A TIDAL DISRUPTION EVENT IN A NEARBY GALAXY HOSTING AN INTERMEDIATE MASS BLACK HOLE

EUVE pointed in the direction of A1795 seven times over 2.5 yr. For all the observations we analyzed the events recorded with the Deep Survey (DS) telescope (Bowyer & Malina 1991) with the Lexan/Boron filter in the 0.0404–0.2816 keV (67–178 Å) energy range. Significant emission was detected only in a ∼70.8 ks observation on 1998 March 27. In an earlier, very long observation (∼90 ks) on 1997 February 3 and in five shorter following pointings (up to ∼25 ks) in 1999, we were able to detect only emission coinciding with the center of the galaxy cluster, and in particular with the galaxy B2 1346+26 (the EUVE bright source J1348+26.5B).

The cluster was the target of all the observations and, consequently, it was always on-axis, allowing us to consider the EUVE PSF as undistorted (the DS focal plane is curved, while the detector is flat). Since the distance of CXO J1348 from the central galaxy is ∼07 and the angular resolution of EUVE is ∼03, the emission from the two objects overlaps.

To estimate the net flux associated with the transient we adopted the following procedure: We extracted the surface brightness profile (counts s⁻¹ arcmin⁻²) for both of the long observations (in 1997 and in 1998) using the reprocessed images downloaded from the HEASARC archive and a series of annular regions centered at the cluster position. The annuli were 04 wide. We used the deadtime-corrected exposure time from the header of the FITS files. A comparison of the two profiles shows that the transient is significant within the first 15 from the cluster center. The counts associated with the cluster extend up to 4'. We extracted the net count rate from the inner 15 region for both epochs and an annular region (4'–5' as inner/outer radii) as background. To generate a response file, we downloaded the effective area for the DS Lexar/Boron filter from the EUVE handbook.

Using Xspec for the observation in 1997, we created a spectrum with one bin covering the 0.0404–0.2816 keV range. Using the APEC thermal model (with kT = 0.13 keV and abundances 0.31 Z_☉; Bonamente et al. 2001) to model the soft cluster emission in the EUVE range, we found an observed flux of (1.36 ± 0.06) × 10⁻¹² erg cm⁻² s⁻¹. We generated a spectrum also for the 1998 observation using a combined model: the thermal model above and a BB component with a temperature of kT = 0.09 keV (see Section 2.1.2). Subtracting the cluster contribution from the total flux, (5.43 ± 0.12) × 10⁻¹² erg cm⁻² s⁻¹, the observed flux from CXO J1348 is (4.07 ± 0.13) × 10⁻¹² erg cm⁻² s⁻¹. Using WebPIMMS, the flux was extrapolated to the Chandra 0.5–7 keV energy range. As a result we found that the absorbed and unabsorbed fluxes are (3.15 ± 0.10) and (3.41 ± 0.10) × 10⁻¹² erg cm⁻² s⁻¹, respectively, at this time.

A1795 is a familiar target for Chandra (Weisskopf et al. 2000): except for the first two pointings (PI: Fabian), the cluster has been used as a calibration source. Up to 2012 March, the telescope observed this field 43 times. Visual inspection of the images reveals that for only 26 pointings the position of CXO J1348 was within the boundaries of the ACIS (I or S) detectors. The list of useable observations is given in Table 1. As mentioned before, only exposures up to 2004 January 18 detect emission from CXO J1348 (see Figure 1 for a snapshot of all the Chandra observations from 1999 to 2004). Running wavdetect we found the following coordinates: $\alpha = 13^{\rm h} 48^{\rm m} 49\buildrel{\mathrm{s}}\over{.}87$; δ = +26°35'576, with a systematic error of 06. As explained in the CIAO 4.4 science thread, after correcting the aspect files we merged the ACIS-I data obtained in 2005 March and all the remaining ACIS-I and ACIS-S observations up to 2012. No significant excess was found at the transient position.

