The nuclear transient AT 2017gge: a tidal disruption event in a dusty and gas-rich environment and the awakening of a dormant SMBH

A dense photometric coverage of the first $\sim$40 days of AT 2017gge have been performed in the $o$ and $c$ filters by the ATLAS survey Smith et al. (2020) and in the $i^{\prime}$ filter by the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS, Huber et al., 2015; Chambers et al., 2016). Additional $w$ images have been obtained by Pan-STARRS between 600 and 700 days from the transient discovery. The spectroscopic monitoring was carried out mainly by using the ESO Faint Object Spectrograph and Camera (EFOSC2, Buzzoni et al., 1984), mounted on the 3.58 m New Technology Telescope (NTT) as part of the ePESSTO ESO public survey Smartt et al. (2015) and the Andalucia Faint Object Spectrograph and Camera (ALFOSC), mounted on the 2.56 m Nordic Optical Telescope (NOT), as part of the NOT Unbiased Transient Survey (NUTS¹¹1http://csp2.lco.cl/not/). Additional optical observations and late time images have been obtained with the Device Optimized for the LOw RESolution (DOLORES) installed at the 3.58 m Telescopio Nazionale Galileo (TNG), the Auxiliary-port CAMera (ACAM) mounted on the 4.2 m William Herschel Telescope (WHT) and the Optical Wide Field Camera (IO:O) at the Liverpool Telescope (LT, Steele et al., 2004), respectively. Higher resolution spectra were obtained with the Gemini North Multi-Object Spectrographs (GMOS-N), mounted on the 8.1 m Frederick C. Gillett Gemini North telescope, with the Optical System for Imaging, the low-Intermediate-Resolution Integrated Spectroscopy (OSIRIS), mounted at the 10.4 m Gran Telescopio CANARIAS (GTC) and the X-shooter spectrograph Vernet et al. (2011), mounted on the UT3 the Very Large Telescope (VLT), located at the ESO Paranal observatory. NIR spectra and images in the $J$, $H$ and $Ks$ bands have been obtained with the infrared spectrograph and imaging camera Son of ISAAC (SofI Moorwood et al., 1998), mounted on the NTT. After the transient light had faded below the detection levels, host galaxy spectra and images were taken for each ground-based facility and with the same observational set-up used during the follow-up.

All the spectroscopic observations have been carried out with the slit oriented at the parallactic angle. In Table 1 the main observing information, such as the observing date, the grism used, the exposure time, the slit width, the airmass and seeing condition, are reported for each instrument. The sequence of the optical spectra is shown in Figure 14, the SofI NIR spectra are shown in the lower panel of Figure 5, while we show the whole X-shooter spectrum (UVB, VIS and NIR arms) in Figure 16.

AT 2017gge was monitored by the Neil Gehrels Swift observatory over a period spanning $\sim$200 days, starting from $\sim$60 days after the transient discovery. A total of 19 XRT and UVOT observations have been obtained. In particular, we were awarded six epochs of Swift ToO observations between 2018 March 09 and 2018 April 07 for late-time follow-up of the source. In Table 2 a list of the Swift observations for AT 2017gge are reported, together with the main properties (date of observation, time after the transient discovery and instruments exposure times). All the XRT observation have been executed in photon counting mode.

In order to analyse the photometric properties of AT 2017gge, we first have carried out differential photometry procedures against the host galaxy for each images of our data-set. This has allowed us to subtract the host galaxy contribution and thus to obtain images where only the nuclear transient’s emission is present. To this purpose we used the pre-transient host galaxy images available in the Pan-STARRS1 data archive (DR2) in the $g^{\prime}$ and $r^{\prime}$ filters, and the host cataloged $u^{\prime}$ image available on the SDSS archive (DR16, Ahumada et al., 2020; Doi et al., 2010). For the data in the NIR taken with NTT/SofI, we have used the host galaxy images obtained with the same instrument at very late times, 1346 days from the AT 2017gge discovery.

For each images we subtracted the host galaxy contribution by using hotpants V5.1.11 software (Becker, 2015). We have then applied aperture photometry on the resulting images by using the sextractor software with apertures of variable size, depending on the seeing conditions. In Table 5 we report the measured host-subtracted optical and IR magnitudes of AT 2017gge, not corrected for foreground extinction and in their common systems (AB for Sloan filters and Vega for Johnson filters). In the case of the UVOT filters, we have used the host contribution as determined from the photometric analysis performed on the very late-time observation (taken at $\sim$1680 days from the transient discovery). The resulting light curves, including also the host-subtracted UVOT, ATLAS and PS1 data, are shown in Figure 2.

