Tidal disruption events (TDEs) refer to the transient phenomena where a star on a close passage to a supermassive black hole (SMBH) is torn apart under tidal stress (Hills 1975). For a star that initially traveled on a parabolic orbit, the disruption unbinds about half of the stellar mass while the bound other half assembles into an accretion disk and feeds the black hole until the debris streams are drained. In scenarios where disk formation is efficient, the onset of accretion resulting from a TDE is set by the fallback time (t fallback ), which corresponds to the time it takes for the most bound debris on highly eccentric orbits to return to the pericenter (e.g., = t M 41 fallback BH,6
1 2 days, where M BH,6 is the black hole mass in units of 10 6 M e ; Lodato et al. 2009). The gravitational potential energy liberated by the infalling gas is then converted to radiation across the electromagnetic spectrum, thus allowing a TDE to be detected by observers (Frank & Rees 1976;Rees 1988).
Observationally, X-ray, ultraviolet (UV), and optical sky surveys have identified nearly five dozen TDEs (e.g., Komossa & Bade 1999;Komossa & Greiner 1999;Esquej et al. 2007;Levan et al. 2011;Gezari et al. 2012;Holoien et al. 2016;van Velzen et al. 2020). However, these events, depending on the wavelength of discovery, exhibit a dichotomy in their properties. While the X-ray-detected TDEs are characterized by thermal emission that is consistent with the accretion model, most optically detected TDEs appear to lack or have very weak (10 -2 L UV ) X-ray emission (e.g., Holoien et al. 2019a;van Velzen et al. 2019van Velzen et al. , 2021)). In particular, the optically detected events tend to have temperatures that are 1-2 orders of magnitude lower than those of their X-ray counterparts (van Velzen et al. 2011;Arcavi et al. 2014;Holoien et al. 2014;Hung et al. 2017). The optical TDEs are also able to maintain roughly the same temperature over a timescale of months.
The discrepancy between the two observational populations of TDEs may arise from the biggest uncertainty in the current TDE framework, which is whether the stellar debris can circularize efficiently following a TDE. Circularization requires the bound debris in highly eccentric orbits to lose a large amount of orbital energy to form a circular disk at twice the pericenter radius, 2R p (conservation of specific angular momentum). In the classical scheme, the energy is dissipated efficiently via shocks produced by self-intersection at the pericenter (Rees 1988;Evans & Kochanek 1989;Phinney 1989;Ramirez-Ruiz & Rosswog 2009). This assumption is challenged by recent simulations and analytical calculations that find circularization to be extremely inefficient in certain regions of parameter space (Dai et al. 2015;Guillochon & Ramirez-Ruiz 2015;Shiokawa et al. 2015;Bonnerot et al. 2016;Hayasaki et al. 2016;Svirski et al. 2017). This possibility lays the foundation for an alternative mechanism in which the UV and optical emission in TDEs is powered by stream-stream collision shocks (Piran et al. 2015;Shiokawa et al. 2015). However, the stream-stream collision model is radiatively inefficient (Mockler & Ramirez-Ruiz 2020), which naturally leads to a slower light-curve evolution that is at odds with the observed t -5/3 decline (theoretical mass fallback rate) in several TDEs (Metzger & Stone 2016).
In TDEs where disk formation is efficient and where radiation is driven by the inner accretion flows, a reprocessing layer at 10-100 R T (where R T is the tidal radius) is often invoked to explain the lower temperature in optical TDEs. Some studies suggest that the bound debris could build up a hydrostatic envelope around the SMBH to reprocess the X-ray and extreme ultraviolet (EUV) radiation released by accretion (Loeb & Ulmer 1997;Guillochon & Ramirez-Ruiz 2013), while others suggest that outflows could be the major reprocessing material in TDEs (Miller 2015;Metzger & Stone 2016;Piro & Lu 2020). Indeed, radiation-driven winds are a natural consequence of TDEs given that a "super-Eddington" phase should be common among stellar disruptions by black holes of mass M BH 10 7 M e (Strubbe & Quataert 2009;Wu et al. 2018). For previous TDEs, the observed temperatures and luminosities generally correspond to an ejecta mass that is greater than 1 M e (Matsumoto & Piran 2021).
Multiwavelength follow-up observations of TDEs have detected winds across a wide range of velocities. X-ray observations of ASASSN-14li revealed highly ionized outflows moving at both low and high velocities, from a few hundred km s -1 to 0.2c (Miller et al. 2015;Kara et al. 2018). The UV spectrum of ASASSN-14li also shows signs of a lowvelocity outflow (Cenko et al. 2016), which has a velocity similar to that of the slower X-ray gas found by Miller et al. (2015). In the same event, radio observations are also supportive of the presence of either a subrelativistic outflow (Alexander et al. 2016) or an off-axis relativistic jet (van Velzen et al. 2016). Including AT 2019qiz, the subject of this paper, blueshifted broad absorption lines (BALs) that correspond to outflow velocities of 5000-15,000 km s -1 are detected in four out of five TDEs that have Hubble Space Telescope (HST) UV follow-up spectroscopy, with the exception of ASASSN-14li (Brown et al. 2018;Blagorodnova et al. 2019;Hung et al. 2019). Orientation effects may explain why BALs are absent from some TDEs (Parkinson et al. 2020).
Contrary to the late-time observational properties of TDEs, which tend to be more uniform as the reprocessing layer becomes transparent and allows one to probe the accretion disk directly (van Velzen et al. 2019;Jonker et al. 2020), early-time observations are expected to exhibit a higher degree of diversity in the flare properties. Early-time multiwavelength observations of TDEs are critical for understanding debrisstream evolution and super-Eddington accretion in TDEs, though they have rarely been obtained. To date, there are only a handful of TDEs with published pre-peak multiband light curves (e.g., ASASSN-18pg and ASASSN-19bt;Holoien et al. 2019bHoloien et al. , 2020)). Among these, ASASSN-19bt has the most densely sampled rising optical light curve because it is located in the TESS Continuous Viewing Zone.
AT 2019qiz is the most well-observed TDE since ASASSN-19bt, with a wealth of early-time multiwavelength data. Here we present the analysis of the UV and optical data of AT 2019qiz, the current record holder for the nearest TDE, at redshift z = 0.0151. We note that Nicholl et al. (2020, hereafter N20) also analyzed the evolution of UV and optical broadband photometry and optical spectroscopy of AT 2019qiz. However, our analysis has a stronger focus on the unique multiepoch HST UV spectra that have not been reported before. In addition, our ugri light curves were obtained independently at a higher cadence than that of N20 at early times.
This paper is structured as follows. The observations, including photometry and spectroscopy at UV and optical wavelengths, are discussed in Section 2. We present the procedures and results of our analysis in Section 3. Specifically, in Section 3.3 we detail the evolution and the identification of broad and narrow UV absorption lines in the HST spectra. We discuss in Section 4 the implications of our results and the origin of the UV absorption lines. Section 5 presents our conclusions. We detail the follow-up observations that are analyzed in this paper in the subsections below. Throughout the paper, we adopt a flat ΛCDM cosmology with H 0 = 67.4 km s -1 Mpc -1 and Ω m = 0.315 as measured by the Planck mission (Planck Collaboration et al. 2020). The time difference (Δt) is expressed in rest-frame time with respect to the peak of the g-band light curve at MJD 58763.93. All of the magnitudes are expressed in the AB system (Oke & Gunn 1983).
