The Structure of Tidal Disruption Event Host Galaxies on Scales of Tens to Thousands of Parsecs

The imaging targets were selected to be the four nearest TDE host galaxies from the sample of eight UV/optical-bright TDEs with broad H/He lines from French et al. (2016): ASASSN-14li (Holoien et al. 2016), ASASSN-14ae (Holoien et al. 2014), iPTF15af (Blagorodnova et al. 2019), and PTF09ge (Arcavi et al. 2014). Of the four TDE host galaxies studied here, one (ASASSN-14li) is a post-starburst, two (ASASSN-14ae and iPTF15af) are quiescent Balmer strong, and one (PTF09ge) is a quiescent early type according to their spectra (French et al. 2016). PTF09ge, however, has star formation missed by the 0.9 kpc aperture spectrum in its outskirts, revealed by the blue ring in our imaging (Figure 1). Our stellar population fitting analysis also revealed a recent short burst of star formation in this galaxy, similar to the star formation history of iPTF15af, although this analysis used the same small-aperture spectrum (French et al. 2017).

Zoom In Zoom Out Reset image size

This sample is representative of the host galaxy types of known TDEs. Using the sample of TDEs with host galaxy data compiled in French et al. (2020; updated from Graur et al. 2018), of the optical/UV-bright TDEs with broad H/He lines (like those identified by Arcavi et al. 2014), 60% are quiescent Balmer strong (including post-starburst galaxies), and 33% are post-starburst. Of the X-ray TDEs with host galaxy data (including those designated by Auchettl et al. 2017 as "likely" or "possible" TDEs), 32% are quiescent Balmer strong and 12% are post-starburst. For the more robust subset of X-ray TDEs, 75% are quiescent Balmer strong and 25% are post-starburst. The post-starburst host of ASASSN-14li is unique among the four galaxies studied here in being the only post-starburst galaxy, but having a sample with 1/4 post-starburst galaxies is consistent with post-starburst galaxies composing 12%–33% of the TDE host population. Similarly, the fraction of quiescent/early-type and quiescent Balmer-strong galaxies in the HST-observed sample is representative of the known TDE host population, spanning the range of likely recent star formation histories.

Given the black hole masses measured by Wevers et al. (2017) using the bulge velocity dispersions and by Mockler et al. (2019) using the TDE light curves, as well as the fitting functions from Stone & Metzger (2016) for all and cuspy galaxies, the radius of influence of the SMBH is 1–3 pc, 0.3–0.9 pc, 2.4–2.7 pc, and 1.0–1.6 pc for the host galaxies of ASASSN-14li, ASASSN-14ae, iPTF15af, and PTF09ge, respectively.

Table 1.  Observed TDE Targets

TDE	R.A.	Decl.	z	Pixel Size (pc)a	References	Peak
ASASSN-14li	12:48:15.23	+17:46:26.44	0.02058	17	Holoien et al. (2016)	2014 Nov 14b
ASASSN-14ae	11:08:40.12	+34:05:52.23	0.0436	33	Holoien et al. (2014)	2014 Jan 25b
iPTF15af	08:48:28.12	+22:03:33.58	0.079	60	Blagorodnova et al. (2019)	2015 Feb 23
PTF09ge	14:57:03.18	+49:36:40.97	0.064	50	Arcavi et al. (2014)	2009 Jun 13

Notes.

^aPhysical size probed by 004 pixels. ^bDiscovery date, discovered after peak.

Download table as:  ASCIITypeset image

Each of the four galaxies was observed using three bands: F438W, F625W, and F814W. Details of the observations are shown in Table 1. The two bluer bands were selected to match those used by Yang et al. (2008) in their study of post-starburst galaxies. The F814W band was then added to best measure the stellar bulge properties. Each galaxy was observed to a depth of ∼26 mag arcsec⁻² in order to measure faint tidal features. The observations were carried out in 2/2017–8/2017 as part of program 14717 (PI: Arcavi) during Cycle 24. Details of the observations are shown in Table 2.

We use the flat-fielded, charge-transfer-efficiency-corrected, ∗.flc files from the HST archive. The individual exposures are then realigned and stacked using SNHST (McCully et al. 2018b). This procedure uses PanSTARRS imaging (Chambers et al. 2016) to refine the astrometric solution to best match the images from multiple visits to each object. The images are then combined using DrizzlePac.¹³ Cosmic rays are removed using AstroScrappy (McCully et al. 2018a).

The resulting three-color images can be seen in Figure 1. In Figure 12 we present high stretch images of the four TDE host galaxies in each of the three bands, optimized to search for low surface brightness tidal features.

We explore three possible solutions for determining the point-spread function (PSF) for each stacked galaxy image in each band: (1) The HST PSF library contains PSF profiles for the F438W and F814W bands.¹⁴ (2) We also consider the PSFs calculated using Tiny Tim¹⁵ (Krist et al. 2011). With Tiny Tim, we assume an A5V star, similar to the stellar population dominating the spectra of these galaxies, and use the central position of the galaxy on the chip. (3) Three of the four galaxies (hosts of ASASSN-14li, iPTF15af, and PTF09ge) are in fields with suitable foreground stars to calculate PSFs, which are bright enough to have high signal-to-noise ratio (S/N) but not saturated. To calculate the best-fit PSFs from stars in each field (except for the ASASSN-14ae field), we first mask the diffraction spikes and then iteratively fit a series of Gaussian profiles using Galfit (Peng et al. 2002, 2010). We follow Stone & van Velzen (2016) and add Gaussian components until the χ²/ν value stops decreasing. For our sample of images, between four and five Gaussian components are chosen.