From the multi-color photometry derived from our analysis, together with the values from the ATLAS and PS1 surveys, we computed the bolometric luminosities for AT 2017gge. We used the python routine superbol Nicholl (2018), all the input magnitudes have been corrected for the Galactic extinction from Schlafly & Finkbeiner (2011), which assumes a reddening law with $R_{V}=3.1$, and K-correction Oke & Sandage (1968) has been applied. When needed, we extrapolated the photometry assuming a constant color evolution for the light curve. Subsequently, we have integrated over the spectral energy distribution (SED) inferred from the multi-band data (for each epoch) and, finally, we fitted the SED with a single black body function. The best-fit black body at each epoch was then used to compute the additional flux bluewards of the UVW2 band and redwards of $K$ band. In Figure 3 we show the results for the first $\sim$300 days from the transient’s discovery: the bolometric luminosity evolution (upper panel, black circles for the luminosity calculated by integrating over observed fluxes only, the red diamonds show the luminosity derived using a single black body model to fit the SED); the black body temperature and radius evolution are shown in the central and lower panel, respectively. While the black body temperature is characterized by no significant evolution over time, being consistent with a constant value of $T_{\rm BB}\sim$1.8$\times$10⁴ K, the black body radius show a first phase of expansion from $R_{\rm BB}$=(7.7$\pm$0.8)$\times$10¹⁴ cm to a maximum value of $R_{\rm BB}\sim$(12.9$\pm$2.8)$\times$10¹⁴ cm, reached at $\sim$25 days from the transient’s discovery, followed by a phase of decline finally reaching a value of $R_{\rm BB}$=(7.3$\pm$2.7)$\times$10¹⁴ after $\sim$300 days form the transient’s discovery.

In Figure 4 we show the bolometric luminosity of AT 2017gge (derived using the black body correction, filled black square) in comparison with the bolometric light curves derived for other TDEs⁵⁵5Photometric data retrieved from The Open TDE Catalog https://tde.space by applying the same method used for AT 2017gge data (colored dashed lines). After PS1-10jh Gezari et al. (2012), AT 2017gge is the second brightest object in the plot, with a luminosity and an evolution notably similar to what observed for AT 2018hyz. Moreover, for the first $\sim$100 days it shows a light curve behaviour comparable to the one observed for AT 2019qiz. From the bolometric luminosity evolution we estimate that the emission reached its peak value of $L_{\rm bol}$=(1.4$\pm$0.5)$\times$10⁴⁴ erg s^-1 at MJD 57 989.26, which corresponds to $\sim$20 days from the transient’s discovery. The rise of the light curve is consistent with a $t^{1/2}$ powerlaw, while in decline it is better represented by a $t^{-1.1}$ powerlaw (red dot-dashed line) rather than the $t^{-5/3}$ trend (light-blue dot-dashed line), when considering $\sim$600 days of emission. We note that during the first $\sim$200 days the light curve decline can be well represented also by the $t^{-5/3}$ power-law. The total energy radiated ($E_{\rm rad}$) is derived by integrating the bolometric luminosity over time and it results $E_{\rm rad}$= (1.0$\pm$0.1)$\times 10^{51}$ erg. In their recent letter on AT 2017gge, Wang et al. (2022) independently derived the bolometric lightcurve, together with the black body temperature and radius by using all the UVOT data but V filter and only the ATLAS $o$ band data. Although their values are consistent with our results, they found a declining trend for the bolometric lightcurve luminosity compatible with the $t^{-5/3}$ powerlaw. This discrepancy can be explained with the use of a more complete and well sampled photometric dataset in our analysis.

The photometric properties, such as the steep rise to the peak luminosity, reached after $\sim$20 days from the transient’s discovery, the powerlaw decay of the bolometric luminosity the constant black body temperature at $T_{\rm BB}\sim$1.8$\times$10⁴ K and the black body radius $R_{\rm BB}$ $\sim$10¹⁵ cm evolving with time, are all commonly observed in TDEs (Hinkle et al., 2020; van Velzen et al., 2020; van Velzen et al., 2021b; Zabludoff et al., 2021, and reference therein), and thus are indicative of the TDE nature of AT 2017gge.

Interestingly, AT 2017gge is listed in the Mid-InfraRed Outburst in Nearby Galaxies (MIRONG) sample Jiang et al. (2021), which includes a total of 137 low-redshift SDSS galaxies that have experienced recent MIR flares of at least an amplitude of 0.5 mag in their WISE light curves. In Figure 5 we show the AT 2017gge WISE light curves for the $W1$ and $W2$ filters (with half a year sampling) and the epochs of the two SofI spectra. The MIR flare is detected at $\sim$197 days from the transient’s discovery in both filters and the first SofI spectrum has been obtained 178 days after the MIR flare, when the MIR emission is still $\sim$0.6 dex higher than during the quiescent phase.

