellan Areal Camera and Spectrograph (IMACS; Dressler et al. 2011) on the 6.5 m Magellan-Baade telescope, the Goodman Spectrograph (Clemens et al. 2004) on the Southern Astrophysical Research (SOAR) 4.1 m telescope, and the Wide Field Reimaging CCD Camera (WFCCD) mounted on the du Pont 100 inch telescope. These observations included seven spectra obtained prior to peak light and four spectra obtained within 3 days of peak light.

The majority of our spectra were reduced and calibrated using standard procedures in Iraf, including bias subtraction, flat-fielding, 1D spectroscopic extraction, and wavelength calibration via comparison to an arc lamp. The IMACS data from 2019 February 7 were reduced using an updated version of the routines developed by Kelson et al. (2014) using He and Hg lamps for wavelength calibration. We flux calibrated our observations using standard star spectra obtained on the same nights as the science spectra and masked prominent telluric features. Details of the spectra in our data set are presented in Table 4.

Table 4.  Spectroscopic Observations of ASASSN-19bt

Date	Telescope	Instrument	Grating	Slit	Exposure Time
2019 Jan 31.20	Magellan Clay 6.5 m	LDSS-3	VPH-All	100 blue	2 × 600 s
2019 Feb 01.26	du Pont 100 inch	WFCCD	Blue	165	3 × 900 s
2019 Feb 07.03	Magellan Baade 6.5 m	IMACS f/2	400 l/mm	120	1 × 300 s
2019 Feb 14.04	Magellan Baade 6.5 m	IMACS f/2	300 l/mm	090	2 × 600 s
2019 Feb 15.03	Magellan Baade 6.5 m	IMACS f/2	300 l/mm	090	2 × 600 s
2019 Mar 02.02	du Pont 100 inch	WFCCD	Blue	165	3 × 600 s
2019 Mar 04.03	du Pont 100 inch	WFCCD	Blue	165	3 × 600 s
2019 Mar 06.10	du Pont 100 inch	WFCCD	Blue	165	3 × 600 s
2019 Mar 24.05	SOAR 4.1 m	Goodman	400 l/mm	095	1 × 1200 s
2019 Mar 29.00	Magellan Clay 6.5 m	LDSS-3	VPH-All	100 blue	4 × 600 s
2019 Apr 10.20	du Pont 100 inch	WFCCD	Blue	165	3 × 600 s
2019 Apr 13.09	du Pont 100 inch	WFCCD	Blue	165	3 × 600 s

Note. Date, telescope, instrument, grating, slit size, and exposure time for each of the spectroscopic observations obtained of ASASSN-19bt for the initial classification of the transient and as part of our follow-up campaign.

Download table as:  ASCIITypeset image

We further calibrated our spectra using our photometric data set. To obtain magnitudes with a similar amount of host contamination as the spectra, which were observed through slits of roughly 10 width, we measured small aperture magnitudes from our Swift and Las Cumbres Observatory data, using a 15 aperture for the Las Cumbres Observatory data and a 35 aperture for the Swift data, due to the larger pixel scale of Swift. For each filter that was completely contained in the wavelength range covered by a given spectrum and for which we could either interpolate the small aperture light curves or extrapolate them by 1 hr or less, we extracted synthetic photometric magnitudes from the spectrum. As the Swift data has a larger PSF and the small aperture magnitudes are more uncertain, we calibrated the spectra using only Las Cumbres Observatory data, except for the three spectra taken prior to 2019 February 10, when we obtained our first observations with Las Cumbres Observatory telescopes. We then fit a line to the difference between the synthetic and observed fluxes as a function of central filter wavelength and scaled each spectrum by the photometric fits. Finally, we corrected each spectrum for Galactic reddening using a Milky Way extinction curve assuming R_V = 3.1 and A_V = 0.336 (Schlafly & Finkbeiner 2011). We also used the procedure to calibrate the archival 6 dF spectrum to the same flux scale as our follow-up spectra using the archival magnitudes shown in Table 1 for calibration.

Figure 5 shows the spectroscopic evolution of ASASSN-19bt as well as the calibrated host spectrum. Prominent telluric bands have been marked in the figure, with the feature from 7550 to 7720 Å and chip gaps (where present) masked to facilitate plotting. Similar to what was seen with PS18kh (Holoien et al. 2018), there is little evidence of broad lines prior to peak light, with the lines becoming more prominent after peak. We further analyze the line emission in Section 3.5.