The results of the three methods for determining the PSF are shown in Figure 2 for each band. We find that the library PSFs are significantly wider than the other measures. The Tiny Tim PSFs are often narrower than the profiles and best-fit models to stars in each frame. We note that the library and Tiny Tim PSFs have not been corrected for the drizzle or co-add steps applied to the science images. For F438W and F625W, the stellar PSFs are very similar for each field, but we observe significant differences for the F814W fields. We adopt the best-fit stellar PSF for each field–band combination. For the ASASSN-14ae field, we test for the best PSF from the other three fields by fitting a Sérsic profile to the ASASSN-14ae host galaxy using each of the three PSFs and choosing the one with the lowest χ²/ν. The best PSF found using this method is from the iPTF15af field, so we adopt this PSF for all of the fits to the ASASSN-14ae host galaxy presented here.

Zoom In Zoom Out Reset image size

From Figure 1, it is clear that the TDE host galaxies have a range of morphologies, although none are grand-design spirals or irregular dwarf galaxies. The host galaxy of ASASSN-14li has a bright central point source that can be seen through its diffraction spikes in the optical image. The host galaxy of ASASSN-14ae is an edge-on disk that also has a bright central bulge. The host galaxies of iPTF15af and PTF09ge are barred lenticular galaxies; both have blue disk components in their outskirts in addition to bright central bulges. The host of iPTF15af has a series of blue clumps in its outer disk to the north. These clumps may be star-forming knots, dwarf galaxies, minor merger remnants, or chance projections along the line of sight. The host of PTF09ge has extended arms reaching from the two blue knots in the disk. The symmetry and blue color of the extended arms indicate that they may be spiral arms, but such features could also be caused by a recent merger. These extended arms also closely resemble the ring-like "${R}_{1}^{{\prime} }$" morphology described by Buta (2017) (see, e.g., UGC 12646 in Buta 2017, Figure 1) and do not require mergers or other nonsecular processes to form. Bar-like structures are seen in two of the host galaxies (iPTF15af and PTF09ge), consistent with ∼40% of low-redshift galaxies having bars (Elmegreen et al. 2004).

No significant dust lanes are seen in the four TDE hosts, aside from the dust seen in the spiral arms around the PTF09ge host galaxy. This is consistent given Poisson errors with the rate of 7/21 seen by Yang et al. (2008) in post-starburst galaxies. Of the seven post-starburst galaxies with dust lanes, Yang et al. (2008) find three instances where the dust lanes cross the bulge of the galaxy. It is not surprising to find no cases of central dust lanes in the TDE host galaxy sample, given the difficulty of detecting a TDE through significant dust attenuation, but a larger sample would be needed to establish a statistically significant difference from the post-starburst sample.

We discuss the galaxy morphologies in more quantitative detail in the remainder of this section by fitting their 2D surface brightness distributions, examining their color gradients, and searching for asymmetric features.

We fit the surface brightness profile of each galaxy in each band using Galfit (Peng et al. 2002, 2010). We use the PSF models derived from stars in each field as described in Section 2.3. We fit a sky level with no gradient across the field in addition to the models discussed below. For galaxies with nearby projected companions, we fit separate Galfit model components. We fit separate Galfit components for the projected nearby edge-on disk galaxy and foreground star near the host galaxy of ASASSN-14li. We fit a separate component for the foreground star projected near the host galaxy of iPTF15af and the foreground star projected near the host galaxy of PTF09ge.

We consider six¹⁶ types of models for each galaxy in each band: (1) a Sérsic profile, with n free to vary; (2) a Sérsic profile plus a central point source; (3) a Sérsic profile plus an exponential disk; (4) a de Vaucouleurs profile, (5) a de Vaucouleurs profile plus a central point source; and (6) a de Vaucouleurs profile plus a disk. Sérsic profiles (Sérsic 1963, 1968) are parameterized using an index such that higher Sérsic indices are more centrally concentrated. A Sérsic index of n = 1 corresponds to an exponential disk and fits typical star-forming disk galaxies, and higher Sérsic indices of n ∼ 4 (with the special case of n = 4 being the de Vaucouleurs profile; de Vaucouleurs 1948) are good models for early-type galaxies. Using Galfit to model combinations of models helps fit galaxies with multiple bulge plus disk components. Post-starburst galaxies are often best fit with high Sérsic indices n > 4 (Yang et al. 2008) and often have central point sources (Yang et al. 2006). Given the long duration of TDE emission (especially in the UV; van Velzen et al. 2019), it is also important to consider possible nuclear point sources if the TDE was still bright 5–8 yr after its peak.

We allow the Galfit models to vary in two dimensions, fitting parameters for axis ratio and position angle in addition to the radial profile parameters. We select the best-fit model by choosing the one with the lowest χ²/ν.

The best-fit models are shown in Table 3 in addition to reduced χ²/ν values and selected parameters from the best-fit models. The host galaxy of ASASSN-14li is best fit by a Sérsic profile plus a central point source for all three bands, and the PTF09ge host galaxy is also best fit by this model for two bands. Both ASASSN-14ae and iPTF15af are best fit by models with disks added for two to three of the bands considered here. The best χ²/ν values are typically between 1 and 2, except for the fits to the PTF09ge host galaxy, which are higher. This indicates the presence of additional unmodeled structure, which we explore further in the subsequent sections.