Notably, in their work, Jiang et al. (2021) propose IR echoes (i.e. optical/UV light absorbed and reradiated by dust) of emission originating from transient accretion onto SMBHs as the main source of the detected MIR outbursts and they derive some physical properties of the dust responsible for the detected emission, such as the temperature, luminosity, mass and its distance from the central heating source. Specifically, for the case of AT 2017gge they infer a distance of the dust of $\log R_{\rm dust}$=-1.51$\pm$0.12 pc or $\log R_{dust}$=-1.01$\pm$0.16 pc, depending on the considered absorption coefficient, which correspond to $R_{\rm dust}\sim$10¹⁷cm. We independently computed the time lags ($\tau$) between the optical and MIR data using two different methods, the interpolated cross correlation function (ICCF, White & Peterson, 1994) and the javelin tool Zu et al. (2013, 2016). From the ICCF method, the time lag $\tau$ for $W1$ band is estimated to be $\tau=204.6^{+77.6}_{-43.5}$ days, which is consistent with that computed applying the javelin method ($\tau\approx 207.6^{+24.7}_{-24.7}$ days). These methods are commonly used in active galactic nuclei (AGN) reverberation mapping studies and are based on the tight correlation between the IR luminosity and the dust radius of ordinary AGNs (size–luminosity relation). Thus, it is possible to infer the luminosity of a possible AGN from the estimated time lag. In particular, by using the size–luminosity relation from Lyu et al. (2019), we found $L_{\rm AGN}\approx 10^{45}$ erg s^-1.

Our estimated peak bolometric luminosity for AT 2017gge from the observed optical and UV fluxes was (1.4$\pm$0.5)$\times$10⁴⁴ erg s^-1, but we can infer that the true peak was larger than this due to the presence of the IR echo. Therefore it is not unreasonable to assume that the intrinsic TDE luminosity was large enough to be consistent with this estimate. The time lag measurements above also provide values for the distance to the dust producing the IR echo, which are $0.161^{+0.061}_{-0.034}$ pc for CCF, and $0.163\pm 0.019$ pc for javelin. These values are similar to the dust sublimation radius of 0.15 pc found for the TDE PTF-09ge by van Velzen et al. (2016), implying that is is reasonable for AT 2017gge to have also sublimated dust out to this radius. The estimates for the dust radius in Jiang et al. (2021) are somewhat smaller than this, and precise measurements require a more involved model. It is possible to use the infrared light echo to estimate the required luminosity of the flare needed to sublimate the dust to the sublimation radius. In particular, we have obtained a flare peak luminosity of $L_{\rm abs}$=1.14$\times$10⁴⁵erg s^-1by using the the Equation 1 reported in van Velzen et al. (2016), the (independently derived) sublimation radius of 0.16 pc and assuming a sublimation temperature of 1850 K for the dust. Notably, this value is consistent with the luminosity inferred by using the AGN size–luminosity relation of Lyu et al. (2019) ($L_{\rm AGN}\approx 10^{45}$ erg s^-1) and thus further support the hypothesis of an intrinsic TDE luminosity larger than what derived from the observed optical and UV fluxes. Interestingly, from the the MIR luminosity analysis, Wang et al. (2022) derived a covering factor of $\sim$0.2 for the dusty environment. As already discussed by the authors, such a high value is comparable with what is usually observed for the AGN torus. However, prior to the transient occurrence, an AGN classification has been excluded given: a) the X-ray non-detection in the RASS observations, b) the quiescent WISE color (W1-W2$\sim$0.6, Stern et al., 2012) and c) the the SDSS spectrum based location in the Baldwin, Phillips & Terlevich (BPT) diagrams. We however note that the pre-transient SDSS values placed the host galaxy on the (theoretical) extreme starburst line and this may indicate the presence of a weak AGN contribution, see Section 8.1 for a detailed discussion).

Wang et al. (2022) estimated the total radiated energy of AT 2017gge in the IR (until July 2021) of 1.3$\times$10⁵¹ erg. Combined with our estimated radiated energy in the optical the total radiated energy detected is $\sim$2$\times$10⁵¹ erg and this is a lower limit for the total radiated energy from the TDE. This is still significantly lower than the total energy of 10⁵³ erg expected to be released by the accretion of a solar mass star (see e.g. Giannios & Metzger, 2011). However, the covering factor of $\sim$0.2 implies that a large proportion of radiation in the UV could still be unobserved in both the optical and IR bands and make up for this discrepancy.

