iPTF16fnl: A Faint and Fast Tidal Disruption Event in an E+A Galaxy

iPTF16fnl was discovered on UT 2016 August 29.4 in an image obtained during the g+R experiment (Miller et al. 2017): during the night, one image each was obtained in the g and R bands; the images are separated by at least an hour in order to filter out asteroids. The event was identified by two real-time difference-imaging pipelines (Cao et al. 2016; Masci et al. 2017), shown in Figure 1. The discovery magnitudes were g = 17.11 ± 0.09 and R = 17.39 ± 0.09 (see Ofek et al. 2012 for photometric calibration of PTF). Given an rms of 05, the coordinates of the source, ${\alpha }_{{\rm{J}}2000}={00}^{{\rm{h}}}{29}^{{\rm{m}}}57\buildrel{\rm{s}}\over{.} 04,\ {\delta }_{{\rm{J}}2000}=+{32}^{{\rm{h}}}{53}^{{\rm{m}}}37\buildrel{\rm{s}}\over{.} 5$ (ICRS) are consistent with a central position of the galaxy, as provided by the SDSS catalog (Alam et al. 2015).

Zoom In Zoom Out Reset image size

The brightness, blue color ($g-R=-0.28$ mag), and central location of the transient in its host galaxy made it a prime candidate for prompt follow-up observation. On the night after discovery, we observed the source with the FLOYDS spectrograph (Sand et al. 2011) on the Las Cumbres Observatory (LCO; Brown et al. 2013) 2 m telescope and the Spectral Energy Distribution Machine (SEDM) on the Palomar 60 inch (P60) telescope. FLOYDS are a pair of robotic low-resolution (R ∼ 450) slit spectrographs optimized for supernova classification and follow-up. The SEDM is a ultra-low resolution (R ∼ 100) integral-field-unit (IFU) spectrograph, dedicated to fast turnaround classification (N. Blagorodnova et. al. 2017, in preparation). The IFU's wide field-of-view of 28'', allows for robotic spectroscopy. Their data reduction pipeline allows rapid reduction with minimal intervention from the user (aperture placing). Figure 2 shows the classification spectra, displaying a blue continuum and broad He ii and Hα emission lines, characteristic of previously observed optical TDE spectra (Arcavi et al. 2014). The fast spectral identification of iPTF16fnl as a TDE candidate allowed us to rapidly inform the astronomical community (ATel #9433; Gezari et al. 2016), which in turn enabled numerous multi-wavelength follow-up campaigns.

Zoom In Zoom Out Reset image size

The host galaxy of iPTF16fnl is Markarian 950 (Mrk 950) located at z = 0.016328 (Cabanela & Aldering 1998; Petrosian et al. 2007). Given the edge-on host inclination and the existence of a peculiar bar and nucleus, the galaxy is classified as Sp (spindle). The luminosity distance is D_L = 66.6 Mpc (distance modulus μ = 34.12 mag) using H₀ = 69.6 km s⁻¹ Mpc⁻¹, Ω_M = 0.29, ${{\rm{\Omega }}}_{{\rm{\Lambda }}}=0.71$ in the reference frame of the 3 K cosmic microwave background (CMB; Fixsen et al. 1996).

The estimated Galactic color excess at the position of the transient is $E(B-V)=0.062\pm 0.001$ mag (from NASA/IPAC Extragalactic Database; NED²⁷ ), after adopting the extinction law of Fitzpatrick (1999) with corrections from Schlafly & Finkbeiner (2011). Assuming ${R}_{{\rm{V}}}=3.1$, the Galactic visual extinction is ${A}_{{\rm{V}}}=0.192$ mag.

The archival host magnitudes are shown in Table 1. The derived K-band luminosity of the galaxy is ${L}_{K}=1.15\times {10}^{10}$ L_⊙. Using the galaxy color $u-g=1.5$, we compute a mass-to-light ratio for the K-band, ${\mathrm{log}}_{10}(M/L)=-0.0755$ (Bell et al. 2003), corresponding to a stellar mass ${M}_{* }\simeq 9.7\times {10}^{9}$ M_⊙. Provided that the size of the PSF in SDSS DR13 is an upper limit for the angular size of an unresolved bulge, we assume that an upper limit for the bulge-to-total ratio (B/T) can be estimated with an average (across all bands) value of psfFlux/cModelFlux ∼ 0.19 ± 0.05. We use this ratio to scale the total galaxy mass and derive ${M}_{\mathrm{BH}}\leqslant {10}^{6.6\pm (0.1+0.34)}$ M_⊙, according to the ${M}_{\mathrm{BH}}-{M}_{\mathrm{bulge}}$ relation (McConnell & Ma 2013), including the 1σ scatter of 0.34 dex.