The magnitude of the central point-source component is 18–19 mag for the cases where the galaxy is best fit with a central point source and 19–23 mag for cases where a central point source did not improve the fit. These point sources are significantly brighter than we expect the residual TDE emission to be. The brightest residual TDE emission is expected from ASASSN-14li, which would have been 19.1 mag in the near-UV at the time of the HST imaging (van Velzen et al. 2019). The g-band magnitude is expected to be 1.5 mag fainter, ∼20.6 mag, which is much fainter than the 18.0–18.5 mag optical point source in the best-fit Galfit model. Thus, the central point sources from the TDE host galaxies are likely persistent and unrelated to the TDE. They may be central star clusters, active galactic nuclei (AGNs), or very cuspy stellar density profiles. High-resolution spectroscopy will be required to differentiate among these possibilities.

We see no evidence for dual nuclei (e.g., Lauer et al. 1993) in the stellar light down to the resolution limit of 004 for any of the TDE host galaxies considered here. The resolution of approximately tens of parsecs constrains any SMBHBs to be smaller than the stage where dynamical friction is affecting their evolution and inspiral (Yu 2002; Kelley et al. 2017). The spatial separation where the TDE rate is expected to be enhanced from SMBHB effects is of order the gravitational radius of influence of the primary black hole r_infl ∼ 0.3–3 pc, smaller than what can be constrained with these data (Chen et al. 2011; Kelley et al. 2017). We explore the constraints we can place on SMBHB effects using larger ∼kiloparsec-scale merger features in Sections 3.6 and 4.2.

We compare the surface brightness profiles measured above for the TDE hosts to post-starburst galaxies and early-type galaxies in Figure 3. Surface brightness profiles for a sample of 260 early-type galaxies from the Atlas-3D survey were measured by Krajnović et al. (2013). These profiles were modeled using two-component Sérsic plus exponential disk models. Surface brightness profiles for 21 post-starburst galaxies were measured by Yang et al. (2008) using two-component de Vaucouleurs plus exponential disk models. We correct all surface brightness profiles for cosmological surface brightness dimming. In order to best compare to the r-band (peak wavelength ∼6231 Å) data used by Krajnović et al. (2013), we use the HST F625W band imaging of the TDE hosts and post-starbursts for this comparison. We note that the Yang et al. (2008) sample of post-starburst galaxies is at slightly higher redshift ($\langle z\rangle =0.09$ vs. $\langle z\rangle =0.05$) than the TDE host galaxies considered here, but the portions of the spectra probed are similar.

Zoom In Zoom Out Reset image size

While we can compare the TDE host galaxies to detailed studies of post-starburst and early-type galaxies, unfortunately there are no similar analyses of quiescent Balmer-strong galaxies in the literature to which to compare ASASSN-14ae and iPTF15af. Compared to galaxies satisfying a stricter post-starburst definition cut on Balmer absorption, quiescent Balmer-strong galaxies could have either longer post-burst ages, lower fractions of stellar mass created in the recent burst, or longer-duration bursts (French et al. 2018). However, smaller bursts are more common in galaxies than larger bursts creating more stars, so we found in French et al. (2017) that most quiescent Balmer-strong galaxies have smaller bursts than post-starburst galaxies, with similar ranges of post-burst ages and burst durations. We thus expect them to have properties in between those of post-starburst and early-type galaxies, due to the less massive population of recently formed (A) stars.¹⁷

Both the TDE host galaxies and post-starburst galaxies have bright, centrally concentrated stellar light profiles. The TDE host galaxy stellar light profiles stay high down to scales of <100 pc and even to ∼30 pc for the host galaxy of ASASSN-14li. The TDE host galaxies have central surface brightnesses comparable to the brightest early-type galaxies with similar stellar mass, and comparable to the dimmer half of the post-starburst galaxies. Even the TDE host galaxies that are not post-starbursts nonetheless have bright central surface brightnesses more typical of post-starburst galaxies.

To consider the case where the central point-source components of the host galaxies of ASASSN-14li and PTF09ge may be nonstellar, we remove the point-source component from the models in the comparison in the right panel of Figure 3. With the central point-source component removed, the host galaxy of ASASSN-14li would still have a bright central surface brightness, similar to post-starburst galaxies, while the host galaxy of PTF09ge would be more similar to early-type galaxies at similar stellar masses.

The surface brightness profiles of the TDE host galaxies appear to be more concentrated than early-type galaxies of similar stellar mass, so we compare their Sérsic indices as a parameterization of their concentration on larger ≳kiloparsec scales. We found in the previous section that the Sérsic-only Galfit fits to the TDE host galaxies have high Sérsic indices. While the Sérsic-only fits are never the best fit of those considered, we use the Sérsic indices measured from these fits to compare to the Sérsic indices measured for a sample of post-starburst galaxies from Yang et al. (2008), as a way to quantitatively compare the typical surface brightness slopes. This parameterization also allows us to compare the TDE hosts to other samples of galaxies, as the variable Sérsic parameter ensures that a range of galaxy morphologies from bulge to disk dominated can be fit with similar accuracy. We plot histograms of the Sérsic indices for each sample in Figure 4. We again use the measurements from the F625W HST imaging for the TDE host galaxies and post-starbursts and the r-band imaging for the early types.