In Figure 6 the XRT light curve in the 0.3-10 keV band, produced by specifying a minimum of 3 counts per bin, is shown. In order to convert the count rates to unabsorbed fluxes, we used the results obtained from the spectral fitting with an absorbed powerlaw model (see subsection 4.2 for a detailed discussion on the spectral analysis), which gives a conversion factor of 0.07$\times$10⁻⁹ erg cm^-2 cts^-1. The RASS pre-transient upper limit of F_0.3-10keV=1.0$\times$10^-12 erg cm^-2 s^-1, corresponding to observation taken at MJD 48102, is also shown before the transient detection for comparison (red cross in Figure 6). The detected X-ray emission has a transient nature, with only upper limits (grey triangles in the Figure) for the first 170 days from the discovery of AT 2017gge and showing the peak around 200 days. Furthermore, no X-rays have been detected in the recent XRT observations taken at $\sim$1684 days (see Wang et al., 2022). Such a delayed X-ray brightening with respect to the optical/UV peak has already been observed in some TDEs, specifically, ASASSN-14li Pasham et al. (2017); Gezari et al. (2017), ASASSN-15oi Gezari et al. (2017); Holoien et al. (2018), AT 2019azh van Velzen et al. (2021b); Liu et al. (2022) and OGLE16aaa Kajava et al. (2020); Shu et al. (2020), which have shown $\sim$30 days, 1 year, 200 and 140 days of delay in the X-ray emission, respectively. In all these cases the authors investigate the different scenarios for the origin and location of the optical/UV emission and the delayed X-ray flare, including the stream-stream collision and reprocessing photosphere/winds hypothesis. Among these, a two-process scenario, where the UV/optical emission is produced by the stellar debris streams during the circularization phase, whereas a delayed formation of an accretion disk is responsible for the X-rays, has been suggested and, in some cases, has been even preferred (i.e. ASASSN-15oi, ASASSN-14li, OGLE16aaa, AT 2019azh Gezari et al., 2017; Pasham et al., 2017; Kajava et al., 2020; Liu et al., 2022).

The spectrum of the source is very soft, with the emission detected only between the 0.3-1 keV energy range. We modelled the data with an absorbed powerlaw by using the model TBABS*ZASHIFT*POWERLAW (solid blue line, left panel) and we obtain a photon index of $\Gamma$=6.0$\pm$0.8 and a c-stat/d.o.f.=11.74/12 The unabsorbed flux in the 0.3-10 keV band results to be $F_{0.3-10}$=(0.18$\pm$0.05)$\times$10^-12 erg cm^-2 s^-1, which corresponds to an X-ray luminosity of $L_{\rm X}$=(2.4$\pm$0.7)$\times$10⁴² erg s^-1, calculated for the distance of AT 2017gge. These spectral features are usually observed in the X-ray emission of non-jetted TDEs, which show very soft spectra well represented either by a single powerlaw with a photon index value $\Gamma$>4 or by a black body model with temperatures between 10-100 eV Auchettl et al. (2017); Saxton et al. (2020). Therefore, we also modelled the X-ray spectrum using a black body model TBABS*ZASHIFT*BBODYRAD (middle panel of Figure 7, magenta solid line). In this case we obtain c-stat/d.o.f.=7.24/12. The inferred temperature and radius are k$T$=70$\pm$10 eV and $R$=(1.4${}^{+0.7}_{-0.4}$)$\times$10¹¹ cm. These values are consistent with a TDE nature for this transient. The unabsorbed flux in the 0.3-10 keV band is $F_{0.3-10}$=(0.15$\pm$0.04)$\times$10^-12 erg cm^-2 s^-1, which corresponds to a luminosity of $L_{\rm X}$=(1.8$\pm$0.5$)\times$10⁴² erg s^-1. The results of the spectral analysis are reported in Table 3 and are shown if Figure 7. These results are all consistent with the X-ray analysis already presented in the recent discovery paper of AT 2017gge Wang et al. (2022). Additionally, in order to investigate the origin of the delayed X-ray flare as produced by a newly formed accretion disk, we have modelled the data with the TBABS*ZASHIFT*DISKBB model (right panel of Figure 7, red solid line). In this case we obtain a c-stat/d.o.f.=7.76/12, a temperature at the inner disk radius of T_in=83$\pm$10 eV, an unabsorbed flux of $F_{0.3-10}$=(0.15$\pm$0.04)$\times$10^-12 erg cm^-2 s^-1, which corresponds to a luminosity of $L_{\rm X}$=(1.9$\pm$0.5)$\times$10⁴² erg s^-1. We note that although it is difficult to discriminate between these two models given the very similar c-stats values, the accretion disk model is able to represent well the data. Thus, given the derived X-ray properties, we can conclude that the AT 2017gge X-ray emission resembles what is usually seen in non-jetted TDEs, as it is characterized by a very soft emission and a steep spectrum well modelled either by an absorbed powerlaw, by a black body or by an accretion disk model.

The first spectra of our data-set (taken $\sim$42-45 days from the transient’s discovery) are characterized by broad H$\upbeta$, He i$\lambda$5876 and H$\upalpha$ emission lines superimposed on a blue continuum. However, from the spectral sequence shown in Figure 14, a clear evolution emerges with time of the broad emission line profiles and, more in general, of the spectral properties, with a number of features, consistent with high ionization coronal emission lines, appearing later in time, following the X-ray flare. In Figure 8, the evolution of the lines profile from a selection of normalised spectra is shown for the H$\upbeta$ and the H$\upalpha$ regions (left and right panel, respectively). The transition of the H$\upalpha$ from a broad-component-dominated profile toward a narrow-component-dominated profile (pink and violet spectra in Figure 8, respectively) is clearly visible, as well as the emergence of the narrow emission lines of [N ii] and [S ii]. A similar behaviour is observed for the emission lines in the H$\upbeta$ region.