AT 2019qiz was simultaneously monitored by the Ultraviolet Optical Telescope (UVOT; Gehrels et al. 2004;Roming et al. 2005) and the X-ray Telescope (XRT) on board the Neil Gehrels Swift Observatory (Gehrels et al. 2004) during its flaring state. The X-ray emission from AT 2019qiz peaks at ∼10 41 erg s -1 in the 0.3-10 keV band, which is 2-3 orders of magnitude weaker than the UV and optical emission (N20).
Given that the Swift XRT data set has already been analyzed by N20 and is independent of our other analysis, we reference their reported luminosity and hardness ratio where appropriate without repeating the data reduction and analysis in this work.
All data for AT 2019qiz have been corrected for Milky Way foreground extinction assuming an extinction curve according to Cardelli et al. (1989) with R V = 3.1 and E(B -V ) = 0.0939 ± 0.0088 mag (Schlafly & Finkbeiner 2011).
We obtained three epochs of UV spectra of AT 2019qiz with the HST Space Telescope Imaging Spectrograph (STIS) (GO-16026; PI Hung) on 2019 October 21, December 12, and 2020 January 15. The spectra were obtained through a 52″ × 0 2 aperture. For the near-UV (NUV) and far-UV (FUV) MAMA detectors, the G140L and G230L gratings were used to cover the spectral ranges 1150-1730 Å and 1570-3180 Å at resolutions of 1.2 Å and 2.2 Å, respectively. During the first two visits, the observation was obtained over a single HST orbit, with three equal exposures of 170 s in the NUV and three equal exposures of 335 s in the FUV. The observations obtained in the last visit consist of three 692 s exposures in the NUV and three 876 s exposures in the FUV, totaling two HST orbits. We used an inverse-variance weighting method to combine the one-dimensional (1D) spectra at the same epoch that were output by the HST pipeline.
Following the classification of AT 2019qiz as a TDE on Δt = -13 days, we triggered photometric and spectroscopic monitoring spanning about six months (between 2019 September and 2020 March) before the TDE became too faint and Sunconstrained. Figure 1 shows the light curves of AT 2019qiz observed by the Swift UVOT and ground-based optical facilities including Las Cumbres Observatory (LCO), ZTF, and the Swope telescope at the Las Campanas Observatory. We list the photometric data in Table A1. Data reduction with each instrument is detailed in the following subsections.
AT 2019qiz was simultaneously observed in the ZTF Mid-Scale Innovations Program field in both g and r with a cadence of 3 days. The ZTF real-time pipeline performs standard image reduction and subtraction with respect to ZTF template images and distributes the events as alert packets on each observing night (Bellm et al. 2019;Masci et al. 2019). We accessed the public alerts of AT 2019qiz via LCO MARSfoot_1 and used the magnitude of the template-subtracted point-spread function to generate the light curves. The last ZTF detection was obtained on 2020 February 25 in g as ZTF stopped monitoring the field containing AT 2019qiz. We measured a signal-to-noise-ratio (S/N) weighted offset of 0 13 ± 0 17 from the ZTF g and r data, confirming that the transient is coincident with the galaxy nucleus, as expected for a TDE.
Optical photometry of AT 2019qiz in ugri was obtained with the 1 m Swope telescope from 2019 September 26 to 2020 March 1 with a cadence of 2-5 days. The images were reduced using the photpipe imaging and photometry pipeline (Rest et al. 2005(Rest et al. , 2014)). We subtracted the bias and flattened each frame using bias and sky-flat images obtained on the same night and in the same instrumental configuration as each AT 2019qiz image. The images were registered and geometric distortion was removed using Two Micron All Sky Survey (2MASS) astrometric standards (Cutri et al. 2003). Using hotpants (Becker 2015), we subtracted pre-discovery Figure 1. Multiwavelength light curves of AT 2019qiz in AB magnitudes. Photometry obtained with Swift, ZTF, LCO, and Swope is marked with diamonds, squares, circles, and crosses, respectively. The triangles indicate the upper limits in the g and r filters. The light curves are color-coded by filters and offset by a constant to aid visualization. The host-galaxy contribution has been removed from all the photometry points obtained with ground-based optical telescopes. Although we did not subtract host-galaxy light from the Swift UV filters, the contamination is negligible compared to the transient light. We indicate the HST epochs with gray vertical lines, whereas the short black vertical lines mark the epochs with optical spectroscopic observations. Pan-STARRS1 (PS1) template images (Flewelling et al. 2020) from each Swope gri frame.
Since we are not yet able to obtain a template image for the u band, we instead extracted photometry within a 5″ radius aperture and subtracted the host-galaxy light in u by modeling the host emission. The u-band emission of the host galaxy is estimated by fitting the PS1 photometry in the grizy bands and the 2MASS photometry in the JHK filters in a circular aperture of 5″ radius with the synthetic stellar population fitting code PROSPECTOR. Our best-fit continuity model of the star formation history with five age bins yields a stellar mass of log 10 (M å /M e ) = 10.43 ± 0.04 and a metal content of log 10 (Z/Z e ) = - -+ 1.11 0.45 0.22 , which are consistent with values derived by N20 and van Velzen et al. (2021).
We also obtained optical photometry of AT 2019qiz in the ugr bands from 2019 September 25 to November 27 with the Sinistro camera mounted on one of the 1 m telescopes of the LCO network in Siding Spring, Australia. Similar to our handling of the Swope photometry, we removed host-galaxy contamination by performing image subtraction for the g and r bands and by modeling the host flux in the same 5″ aperture for the u band.
We extracted UV light curves from a series of 39 Swift UVOT observations with the package HEASoft v6.27 and CALDB version 20200305. We estimated the counts of the source from a circular aperture of 5″ radius and the background from a circular aperture of 40″ radius using the task UVOTSOURCE. These were then converted to flux and magnitude with the Swift photometric calibration data (Poole et al. 2008;Breeveld et al. 2011).
The Swift target-of-opportunity (ToO) observations covered the evolution of AT 2019qiz from Δt = -11 to 168 days. Although the observations were made in all six UVOT filters (UVW2, UVM2, UVW1, U, B, and V ), in Figure 1 we only show the data from the three bluest filters (UVW2, UVM2, and UVW1), where the host-galaxy contributions are negligible.
These are also the only Swift filters used in our data analysis. The nondetections in the archival GALEX All-Sky Imaging Survey place an upper limit of FUV > 20 mag and NUV > 20.8 mag on the host-galaxy light. Given that the flux in the Swift UV filters is highly dominated by the TDE at the time of the observations, we did not attempt to subtract the host-galaxy flux from these bands.
We obtained a total of 18 spectroscopic observations with the Kast spectrograph (Miller & Stone 1993) on the Lick 3 m Shane telescope, the Goodman spectrograph on the SOAR telescope (Clemens et al. 2004), and LRIS on the Keck I 10 m telescope. Detailed instrumental configurations are listed in Table 1. We performed 1D spectrum extraction and flux calibration with standard PyRAFfoot_2 routines. Observations of standard stars BD+174708 and BD+284211 were used to determine the relative flux calibration and remove telluric features (e.g., Foley et al. 2003;Silverman et al. 2012;Dimitriadis et al. 2019). All of the spectra presented in this paper have been corrected for Galactic extinction. We calibrated each spectrum's absolute flux by comparing its gband synthetic photometry to the photometry from Swope and LCO imaging data (including host contribution), interpolated to each spectroscopic epoch.