Zoom In Zoom Out Reset image size

We compare the Sérsic indices of the TDE host galaxies to measurements of post-starburst galaxies from Yang et al. (2008) and early-type galaxies from Krajnović et al. (2013). Krajnović et al. (2013) measured the mean Sérsic-only fit Sérsic index for the Atlas-3D sample to be $\langle n\rangle =3.6$, with a standard deviation of 2.0. Most post-starburst galaxies have higher Sérsic indices n > 4. The TDE host galaxies have high Sérsic indices, similar to the post-starburst sample and higher than typical early-type galaxies. The TDE host galaxy and post-starburst galaxy distributions cannot be distinguished statistically using a Kolmogorov–Smirnov test.

The host galaxy of ASASSN-14li has a Sérsic index close to the typical value for post-starburst galaxies and a bright central surface brightness profile also consistent with this sample. The host galaxy of iPTF15af also has a high Sérsic index, higher than typical early-type galaxies and consistent with post-starbursts; the other quiescent Balmer-strong host of ASASSN-14ae has a Sérsic index consistent with early-type galaxies. Both hosts have central surface brightnesses similar to the lower half of the post-starburst galaxies and higher than most of the early-type galaxies. This is not inconsistent with what we might expect for quiescent Balmer-strong galaxies. The host of PTF09ge has the highest Sérsic index, significantly higher than early-type galaxies, and higher than most of the post-starbursts. Given that most star-forming galaxies have Sérsic indices of ∼1 (exponential disks), the high Sérsic index is unexpected considering the star-forming ring.

In several cases, the best-fit Galfit model indicates that the high Sérsic indices in the TDE hosts are driven by a central point source. Bright nuclei are also seen in post-starburst galaxies, and Yang et al. (2008) find that most post-starburst galaxies have lower Sérsic indices after masking the central 3 pixels.

We consider the TDE host galaxy color gradients in Figure 5. We use the F438W and F625W band images, as well as the best-fit centers and sky levels, to determine the B–R color radial profiles. Three of the four TDE host galaxies (all except for PTF09ge) have blue color gradients, such that their centers are bluer than their outskirts. Blue color profiles are often seen in post-starburst galaxies (Yang et al. 2008) because of the centrally concentrated young/intermediate stellar populations but are the reverse of the red central bulge—blue outer disk profiles often seen in star-forming galaxies and the flat or slightly red profiles seen in early-type galaxies (Wise & Silva 1996).

Zoom In Zoom Out Reset image size

The blue color gradient (bluer at the center and redder toward higher radii) in the host galaxy of ASASSN-14li is consistent with those seen in most post-starburst galaxies. The quiescent Balmer-strong TDE host galaxies also have blue color gradients. Early-type galaxies have a narrower range of color gradients than post-starburst galaxies, with most having flat or slightly red color gradients (Yang et al. 2008). Only the host galaxy of PTF09ge has a red color gradient, consistent with the early-type sample. Even though not all of the TDE host galaxies are post-starburst, their color gradient distribution is consistent with the Yang et al. (2008) study of post-starburst galaxies, where 60% had blue color gradients, 20% had red color gradients, and 20% had flat color gradients.

In order to quantitatively compare the color gradients to the post-starburst galaxies studied by Yang et al. (2008), we fit both single and double power-law slopes to the TDE host B–R color gradients. The double power-law slopes allow us to separately parameterize the inner and outer slopes, using a break radius that is allowed to vary. Using either a Kolmogorov–Smirnov or Anderson–Darling test, we cannot rule out the possibility that the color gradient slopes, inner slopes, or outer slopes for the TDE host galaxies and post-starbursts are drawn from the same distribution.

The two galaxies best fit with central point sources (ASASSN-14li and PTF09ge) show different behaviors, with ASASSN-14li having a blue central point source and PTF09ge having a point source with similar color to the rest of the galaxy. These colors may point at different physical origins for the central point sources. While a blue central point source could be caused by a central star cluster, a concentrated stellar light profile, or an AGN, a red central point source could be a concentrated older stellar population or a dust-obscured blue source. In the case of the PTF09ge host galaxy, however, we observe no dust lanes near the core, and indeed a TDE was observed, limiting the possibilities for heavy nuclear dust extinction.

Because bright central surface brightnesses could be due to either high stellar densities or low mass-to-light (M/L) ratios from younger stellar populations, we consider next the stellar mass surface densities of the TDE host galaxies. We use the F438W − F625W color profiles of the TDE host galaxies and the relations from Bell & de Jong (2001) to estimate the M/L profiles. We use a rolling median of the radial color profiles, smoothed over 4 pixels to reduce noise and account for the width of the PSF. We use the SDSS spectra and Keck spectra (PTF09ge host) used to initially classify the galaxies in French et al. (2016) to convert the observed HST colors to rest-frame B–R colors. Bell & de Jong (2001) find that uncertainties in color-based M/L values are 0.1–0.2 dex, except in the case of a recent ∼10% starburst, where uncertainties are ∼0.25 dex. We adopt a 0.25 dex uncertainty for the post-starburst host of ASASSN-14li and 0.15 dex uncertainty for the other three host galaxies. The uncertainty in the M/L ratio dominates the error in the central stellar mass surface densities. We assume that all the light is stellar in origin, with no contribution from a central persistent AGN.

For the comparison early-type galaxy sample, we use the M/L values from Cappellari et al. (2013). Because the early-type galaxies have less radial variation in color, we assume these M/L values for the entire radial extent of the early-type galaxies.