Of particular interest is the emergence of the broad He ii $\lambda$4686, which is not present in the first spectrum of the sequence nor in the pre-transient SDSS spectrum Smee et al. (2013), as shown in Figure 8 (pink and black spectra, respectively). This feature is first detected in the spectrum taken after 193.9 days from the transient’s discovery and it is characterized by a line profile time evolution, with the gradual narrowing of the broad component until only a narrow core remains, which is well visible in the Gemini spectrum taken at day 407.9 (dark green in Figure 8). Notably, the phase in which the broad He ii $\lambda$4686 first appears corresponds to the detection epoch of the soft X-ray flare, as shown in the X-ray light-curve in Figure 6.

In Figure 9 we show a comparison between the late-time Gemini and X-shooter spectra (taken at days 408, 572 and 1698, respectively) and the pre-transient SDSS spectrum. A number of high ionization coronal narrow lines compatible with some iron transitions (i.e. ([Fe xiv] $\lambda$5303, [Fe vii] $\lambda\lambda$5720,6087, [Fe x] $\lambda$6374) are well visible in the Gemini and X-shooter spectra, with the exception of the [Fe xiv] $\lambda$5303, which is not detected anymore in the X-shooter spectrum. We note that the derived X-ray temperature of $\sim$70 eV, imply that these lines arise from photoionization, rather than from collisional ionization equilibrium Lundqvist et al. (2022); Arnaud & Raymond (1992). As most of these coronal lines are first detected 218 days from the transient’s discovery and are not detected at the resolution of the pre-transient SDSS spectrum, we conclude that these features have a transient nature. These findings are remarkable as, together with the TDE AT 2019qiz Nicholl et al. (2020), they represent the first time in which high ionization coronal emission lines are detected in the optical spectra of a TDE soon after the occurrence of the X-ray flare (see Short et al., in prep., for an accurate study on the detection of the high ionization coronal emission lines in the optical spectra of AT 2019qiz). This strongly suggests a close connection between the TDE X-ray outburst and the presence of high ionization coronal emission lines in the TDE late-time spectra and host galaxy.

In the lower panel of Figure 5 we show the NIR spectra taken with SofI at two different epochs from the transient’s discovery, specifically at 374.7 days (in black) and 1342 days (in orange) in comparison with a part of the NIR X-shooter spectrum taken after 1698 days (in red). Although our NIR data-set is composed of only three spectra, a spectral time evolution is visible also in this wavelength range. Indeed, the first SofI spectrum shows a prominent and composite emission line in the He i $\lambda$10830 region, with a very broad component and a narrower one (FWHM=7660$\pm$1500 km s^-1 and FWHM=1639$\pm$330 km s^-1, respectively), blended with the Pa$\upgamma$ in which we also detect an intense narrow emission line consistent with the high ionization coronal line [Fe xiii] $\lambda$10798 (see the upper right panel of Figure 5). This finding is particularly notably as it represents the first detection of a transient high-ionization coronal NIR line in a TDE. Indeed, such a broad and complex feature becomes less prominent in the second SofI spectrum, where the broad component becomes fainter, the Pa$\upgamma$ starts to emerge and the narrow [Fe xiii] $\lambda$10798 is not detected anymore. Finally, this broad feature has completely disappeared in the X-shooter spectrum, the He i $\lambda$10830 and the Pa$\upgamma$ are well detected and show only narrow components. Also in this case, the narrow [Fe xiii] $\lambda$10798 is not detected.

In order to investigate the properties and evolution of the main spectral features observed in the host-subtracted spectra of AT 2017gge (see Figure 15 for the entire sequence), we modelled them with Gaussian functions by using the python packages curvefit and leastsq. In the case of simultaneous presence of broad and narrow components in the line profiles, a multi-component Gaussian fit has been applied. During the fitting procedure, a wavelength window of $\sim$1100 Å has been selected, in order to include both the spectral features of interest and the local continuum. A narrower fitting window has been selected in late time spectra, when the width of the broad emission lines became smaller. The same fitting procedure has been applied also in the case of not host-subtracted spectra, such as the SDSS, SofI, Gemini and X-shooter spectra. In Figure 11 the application of the multi-component Gaussian fit for the H$\upbeta$ and H$\upalpha$ region in the Gemini and X-shooter spectra is shown, while in the upper right panel of Figure 5 it is shown for the He i $\lambda$10830 region of the SofI spectrum of day 374.