We have not yet been able to obtain high-resolution spectra to measure the stellar velocity dispersion of the host galaxy. Nevertheless, N20 measured a velocity dispersion of σ = 69.7 ± 2.3 km s -1 from their late-time X-shooter spectrum. Using the scaling relation derived for a sample of low-mass galaxies from Xiao et al. (2011), this velocity dispersion corresponds to a black hole mass of M log 10 BH ( /M e ) = 6.16 ± 0.43. Different M-σ relations generally agree with an upper limit of M log 10 BH ( / M e ) 6.5 (Nicholl et al. 2020). As detailed below, the black hole mass derived from light-curve fitting is also consistent with this value. Therefore we use M BH = 1.4 × 10 6 M e when estimating relevant scales throughout the paper.
We also used the Python package Modular Open-Source Fitter for Transients (MOSFiT; Guillochon et al. 2018) to simulate the observed UV and optical light curves in Figure 1 to derive the physical parameters of the TDE, including the black hole mass. The TDE model implemented in MOSFiT estimates the bolometric luminosity of the TDE by converting the mass fallback rates from hydrodynamic simulations (Guillochon & Ramirez-Ruiz 2013) with a constant efficiency parameter (Guillochon et al. 2018;Mockler et al. 2019). The program then tries to match the observed flux in each band by reprocessing the bolometric flux with a blackbody photosphere, assuming that the blackbody photosphere evolves as a power law of the mass fallback rate.
The TDE model in MOSFiT has eight parameters: black hole mass (M BH ), stellar mass (M * ), scaled impact parameter (b), photosphere power-law exponent (l), photosphere radius normalization constant (R ph0 ), efficiency (ò), viscous delay time (t viscous ), and the fallback time of the most bound debris (t fallback ). We present the best-fit values from the MOSFiT run for AT 2019qiz in Table 2. The only constraint we imposed on the fitting parameters is for the stellar mass (M * ) to stay below 3 M e . This is because M * is not always well-constrained by MOSFiT owing to its degeneracy with the efficiency parameter, and it is physically unlikely for M * to exceed 3 M e since higher-mass stars tend to have shorter lifetimes.
Our derived parameters are generally consistent with those of N20, though the fitting was performed on a different set of optical data. Our results suggest a negligible viscous time, therefore t fallback approximates the start of the flare. We derived t fallback = -25 days relative to the time of peak light while N20 found t fallback = -27 days. This value is also consistent with that estimated from a photosphere expanding at constant velocity, where Δt = -31 ± 2 days (see Section 3.2). The black hole mass of M log 10 BH ( /M e ) = 6.14 ± 0.1 derived from MOSFiT is in good agreement with that estimated from the M-σ relation.
As a standard TDE analysis procedure, we model the spectral energy distribution (SED) of AT 2019qiz with a single-temperature blackbody. To do so, we first construct the SEDs at epochs having Swift observations in all three UV filters (UVW2, UVM2, and UVW1) and interpolate the photometry in the ugri filters measured from ground-based observatories. We show the best-fit blackbody temperature (T bb ) and the 10%-90% confidence level in Figure 2. We derive the bolometric luminosity (L bol ) at each Swift epoch by integrating over the best-fit blackbody spectra and calculate the emitting radius of the blackbody with the Stefan-Boltzmann law, where
4 . The evolution of L bol and R bb is also given in Figure 2. Our measurements are in good agreement with those of N20.
The blackbody temperature of AT 2019qiz initially had a constant value of T bb ≈ 1.9 × 10 4 K as the light curves approached maximum brightness (t peak ). From day 0 to day 25, it started to cool significantly down to T bb ≈ 1.4 × 10 4 K. Afterward, T bb slowly recovered back to ∼1.6 × 10 4 K and remained roughly constant out to day 100. This initial decline in T bb was also observed in ASASSN-14ae (Holoien et al. 2014) shortly after discovery around peak light, and in ASASSN-19bt while the light curves were rising (Holoien et al. 2019b). This cooling phase is short enough to be missed by previous TDE discoveries, where the classification and UV follow-up observations typically came around or after maximum light; hence, only the constant-temperature phase was observed.
Our derived peak luminosity, L peak = 4.9 × 10 43 erg s -1 , corresponds to an Eddington ratio of 0.27 assuming a black hole mass of 1.4 × 10 6 M e estimated from the M-σ relation. We determine a total radiated energy of 1.3 × 10 50 erg by extrapolating the luminosity linearly out to t = ±∞. If entirely powered by accretion, this energy would imply an accreted mass of
´-- 6.9 10 0.1 4 1
) M e , where ò is the accretion efficiency. The small accreted mass compared to the bound stellar mass could imply a low radiative efficiency or a partial disruption, or that we are underestimating the TDE energetics in other wavelengths (e.g., EUV or infrared). In fact, we see indications that the best-fit blackbody tends to underestimate the FUV continuum in our HST spectra. The luminosity evolution of AT 2019qiz places it in an unoccupied strip on the luminosity-phase plot that is between the "fast and faint" TDE iPTF16fnl and the slower but brighter TDE population (see Figure 1 The blackbody radius increased monotonically with time toward light-curve peak. We estimate an expansion velocity of 2700 ± 200 km s -1 for the photosphere. It is worth mentioning that the derived photosphere expands at a much slower rate than the BALs (10,000 km s -1 ) in the UV spectra, which suggests either a difference in the physical location of the outflow or that the two kinematic components are unrelated (see additional discussion in Section 4.1). If the photosphere was expanding at this constant velocity during the entire rise, this implies that the most bound debris fell back at t 0 = -30.6 days, which is well before the most recent ZTF preflare upper limit at Δt = -21.7 days. We estimated a similar reference time of -26 days by fitting the luminosity evolution during the rise time with a quadratic function µ -
) . The prediction of a quadratic rise traces back to the "fireball" model that is frequently used to describe the light curves of Type Ia supernovae (Riess et al. 1999). The TDE , which has a densely sampled TESS rising light curve, also exhibits a power-law rise with an exponent of ∼2 (Holoien et al. 2019b).
A sequence of three HST UV spectroscopic observations of AT 2019qiz is shown in Figure 3. Among the three HST epochs, the spectrum from the first epoch exhibits signatures distinct from those in previous TDEs. The apparent reduction in outflow velocity is also seen in the UV spectra for the first time. Below we describe these differences qualitatively and defer the quantitative details to the following subsections.
At all three epochs, high-ionization broad absorption lines (HiBALs) blueward of the rest wavelengths of N V, Si IV, and C IV can be readily seen. The HiBAL absorption troughs are blueshifted more in the first HST epoch (v ≈ 15,000 km s -1 on Δt = 13 days) and decelerated to v ≈ 10,000 km s -1 in the later two epochs (Δt = 65 and 98 days). In addition, the first HST spectrum is characterized by broad structures in the NUV (λ rest 1650 Å). We later identified these broad features to be associated with Al III λ1857 and iron (Fe II and Fe III) absorption, making AT 2019qiz the first TDE to be detected with Fe and low-ionization broad absorption lines (FeLo-BALs). In fact, we find the BAL pattern in AT 2019qiz to share more similarities with that in a superluminous supernova (SLSN) than with those of BAL quasars (BALQSOs), which we discuss in more detail in Section 3.3.2.