Using the M/L profiles derived from the color gradients and the best-fit Sérsic + disk models for the TDE hosts, we calculate the stellar mass surface density profiles. We plot the stellar mass surface density profiles in Figure 6(a). If we assume that all of the central light from the TDE host galaxies is from stars, even after correcting for the lower M/L of the younger stars, the TDE host galaxies have higher stellar mass densities in their central 30–100 pc than most early-type galaxies with similar total stellar mass.

Zoom In Zoom Out Reset image size

We note that these results do not depend on the PSF modeling method, and we find that the stellar mass surface densities do not change more than the uncertainties if we vary our PSF modeling assumptions or do not PSF-correct the models.

We compare the mean stellar mass surface densities within 100 pc for the TDE host galaxies and the comparison early-type galaxies (Figure 6(b)). The TDE host galaxies have stellar mass surface densities on these scales higher than 2/3 of the early-type galaxies with similar total stellar mass. We discuss the implications for how this increased central stellar density could raise the TDE rate in these galaxies in Section 4.1.

Recent major or minor mergers will result in tidal features and disturbed morphologies in galaxies on large ∼kiloparsec scales, which can be quantified through the asymmetric components of the stellar light (Conselice et al. 2000). Using a galaxy image rotated 180° as a model for the symmetric stellar light, the residual light can reveal asymmetric features indicative of a recent merger.

We quantify the presence of asymmetric components using the asymmetry measure defined by Conselice et al. (2000):

$$\begin{eqnarray}&&A=\min \left(\displaystyle \frac{{\rm{\Sigma }}| {I}_{0}-{I}_{180}| }{{\rm{\Sigma }}| {I}_{0}| }\right)-\displaystyle \frac{{\rm{\Sigma }}| {B}_{0}-{B}_{180}| }{{\rm{\Sigma }}| {I}_{0}| },\end{eqnarray} \tag{ 1 }$$

where I is the flux of the image (I₀) and the image rotated 180° (I₁₈₀). Each sum is over a circular aperture of radius 2× the Petrosian radius. B₀ represents background pixels and their rotated image (B₁₈₀) summed over the same size aperture. Because the galaxy centers may not be at the exact center of our pixels, we iterate the rotation center in a small 7 × 7 pixel grid and choose the center in which the sum of the squared asymmetric residuals is the least.

We compare this value to the concentration

$$\begin{eqnarray}&&C=5\mathrm{log}\left(\displaystyle \frac{{r}_{80}}{{r}_{20}}\right),\end{eqnarray} \tag{ 2 }$$

where r₈₀ and r₂₀ and the radii containing 80% and 20% of the flux, respectively, as in Conselice (2003). The C–A plane separates out early-type, late-type, and starburst galaxies owing to the presence of central bulges, spiral arms, and merger companions. Yang et al. (2008) found post-starburst galaxies to lie at slightly higher concentrations and much higher asymmetries than early-type galaxies.

We compute C and A for the TDE host galaxies and compare them to early-type, late-type, and starburst galaxies from Conselice (2003) and post-starburst galaxies from Yang et al. (2008) in Figure 7.

Zoom In Zoom Out Reset image size

We use the F625W band to be most consistent with the post-starburst analysis. The Yang et al. (2008) sample of post-starburst galaxies is at slightly higher redshift ($\langle z\rangle =0.09$ vs. $\langle z\rangle =0.05$) than the TDE host galaxies considered here, and the observations of the TDE host galaxies are deeper, at 26.3–27 mag_AB arcsec⁻² versus 25.3 mag_AB arcsec⁻² for the Yang et al. (2008) sample. Thus, the TDE host galaxy observations presented here should be at least as sensitive to faint tidal features as Yang et al. (2008).

Tidal features will fade with time after the merger, but the Yang et al. (2008) sample has a similar post-starburst age distribution to the TDE host galaxies. The Zabludoff et al. (1996) method used to select the sample used by Yang et al. (2008) selects similar galaxies to the age-dated sample considered by French et al. (2018). French et al. (2017) test whether the TDE host galaxies have a distinguishable distribution of post-burst ages compared to the post-starburst galaxies selected from the SDSS and find no statistically significant difference. Thus, it is unlikely that the lower asymmetries seen in the TDE hosts compared to the Yang et al. (2008) sample of post-starburst galaxies are due to older post-merger ages in the TDE host galaxies.

The imaging data used by Conselice (2003) are from the Frei et al. (1996) sample of nearby galaxies. Because the physical spatial resolution and S/N in each of the data sets considered here are high, the uncertainty in concentration and asymmetry due to variations in data quality will be low. Using the tests done by Lotz et al. (2004), Yang et al. (2008) expect ΔA ≲ 0.05 and ΔC ≲ 0.1 if the S/N is ≳7 and resolution is ≲100 pc pixel^–1. These criteria are also satisfied by the TDE host galaxy imaging, so the values of A and C can be directly compared between each of the samples considered here.

We calculate the asymmetry within 2× the Petrosian radius to be consistent with Yang et al. (2008), but we note that if we used 1.5× as done by Conselice (2003), the asymmetries would be higher by only 0.002–0.004.

The four TDE host galaxies considered here have high concentrations like the post-starburst and early-type galaxies but low asymmetries like the early types. Low asymmetries in TDE host galaxies have also been noted by Law-Smith et al. (2017) using SDSS imaging, though we note that that analysis uses residual asymmetries, where smooth models are first subtracted from the data before rotating.