Zoom In Zoom Out Reset image size

We measured the position of ASASSN-19bt using our initial V-band image and a V-band image taken near peak from the Las Cumbres Observatory 1 m telescopes. Using the early image as a subtraction template, we generated a subtracted image of the TDE. While some TDE flux was likely removed along with the host flux, this allowed us to measure a centroid position of only the TDE signal. Using the Iraf task imcentroid, we measured the centroid position of the flux in the subtracted image as well as the centroid position of the nucleus of the host galaxy in the early V-band image that was used as the subtraction template, which is likely host-dominated. From this method, we obtain a position for ASASSN-19bt of R.A. = 07:00:11.41, decl. = −66:02:25.16. This is offset by 014 ± 015 from the position of the host nucleus measured from the early image, which corresponds to a physical offset of 78.2 ± 83.8 pc.

The redshift of the host galaxy in the 6 dF catalog is reported as z = 0.0262. In order to verify this, we downloaded the 6 dF spectrum and measured the redshift using the narrow Hα and [O iii] 5007/4959 Å emission lines, finding z = 0.026. As this is consistent with the reported redshift, we adopt z = 0.0262 and the corresponding luminosity distance of d = 115.2 Mpc throughout our analysis.

To obtain an estimate of the time of peak light, we fit a parabola to the host-subtracted ASAS-SN g-band light curve prior to MJD = 58,560, as the decline of the light curve is flatter than the rise, making a parabolic fit to the entire light curve inaccurate. To estimate the uncertainty on the peak date, we generated 10,000 realizations of the g light curve prior to MJD = 58,560 with each magnitude perturbed by its uncertainty assuming Gaussian errors. We then fit a parabola to each of these light curves and calculated the 68% confidence interval and median t_peak values from these realizations. Using this procedure, we find t_g,peak = 58546.9 ± 0.2 and m_g,peak = 14.9. We also performed the same analysis for all of our photometric filters and found that there is some evidence that the bluer filters peaked earlier, with ${t}_{{UVW}2,\mathrm{peak}}={58544.2}_{-1.9}^{+2.6}$ and ${t}_{i,\mathrm{peak}}={58548.9}_{-1.5}^{+2.4}$, not unlike what has been seen in some other TDEs (e.g., Holoien et al. 2018). Due to the high cadence of the ASAS-SN light curve, its peak is much better constrained than those of the other filters, and we adopt the g-band peak of t_g,peak = 58546.9, corresponding to 2019 March 04.9, throughout this paper.

We characterized the early-time rise of ASASSN-19bt with a power-law model

$$\begin{eqnarray}&&f\,=\,z\,{\rm{when}}\,t\lt {t}_{1},{\rm{and}}\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&f\,=\,z\,+\,h{\left(\displaystyle \frac{t-{t}_{1}}{\mathrm{days}}\right)}^{\alpha }{\rm{when}}\,t\gt {t}_{1}.\end{eqnarray} \tag{ 2 }$$

for the TESS flux, described by a residual background z, the start of the rise t₁, a flux scale h, and the power-law index α. We use the scipy.optimize.curve_fit package's Trust Region Reflective method to obtain a best-fit model with parameters z = −0.22 ± 0.22 μJy, h = 2.44 ± 0.90 μJy, t₁ = MJD = 58504.61 ± 0.42, and α = 2.10 ± 0.12. This fit is shown as the red curve in Figure 6.

Zoom In Zoom Out Reset image size

This power-law index is consistent with the "fireball" model used to model the early flux from SNe where

$$\begin{eqnarray}&&{L}_{\nu }\propto {r}^{2}T\propto {v}^{2}{\left(t-{t}_{1}\right)}^{2}T,\end{eqnarray} \tag{ 3 }$$

which assumes a band on the Rayleigh–Jeans tail of the SED and homologous expansion. For the early phases of SNe, this power-law rise with L_ν ∝ (t−t₁)² is due to the fact that the velocity v and temperature T are nearly constant (e.g., Riess et al. 1999; Nugent et al. 2011). However, as we shall see in Section 3.3, the temperature of ASASSN-19bt does not appear to be constant in these early phases, so the consistency of the exponent α with the fireball model appears coincidental.