At the first HST epoch, we also detect narrow absorption lines with a dispersion of 240 km s -1 at the host redshift z abs ≈ z gal . The narrow absorption lines are only marginally detected at the later HST epochs owing to lower S/N. However, for the stronger absorption lines in the FUV, the line strengths do not seem to vary significantly between the first and second HST epochs.
We identified and measured the narrow absorption of both low-and high-ionization lines in the Δt = 13 days HST spectrum and tabulated them in Table 3. To interpret these narrow lines, we model them by defining the "effective continuum," which consists of the continuum emission and the broad TDE features (emission or absorption), and create a continuum-divided spectrum. This step is done by masking the parts of the spectrum containing narrow absorption lines, then smoothing and interpolating over the masked wavelengths that contain the narrow absorption lines using a Gaussian process with a squared exponential covariance function. We then divided the HST spectrum by this effective continuum, leaving only the narrow absorption features. Finally, we modeled these absorption lines with Gaussian profiles and measured their line widths and the equivalent widths (W r ) in the rest frame. Since the C IV λλ1548, 1551 resonance doublet is not resolved in the spectrum, we modeled it by requiring the two components to have the same width. We also fix the width of O I λ1302 and Si II λ1304 to be the same due to line blending. The smoothed effective continuum, the normalized spectrum, and the best-fit models are shown in Figure 4. From the line fit, we measured an S/N-weighted velocity offset of 70 ± 90 km s -1 that is consistent with the systemic host velocity using only the isolated and unblended absorption lines. We further place an upper limit of 200 km s -1 on the outflow velocity based on the weak Si II λ1265 feature. The average dispersion of the absorption lines is 240 ± 70 km s -1 .
We adopt the multi-ion single-component curve of growth (CoG) analysis to derive the ionic column densities from the measured W r of the narrow absorption lines assuming that the Doppler broadening parameter b is the same for each ionic species. This assumption holds if all the absorption, regardless of high-or low-ionization lines, took place in a relatively small region that cannot be resolved by our HST spectrum. We solve for b iteratively using the unsaturated, unblended ionic lines with the same lower energy states (Si IV, Fe II UV1, and Mg II). In our bestfit model (Figure 5), b = 95 km s -1 . The column densities for the uncontaminated and unblended ionic species are listed in Table 3.
We note that one caveat of applying the CoG analysis to a low-resolution spectrum, such as our STIS observation, is that the derived metal column densities can be underestimated by as much as 1 dex (Prochaska 2006). High-quality and highresolution spectroscopy of the afterglows of gamma-ray bursts (GRBs) suggests that the absorption-line structures around strong Fe II lines are characterized by multiple saturated components spreading over a few hundred km s -1 . In lowerresolution spectra, these absorption structures may be viewed as a single component with greater effective b. Although the larger b value does not necessarily lead to erroneous inference of the column density, it can sometimes affect the measurements by forcing weaker transitions to be optically thin (τ 0 ∝ b -1 ) even when they are saturated. Fine structures may also be present in the absorption lines in AT 2019qiz. Therefore, we emphasize that the metal column densities displayed in Table 3 should be considered as lower limits.
Before delving into the identification of broad features in AT 2019qiz, we remind readers that the continuum placement, f The transition may be subject to contamination by foreground absorption.
g Blended with neighboring lines in the same absorbing system. Therefore, we do not use the measured W r to infer the metal column density.
especially in the FUV, may be quite uncertain. Traditionally, we find that the NUV and optical photometry can be described well by a single-temperature blackbody with T bb ≈ a few × 10 4 K. This has been confirmed to be generally true by past HST observations, where the NUV spectra of most TDEs are characterized by a featureless continuum with a slope that is consistent with the Swift photometry in each event (Cenko et al. 2016;Brown et al. 2018;Blagorodnova et al. 2019;Hung et al. 2019). At λ rest < 1600 Å, the common presence of HiBALs often obscures a large amount of flux in the FUV continuum and broad emission lines in TDEs, making the FUV continuum placement especially difficult. It is noticed that the FUV continuum can be significantly underestimated by extrapolating the NUV blackbody spectrum, as was found in AT 2018zr (Hung et al. 2019) and also in AT 2019qiz (Figure 3). Therefore, in the analysis we avoid measuring line properties that are dependent on the continuum, such as the line flux or balnicity index (BI; a measure of the strength of the BAL features defined in Weymann et al. 1991), except for the narrow absorption lines, where the effective continuum can be determined relatively accurately. Despite the fact that the blackbody spectrum extrapolated from NUV and optical observations is clearly underestimating the FUV continuum, we keep them in Figure 3 to guide the eye. Broad emission and absorption features are detected at all three epochs. In order to identify the lines, it is natural to compare the HST spectra of AT 2019qiz with a BALQSO since both TDEs and BALQSOs are phenomena driven by accretion and winds. Furthermore, similarities between the rest-frame UV spectra of TDEs and BALQSOs have been drawn in the past (e.g., Blagorodnova et al. 2019). We find broad structures in both the FUV and NUV sides of the spectrum on day 13. Broad emission lines seem to have formed around the rest wavelengths of the high-ionization lines N V, Si IV, and C IV. Blueward of these emission lines are the absorption troughs associated with these high-ionization transitions. We estimate For each wavelength segment, the original spectrum (blue) and the smoothed effective continuum (red) are shown in the upper panel while the normalized spectrum is shown in the lower panel. We identified the line transitions (labeled in black) and modeled the normalized narrow absorption lines with Gaussian profiles, where the best-fit models are shown in orange. In the panels with normalized spectra, the gray ticks mark the wavelengths of common Galactic absorption lines.
the blueshift velocity of the broad absorption features (v w ) with respect to the minimum of the troughs. The broad features associated with Si IV and C IV are blueshifted by v w ≈ 15,000 km s -1 , whereas the velocity offset of the N V absorption trough cannot be precisely measured owing to overlap with the geocoronal emission.
A subclass of BALQSOs, the low-ionization broad absorption line (LoBAL) QSOs, have UV spectra imprinted by broad absorption of Mg II λλ2796, 2803, Al III, and AlII at λ rest > 1750 Å. In even rarer cases, iron absorption lines are present in the QSO spectra, and thus these QSOs are termed "FeLoBAL" QSOs. Given the presence of similar ionic species, we compare AT 2019qiz with the spectrum of an SDSS FeLoBAL QSO at z ≈ 2.1 in Figure 6. From this comparison, we infer that the absorption around 1800 Å in AT 2019qiz is likely due to a blueshifted (v w ≈ 9000 km s -1 ) Al III λ1857 line. At λ rest > 2000 Å, the Δt = 13 days spectrum of AT 2019qiz is characterized by several broad peaks and valleys around the most prominent Fe II and Fe III lines. The iron lines that are possibly associated with the broad NUV features are labeled by dashed lines in Figure 6. If we attribute the absorption features at 2000 and 2240 Å to Fe III and Fe II UV2, the measured wavelengths correspond to v w ≈ 12,000-16,000 km s -1 , which is in good agreement with that derived from the HiBALs. Unlike the BALQSOs, the HST spectra of AT 2019qiz did not show any significant broad, blueshifted Mg II λλ2796, 2803 absorption at any phase (Figure 6).