The host galaxy of ASASSN-14li has the highest asymmetry of the TDE hosts yet is similar to the lowest-symmetry post-starburst galaxies, with 80% of the Yang et al. (2008) post-starburst galaxies having higher asymmetries. In contrast to the relatively undisturbed morphology seen in the stellar light, this galaxy has extended regions of ionized gas observed using MUSE (Prieto et al. 2016). Extended ionized regions may be caused by past AGN activity and have been seen around other galaxies in large imaging surveys ("voorwerps"; Lintott et al. 2009; Keel et al. 2017). We compare the morphology of the extended [O iii] λ5007 bright filaments with the asymmetric residuals in Figure 8 to search for corresponding features in the stellar light. A faint residual feature seen to the upper right of the galaxy in the F625W image is aligned with the extended [O iii] λ5007 arm. This feature is not seen in the F814W or F438 images. Thus, the feature in the F625W image may not be due to stellar light, but instead Hα and/or [N ii]6584 emission from extended ionized gas, as these lines are in the F625W band at the redshift of the ASASSN-14li host galaxy. However, the F625W image is deeper than the other two images by ∼0.3 mag, so the observed feature could be faint stellar light. The sum of the F625W residuals nonetheless results in an asymmetry below most other post-starburst galaxies. Thus, the recent starburst in this galaxy may have been triggered by a more minor merger than typical for other post-starburst galaxies. We explore this possibility in more detail in Section 4.2.

Zoom In Zoom Out Reset image size

The host galaxies of ASASSN-14ae and iPTF15af are quiescent Balmer strong and have asymmetries consistent with early-type (quiescent) galaxies. The only features that contribute to their low asymmetries are two faint features extending from the nucleus of the host galaxy of ASASSN-14ae (Figure 9) and the blue clumps surrounding the host galaxy of iPTF15af (which can be seen in Figure 1). Despite these features, the asymmetries we calculate are consistent with early-type galaxies. In the absence of an HST-observed sample of quiescent Balmer-strong galaxies, we expect them to be similar to both early-type and post-starburst galaxies in C–A space, as they have similar post-burst ages to the post-starburst sample, with generally lower fractions of stellar mass formed in the recent starbursts (French et al. 2017). The lack of strong asymmetries in the two quiescent Balmer-strong TDE hosts may indicate that their recent bursts of star formation were triggered by minor mergers, other interactions, or internal instabilities that were sufficient to drive gas to their centers and form stars, but not enough to create extended tidal features typical after major mergers.

Zoom In Zoom Out Reset image size

The host galaxy of PTF09ge was classified as quiescent, although the blue outer ring would have been partially missed by the SDSS fiber used in classification and likely has more star formation. The extended arms seen in Figures 1 and 12 are nearly symmetric with one another. Despite the blue star-forming ring, the host galaxy of PTF09ge has an asymmetry and concentration more similar to early-type galaxies than star-forming galaxies.

We consider the implications of the observed morphologies on the recent merger histories and possibilities for rapidly producing a bound SMBHB in Section 4.2.

The shapes of the central isophotes of the TDE host galaxies reveal information about the distribution of nuclear stellar orbits. In order to further quantitatively explore the host galaxy morphologies, we use the astropy package photutils (Bradley et al. 2019) to perform elliptical isophotal fitting on the TDE host galaxies. The resulting best-fit ellipticity and position angles as a function of semimajor axis are shown in Figure 10. The radial trends in position angle vary, with shifts of 10°–100° seen over ∼100 pc changes in radius. The ellipticities do not appear to vary with wavelength, with the exception of the F625W band at the center of the ASASSN-14li host galaxy. Due to the presence of Hα and N ii lines in this band, this measurement may also include nonstellar light. There are several discrepancies by band in the position angle measurements, as the F438W band traces the different features from ongoing star formation and its associated bluer light, especially for the host galaxy of PTF09ge.

Zoom In Zoom Out Reset image size

We compare the central ellipticities of the TDE host galaxies to those obtained by Lauer et al. (2005) for the Nukers sample of early-type galaxies using similar methods. We use the F625W filter images for the TDE hosts to be most consistent with the F555W filter images used by Lauer et al. (2005). While Lauer et al. (2005) calculate the inner ellipticities for the early-type galaxies using the luminosity-weighted mean over the central region inside the break radius, we calculate the inner ellipticities for the TDE host galaxies inside the region with radius <140 pc (the mean break radius), due to the lack of good Nuker profile fits to the TDE host galaxies. The inner ellipticities for the TDE host galaxies and early-type galaxies are shown in Figure 11.

Zoom In Zoom Out Reset image size

The TDE host galaxies have relatively small inner ellipticities compared to the early-type sample. These ellipticities are smaller than e ∼ 0.5, where stars at the inner disk will start to be lost to the SMBH at an increased rate (Madigan et al. 2018), though we cannot rule out an eccentric disk at smaller scales closer to the radius of influence r ∼ 0.3–3 pc.

While low inner ellipticities can be related to high central concentrations in galaxies (due to the formation of stars after gaseous dissipation in a gas-rich merger; Cox et al. 2006), the low inner ellipticities represent additional information about the galaxy centers. Post-merger early-type galaxies with extra central light in their surface brightness profiles are more likely to have lower ellipticities, but with significant scatter (Hopkins et al. 2009). A central unresolved point-source component could also contribute to a low central ellipticity, but the PSF scale is only FWHM ∼ 2 pixels, and the low central ellipticities in the two TDE host galaxies best fit with central point-source components in Section 3.2 (ASASSN-14li and PTF09ge) extend to major-axis lengths of ∼8 pixels.