Table 1.  Archival Photometry of Mrk 950

Survey	Band	Magnitude	Reference
		(mag)
GALEX	FUVAB	21.22 ± 0.39a	[1]
GALEX	NUVAB	20.191 ± 0.13a	[1]
SDSSDR12	u	17.005 ± 0.012b	[2]
SDSSDR12	g	15.491 ± 0.003b	[2]
SDSSDR12	r	14.913 ± 0.003b	[2]
SDSSDR12	i	14.585 ± 0.003b	[2]
2MASS	J	13.212 ± 0.040	[3]
2MASS	H	12.545 ± 0.052	[3]
2MASS	K	12.360 ± 0.069	[3]
WISE	W1	13.075 ± 0.029	[4]
WISE	W2	13.086 ± 0.034	[4]
WISE	W3	12.331 ± 0.285	[4]
WISE	W4	>9.136	[4]
NVSS	1.4 GHz	<1.7 mJy	[5]
ROSAT	0.1–2.4 keV	$\lt 1.04\times {10}^{-12}$ erg s−1	[6]

Notes.

^aMeasured within 75 diameter aperture. ^bModel magnitude.

References. [1] Bianchi et al. (2011), [2] Alam et al. (2015), [3] Jarrett et al. (2000), [4] Wright et al. (2010), [5] Condon et al. (1998), [6] Voges et al. (1999).

Download table as:  ASCIITypeset image

The age and metallicity of the host galaxy were determined by fitting a grid of galaxy models from Bruzual & Charlot (2003), using stellar evolutionary models from (Chabrier 2003). The fit was done using the pPXF code (Cappellari 2017). The best model was in agreement with a single burst of star formation with an age of 650 ± 300 Myr and a metallicity of Z = 0.18 (here, Z = 0.2 corresponds to that of the Sun).

We use the high-resolution spectrum taken with VLT/UVES taken two weeks after discovery (see log in Table 2; PI: P. Vreeswijk) to measure the host velocity dispersion. We obtain ${v}_{\mathrm{host}}=89\pm 1$ km s⁻¹, from the Ca ii λλ8544,8664 absorption lines. According to the M–σ relation (McConnell & Ma 2013), this corresponds to a ${M}_{\mathrm{BH}}={10}^{6.33\pm 0.38}$ M_⊙, consistent with our previous estimate.

The field of iPTF16fnl was extensively observed for the last six years by PTF/iPTF. No prior activity in the host is detected with upper limits of ∼20–21 mag. The most recent non-detection is from MJD 57432.6, around 194 days before discovery. The host galaxy was also monitored by The Catalina Real-Time Transient Survey (CRTS; Drake et al. 2009) during the 2006–2016 period. The magnitude of the galaxy was stable within the errors, with an unfiltered average magnitude of 14.88 ± 0.05 mag.

Following spectroscopic identification of iPTF16fnl as a TDE candidate, the source was monitored at Palomar and by the Ultraviolet and Optical Telescope (UVOT; Roming et al. 2005) on board the Swift observatory (Gehrels et al. 2004). The UVOT observations were taken in UVW2, UVM2, UVW1, U, B, and V; see Table 3. The data were reduced using the software UVOTSOURCE using the calibrations described in Poole et al. (2008) and updated calibrations from Breeveld et al. (2010). We use a 75 aperture centered on the position of the transient.

At Palomar, photometry in the g and Mould-R bands were obtained with the iPTF mosaic wide-field camera on the Palomar 48-inch telescope (P48; Rahmer et al. 2008). Difference-image photometric measurements were provided by the IPAC Image Subtraction and Discovery Pipeline developed for the iPTF survey (PTFIDE; Masci et al. 2017). Difference-imaging photometry in the $u^{\prime} g^{\prime} r^{\prime} i^{\prime}$ bands, obtained with the SEDM, was computed using the FPipe software (Fremling et al. 2016). The zero-points were calibrated using stars in the SDSS footprint. Table 3 reports the measured Swift aperture photometry magnitudes and the difference-imaging photometry for the Palomar data.