Zoom In Zoom Out Reset image size

Before performing the data analysis of those observations with a detection, we re-generated the event 2 files using the chandra_repro script. We checked for the presence of flaring activity in the ACIS background: only very short time intervals were excluded from the following analysis. We selected the larger energy range 0.3–8 keV to estimate the net count rate and the source significance, while we limited the spectral analysis to the artifact-free 0.5–7 keV range. The spectral files for the source and the background as well as the response files were generated by the tool specextract, using a circular source extraction region with radius of 2'' for the 1999 observation, when the source was brightest, 15 for the observation in 2000, and 10 for the remaining observations in 2002 and 2004. Smaller extraction regions were chosen in later epochs to reduce the contamination from the cluster X-ray light. A smoothed image shows that the cluster emission still has a high gradient at the position of CXO J1348. For this reason, we selected as background regions 2 boxes along the isophotes of the smoothed emission, positioned on the two sides of the transient. The boxes are 4'' wide and 10'' long. The source spectrum was grouped to a minimum number of 15 counts per channel when the source was bright (in 1999 and 2000).

The 0.5–7 keV spectrum of CXO J1348 in 1999 can be adequately fitted (χ² = 15.5 for 12 dof) by a single BB model with kT = 0.10 ± 0.01 keV and absorption fixed at the Galactic level (left panel of Figure 2). The unabsorbed flux is (2.8 ± 0.4) × 10⁻¹⁴ erg cm⁻² s⁻¹. There is a hint of an excess above 1 keV and we added a second BB component to improve the fit. We found that the new fit (χ² = 10.3 for 10 dof) is obtained with kT₁ = 0.08 ± 0.02 keV and kT₂ = 0.25 ± 0.19 keV. The second component is not statistically significant as the probability obtained with the F-test is only 13%. Using more complicated spectral models (diskbb or diskpbb) does not improve the fit, since they still do not compensate for the excess above 1 keV. Alternatively, a good fit can be obtained using a bremsstrahlung model. We found a temperature of 0.19 ± 0.03 keV (χ² = 13.6 for 12 dof) and an absorbed flux of (2.6 ± 0.4) × 10⁻¹⁴ erg cm⁻² s⁻¹. A fit with a power-law with a photon index of Γ = 4.35 ± 0.27 is also a reasonable representation of the X-ray spectrum (χ² = 11.6 for 12 dof). Such a steep spectrum typically mimics thermal emission over the limited Chandra bandpass. Similar results have been found using an un-grouped spectrum and the C-statistic.

Zoom In Zoom Out Reset image size

The observation obtained in 2000 was well fit (χ² = 7.8 for 7 dof) by an absorbed BB model (right panel of Figure 2). The temperature is kT = 0.09  ±  0.01 keV and the unabsorbed flux is (2.1 ± 0.4) × 10⁻¹⁴ erg cm⁻² s⁻¹. No emission associated with the transient is observed above 2 keV. The use of a bremsstrahlung model does not fit the spectrum better: although the temperature (kT = 0.16 ± 0.03 keV) is very similar to the 1999 observation, the residuals are higher (χ² = 9.8 for 7 dof).

In 2002 the source faded significantly and only 27 net counts in the 0.3–8 keV range are possibly associated with CXO J1348. The detection significance is 4.1σ (that increases to 5.3σ in the 0.3–2 keV range). A spectral analysis using the C-statistic still shows the presence of the thermal BB component (kT = 0.07 ± 0.03 keV) but at lower intensity (2$^{+2}_{-1} \times 10^{-15}$ erg cm⁻² s⁻¹ in the 0.5–7 keV range).

In all these trials we allowed an additional absorption at the cluster distance to vary, but its value was either small/unconstrained or did not improve the fit.