Given the distance to the source and the effective wavelength of the TESS bandpass, the flux scale given by the fit implies a radius of

$$\begin{eqnarray}&&r={10}^{13.9}{T}_{4}^{-1/2}{\left(t-{t}_{1}\right)}^{1.05}\,\mathrm{cm}\end{eqnarray} \tag{ 4 }$$

for a temperature of T = 10⁴T₄ K, the central values of h and α and assuming Equation (3). For our estimated black hole mass, this corresponds to $100{T}_{4}^{-1/2}$ gravitational radii at t−t₁ = 1 day. Since the emission region is probably not spherical, it is probably more accurate to say that the surface area producing the UV/optical emission is 4πr².

The TESS light curve begins to deviate from the initial power-law rise approximately 15 days after first light, with the rise slowing as it approaches peak brightness. Based on the inferred time of first light taken from the power-law fit and the time of peak light measured from the ASAS-SN light curve, we constrain the rise time to 41.2 ± 0.5 rest-frame days.

The photometric precision of the TESS light curve, particularly with the addition of the Sector 9 data, allows us to see clearly that the TDE light curve is very smooth, with little to no short-term variability. This is in contrast to what is often seen during AGN flares (e.g., Peterson 1993; Peterson et al. 2004; Shappee et al. 2014). While the light curves from ASAS-SN and Las Cumbres Observatory are also quite smooth, the TESS cadence provides us with a unique ability to see exactly how smooth the light curve is. To take advantage of this, we fit the TESS Sector 9 light curve taken after peak light with a power-law decline of the form

$$\begin{eqnarray}&&f=z-h{\left(\displaystyle \frac{t-{t}_{\mathrm{peak}}}{\mathrm{days}}\right)}^{\alpha }\end{eqnarray} \tag{ 5 }$$

with t_peak being the peak date inferred from the ASAS-SN light curve.

The best-fit power law has the parameters z = 2.68 μJy, h = 0.006 μJy, and α = 1.53, with a reduced chi-squared value of ${\chi }_{\nu }^{2}=1.05$. The TESS declining light curve and best-fit model are shown in Figure 7. The fit residuals are roughly 1% or less of the actual flux, with an rms value of 0.01, implying that deviations from the power-law fit are likely systematic. The fact that the light curve is very well fit by a simple model such as this demonstrates the smoothness of the light curve in a way that was not previously possible.

Zoom In Zoom Out Reset image size

To better understand the physical parameters of the transient, we modeled the UV and optical SED of ASASSN-19bt for epochs where Swift data were available as a blackbody. We used Markov Chain Monte Carlo methods to fit the blackbody SED, using a flat prior of 10,000 K ≤ T ≤ 55,000 K so as not to overly influence the fits. From the blackbody fits, we estimate the bolometric luminosity, temperature, and radius of ASASSN-19bt in each epoch.

To take better advantage of the high-cadence light curve from ASAS-SN, we used the Swift blackbody fits to calculate bolometric corrections for g-band magnitudes by linearly interpolating between the previous and next g-band measurements bracketing each epoch of Swift observations. We then estimated the bolometric luminosity of ASASSN-19bt from the ASAS-SN light curve, by linearly interpolating bolometric corrections calculated for the Swift epochs to each epoch of g data. For times prior to our first Swift observation, we use the bolometric correction from the first Swift sED fit. The luminosity evolution calculated from the Swift sED fits and estimated from the g-band light curve is shown in the left panel of Figure 8.

Zoom In Zoom Out Reset image size

The luminosity fits and corrected g-band light curve indicate that both the rise to peak and then initial decline after peak are relatively smooth. However, the first Swift sED fit indicates that the luminosity was higher in the first epoch than it was in the second, resulting in a short decline at t ≃ −32 rest-frame days before peak. This corresponds to a similar drop in the early temperature (see below). Such behavior has not been seen in TDEs prior to ASASSN-19bt. However, Swift UV data has not been obtained at such early times prior to peak for previous TDEs. This makes it unclear whether an early, rapid drop in temperature and luminosity is common in TDEs, or unique to ASASSN-19bt, and further highlights the need for early detection and prompt scheduling of follow-up observations.

The early luminosity spike is not seen in the uncorrected g-band data, and the rapid rise to this early peak shown in Figure 8 is driven by the fact that we used the first Swift luminosity fit to calculate the bolometric correction for all of the previously obtained epochs of g-band data. Due to the 10 day gap between our second and third Swift observations, it is also difficult to determine precisely when the luminosity begins to rise again. However, we can say that the early luminosity spike must have lasted for at least 2 days and could have been as long as roughly 15 days, depending on how rapid the rise actually was and when it stopped declining and began to re-brighten. We also note that while the luminosity began to re-brighten sometime between −30 and −20 days prior to peak, its temperature continued to drop during this time, resulting in an early temperature decline that lasted at least 12 rest-frame days.