Intriguingly, we find the Δt = 13 days UV spectrum of AT 2019qiz to bear greater resemblance to the SLSN Gaia16apd (Yan et al. 2017) than to any type of BALQSOs (Figure 6). The similarity between the two spectra possibly stems from the fact that both events had a similar blackbody temperature (T bb ≈ 17,000 K) and an expanding photosphere, as probed by the SED fit in Section 3.2, during the early phases.
The absorption troughs at 1750 Å < λ rest < 2400 Å aligned particularly well with Gaia16apd after redshifting the SLSN spectrum by 12,000 km s -1 . Using the syn++ model that generates synthetic spectra for homologously expanding atmospheres, Yan et al. (2017) identified the ionic species that contribute to each absorption line in Gaia16apd. This new comparison makes a difference in the identification of the features at 1800 and 2240 Å. If the ionization state is also similar in AT 2019qiz and Gaia16apd, the 2240 Å feature would instead be attributed to the absorption of C II with v w ≈ 11,000 km s -1 . In addition, the 1800 Å feature would not be uniquely associated with Al III, but also with Si II and Ti II. The absorption troughs around the Si IV and C IV lines may also be contributed by other low-ionization species such as C II and Si II. While many LoBALs are detected in the day 13 UV spectrum, the lack of Mg II λλ2796, 2803 in AT 2019qiz would possibly need to be explained by the composition.
The broad UV features evolved significantly from day 13 to day 65. The most remarkable difference is that the absorption troughs associated with the high-ionization Si IV and C IV λλ1548, 1551 lines became redder, decelerating from v w ≈ 15,000 km s -1 to v w ≈ 10,000 km s -1 (Figure 7). We also notice that the NUV spectrum became featureless and the UV spectra resemble those of previously observed TDEs more at later epochs. The spectra of day 65 and day 98 are very similar except that the bluest Si IV absorption edge seems to have moved to a slightly lower velocity on day 98 (Figure 7).
We show a sequence of 18 optical spectra of AT 2019qiz in Figure 8. The spectroscopic features in AT 2019qiz underwent significant changes on a timescale of days to weeks. This is best illustrated by the evolution in the shape of the Hα emission (Figure 9). The pre-peak and early post-peak spectra from day -13 to day 13 are characterized by a blue continuum with very broad emission features around Hα and He II λ4686. The latter is likely contaminated by the emission from Hβ because the entire broad feature spans rest wavelengths 4400-5200 Å. During this stage, the Hα emission structure has an FWHM intensity of 2.6 × 10 4 km s -1 . We fit this broad Hα feature with a double Gaussian function and derived two components with similar FWHMs of ∼1.5 × 10 4 km s -1 centered roughly at the rest wavelength of Hα and at a redshift of 1.6 × 10 4 km s -1 . The red component has a weaker line flux, about half that of the blue component.
On day 22, the Hα emission-line profile changed dramatically, with a narrow component appearing around the rest-frame Hα wavelength. We then modeled the Hα emission with three Gaussians simultaneously while requiring the narrow component to have FWHM < 6000 km s -1 . The best-fit model consists of a broad Hα base, which is the sum of two Gaussians as in the earlier epochs, with FWHM of 2.1 × 10 4 km s -1 and a narrow Hα component with FWHM of 3000 km s -1 centered at zero velocity. During this time, the blue and red components of the broad Hα emission became almost equally strong while the velocity offset of the red component decreased to 1.1 × 10 4 km s -1 .
The broad Hα base continues to decay while the narrow Hα remains strong in all later epochs. The broad Hα showed not only showed a decrease in line flux but also a change in the line shape. From day 71 onward, the broad Hα shifted to a single Gaussian centered around the rest Hα wavelength, where the red shoulder detected with v > 10,000 km s -1 at previous epochs disappeared. The FWHM of the broad Hα dropped to ∼12,000 km s -1 . The FWHM of the narrow Hα emission fluctuates between 3000 and 4000 km s -1 since it first appeared around day 22.
Narrow Balmer emission lines from higher excited states and narrow Bowen emission (He II + N III) likely also appeared from day 22 onward. However, only after day 49 did the broad emission and continuum fade enough for the narrow lines to be detected. The presence of both the Balmer lines and He II + N III makes AT 2019qiz a TDE-Bowen, which is a spectral class that makes up about half of the optical TDE population defined by van Velzen et al. (2021). On day 71, when the narrow emission lines are the strongest, we measured an FWHM of 4000 km s -1 for Hβ and an FWHM of 3100 km s -1 for Hγ. The line intensity ratio Hα/Hβ ≈ 1.2 corresponds to a flat Balmer decrement, which has also been reported in several TDEs (e.g., AT 2018hyz; Hung et al. 2020;Short et al. 2020). We also measured intensity ratios Hγ/Hβ ≈ 0.6 and (He II + N III)/Hβ ≈ 1.
The host galaxy of AT 2019qiz is 2MASX J04463790-1013349 (WISEA J044637.88-101334.9) at z = 0.0151. The r-band Pan-STARRS image reveals that it is a spiral galaxy with a clear bar Figure 6. UV spectra of AT 2019qiz on day 13 and day 65, compared to the SLSN Gaia16apd (Yan et al. 2017) and the FeLoBAL QSO SDSS J090152.04+624342.6 (Wang et al. 2017). The first UV spectrum of AT 2019qiz bears a stronger resemblance to that of Gaia16apd. The orange labels mark the potential contributing species identified by Yan et al. (2017).
Figure 7. Three epochs of HST STIS spectra in velocity space with respect to the rest wavelengths of the high-ionization lines N V, Si IV, and C IV. The region affected by geocoronal airglow emission indicated by a circled plus symbol is truncated. The UV spectrum from the first HST epoch (Δt = 13 days) is shaded in pale blue while the spectra from the second and the third epochs are plotted as orange and green curves, respectively. It can be seen that the absorption troughs shifted to lower velocities as time progressed from 13 to 65 days, which makes the emission lines seem more redshifted.
structure. N20 measured a Sérsic index in the range 5.2-6.3, which is considered high compared to the galaxy sample in the same black hole mass bin (Law-Smith et al. 2017). Like many other TDEs, AT 2019qiz is in a galaxy with a more concentrated population of stars (French et al. 2020).
In the late-time optical spectra of AT 2019qiz, we detect nebular emission lines [O III] λλ4959, 5007 and [N II] λλ 6548, 6583 that are indicative of preflare activity in an active galactic nucleus (AGN). Given that the TDE light has faded substantially in the Δt = 340 days spectrum, we fit its stellar and gas kinematics simultaneously using the Penalized Pixel-Fitting (pPXF) package (Cappellari 2017) with the MILES stellar library (Figure 10). In addition to the narrow nebular emission lines, we added a second gas component in our fit to account for the TDE Balmer emission at late times, which has also been observed in previous TDEs (e.g., AT 2018zr; Hung et al. 2019). Our best-fit width for the TDE Balmer lines (FWHM ≈ 2000 km s -1 ) is consistent with the width of the narrow Hα component in earlier TDE spectra (22 days Δt 115 days). We measured the flux of the host-galaxy emission lines and plot the line ratios in a Baldwin, Phillips, & Terlevich (BPT) diagram (Baldwin et al. 1981) (Figure 10). The line ratios are in good agreement with the nebular emission lines being produced by AGN photoionization.