In 2004, the source was observed three times in a four day time span and it showed some signs of flaring activity: on January 14, the source was not significantly detected on the ACIS-S detector (0.9σ, with an upper limit on the count rate of <4.7 × 10⁻⁴ in the 0.3–8 keV range), but it re-appeared on January 16 with a detection significance of 3.5σ (ACIS-S net count rate (1.4 ± 0.4) × 10⁻³ counts s⁻¹). The transient was visible also on January 18 with a significance of 3.5σ (ACIS-I net count rate (1.3 ± 0.3) × 10⁻³ counts s⁻¹). Due to the extremely poor statistics, no spectral analysis was performed. The unabsorbed flux in the 0.5–7 keV range are (4 ± 1) × 10⁻¹⁵ erg cm⁻² s⁻¹ and (1.5 ± 0.5) × 10⁻¹⁴ erg cm⁻² s⁻¹, using the Chandra PIMMS tool¹⁰ and assuming a BB model with a temperature of 0.09 keV (similar to the values found in the 1999–2002 period). For the observation on January 14, we estimated a 3σ upper limit of <1.5 × 10⁻¹⁵ erg cm⁻² s⁻¹.

Visual inspection of all the pointings starting in 2005 does not reveal the presence of the transient. We estimated an upper limit from the first pointing obtained on 2005 March 20 because the most sensitive detector ACIS-S was used. Since the cluster is the dominant source of emission, the upper limit remains high at <1.4 × 10⁻¹⁵ erg cm⁻² s⁻¹.

The field of A1795 has been extensively observed with the NRAO¹¹ VLA due to interest in its central bright radio galaxy, 1356+268 (e.g., Ge & Owen 1993). Considering the time span after the detection of the transient X-ray source, we selected and analyzed archival VLA data of the field consisting of observations obtained at three epochs from 2000 October to 2005 October in the ∼5–8 GHz range. All on-source exposures were single snapshots lasting 1–10 minutes, using various VLA configurations. The data were calibrated in AIPS using standard procedures and self-calibration and imaging were performed with DIFMAP (Shepherd et al. 1994). No significant radio emission was detected at the position of CXO J1348 in any of the VLA observations with point source limits ranging from <0.10 to <0.32 mJy (see Table 2).

Observatorio Astronoḿico Nacional/San Pedro Mártir (OAN/SPM) Johnson Telescope. Data were obtained with the multi-channel Reionization And Transients InfraRed camera (RATIR; Butler et al. 2012; Watson et al. 2012) mounted on the 1.5 m OAN/SPM telescope in Baja California, México. On 2013 February 12 and 19, we took a series of 60 s exposures with dithering between them in various filters (the number of exposures are indicate in parenthesis): g (80), r (80), i (160), Z (120), J (120), and H (22). Given the small galaxy size, a sky frame was created from a median stack of all the images in each filter. Flat-field frames consist of evening sky exposures. Due to lack of a cold shutter in RATIR's design, IR darks are not available. Laboratory testing, however, confirms that dark current is negligible in both IR detectors (Fox et al. 2012). The photometric images were reduced and co-added using standard CCD and IR processing techniques in IDL and Python.

Nordic Optical Telescope (NOT). We obtained three images in the B and i filters on 2012 March 20 with the ALFOSC camera and five images in the U filter on 2013 March 14 with the MOSCA camera mounted on NOT (Karttunen 1993). Exposure times were 500, 300, and 600 s, respectively. On both occasions, the sky conditions were photometric, however the seeing was variable. The frames were reduced and co-added using standard IRAF procedures (de-biasing, flatfield correction).