The multi-band Swift and optical light curve, corrected for Galactic extinction, is shown in Figure 3. To correct the UV bands for host light contamination, we used the average of the last four epochs of Swift data, from MJD 57712.7 to 57724.7 (>80 days), to subtract from the early part of the light curve. The measurements were constant, with an rms of ∼0.05 mag, comparable to the measurement error. Optical Swift data were excluded from the analysis, as they were dominated by the host, not by the central point source.

Zoom In Zoom Out Reset image size

In order to estimate the extinction in the host galaxy, we use all available Swift UV data and difference-imaging photometry in the ugri bands to fit blackbody emission curves, as detailed in Section 4. For each epoch, the photometry is corrected for both Galactic (fixed) and additional host extinction E(B − V), from 0 to 0.25 mag in 0.05 steps. The likelihood between the de-reddened photometry and a blackbody model is computed for each epoch. In a final step, we marginalize over all epochs to derive the final value. We find that the best fit corresponds to E(B − V) = 0, with an upper limit of E(B − V) = 0.05. The selection of different extinction laws for host extinction (Fitzpatrick 1999; Calzetti et al. 2000) does not change our conclusions. Given the assumption that the emission from iPTF16fnl in fact follows a distribution, from now on we will assume the reddening in the host to be negligible.

Radio follow-up observations of iPTF16fnl were taken with the Jansky Very Large Array (VLA; PI A. Horesh), the Arcminute Microkelvin Imager (AMI; PI K. Mooley) and the James Clerk Maxwell Telescope and the Submillimetre Common-User Bolometer Array 2 (JCMT/SCUBA-2; PI A. K. H. Kong). The upper limits corresponding to the observations are shown in Table 4. The limits for the monochromatic radio luminosities corresponding to the first VLA epoch are $\nu {L}_{\nu }\lt 2.3\times {10}^{36}$ erg s⁻¹ and $\nu {L}_{\nu }\lt 1.7\times {10}^{37}$ erg s⁻¹ at 6.1 and 22 GHz. These limits are respectively two and one order of magnitude deeper than the VLA detection of ASASSN-14li at the peak for a similar frequency range (van Velzen et al. 2016; Alexander et al. 2016).

We can use the limits reported here to argue against the presence of an on-axis relativistic outflow, or at least constrain the energy of the jet, E_j. We compare our limits in 6 and 15 GHz with the analytical light curves for on-axis TDE radio emission from Generozov et al. (2017). Figure 4 shows that our early-time observations suggest jet energies with ${E}_{j}\lt {10}^{49}$erg s⁻¹.

Zoom In Zoom Out Reset image size

Additionally, we contrast our measurements, scaled to the redshift of PS1-11af, with the GRB afterglow models of van Eerten et al. (2012) presented in Chornock et al. (2014; their Figure 13). Based on our non-detections, we can rule out the existence of a relativistic jet, as viewed 30° off-axis. For larger angles, the radio emission is expected to arise at later times (>1 year after disruption). Therefore, continuous monitoring of the event at radio wavelengths is encouraged.

We observed the location of iPTF16fnl with the X-Ray Telescope (XRT; Burrows et al. 2005) on board the Swift satellite beginning at 19:32 UT on 30 August 2016. Regular monitoring of the field in photon counting mode continued over the course of the next four months (PIs T. Holoien and B. Cenko).

No significant emission was detected in individual epochs (typical exposure times of ≈2 ks). Using standard XRT analysis procedures (e.g., Evans et al. 2009), we placed 90% confidence upper limits ranging from (3.8–12.1) ×10⁻³ counts s⁻¹ in the 0.3–10.0 keV bandpass over this time period.

Stacking all the XRT data obtained over this period together (58 ks of total exposure time), we find evidence for a weak (≈4σ significance) X-ray source at this location with a 0.3–10.0 keV count rate of $(2.6\pm 1.2)\times {10}^{-4}$ counts s⁻¹.