We analyzed the evolution of the emission line properties, such as the full width at half maximum (FWHM), the equivalent width (EW) and the EW ratio of the main broad emission lines detected in the host-subtracted spectra of AT 2017gge. Specifically, we focus on the properties of the He ii $\lambda$4686, H$\upbeta$, He i $\lambda$5876 and H$\upalpha$ emission lines, as derived from the fitting procedure described in section 7.1. In Figure 12 the results from this analysis are shown. Broad components in the H$\upbeta$, He i $\lambda$5876 and H$\upalpha$ are detected starting from the first spectra of our data-set, taken 41.7 and 44 days from the transient’s discovery, and they last until phase 291 (for He i), 408 (for H$\upbeta$) and 572 (for H$\upalpha$). As shown in the upper panels of Figure 12, all these three features have high values of the FWHM in the initial phases of the transient emission with FWHM$\sim$(1.3-2.0)$\times$10⁴ km s^-1, which are consistent with the broad emission lines widths typically observed in TDEs Arcavi et al. (2014); van Velzen et al. (2020); Charalampopoulos et al. (2022). In this picture, the He ii $\lambda$4686 broad emission line is an exception: it is characterized by a late-time development, a FWHM one order of magnitude smaller (FWHM$\sim$10³ km s^-1) than what measured for the H and He i $\lambda$5876 lines and it shows a completely different time evolution.

We investigated on the activity status of the AT 2017gge host galaxy by using the BPT diagrams (Baldwin et al., 1981) and the EW of the narrow H$\upalpha$, H$\upbeta$, [O iii] $\lambda$5007, [N ii] $\lambda$6583 and [S ii] $\lambda\lambda$ 6716,6731 lines detected in these spectra. The lines profiles have been modelled following the method described in Section 7.1.

The line ratios derived from the SDSS spectrum are log([N ii]/H$\upalpha$)=$-$0.27$\pm$0.0), log([O iii]/H$\upbeta$)=$-$0.40$\pm$0.33 and log([S ii]/H$\upalpha$)=$-$0.45$\pm$0.10. Despite the faint [O iii] $\lambda$5007 emission line, these values place the galaxy on the boundary between the HII and composite regions in the BPT diagram shown in the left panel of Figure 13 and well inside the star–forming region in the BPT diagram shown in the right panel of Figure 13. We note that the location on the extreme starburst line may indicate that an additional ionising component is required to explain the line ratios. The line ratios obtained from the Gemini spectrum of day 408 are log([N ii]/H$\upalpha$)=$-$0.04$\pm$0.08, log([O iii]/H$\upbeta$)=$-$0.08$\pm$0.11 and log([S ii]/H$\upalpha$)=$-$0.09$\pm$0.12. While the line ratios obtained from the Gemini spectrum of day 572 are log([N ii]/H$\upalpha$)=$-$0.03$\pm$0.06, log([O iii]/H$\upbeta$)=$-$0.2$\pm$ 0.10 and log([S ii]/H$\upalpha$)=0.01$\pm$ 0.07. In this case, the derived line ratios values are consistent between each other and place the galaxy well inside the LINER region and on the boundary between the composite and the AGN regions (cyan square and magenta triangle in Figure 13), indicating an increase in the activity of the galaxy nucleus after the nuclear transient took place. Finally, the line ratios obtained from the X-shooter spectrum (taken at 1698 days) are log([N ii]/H$\upalpha$)=$-$0.36$\pm$0.03, log([O iii]/H$\upbeta$)=$-$0.08$\pm$0.03, and log([S ii]/H$\upalpha$)=$-$0.54$\pm$0.04 and place the galaxy inside the composite and the starforming regions in the BPT diagrams of Figure 13 (green star in the left and right panels, respectively), indicating a further variation in the host galaxy activity, which come back to values in agreement with what derived from the pre-transient spectrum of SDSS. A similar variation in the activity status of a TDE host galaxy has been observed also in AT 2019qiz (Short et al, in prep.).

The discovery of AT 2017gge in a starforming galaxy make this TDE particularly interesting as it probes a host galaxy population (the blue galaxies) poorly represented in most optically and X-ray selected TDE host galaxies samples which result to be dominated by green valley galaxies Hammerstein et al. (2021); Sazonov et al. (2021).

The detection of AGN-like features in the Gemini optical spectra, their location in the BPT diagnostic diagrams (see Figure 13) and the evolution of the H$\upalpha$/H$\upbeta$ line ratio, suggest an AGN-like behaviour of the AT 2017gge host galaxy after the TDE occurrence. This led us to use these two spectra and a single epoch (SE) relation to test a different and independent method for the SMBH mass derivation, beside the commonly used $M-\sigma_{\star}$ scaling relations. Recently, Ricci et al. (2017) derived a SE relation based on the AGN X-ray luminosity in the 2-10 keV band and on the width of H$\upbeta$ and/or H$\upalpha$ broad components. If the broad H emission lines detected in the Gemini spectra are emitted in a virialized photosphere (similar to an AGN-like BLR), then we can use this SE relation to estimate the mass of the black hole.