Large Binocular Telescope (LBT). On 2013 April 2 we obtained an optical spectrum with the Multi-Object Double Spectrograph (MODS1) instrument at the focus of the two 8 m mirrors of the LBT (Hill et al. 2000) using a 10 wide slit and the G400L and G670L grisms for the blue and red channels, respectively, covering the 3200–5800 Å and 5800–10000 Å range. The Clear filter was used for both grisms. The spectral resolution is δλ*c/λ ∼ 200–400 km s⁻¹, depending on the spectral region. Since MODS1 does not have an atmospheric dispersion corrector, the slit was oriented along the mean parallactic angle (P.A. = 70°). Conditions were clear with an average seeing always better than 15 FWHM and the observations were done in a sequence of four 1800 s exposures for a total integration time of 2 hr. The data were reduced by the LBT data center. Since the optical host is a dim object, we were able to detect only a weak continuum emission, with signal-to-noise ratio (S/N)  ≈  3 per pixel (≈8 per resolution element) in the range from 6000–8000 Å. Due to the red galaxy color, the S/N decreases as a function of wavelength and no significant signal is detected below ≈4000 Å. No significant features are observed over the range from 4000–9500 Å, neither in absorption nor emission. Specifically, for the region from ≈6000–8000 Å, we limit the flux from any emission line to be f ≲ 10⁻¹⁷ erg cm⁻² s⁻¹ (assuming a line width of several hundred km s⁻¹, corresponding to our instrumental resolution).

Hubble Space Telescope (HST). Although A1795 has been the target of many HST pointings, due to the very small field of view CXO J1348 fell on a detector only in a single set of observations. On 1999 April 11 the telescope observed the source position using the Wide Field Planetary Camera 2 (Holtzman et al. 1995) for 300 s, both with the F555W and the F814W filters. Unfortunately, at that time the drizzling technique to remove cosmic rays was not adopted and single images (per filter) were taken. An optical counterpart at the position of the X-ray transient is detected on the WF2 chip in both filters. A visual inspection of the processed images revealed the presence of cosmic rays very close to the counterpart in the F814W image. The photometry in this filter is, thus, unreliable. The estimated flux in the F555W filter is corrected for the finite aperture using Table 2 in Holtzman et al. (1995).

We also check the morphology of the optical source by comparing it with an artificial PSF. We used the web interface of the PSF modeling tool Tiny Tim (Krist et al. 2011) to generate a PSF located at the same position on the WF2 chip of the F555W filter, and we assumed a spectrum described by a power-law with index −1 (changing the spectral slope does not alter the shape of the PSF significantly). We selected the F555W filter because there were no identifiable cosmic rays close to the source. Since the object is very dim and the PSF is under-sampled, we smoothed the image using a Gaussian function with a 2 pixel kernel radius. We extracted a brightness profile along the east–west direction (the diffraction spikes do not contribute since they are tilted by 10°). The profile of the source was compared with that of the generated PSF, and smoothed with the same kernel function. The Gaussian function that describes the artificial PSF has a FWHM of 028, while the source brightness profile has a larger width (FWHM = 040). This indicates that the source is spatially resolved, suggesting a galaxy-like morphology. A comparison between the images in the two filters shows that there is some additional emission above the PSF and 3σ above the local background located at P.A. = 215° and with extension of 06. The source measures 105 along this angle and 07 perpendicularly. The short exposure of the two images did not allow us to speculate on the nature of this feature.

Very Large Telescope (VLT) and Canada–France–Hawaii Telescope (CFHT). The optical counterpart at the position of CXO J1348 has been observed with the FORS1 camera on the VLT (Nicklas et al. 1997) on 2002 June 8 and July 19, and the MegaPrime/MegaCam on CFHT (Boulade et al. 2003) on 2008 August 5 and 2009 June 25. Frames from the two telescopes have been corrected by means of bias or dark frames and response was normalized by means of flat-field frames. Given the lack of variability due to the temporal proximity of the two sets of VLT data, we decided to sum the observations to increase the signal-to-noise ratio, in particular in the U filter where the source was barely detected.

Isaac Newton Telescope (INT). Querying the online catalogs at HEASARC we found that photometric measurements for an object consistent with the Chandra source were already available. The optical source was observed as part of the WIde-field Nearby Galaxy-cluster Survey (WINGS; Varela et al. 2009) between 2000 and 2001. The automatic software that run the analysis in the B and V filters determined that the object (WINGS J134849.88+263557.5) can be classified as a galaxy. The galaxy cluster has been extensively observed through the years, from 1992 to 2010, using the Wide Field Camera on the INT (Lewis et al. 2000). While the images in the r and i filters suffer from bad fringing, the observations in the U, B, and V can be used to extract valuable photometry. The source does not show any sign of variability. In Table 3 we report the values obtained in the 2010 May campaign, a period not covered by other ground-based telescopes.