During the flare, N20 measured a weak X-ray luminosity of L X = 5.1 × 10 40 erg s -1 in AT 2019qiz that is 2-3 orders of magnitude below the UV and optical luminosity. N20 associated the X-ray emission with the TDE rather than the AGN owing to variability in the X-ray flux and the evolution of the hardness ratio. However, the hardness ratio of -0.1 ± 0.04 from the merged XRT observations is at the higher end of the TDE distribution when compared to X-ray-bright TDEs such as ASASSN-14li, and is more similar to those seen in AGNs (Auchettl et al. 2018). Therefore, the possibility of AGN-driven X-ray emission in AT 2019qiz cannot be completely ruled out.
The UV and optical photometry reveals the evolution of the blackbody radius from the pre-peak phase out to Δt ≈ 100 days. The blackbody radius was initially expanding linearly with a velocity of 2700 km s -1 up to the time of peak light in the UV and optical. The constant-velocity phase was then followed by a constant-radius phase for ∼25 days before the radius began to shrink with time like that observed in previous TDEs, as a result of the combination of decreasing luminosity and constant color temperature.
Our intensive follow-up observations of the optical photosphere and the BALs in the UV spectra strongly support the presence of an outflow that evolved significantly with time in AT 2019qiz. Various models have been proposed to explain the origin of TDE outflows (Strubbe & Quataert 2009;Lodato & Rossi 2011;Dai et al. 2018;Lu & Bonnerot 2020). In such models, the wind typically has a nonspherical configuration, and the velocity can have some variation depending on the wind-launching radius and observer inclination. However, when strong optical emission is observed, one can assume that the wind is mass-loaded and optically thick along the line of sight, thus justifying the adoption of a quasispherical model for studying the wind physics (Metzger & Stone 2016;Roth et al. 2016). Here we construct a simple 1D wind model, where the mass outflow rate is (2011), stating that a stronger wind is launched at earlier phases in TDEs when the mass fallback rate is higher. In the analytical model developed by Metzger & Stone (2016), only a small fraction of the bound debris accretes onto the black hole. The binding energy of the accreted mass naturally gives rise to a quasispherical outflow that is massive enough to absorb and reprocess the hard EUV/X-ray photons into UV and optical emission. One prediction of the model of Metzger & Stone (2016) is that the TDE light curve can deviate from the mass fallback rate (∝ t -5/3 ) at early times owing to photon trapping. They find that the observed luminosity can be suppressed at first because of adiabatic losses in the inner wind, resulting in a flatter light curve than ∝ t -5/3 . The suppression continues until the trapped photons are advected to a radius where the photon diffusion time is short compared to the outflow expansion time on a timescale of
3 5 / / / / . At this point, photons can diffuse out of the ejecta freely and thus allow the photosphere to cool. The effect of photon trapping aligns with the observed T bb evolution of AT 2019qiz, where the measured temperature started to decrease at around one fallback time. This may also explain why the cooling in ASASSN-14ae and ASASSN-19bt occurred at different phases, since the trapping time is dependent on other parameters.
The balance between decreasing luminosity and decreasing temperature gives rise to the plateau in the blackbody radius from Δt = 0 to 25 days. In practice, the photosphere velocity does not need to follow the outflow velocity if the density or ionization of the photosphere is changing. That is, even if the gas producing the UV and optical continuum continues to expand at a constant velocity, a reduction in the electron density can move the electron-scattering photosphere inward, and can lead to a plateau or even a decrease in the derived blackbody radius as observed in AT 2019qiz.
Blueshifted broad absorption features detected in AT 2019qiz signify the presence of outflowing gas at a phase as early as Δt = 13 days. The ejecta producing the BALs would have reached a distance of ∼5 × 10 15 cm if they were moving at the same velocity since the most bound debris returned to the pericenter (assuming t fallback = -25 to -30 days). BALs have also been detected at different stages in almost all of the TDEs observed with UV spectroscopy, including PS1-11af, iPTF15af, iPTF16fnl, and AT2018zr (Chornock et al. 2014;Brown et al. 2018;Blagorodnova et al. 2019;Hung et al. 2019). the HST TDE sample, ASASSN-14li is the only object that exhibits pure broad emission lines (Cenko et al. 2016).
To date, iPTF16fnl (Δt = 7, 22, 44 days), AT 2018zr (Δt = 23, 36, 41, 59, 62 days), and AT 2019qiz (Δt = 13, 65, 98 days) are the only TDEs with a sequence of HST UV We used models with different numbers of Gaussian components to fit the line at different phases. The Hα line profile in the earliest spectra can be characterized by two broad Gaussians (blue). In the post-peak period, a narrow component became apparent, which motivated us to adopt a three-component model (pink) to describe the line shape. In later spectra, the red shoulder in the broad Hα disappeared; thus, we use a narrow and a broad component centered around Hα to model the emission line (green).
spectra (Brown et al. 2018;Hung et al. 2019). Among these, AT 2019qiz has the longest observational baseline. Each of these TDEs follows a very different evolutionary path. In the earliest HST epoch, iPTF16fnl was observed with weak HiBALs (FWHM ≈ 6000 km s -1 ) that diminished over time. AT2018zr started resembling ASASSN-14li with a UV spectrum characterized by broad emission lines. Broad absorption troughs blueward of both high-and low-ionization species with a velocity offset of ∼0.05c started to appear from day 59 onward. AT 2019qiz exhibited a rare FeLoBAL spectrum at the first epoch, which then transitioned into a HiBAL spectrum at the two later epochs. The HiBALs remained strong in AT 2019qiz though the outflow velocity decreased from 15,000 to 10,000 km s -1 in ∼50 days (Figure 7). The decrease in outflow velocity is consistent with the theoretical prediction, where the wind becomes weaker with time as the mass fallback/accretion rate declines.
We select the most representative phases from the multiepoch HST spectra of iPTF16fnl, AT2018zr, and AT2019 qiz and compare them with the single-epoch spectra obtained for all the rest of the TDEs shown in Figure 11. While we arrange the spectra in Figure 11 with respect to phase, we find that any two spectra obtained at comparable phases do not guarantee similarity.
The spectrum that stands out in the current UV TDE population is the UV spectrum of AT 2019qiz at Δt = 13 days (Figure 11). The presence of FeLoBALs in the UV spectrum of AT 2019qiz is in accordance with the reprocessing scenario. Photoionization models of FeLoBAL QSOs, which are typically X-ray-faint, suggest that high column densities are required in order to produce the iron features (e.g., N H ≈ 10 20.6 cm -2 ; Korista et al. (2008)). Analogously, the BAL clouds need to be shielded from the hard X-ray photons to form the FeLoBAL in AT 2019qiz. The shielding gas may be an inner wind consisting of material that has fallen back more recently, as proposed in the model of Metzger & Stone (2016). In a few fallback times, the column density drops sufficiently low such that the ejecta become transparent to the X-ray and EUV photons. This likely explains why the iron and lowionization features disappeared, leaving only the HiBALs in later HST spectra of AT 2019qiz.