We consider the impact of these observations on the various dynamical effects that could increase the TDE rate in Section 4.3.

In a simple galactic nucleus (i.e., a quasi-spherical star cluster surrounding a single SMBH), TDE rates are set by diffusion of stars into a phase-space loss cone. Diffusion of stellar orbital parameters is driven by two-body scatterings, the rate of which scales with stellar density. Holding all else equal, the TDE rate will therefore increase if the stellar density increases. Because the bulk of the loss cone flux comes from near the black hole gravitational influence radius (r_infl ∼ 0.3–3 pc for the black hole masses of observed TDEs), galaxies with a density enhancement on this scale are expected to have a high TDE rate.

Using SDSS imaging and multiple parameterizations of stellar concentration, TDE host galaxies have been observed to have high central surface brightnesses on kiloparsec scales (Law-Smith et al. 2017; Graur et al. 2018). While SDSS imaging alone cannot resolve scales near r_infl in distant TDE hosts, if one extrapolates kiloparsec-scale properties inward, the dependence of TDE rates on concentration found by Graur et al. (2018) is consistent with theoretical predictions by Wang & Merritt (2004). HST observations of the nearby post-starburst galaxy NGC 3156 find an unusually steep central surface brightness profile on scales close to r_infl;¹⁸ loss cone modeling implies an elevated TDE rate matching that observed in post-starburst galaxies (Stone & van Velzen 2016). Simulations also show an increased TDE rate after a galaxy–galaxy merger due to stellar overdensities caused by the merger (Pfister et al. 2019).

In this work, we have shrunk to 30–100 pc the spatial scales at which we observe TDE host galaxies to have high surface brightnesses. These scales are still above r_infl but nevertheless demonstrate that the unusually high surface brightnesses seen on ∼kiloparsec scales (in ground-based imaging) are correlated with smaller-scale high-concentration effects. In our analysis of the central stellar surface densities (Section 3.5), we also find that—in comparison to early-type galaxies of similar mass—the TDE host galaxies have high stellar surface densities on 30–100 pc scales, using the TDE host color gradients to account for variations in M/L with radius. The radial color gradients identify a bluer, younger central population in three out of four of the TDE host galaxies we have studied (Figure 5), providing further evidence that it was the past star formation episode that created an unusually dense central stellar population. While there are not yet post-starburst TDE hosts close enough for HST to resolve down to r_infl, this result is consistent with HST observations of the nearby post-starburst NGC 3156 (Stone & van Velzen 2016). Thus, it is likely that the TDE host galaxies have a greater number of stars available in orbits where a star could be tidally disrupted.

The TDE rate can be greatly enhanced during the late inspiral phase of an SMBHB (e.g., Ivanov et al. 2005; Chen et al. 2011), as the two black holes reach the nucleus and form a bound pair. Because mergers can both produce an SMBHB and trigger a starburst, establishing whether the TDE host galaxies have experienced a recent merger (major or minor) is important for assessing whether the TDE rate is impacted by SMBHB effects.

The recent merger history of galaxies can be probed by the presence of tidal features and other asymmetries. We explore the asymmetric features of the TDE host galaxies in Section 3.6 and compare them to normal early-type and star-forming galaxies, starbursts, and post-starbursts. While high asymmetries caused by merger features are common in post-starburst galaxies (Yang et al. 2008), the one post-starburst TDE host considered here (ASASSN-14li) is consistent with only the low-asymmetry tail of the post-starburst distribution. The other three TDE host galaxies show even less asymmetry, consistent with early-type galaxies.

A recent gas-rich major merger (with mass ratio ∼1:1–1:3) consistent with producing a post-starburst signature (Snyder et al. 2011) will produce tidal features with A > 0.35 lasting for hundreds of millions of years (Lotz et al. 2010) before and during the starburst phase, and elevated above the level of early-type galaxies throughout the post-starburst phase (Yang et al. 2008). The lack of high asymmetries seen in the TDE hosts considered here implies that they have not experienced recent gas-rich major (∼1:1–1:3) mergers. Indeed, the low mass fractions produced in the recent starburst for the quiescent Balmer-strong hosts are more consistent with being triggered by a more minor merger, and even the extended ionized features seen in the host galaxy of ASASSN-14li could be caused by a ∼1:3–1:5 mass ratio merger since even minor mergers can create tidal debris (Tal et al. 2009).

The mass ratio of the recent merger is important in the context of the TDE rate, because it sets the dynamical friction timescale from the start of the merger to the formation of a bound SMBHB. A starburst episode is likely to be triggered on the first passage of the two galaxies and at their final coalescence (Mihos & Hernquist 1994; Cox et al. 2008; Renaud et al. 2015). The dynamical friction timescale from N-body simulations is ∼1.4 Gyr for a 1:1 mass ratio major merger¹⁹ and scales with the mass ratio such that the dynamical friction timescale will be ∼2× longer for a 1:3 merger and ∼3× longer for a 1:5 merger (Boylan-Kolchin et al. 2008). Thus, even allowing for a gigayear separation between the start of the merger and the triggering of the starburst, unequal-mass mergers are unlikely to have formed an SMBHB in the 500 Myr required by the post-burst ages (French et al. 2017) of the three post-starburst/quiescent Balmer-strong host galaxies.