Spitzer. Only one imaging data set covering the location of CXO J1348 is available in the Spitzer archive. It was taken on 2010 August 8 with the IRAC instrument (Fazio et al. 2004) in both the 3.6 μm and 4.5 μm bandpasses. A visual inspection reveals that only in the 3.6 μm (a 380 s exposure) mosaic image there is a source at the X-ray transient position detected at the 2σ level, while in the 4.5 μm bandpass the significance is 1σ. Unfortunately, the relatively short exposures and the contamination from the galaxy at the center of the cluster do not allow us to extract a reliable and meaningful photometric measurement: using circular regions for both the source and the background with a 2 pixel radius and applying the aperture corrections listed in Section 4.10 of the IRAC instrument handbook, we obtained flux densities of 12.8 ± 7.7 μJy and 3.4 ± 3.0 μJy in the 3.6 μm and 4.5 μm bandpasses, respectively. Due to the large errors, we do not use these values in the analysis.

Swift Ultra-Violet/Optical Telescope (UVOT). A1795 was observed with UVOT (Roming et al. 2005) in the UVW1 filter on board the Swift satellite on 2005 November 12. We used uvotimsum to combine the total of ∼24 ks exposure obtained over three orbits. Running uvotsource we found that no significant emission above the background was observed at the X-ray position with an upper limit, m_UVW1 < 22.4 (in the Vega system).

Since no spectroscopic redshift of the host galaxy is available, we combined photometry from archival VLT, CFHT, INT, and HST data with new observations from RATIR and NOT to produce a detailed SED of the host galaxy (right panel of Figure 3).

We fitted the SED using the EaZy code (Brammer et al. 2008) and models from Bruzual & Charlot (2003): We considered different kinds of galaxy spectral templates (elliptical, early and late spirals, irregular and starburst galaxies), set at different ages and with various metallicities and star formation rates. Although many templates provide a reasonable fit of the putative host galaxy SED, the best fit is obtained by either an elliptical or S0 template, both dominated by an old, evolved stellar population, located at a redshift of $z = 0.13^{+0.18}_{-0.05}$ (68% confidence level). The derived photometric redshift is compatible with the average value of the cluster redshift (z = 0.062476) and the dispersion of the velocities of the cluster brightest galaxies, whose redshifts range from 0.054 and 0.068 (Smith et al. 2004).

The inferred de-reddened magnitude at the cluster redshift is M_B ∼ −13.8 (M_R ∼ −15.1) and the estimated scale is 1.188 kpc arcsec⁻¹ (Wright 2006). Based on the result of the spatial analysis of the HST data, this corresponds to a minimum radial extension of 0.4 kpc (and up to 0.7 kpc along the putative feature observed to the south-west of the source). Thus, both the source brightness and its radial extent suggest that we are observing either a compact elliptical galaxy or a spiral galaxy with a small bulge and even dimmer spiral arms. Both scenarios are supported by the reasonable fit obtained using an elliptical or S0 template for the estimate of the photo-redshift.

The uncertainties on the photometric redshift are somewhat large in the upper end side: the 68% error puts the host galaxy at z = 0.31, an increase by a factor of 5.8 in the luminosity distance and of 3.8 in the angular scale. Thus, the source would be 3.8 magnitude brighter and 1.5 kpc in size.