With only 15 source counts we have a limited ability to discriminate between spectral models, however, with several photon energies detected above 1 keV. We derive response matrices for the stacked XRT observations using standard Swift tools. Adopting a power-law model for the spectrum with a photon index of 2 and accounting for line-of-sight absorption in the Milky Way (Willingale et al. 2013), we find the measured count rate corresponds to an unabsorbed flux of 0.3–10.0 keV flux of ${4.6}_{-2.0}^{+3.7}\times {10}^{-15}$ erg cm⁻² s⁻¹.

At the distance of Mrk 950, this corresponds to an X-ray luminosity of ${L}_{X}={2.4}_{-1.1}^{+1.9}\times {10}^{39}$ erg s⁻¹. Without additional information (e.g., variability and/or spectra), we cannot determine conclusively if this X-ray emission is associated with the transient iPTF16fnl, or if this is unrelated X-ray emission from the host nucleus (e.g., an underlying active galactic nucleus) or even a population of X-ray binaries or ultra-luminous X-ray sources. However, the lack of evidence for ongoing star formation or AGN-like emission lines in the late-time optical spectra of Mrk 950 (Section 5) suggest an association with the TDE.

If this is indeed the case, the implied X-ray emission would be extremely faint, both in an absolute and a relative sense. We can contrast, for example, with the X-ray emission observed from ASASSN-14li (Holoien et al. 2016b; van Velzen et al. 2016), with a peak luminosity approximately four orders of magnitude above that seen for iPTF16fnl. Even sources with much fainter X-ray emission, such as ASASSN-15oi (Holoien et al. 2016a), still outshine iPTF16fnl by more than a factor of 100.

Spectroscopic follow-up observations of iPTF16fnl have been carried out with numerous telescopes and instruments, summarized in the spectroscopic log in Table 2. Figure 5 shows the spectral sequence for iPTF16fnl, spanning three months. The spectroscopic data are made public via WISeREP (Yaron & Gal-Yam 2012).

Zoom In Zoom Out Reset image size

Given the brightness of the galactic bulge relative to the TDE, the interpretation of the TDE spectrum poses some challenges. A notable feature is the strong host component in all the available spectra. Different slit widths, the variable seeing, different orientations of the slit during the acquisition of the data (generally taken at the parallactic angle), and the highly elongated geometry of the host galaxy, contribute to create a strong variation in the contribution from the host component, which appears to vary from one instrument to another.

In early epoch spectra (<50 days), the most prominent emission lines correspond to broad He ii and Hα, shown in Figure 8. These lines have average FWHMs of ∼14,000 km s⁻¹ and ∼10,000 km s⁻¹, respectively. The analysis and evolution of their profiles is further explored in Section 4.2.

Several narrow absorption lines, associated with the host galaxy, were identified. The region around Hα contains an emission line that can be associated with [N ii] at λ6583. We also observed narrow absorption lines corresponding to Ba ii at λλ 6496, the Na i D doublet at λλ 5889, 5896, Mg i λλ 5167, 5173, 5184, Fe i λλ 5266, 5324 Ca ii is detected at λλ 3934, 3968 and as strong NIR triplet absorption at λλ 8498, 8542, 8662 Å.

Swift host-subtracted UVW1, UVM2, UVW2 photometry, and Palomar data were used to fit the object blackbody temperature and radius. We fit the fluxes derived from each band with spherical blackbody emission models using Markov Chain Monte Carlo simulations, based on the Python package emcee (Foreman-Mackey et al. 2013). The blackbody bolometric luminosity, and the best fits for the temperature and the radius, are shown in Figure 6. Because only g-band measurements were available for our first detection epoch, we assumed that the luminosity follows a blackbody emission with a temperature of 21,000 K (average for the first 2 weeks) and scaled the flux to match the g-band magnitude.