As shown in Figure 6, this spectrum is more like that observed in SLSNe instead of in TDEs or BALQSOs. According to the model of Metzger & Stone (2016), the outflow properties of TDEs may be similar to those in enginedriven supernovae as black hole accretion continues to inject energy into the ejecta. The observed similarities between the early-time UV spectral features in AT 2019qiz and those in the SLSN Gaia16apd are strong evidence supporting this connection.
The HiBAL TDE spectra, which are seen in at least half of the TDEs at a later phase, all seem to be very similar except for small differences in outflow velocities (6000-15,000 km s -1 ). The reduction in the mass fallback rate and the expanding ejecta likely facilitated changes in the ionization structure and the column densities that cause the UV spectral features to transition from SLSN-like to more BALQSO-like. It is worth noting that the C IV and Si IV emission always appears to be weaker than the N V emission in TDEs while the same trend is not observed in AGNs and BALQSOs. This may be attributed to the stellar debris having a higher N/C ratio owing to CNO processing in the stellar core (Cenko et al. 2016;Kochanek 2016;Yang et al. 2017;Gallegos-Garcia et al. 2018;Law-Smith et al. 2019, 2020).
The diverse properties of the TDE UV spectra perhaps reflect the intricate processes involved in a TDE, such as accretion disk formation and wind launching. In recent theoretical developments, studies have shown that orientation effects can, at least in part, explain the diverse broadband and lineemission/absorption properties seen in TDEs (Dai et al. 2018;Parkinson et al. 2020). Parkinson et al. (2020) simulated spectra with broad emission lines (BELs) and BALs with a disk and wind model for TDEs. In particular, they rendered spectra that consist of purely BELs on sightlines that do not intersect the disk wind. While viewing angle may be responsible for the dichotomy of BELs and BALs in TDE spectra, it does not explain how one object can transition from BEL to BAL, as seen in AT 2018zr during the first 60 days (Hung et al. 2019). The SLSN-like early-time spectrum of AT 2019qiz further calls for a timedependent, multizone wind model to be considered in future TDE simulations. Including AT 2019qiz, the high occurrence of BALs in TDE spectra (c.f. QSOs) continues to support a wide-angle wind geometry in TDEs (Hung et al. 2019;Parkinson et al. 2020).
The accretion disk, outflow, and bound stellar debris are all possible production sites of Hα emission in a TDE. At a glance, the broad Hα component in AT 2019qiz may be reminiscent of a disk line profile with an intermediate inclination angle (i) such that the peaks corresponding to material moving away and toward the observer are not resolved (e.g., Liu et al. 2017;Holoien et al. 2019a). If due to rotation, the measured separation of ∼15,000 km s -1 between the two fitted peaks in the pre-peak epochs implies a disk size of 400 i sin 0.5 2 ( ) R g . However, the fast evolution of this broad component, with FWHM shrinking by as much as ∼50% in roughly 80 days, is hard to reconcile with the disk origin since changes in the disk geometry are expected to occur on longer timescales. In the double-peaked TDE AT 2018hyz, the Hα line widths remained roughly constant for at least 100 days (Figure 12; Hung et al. 2020).
The width of the broad Balmer emission in TDEs is typically of the order of 10 4 km s -1 . This value is comparable to the expected outflow velocity (e.g., Strubbe & Quataert 2009;Metzger & Stone 2016) in a TDE and therefore hints at the association. Roth & Kasen (2018) first proposed that Figure 11. A compilation of TDE UV spectra normalized to rest wavelength 1700 Å. Most of the TDE UV spectra show fast-moving (v ≈ 10 3 -10 4 km s -1 ) BAL features that signify the presence of outflows at some point of their evolution. The BAL features in TDEs are typically seen around high-ionization lines such as C IV λλ1548, 1551 and Si IV λ1397 yet are usually absent from their NUV spectra. The Δt = 13 days spectrum of AT2019qiz is distinct from other known TDEs by having broad Fe features in the NUV. Note that the scale on the abscissa has been compressed for the NUV segment (1800-3000 Å) to allow more detailed examination of the FUV portion of the spectra. asymmetric emission lines may result from an optically thick, outflowing gas that is expected to be common in TDEs. The resulting line lacks the blueshifted absorption portion of a P Cygni line profile when the excitation temperature of the line is higher than the brightness temperature of the photosphere. Indeed, progressively more TDEs are found to have emissionline shapes deviating from a simple Gaussian (e.g., ASASSN-14ae, AT2018 zr, AT2018 hyz). Hung et al. (2019) found a good fit to the flat-topped Hα emission in AT2018 zr with this model, though an elliptical disk within a certain parameter space also produces such line shapes (Holoien et al. 2019a). N20 explored this possibility by comparing the model of Roth & Kasen (2018) with the spectra of AT 2019qiz. In summary, N20 found the model with an outflow velocity of 5000 km s -1 to be the best match to the Hα line profile at earlier and later epochs in AT 2019qiz. However, at the epochs where the Hα emission is the strongest, the model of Roth & Kasen (2018) fails to replicate the strong red wing of the line profile. The discrepancy may be resolved by adopting a higher opacity, changing the photosphere radius, or allowing for asymmetries in the outflow.
The potential correlation between the HST UV spectral features and the broad Hα component in AT 2019qiz could lend further support to the outflow origin of the broad Hα emission. We observed a reduction in the velocity of both the HiBALs in the UV and in the broad Hα component in the optical (Figure 12). The optical spectra with a higher cadence than the HST UV spectra find the FWHM to decrease from ∼22,000 km s -1 on Δt = 13 days to 15,000 km s -1 on Δt 48 days. This change is well reflected in the Δt = 65 days HST spectra, which decelerated by ∼5000 km s -1 since the first HST observation on Δt = 13 days.
As mentioned, we are unable to constrain the variability in the narrow absorption lines confidently owing to the lower S/N at the later HST epochs. However, the FUV spectrum from the second HST epoch does have sufficient S/N to reveal stronger absorption lines with W r 1.0. We find these lines to have comparable strengths in the day 13 and day 65 spectra.
GRB afterglow spectra have long been exploited to study the intervening interstellar medium (ISM) and circumgalactic medium along the line of sight (e.g., Prochaska et al. 2007;Gatkine et al. 2019). To shine light on the origin of the narrow UV absorption lines in AT 2019qiz, we compare its spectrum with that of the GRB afterglow composite (Figure 13; Christensen et al. 2011). In AT 2019qiz, we derived N(H I) = 5 × 10 17 cm -2 from the CoG analysis. This column density is consistent with a Lyman-limit system, where the cloud starts to be dense enough to shield itself against the EUV background and remain neutral. GRB afterglow spectra, on the other hand, are often associated with a damped Lyα (DLA) system with a high neutral hydrogen column, > N log H 20.3 I ( ( )) cm -2 . The DLA line is dominated by the Voigt profile, which is very different from the Gaussian-like Lyα line profile in AT 2019qiz (Figure 13).