The assumption that the host galaxy belongs to the cluster A1795 is supported by its projected location, very close to the main galaxy of the cluster, and by the richness of the cluster. Querying the SDSS catalog for an area with a radius of 20', where previous works (see, e.g., Oegerle & Hill 2001) have found the majority of the cluster components, we find that more than 150 galaxies (corresponding to three out of four galaxies with a spectroscopic measurement) have a redshift compatible with the cluster. We cannot rule out entirely the possibility that the host galaxy belongs to a more distant group or is a background object. Of the remaining galaxies with an estimate of their distance, 20 sources have a redshift of z = 0.11, indicating that another cluster may be located behind A1795, ∼15 objects have redshift more evenly distributed in the ranges 0.15–0.20 and 0.24–0.31, while a few other sources are located at cosmological distances.

The radio emission observed in these objects is produced by accelerated particles that travel in collimated, relativistic jets and emit synchrotron radiation. As mentioned earlier, the object is not found in VLA archival images down to a 3σ flux limit of <0.1 mJy. At a distance of z ∼ 0.07, the radio luminosity is L_8.5 GHz < 10³⁸ erg s⁻¹. This makes the putative AGN a radio-dim (or even a radio-quiet) source. Indeed, observations of the Hubble deep fields at 8.5 GHz show that all the radio sources have fluxes below the above upper limit and they are not identified with quasars but rather with star-forming galaxies, bright field elliptical and late-type galaxies with evidence of nuclear activity (Seyferts) and field spiral galaxies (Richards et al. 1998). The upper limit excludes not only radio-loud galaxies but also radio-loud quasars (blazars): their radio luminosities start two orders of magnitude higher (Donato et al. 2001; Giommi et al. 2012) and the broadband radio-to-optical spectral index (α_ro) between 8.46 GHz and 5500 Å is higher than the upper limit of 0.28 found for CXO J1348, a value that places the putative AGN in the radio-quiet regime.

Furthermore, blazars typically occur in very massive, luminous galaxies (see, e.g., Urry et al. 2000) and there is some evidence that they might be considered quasi-standard candles, with optical absolute magnitude greater than −24, many orders of magnitude brighter than what is observed in this host galaxy.

Among the radio-quiet AGNs, like quasars, Seyfert galaxies, and low-ionization nuclear emission-line regions, narrow-line Seyfert 1 (NLSy1) galaxies are the most variable, although the changes are not as dramatic as in CXO J1348. In NLSy1 galaxies, the variability can be explained as due to changes in the properties of the absorber surrounding the inner BH such as column density, ionization parameter, and covering factor (e.g., Grupe et al. 2012). The highest variability was seen in WPVS 007 (Grupe et al. 1995) with a decrease in flux by a factor of 400 in the soft band in a 3 yr time span, while fast variability has been observed in NGC 4051 (McHardy et al. 2004) with a change of the X-ray flux by a factor of 10 in a few days. This could explain changes seen in Chandra data alone, but not if we assume that the EUVE transient is the same source.

One difference between CXO J1348 and the radio-quiet AGNs is the shape of the X-ray spectrum: radio-quiet AGNs have their spectra described by a power-law with spectral index in the range 1.5–3.5, irrespective if the source harbors a SMBH or an IMBH, and only the more luminous (and more massive) objects show the presence of the soft excess with temperature above 0.1 keV (Brandt et al. 1997; Grupe et al. 1995; Porquet et al. 2004; Miniutti et al. 2009; Pian et al. 2010; Dong et al. 2012). The NLSy1 galaxies do show a steep spectrum in the soft band (some of the ROSAT spectra have slopes above 4), but they are detected at higher energies as well: the ASCA spectra can be fitted in first approximation with a combination of a power-law model and a thermal BB component with temperature in the range 0.1–0.2 keV (Leighly 1999). On the contrary, the spectrum of CXO J1348 does not show any emission above 3 keV and, consequently, no additional power-law model is needed.

At optical wavelengths, Zhou et al. (2006) found that in a sample of ∼2000 optically selected NLSy1 in the SDSS, their absolute magnitude in the g' filter is in the range $-18 \lesssim M_{\rm g^{\prime }} \lesssim -26$, while the host galaxy of CXO J1348 is much dimmer ($M_{\rm g^{\prime }} \sim -14.1$).