Zoom In Zoom Out Reset image size

The bolometric luminosity at peak is ${L}_{p}\simeq (1.0\pm 0.15)\,\times {10}^{43}$ erg s⁻¹. This is one order of magnitude lower than most of the optical TDEs (PS1-10jh, ASASSN-14ae, ASASSN-14li, and ASASSN-15oi; see Figure 6). If compared to ASASSN-15lh, a TDE candidate from a rotating high-mass SMBH (Leloudas et al. 2016),²⁸ the peak is two orders of magnitude fainter. Based on our M_BH estimate, we derive its Eddington luminosity as ${L}_{\mathrm{Edd}}={2.7}_{-1.6}^{+3.7}\times {10}^{44}$ erg s⁻¹, implying that at peak, iPTF16fnl shines only 2%–10% of ${L}_{\mathrm{Edd}}$. Assuming a radiative efficiency of $\eta =0.1$, this translates into a peak accretion rate of ${\dot{M}}_{\mathrm{peak}}\sim 1.8\times {10}^{-3}$ M_⊙ yr⁻¹ ($L=\eta \dot{M}{c}^{2}$).

Integrating the available bolometric luminosity (see Figure 6), we find the radiated energy to be ${E}_{R}\,=(2.0\pm 0.5)\times {10}^{49}$ erg. The accreted mass for this interval is ${M}_{{acc}}\sim 7.3\times {10}^{-5}$ M_⊙.

The rest-frame blackbody bolometric luminosity (Figure 6) is fit with the characteristic power law $L\propto {((t-{t}_{0})/\tau )}^{-5/3}$ and an empirically motivated exponential profile, $L\,\propto {e}^{-(t-{t}_{0})/\tau }$, where t₀ and τ are free parameters. For the decaying part of the light curve, the exponential model fits the data better (${\chi }^{2}=17$ versus 110), and best-fit parameters are ${\tau }_{0}\simeq 15$ and ${t}_{0}\simeq -6$. For comparison, the decay for other optical TDEs is slower: ASASSN-14ae had τ = 30 days, whereas ASASSN-15oi and ASASSN-14li faded on timescales of 46.5 and 60 days, respectively. Continued, high-quality photometric monitoring would be required to draw conclusive results on long-term evolution, beyond the initial fading stage.

We estimate a lower limit for the time from disruption to peak light ${t}_{\mathrm{peak}}\geqslant 11$ days from the bolometric light curve. We select the measurements at ±10 day from the peak in the g-band and fit the luminosity with a 2° polynomial. For the raising part of the light curve, we use our only available g-band measurement. While this approach is widely used for estimating the explosion time for supernovae, the emission mechanism for TDEs is different and therefore it only yields to lower limits, as seen when applied to PS1-10jh.

Assuming that our bolometric light curve traces the rate of mass falling into the black hole, $\dot{M}(t)$, we use the light curve models from Guillochon & Ramirez-Ruiz (2013; hereafter GR13), scaling them to the peak accretion mass rate and time to peak. The light curves are defined for a range of impact parameters 0.5 ≤ β ≤ 2.5, where β is defined as the depth of the encounter $\beta ={R}_{T}/{R}_{p}$, R_T is the tidal disruption radius, and R_p is the pericenter radius. We impose a ${M}_{\mathrm{BH}}=2\times {10}^{6}$ M_⊙, but we leave the mass and radius of the disrupted star as free parameters. Our best fit corresponds to a star with a polytropic index $\gamma =4/3$ (fully radiative), with ${M}_{* }\sim 0.03$ M_⊙, ${R}_{* }\sim 0.3$ R_⊙, and a depth of the encounter comparable to the disruption radius ($\beta \simeq 1$). The light curve for the best-fit model is shown in Figure 6. We obtain ${t}_{\mathrm{peak}}\sim 11$ days, comparable with our previous naive estimate. If we impose that low-mass stars are fully convective and fit for an object with $\gamma =5/3$, we find a relatively good fit for a partial disruption ($\beta \sim 0.6$) and a similar value of ${M}_{* }\sim 0.06$ M_⊙, although in this case t_peak is shorter than our observations suggest. Although these values illustrate that the disrupted object was likely a low-mass star, detailed modeling of the event would be required to draw quantitative results.