We observe stronger Si IV and C IV absorption in AT 2019qiz than in the GRB composite. The strengths of the low-ionization species (Fe II and Mg II) are comparable to those in GRB afterglows. While the N V λλ1238, 1242 absorption lines are well detected in AT 2019qiz, they are harder to identify in GRB spectra owing to blending with the red wing of Lyα. The high ionization potential of N 4+ (77 eV) makes it difficult to produce in stellar radiation fields. Galactic halos and disks do have N V absorption, but none of them show a column density >10 14.5 cm -2 (Prochaska et al. 2008, and references therein). Our measured N 4+ column density of 10 15 cm -2 thus likely indicates a hard ionizing spectrum, which can be attributed to the TDE accretion disk or in part to the underlying weak AGN. The strong ionizing source may also explain the stronger Si IV and C IV W r in AT 2019qiz than in the GRB composite. These high-ionization lines are likely to have formed in regions close to the SMBH rather than in the ISM. Our HST spectra cannot resolve fine-structure transitions such as Si II * and Fe II * in AT 2019qiz. In GRB afterglows, these fine-structure lines are thought to be produced by UV pumping and uniquely trace gas in the vicinity (100-1000 pc) of the burst (Prochaska 2006). We expect these lines to be detected in high-S/N, high-resolution TDE spectra since TDEs can also create a temporary UV radiation field.
Given that the redshift of the absorption lines is close to the host redshift, we consider both circumnuclear gas and the bound stellar debris as the candidate absorbing system. In ASASSN-14li, Cenko et al. (2016) also detected narrow absorption lines with a width of ∼500 km s -1 , comparable to that in AT 2019qiz. However, the absorption lines in ASASSN-14li are blueshifted by 250-400 km s -1 while the lines in AT 2019qiz do not appear to have a systemic velocity shift with respect to the host. The abundance pattern in ASASSN-14li is also very different. The low neutral hydrogen column of
cm -2 is somewhat expected in host galaxies with an old stellar population (Cenko et al. 2016). However, ASASSN-14li lacks the low-ionization Mg II λλ2796, 2803 absorption that is commonly observed in cold ISM in the galaxy while having well-detected highionization lines. Together with the blueshift velocity being consistent with the outflow velocity measured from X-ray spectroscopy, Cenko et al. (2016) suggest that the low-velocity absorber is more likely the bound stellar debris on an elliptical orbit (Miller et al. 2015).
The absorber in AT 2019qiz has several properties different from those in ASASSN-14li. The low-ionization species, such as Fe II and Mg II, are readily detected in AT 2019qiz with W r comparable to those observed in GRB afterglow and QSO spectra, where the absorbing systems have been associated with galaxies. The likely lack of variability also counters the scenario in which the absorbing gas is the bound stellar debris. Clumpiness or bulk motion of the debris stream could easily alter the gas density and therefore lead to variability in the absorption lines on the fallback timescale t fallback ≈ 40 days. Therefore, we conclude that the absorption lines in AT2019 qiz are more likely to be probing the host ISM and circumnuclear gas instead of the stellar debris as in ASASSN-14li.
The early classification of AT 2019qiz and its proximity allow us to perform a detailed, multiwavelength study of a TDE. Our UV and optical follow-up observations in both photometric and spectroscopic modes offer new insights for optically selected TDEs. AT 2019qiz is located in a galaxy with a weak AGN, as evidenced by the line ratios in the BPT diagram. The evolution timescale and luminosity of AT 2019qiz are fast and faint, making it an intermediate event between iPTF16fnl and the rest of the TDE population.
Both the photosphere evolution and the BALs in the HST spectra strongly corroborate the outflow scenario in TDEs. We find the photosphere to be expanding at a constant velocity v ≈ 2700 km s -1 before reaching t peak . The early luminosity evolution in AT 2019qiz follows a power law (∝t 2 ), consistent with the prediction of an expanding fireball. The BALs are detected at all three HST epochs with a decrease in outflow velocity from 15,000 to 10,000 km s -1 . This deceleration is expected as the mass fallback rate declines at later times.
The UV spectrum at the first HST epoch is unlike that of all the TDEs observed in the past, because it contains both HiBAL and FeLoBAL features such as Al III and Fe II. The UV spectrum of AT 2019qiz resembles that of the SLSN Gaia16apd at maximum light more than that of any known FeLoBAL QSO. However, the broad Mg II λλ2796, 2803 doublet, which is readily found in both Gaia16apd and FeLoBAL QSOs, is not detected in AT 2019qiz. The disappearance of the FeLoBAL features at the two later HST epochs of AT 2019qiz is likely due to the thinning of the X-ray/EUV shielding material as the mass fallback/accretion rate decreases with time.
The detection of He II λ4686 and the Balmer emission since the beginning of the monitoring up to day 89 makes AT 2019qiz a TDE-Bowen in the spectral classification scheme proposed by van Velzen et al. (2020). We carefully studied the Hα emission and found the line profile to be characterized by a very broad component with FWHM 10 4 km s -1 and a narrower component with FWHM ≈ 3000 km s -1 that only appeared after Δt 22 days. The broad Hα component narrows with time as observed in other TDEs and becomes progressively more symmetric around the rest wavelength of Hα. We inferred that the broad Hα evolution is most likely to be driven by the outflow given the similar value of the Hα line width and the BAL velocity, and the fact that the timing of the decrease in broad Hα width is seen to be coincident with the decrease in wind velocity probed by the BALs.
We also analyzed the narrow absorption lines in the first HST spectrum and derived ionic column densities for several species using the CoG method. The measured Lyα W r corresponds to a Lyman-limit system, where the highionization species have stronger W r and the low-ionization lines have similar W r to those in GRB afterglows. We concluded that the narrow absorption lines are more likely to be probing the gas in the host galaxy rather than the bound stellar debris as in the case of ASASSN-14li. The UV spectra of AT 2019qiz demonstrate the potential of using TDEs to probe circumnuclear gas and the ISM in TDE host galaxies. Future high-S/N and high-resolution UV spectra will allow us to resolve the fine-structure lines produced by UV pumping in TDEs and study the gas in the vicinity of the SMBHs.
We thank anonymous referee for helpful suggestions that greatly improved the quality of the paper. T.H. is grateful to Ari Laor, Alexei Baskin, and Frederick Hamann for generously sharing their insightful thoughts on the UV spectral signatures in AT 2019qiz and providing BALQSO templates for comparison. She also thanks Dan Perley for helpful discussion on the similarities between AT 2019qiz and Gaia16apd. T.H. thanks Elizabeth Nance and John Debes for their help with scheduling the HST ToO observations. The UCSC transient team is supported in part by National Science Foundation (NSF) grant AST-1518052, National Aeronautics and Space Administration (NASA) Swift grant 80NSSC19K1386, the Gordon & Betty Moore Foundation, the Heising-Simons Foundation, and by a fellowship to R.J.F. from Note. All the data presented in this table have been corrected for Galactic extinction. We did not perform host subtraction on the Swift UV data since the host contribution is negligible. On the other hand, the ZTF, LCO, and Swope data are all host-subtracted.
The Astrophysical Journal, 917:9 (22pp), 2021 August 10 Hung et al.
https://Mars.lco.global
http://www.stsci.edu/institute/software_hardware/pyraf