To this scope we first have derived the unabsorbed X-ray luminosity in the 2-10 keV band from the Swift/XRT spectral analysis of $L_{\rm X}$=(6$\pm$3)$\times$10³⁹ erg s^-1. Then we have fitted the H broad emission lines detected in the Gemini spectra. In particular, the broad components detected in the H$\upbeta$ and H$\upalpha$ emission lines of the Gemini spectrum taken at 408 days results to be centered at $\lambda_{c}$ = 4857.30$\pm$1.40 Å and $\lambda_{c}$= 6559.10$\pm$0.62 Å and are characterized by a FWHM=2318$\pm$268 km s^-1 and FWHM=2760$\pm$69 km s^-1, for the H$\upbeta$ and the H$\upalpha$ respectively. In the Gemini spectrum of 572 days only the H$\upalpha$ broad component is detected, which results centered at $\lambda_{c}$=6556.08 $\pm$ 0.49 Å and with a FWHM=2274$\pm$38 km s^-1.

By using the virial estimator presented in Table 4 of Ricci et al. (2017) with the zero points constants valid for all the classes of host galaxies we derived a black hole mass of $\log M_{\rm BH}$=5.5$\pm$0.3 when using the width of the broad H$\upbeta$, a $\log M_{\rm BH}$=5.8$\pm$0.3 and $\log M_{\rm BH}$=5.4$\pm$0.3 when using the width of the broad H$\upalpha$ measured in the Gemini spectra of 408 days and 572 days, respectively. Finally, we derive the BH mass of the host galaxy by using the $M_{\rm BH}-\sigma_{\star}$ scaling relation of McConnell & Ma (2013) and Gültekin et al. (2009), following the method illustrated in Wevers et al. (2017, 2019a) for a sample of TDEs. To this purpose, we measured the stellar velocity dispersion of the host galaxy of AT 2017gge by applying the ppxf fitting procedure to the X-shooter UVB and VIS spectra. We derive a stellar velocity dispersion of $\sigma_{\star}$=(97$\pm$3) km s^-1, which correspond to a $\log M_{\rm BH}$=6.55$\pm$0.45 or to a $\log M_{\rm BH}$=6.80$\pm$0.43, when using the $M_{\rm BH}-\sigma_{\star}$ scaling relation of McConnell & Ma (2013) or the relation of Gültekin et al. (2009), respectively. We note that by using the SE relation of Ricci et al. (2017) we obtain values for the BH mass that are $\sim$ one order of magnitudes lower than what is derived with the $M_{\rm BH}-\sigma_{\star}$ relation. Moreover, given the bolometric luminosity of $L_{bol}$=4.6$\times$10⁴⁴ at peak, the BH masses derived with the SE relation would imply a super Eddington accretion (with an Eddington ratio $\sim$10). Instead, a Eddington limited accretion would requires a BH mass of $\sim$4$\times$10⁶ M_⊙, which is consistent with the values obtained with the $M_{\rm BH}-\sigma_{\star}$ relations. Thus we favour the BH mass as derived from the $M_{\rm BH}-\sigma_{\star}$ scaling relation for AT 2017gge and we suggests that the region emitting the broad lines in this TDE is not virialized as in the case of the AGN BLR.

The detection of very broad (FWHM $\sim$ 10⁴ km s^-1) H emission lines in the optical spectra place AT 2017gge among the H-rich TDEs subclass Arcavi et al. (2014); van Velzen et al. (2020); Charalampopoulos et al. (2022). Notably, in correspondence of a delayed X-ray flare, a broad (FWHM $\sim$10³ km s^-1) He ii $\lambda$4686 emerges after 194 days from the transient discovery, indicating a transition toward a TDE H+He subclass. Together with the He ii $\lambda$4686, also a number of high ionization coronal emission lines appear. These spectral features and the IR echo, strongly indicate the presence of a gas-rich and dust-rich environment surrounding the source, consistent with the starforming nature of the host galaxy.

The MIR reverberation signal is detected with a delay of $\sim$200 days from the optical peak emission and corresponds to a distance for the dust producing the IR echo of $\sim$0.16 pc, as derived in this work by using two cross correlations methods. This value is similar to the sublimation radius obtained for the case of the TDE PTF-09ge van Velzen et al. (2016) and suggests that AT 2017gge may have sublimated the pre-existing in-situ dust out to this radius. Our estimation of the the total (optical/UV+IR) radiated energy of 2$\times$10⁵¹ erg is significantly lower that the 10⁵³ erg that has been theoretically predicted for the case of the disruption of a solar mass star, despite the large covering factor of $\sim$0.2 derived by Wang et al. (2022). This could be explained by a large proportion of far-UV emission remaining unobserved, as potentially indicated by the large intrinsic luminosity required to produce the delay time we observe in the IR. This can be compared with TDEs in galaxies with a very high covering factor for the SMBH such as the IR-luminous, highly obscured TDE Arp 299-B AT1, where a direct integration over time of the IR SED yielded a total radiated energy of $>$2$\times$10⁵² erg Mattila et al. (2018); Reynolds et al. (2022).