In the right panel of Figure 8, we show a comparison of the light-curve evolution of ASASSN-19bt to that of several other TDEs from the literature: ASASSN-14ae (Holoien et al. 2014), ASASSN-14li (Holoien et al. 2016b), ASASSN-15oi (Holoien et al. 2016a), iPTF16fnl (Brown et al. 2018), iPTF16axa (Hung et al. 2017), PS18kh (Holoien et al. 2018), and ASASSN-18pg (T. W.-S. Holoien et al. 2019, in preparation; Leloudas et al. 2019). The luminosity rise of ASASSN-19bt is similar to those of ASASSN-18pg and PS18kh with the exception of the spike at t = −32 days, and it begins to decline more rapidly after peak than ASASSN-18pg, which has a plateau-like phase at peak. Both the decline rate shortly after peak and the total luminosity are similar to most of the other objects in the sample, and ASASSN-19bt peaked at a luminosity of L ≃ 1.3 × 10⁴⁴ erg s⁻¹.

Figure 8 also shows the X-ray luminosities calculated from the Swift XRT and XMM-Newton observations, scaled up by a factor of 100 to make them comparable to the UV/Optical luminosities. We do not detect X-ray emission in most epochs, and all detected luminosities are three or more orders of magnitude weaker than the UV/optical emission. We also see some evidence that X-rays are only detected near the peak of the UV/optical light curve. We further analyze and discuss the X-ray results in Section 3.4 below.

We show the SEDs and blackbody fits for the first three Swift epochs in Figure 9. The SED for the median luminosity and the range corresponding to the 16%–84% confidence interval on the luminosity are also shown. The emission clearly becomes redder over these 12 days, with the UVW2 flux becoming fainter relative to the UVM2 flux over time. After −20 days, the TDE continues to exhibit the highest luminosity in the UVM2 filter, which drives the cooler temperature fits up to and shortly after peak.

Zoom In Zoom Out Reset image size

Integrating the entire rest-frame bolometric light curve, including both the Swift blackbody fits and the converted g-band data, we obtain a total radiated energy of E = (5.92 ± 0.06) × 10⁵⁰ erg. Of this, (3.17 ± 0.05) × 10⁵⁰ erg are released during the rise to peak, indicating that a large fraction of the energy radiated by TDEs can be emitted prior to peak light. The accreted mass required to generate the emitted energy is M_Acc ≃ $0.003{\eta }_{0.1}^{-1}$ M_⊙, where the accretion efficiency is η = 0.1η_0.1. This low implied accreted mass is similar to what has been seen in other TDEs and again indicates that either only a small fraction of the bound stellar material is actually accreting onto the SMBH, or that the material accretes with very low radiative efficiency.

The blackbody temperature evolution of ASASSN-19bt is shown in Figure 10, along with that of the same TDE comparison sample shown in Figure 8. Unique among the TDEs shown in the figure, ASASSN-19bt exhibits a steep temperature decline in the first three epochs, corresponding to the luminosity decline over the same period before leveling off and exhibiting the relatively constant temperature evolution that is expected. None of the comparison objects have UV data as early as ASASSN-19bt, so it is possible that an early drop in temperature is common but not observed due to discovering the TDE flare too late. After the initial decline, the temperature remains fairly steady at T ≃ 16,000–17,000 K, on the low end of the temperature range observed for TDEs.

Zoom In Zoom Out Reset image size

Figure 11 shows the blackbody radius evolution of ASASSN-19bt taken from the Swift fits compared to the radius evolution of the other TDEs in our comparison sample. ASASSN-19bt exhibits a rapidly growing radius in the first three epochs to match the drop in temperature over the same timeframe. It then appears to peak and hold relatively steady for the remainder of our period of observation at one of the largest radii of the TDEs in our sample. While there does not seem to be a strong correlation between size and rate of decline in radius, it appears that the hotter TDEs tend to have smaller emitting regions, which is not unexpected given that they are have similar luminosities. ASASSN-19bt is a very close match to PS18kh in both temperature and radius evolution, and it will be interesting to see if ASASSN-19bt exhibits a similar "re-brightening/plateau" phase later in its evolution.