The blackbody model has an average temperature of ${T}_{{BB}}=19,000\pm 2000$ K, which does not vary significantly over time, as shown in Figure 6. At later epochs, the increased uncertainties in the host subtraction lead to increased scatter in T_BB. Given the uncertainty of the extinction in the host, these values can be assumed as lower limits. The model blackbody radius, R_BB, starts at ∼2.5 × 10¹⁴ cm, linearly declines for the first 20 days and then flattens to 5 × 10¹³ cm. These radii are much larger than the Schwarzschild radius ${r}_{\mathrm{Sch}}\sim 6\times {10}^{11}$ cm of the nuclear black hole. In comparison, we note that the tidal disruption radius of such SMBH for a main-sequence solar-like star is ${R}_{T}\simeq 1\times {10}^{13}$ cm. Photospheric emission at radii larger than R_T is commonly observed for the optical sample of TDEs. This has been attributed to the existence of a reprocessing layer at larger radii, which re-emits the X-ray and UV in optical bands (Loeb & Ulmer 1997; Guillochon et al. 2014). Alternatively, the emission mechanism may originate from the energy liberated by shocks between streams in the apocenter, during the formation of the accretion disk (Piran et al. 2015). Such an optically thick layer, mainly comprised of stellar debris, is associated with the origin of the emission line signature for optical TDEs (Metzger & Stone 2016; Roth et al. 2016).

The early time spectrum of iPTF16fnl is dominated by blue continuum radiation and the characteristic broad He ii λ 4686 line. The emission around 6500 Å can be attributed to Hα, although the He i λ6678 line is also detected. In the region around He II, we also detect Hβ in emission. However, the strong host contribution makes its identification challenging at late times. In our analysis, we use the late-time host spectrum (+119.2 days) as our template. We select the highest signal-to-noise (S/N) spectra, and fit them with a combination of a host and a blackbody continuum, as shown in Figure 7. On the residual spectrum, we fit a line model using the python package lmfit (Non-Linear Least-Squares Minimization and Curve-Fitting for Python). After masking the regions affected by telluric absorption, He ii+Hβ and Hα + He i lines are fit using two component Lorentzian model, in order to derive the width (FWHM) and central location of the emission. The results, plotted as insets, are shown in Figure 8. The fluxes for each line are derived from the best-fit model and shown in Table 5. As discussed in Brown et al. (2017), if all the flux in the He ii line would be attributed to recombination produced by blackbody photoionizing radiation, the observed flux of $\gt {10}^{40}$ erg s⁻¹ would require blackbody radiation with a temperature ∼4 × 10⁴ K, which is higher than our fit, requiring an additional energy source to power this line.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Around the peak, the He ii lines appear to have higher velocity, showing an average value of ${\mathrm{FWHM}}_{\mathrm{HeII}}\,\simeq \mathrm{14,000}\pm \mathrm{3,000}$ km s⁻¹, in contrast to the Hα line, with ${\mathrm{FWHM}}_{H\alpha }\simeq \mathrm{10,000}\pm 500$ km s⁻¹. As shown in Figure 9, at +30 and +45 days after peak, the FWHM narrows down to 8,500 ± 1500 km s⁻¹ for He ii and 6,000 ± 600 km s⁻¹ for Hα. The centers of the lines appear constant within the scatter for the first 90 days: for He ii, the lines appear marginally blueshifted with velocity of −700 ± 700 km s⁻¹, while the Hα lines appear to be consistent with the reference wavelength, with a shift in velocity of −800 ± 1200 km s⁻¹.

Table 5.  Flux Values for Hα and He ii 4868 Å for the Lines Shown in Figure 8

MJD	Phase	Hα	He ii
(d)	(d)	(erg s−1)	(erg s−1)
57630.4	−1.7	$2.1\times {10}^{40}$	$8.9\times {10}^{40}$
57631.0	−0.8	$3.8\times {10}^{39}$	$1.5\times {10}^{40}$
57631.3	0.0	$6.6\times {10}^{39}$	$8.3\times {10}^{39}$
57661.4	29.3	$3.0\times {10}^{39}$	$3.0\times {10}^{39}$
57661.4	29.3	2.1 × 1039	$2.5\times {10}^{39}$
57687.3	55.2	⋯	$1.1\times {10}^{40}$
57694.4	62.3	⋯	$2.7\times {10}^{39}$
57720.3	88.2	⋯	$1.1\times {10}^{39}$

Note. The values were derived from the best model fit line profile. From the fit uncertainties, we estimate errors of 40% and 30% of the total flux for Hα and He ii lines, respectively.

Download table as:  ASCIITypeset image