The lack of strong merger features can, in principle, allow us to rule out the possibility of a major merger recent enough to have formed an SMBHB capable of a high TDE rate. A detailed morphological comparison of our four galaxies with predictions (from cosmological simulations) for the duration of post-merger features, the inspiral time to form an SMBHB, and the evolution of the stellar populations is beyond the scope of this paper.

To summarize, we have found that the post-starburst and quiescent Balmer-strong nature of these TDE host galaxies does not generally arise from recent major (∼1:1 mass ratio) mergers. This result qualitatively disfavors the hypothesis that elevated TDE rates are due to SMBHBs and may be quantifiable in the future through morphological comparisons to galaxy simulations.

A variety of other effects relating to the dynamics of the central stars have been proposed to increase the TDE rate, such as a radial anisotropy (Stone et al. 2018) or an eccentric disk (Madigan et al. 2018). As discussed in Section 3.7, we observe low central ellipticities in the TDE host galaxies, but the resolution of these observations is not sufficient to observe the scale of the eccentric disk (∼r_infl) proposed by Madigan et al. (2018).

On larger scales, an excess of stars with eccentric orbits can still contribute to a high TDE rate through the radial anisotropy scenario considered by Stone et al. (2018). Velocity anisotropy is often quantified with the β parameter, which ranges from purely tangential orbits ($\beta =-\infty$) to purely radial orbits (β = 1); an isotropic galaxy model has β = 0. Because mild radial velocity biases wash out (due to relaxation) fairly quickly, nuclear starbursts must create a large initial bias (β ≳ 0.5) in order to explain the observed rate enhancement in post-starburst galaxies (Stone et al. 2018). While this mechanism can operate even in a quasi-spherical geometry, the requisite β values are so large as to be near the onset for the radial orbit instability (Henon 1973). The nonlinear outcome of this instability is to convert initially spherical geometries into highly flattened ones (Polyachenko & Shukhman 1981) that may be nearly prolate or (for the largest β parameters) significantly triaxial (Merritt & Aguilar 1985).

Although most TDEs are sourced from scales of r_infl in a quasi-isotropic galactic nucleus, this is not the case in a radially biased galaxy; for large β values, the peak of the loss cone flux can be up to an order of magnitude beyond r_infl, a distance that may be resolved for the host of ASASSN-14li (and is only unresolved by a factor of a few for our other galaxies). It is therefore significant that we observe unusually low values of nuclear ellipticity in the centers of all TDE host galaxies. If elevated TDE rates in post-starburst galaxies were due to the production of β ≳ 0.5 anisotropies in nuclear starbursts, many TDE host galaxies should undergo the radial orbit instability and acquire prolate-to-triaxial nuclear geometries. The absence of evidence for this in four TDE host galaxies is a significant indication that radial orbit biases do not explain the TDE host galaxy preference.

The low central ellipticities we observe are not inconsistent with triaxial geometries, however. A central bar (on <100 pc scales) or many other nonaxisymmetric central geometries²⁰ can elevate TDE rates by an order of magnitude (Merritt & Poon 2004). Triaxialities can be observed through changes in the position angle with radius. We examine the position angles in Section 3.7, finding a range of variations in the PA with radius at higher radii (>100 pc). However, we do not have the resolution to be sensitive to position angle twists on scales ≪100 pc, as studied by Merritt & Poon (2004). Triaxiality will increase the TDE rate on scales smaller than the radius of gravitational influence (Vasiliev et al. 2014), to which our observations are not sensitive.

The observations of the surface brightness profiles and ellipticity profiles also allow us to test the possibility proposed by Cen (2020), in which a massive ∼100 pc disk with >50% of the stellar mass produces a high rate of TDEs around the secondary SMBH as it inspirals through the disk. There are two separate observational consequences, depending on whether the massive disk is closer to being edge-on or face-on. If the disk is closer to face-on, we would observe a dominant central disk component in our Galfit modeling (Section 3.2). However, the four TDE host galaxies considered here are better fit by a high Sérsic index model than a low Sérsic index or exponential disk.²¹ If the massive disk were closer to edge-on, we would observe elliptical isophotes in the central ∼100 pc, inconsistent with the low central ellipticities observed by the four TDE host galaxies considered here. However, the discovery and identification of TDEs around the secondary SMBHs considered by Cen (2020) may be selected against as possible TDEs, as they will be offset from the primary nucleus, and the primary SMBH could be an AGN. Cen (2020) predicts a high instance of repeating TDEs in such systems; studies of the host galaxies of any repeating TDEs would be greatly informative in constraining this possibility.

The presence of bars can also have an impact on the funneling of gas toward the nuclei of galaxies and the formation of central eccentric disks (Hopkins & Quataert 2010a, 2010b), which may be related to the TDE rate (Madigan et al. 2018; Wernke & Madigan 2019). However, larger-scale bars such as those seen in the host galaxies of iPTF15af and potentially PTF09ge are common among low-redshift galaxies (Elmegreen et al. 2004), and there is no evidence that barred galaxies have a higher TDE rate. Further study of the kinematics of TDE host galaxies on scales close to r_infl ∼ 0.3–3 pc will be needed to determine the possible impact of these mechanisms. Depending on the SMBH mass, this will require a TDE within ∼50 Mpc (to resolve 10 pc scales) or within ∼5 Mpc (to resolve 1 pc scales).