Zoom In Zoom Out Reset image size

We can also combine the radius estimates from the SED fits with the results from fitting the TESS light curve in Section 3.2, as shown in Figure 12. Here, we work in terms t−t₁ based on the onset time found in the TESS fits. The TESS radius estimate depends on the temperature, and we show two simple assumptions. We either assume the temperature found for the first Swift epoch (log(T/K) = 4.60) or the power law in temperature defined by the first two Swift epochs (log(T/K) = 6.66–2.06 log(t−t₁)) in order to illustrate the uncertainties. Where they overlap, the two sets of radius estimates agree reasonably well (factor of ∼2). At the time of our earliest TESS detection, when t−t₁ ∼ 2 days, the emission region probably had an effective size of some tens of gravitational radii.

Zoom In Zoom Out Reset image size

Figure 12 also shows the escape velocity corresponding to a given radius assuming a BH mass of 10^6.8M_⊙. For the constant-temperature case, the TESS radius evolution corresponds to a velocity of ∼2700 km s⁻¹, which is well below the scale of the escape velocity and well above the sound speed implied by the temperatures. With the rapidly evolving temperature, the velocity implied by the TESS radius is 870 (t−t₁)^1.08 km s⁻¹, so accelerating from 870 km s⁻¹ on day 1 to 10,500 km s⁻¹ on day 10. Similarly, the ∼10^14.5 cm sizes implied by the first few Swift epochs also only require velocities of 3000–4000 km s⁻¹. Thus, if the early-time radii implied either by the TESS flux evolution or the first few SED fits are related to the distance from the black hole, the apparent photosphere is expanding very slowly compared to the local escape speed. If these velocities are related to dynamical velocities, then the emission region must be at a distance near 10^16.3 cm from the BH and the emission region is very small compared to the distance.

During the first ∼two weeks of its evolution, our Swift XRT observations of ASASSN-19bt showed no evidence of X-ray emission. We first detect the source approximately two weeks prior to the peak, the first such detection. For those events that do show evidence of X-ray emission (e.g., ASASSN-14li, Holoien et al. 2016b; Brown et al. 2017; ASASSN-15oi, Holoien et al. 2016a, 2018; Gezari et al. 2017; ASASSN-18ul, Wevers et al. 2019; and those in Auchettl et al. 2017), the X-ray emission is first detected at or after peak. However, some of these events may have had emission at or before peak, as most X-ray observations of these sources only commenced as the associated optical/UV transient began to decline. For the other two events that were detected on the rise in the optical/UV, Swift XRT observations were taken after peak and showed that PS18kh (Holoien et al. 2019b; van Velzen et al. 2019) exhibited weak X-ray emission. Pre-peak Swift observations of ASASSN-18pg only yielded upper limits on the presence of X-ray emission (Leloudas et al. 2019, T. W.-S. Holoien et al. 2019, in preparation).

There are too few counts in the XRT detections to extract a spectrum, but we could determine the hardness ratio, HR = (H− s)/(H+S), where S is the number of counts in the soft 0.3–2.0 keV energy band and H is the number of counts in the 2.0–10.0 keV energy band. A source is considered soft if it has an HR = −1, while it is considered hard if it has an HR = 1. Interestingly, we find an average hardness ratio of HR ∼ −0.2, which is much harder than found for non-jetted, thermal TDEs like ASASSN-14li and ASASSN-15oi, which have HR ∼ −0.7, and more consistent with those seen from the jetted TDEs Swift J1644+57 and Swift J2058-05 (Auchettl et al. 2017; Holoien et al. 2018).

Due to the hardness of this X-ray emission, we triggered two deep XMM-Newton TOO observations of the source to better constrain its nature. The first observation was taken ∼4 days before peak, while the second observation was take ∼42 days after peak, and in both observations, the source is significantly detected. The shallower Swift observations taken around this time were only able to determine upper limits due to the faintness of this emission. In Figure 13, we show the resulting PN and MOS1+MOS2 spectra. In the first observation (Figure 13 top panel), the spectra are well fit by an absorbed power-law model with a photon index of Γ = 1.47 ± 0.3 and the Galactic H i column density in the direction of ASASSN-19bt (Kalberla et al. 2005). Letting the column density (N_H) vary does not significantly improve the fit. This photon index is consistent with the best-fit photon index for the jetted TDE Swift J1644+57 and Swift J2058+05 at late times (e.g., Burrows et al. 2011; Cenko et al. 2012b; Saxton et al. 2012; Levan et al. 2016; Auchettl et al. 2017) and those found from AGN (e.g., Ricci et al. 2017; Auchettl et al. 2018), implying that the X-ray emission may indicate the presence of a jet. The X-ray emission of several TDEs is well-modeled by a cool blackbody. A blackbody fit to the spectrum of ASASSN-19bt gives a temperature of 0.48 ± 0.1 keV. This is significantly higher than the temperatures found for ASASSN-14li (e.g., Brown et al. 2016; Kara et al. 2018) or ASASSN-15oi (Holoien et al. 2018), which have temperatures of ∼50 eV. Due to the poor signal-to-noise of the spectra, models combining a blackbody with a power law do not improve significantly on our best-fit power-law model. In the second observation (Figure 13 bottom panel), we find that the X-ray emission has softened considerably after peak and is now well fit by an absorbed power-law model with a photon index of ${\rm{\Gamma }}\,={2.34}_{-0.6}^{+0.8}$ but still consistent with those found from jetted TDES (e.g., Burrows et al. 2011; Cenko et al. 2012b; Saxton et al. 2012; Auchettl et al. 2017). If we fit the spectra using a blackbody instead, we find a slightly lower temperature of 0.20 ± 0.1 keV compared to the first observation, but it is still higher than that found for other TDEs. Similar to the first observation, fitting the spectra with a blackbody plus a power law does not significantly improve the fit.

Zoom In Zoom Out Reset image size

ASASSN-19bt exhibits several broad emission lines that begin to emerge following the 2019 February 07 spectrum, prior to which there are no discernible features. We identify strong emission from Hα, He i 5875 Å, and a broad blue feature that spans the Hβ, He ii 4686 Å, and Hγ lines. In later epochs, the broad blue feature can be differentiated into likely Hβ and Hγ lines, and it is apparent that there is little to no emission from the He ii 4686 Å line, which is a rarity among TDEs. There is no sign of emission from any of the nitrogen lines identified by Leloudas et al. (2019) in any epoch, meaning ASASSN-19bt is not among the subset of nitrogen-rich TDEs. The Hα and He i 5875 Å lines are extremely broad once they begin to emerge, with maximum FWHM of FWHM_Hα ≃ 2.7 × 10⁴ km s⁻¹ and FWHM_{He i} ≃ 2.1 × 10⁴ km s⁻¹. The lines continue to exhibit similar widths throughout the period of observation and begin to develop significantly non-Gaussian shapes in later epochs.

We also measured the luminosities of the prominent emission lines in the epochs after the 2019 February 7 spectrum. For each spectrum, we subtracted a local continuum estimate for the Hα and He i 5875 Å lines, then measured the integrated line flux using the IRAF task splot. We also used the same procedure to measure the broad blue feature as a single emission line when possible, as this feature is difficult to break down into its component lines, making it difficult to measure the lines individually. The measured luminosities are given in Table 5 and are shown in Figure 14. Estimating the true error on the line fluxes is difficult given their complex shape, and we assume 30% errors on the emission fluxes calculated from each epoch.

Zoom In Zoom Out Reset image size

The broad lines appear roughly 18 days prior to peak and continue to grow stronger throughout the period of observation, with the Hα luminosity roughly tripling between the first detections and the latest spectrum from 2019 April 10. The emergence of spectroscopic emission lines prior to peak light and the continuing strengthening of those lines after peak is similar to what was seen with the TDE ASASSN-18pg (Leloudas et al. 2019), although in that case the lines were present more than 30 days prior to peak. Increasing line luminosities after peak were also seen in the TDE PS18kh; though in that case, the lines were not present until roughly peak light (Holoien et al. 2018). ASASSN-19bt thus joins a growing sample of TDEs with early spectra that exhibit increasing line luminosities near peak light, in contrast to early TDE discoveries that were found after peak and showed lines that became less luminous in the epochs following discovery (e.g., Holoien et al. 2016b; Brown et al. 2017).

The cause for the early, featureless spectra we observe in ASASSN-19bt is not well understood at this time, in part due to the small number of TDEs with spectra obtained this early. But one possible explanation is that at very early times, the material emitting the lines may be moving at very high speeds, as indicated in Figure 12, which would have the effect of suppressing these features. Further analysis on a larger sample of pre-peak TDE spectra is needed to determine whether other physical effects could play a role in suppressing early line emission, particularly if different TDEs are producing emission via different physical mechanisms, as some of the observations seem to indicate.