We have presented three NIR spectra taken after the IR echo and we report the detection of a broad feature in the He i $\lambda$10830 and a intense Fe xiii coronal emission line, both gradually disappearing in the subsequent spectra. This is the first time that a broad feature and a high-ionization coronal line are detected in the NIR spectra of a TDE taken following an IR echo from surrounding dust (but see the case of the candidate TDE AT 2017gbl in which NIR spectra have been presented, Kool et al., 2020).

The presence of the luminous and transient high ionization coronal emission lines in the spectra obtained soon after the detection of the soft X-ray emission implies a strong correlation with the soft X-ray emission and a gas-rich environment of the TDE, as expected from the starforming nature of the host galaxy. These high ionization coronal lines are long-lasting as, together with a narrow He ii $\lambda$4686, are still present in the X-shooter medium resolution spectrum taken after 1698 days from the transient discovery. Together with the TDE AT 2019qiz (Short et al., in prep.), this is the first time that high ionization coronal emission lines are detected in the optical spectra of TDEs following the X-ray outburst. This strongly indicates a close connection between the two phenomena and support the hypothesis that the extreme coronal emission lines detected in the spectra of a sample of non active galaxies could be a signature of the occurrence of a TDE in the past (as suggested by Wang et al., 2012).

We study the evolution of the EW, the FWHM and the line ratios of the broad H$\upbeta$, He ii $\lambda$4686, H$\upalpha$ and He i $\lambda$5876 emission lines detected in the host-subtracted spectra of AT 2017gge. With the exception of the He ii emission line, all the other broad lines are detected starting from the very first spectra (at day $\sim$40) with FWHM$\sim$10⁴ km s^-1and show a declining trend with time. In particular, the H$\upbeta$ have a different behaviour with respect to the H$\upalpha$ and He i $\lambda$5876, being characterized by a faster decline before becoming undetectable after $\sim$400 days from the transient discovery. Instead, the broad H$\upalpha$ is characterized by a slow declining trend and it is long-lived as it is still detected in the X-shooter spectrum, after 1698 days from the transient discovery, when all the other broad lines are no longer detectable. Contrary to what observed in the H$\upalpha$, the broad He i $\lambda$5876 keeps constant high values of the FWHM for $\sim$250 days and soon after this phase it starts a very rapid decline, which last only $\sim$50 days. Notably, this almost coincides with the appearing of the broad He ii $\lambda$4686 emission line. This feature it is first detected at 194 days, soon after the delayed X-ray flare, and it does not show a particular declining trend. Instead it keeps an almost constant width of FWHM$\sim$6$\times$10³ km s^-1 (one order of magnitude smaller than what measured for the broad H and He i) until its last detection at 338.6 days.

The observed different behaviour of the various broad emission lines is indicative of a stratified photosphere where different lines are produced at a different distance from the continuum source. Moreover, the time evolution of the various lines ratios can give us more insight in the properties of the region emitting the broad lines. In particular, the observed rising trend of the He ii/He i line ratio can be ascribe to an increase in the photoionization of the broad Helium emitting region. Given that the He ii line is known to be produced by photoionisation due to (soft) X-ray photons, we suggests that the observed delayed X-ray flare could be responsible for this phenomenon. Furthermore, the observed evolution of the H$\upalpha$/H$\upbeta$ line ratio, which experiences variation from a values consistent with the Case B recombination to values predicted for homogeneous AGN BLR models, is mainly due to the faster decay of the broad H$\upbeta$ with respect to the H$\upalpha$.

The X-ray peak brightening detected with a delay of $\sim$200 days with respect to the UV/optical peak emission, followed by the development of a broad He ii $\lambda$4686, suggests a different origin for the two signals in AT 2017gge. A delayed X-ray emission is not new in the TDE field. It has been predicted to eventually occur in some models of TDE emission (i.e. reprocessing envelope model, the stream-stream collision scenario and the TDE unified model, Guillochon et al., 2014; Piran et al., 2015; Dai et al., 2018) and it has been already observed in some TDEs (i.e. ASASSN-14li, ASASSN-15oi, AT 2019azh and OGLE16aaa, Pasham et al., 2017; Gezari et al., 2017; Holoien et al., 2018; Liu et al., 2022; Kajava et al., 2020; Shu et al., 2020). Recently, Hayasaki & Jonker (2021) proposed an interesting two emission scenario in order to explain such a time delay between the two signals, observed in a sub-sample of optically selected TDEs. In particular, in this picture, the optical emission is produced by the collision of intercepting stellar debris streams during the initial phase of the stellar disruption, while the X-ray flare is emitted following the formation of an accretion disk, after the end of the circularization process.

As suggested also by Wang et al. (2022), the AT 2017gge observed emission can be well described by this scenario. Indeed, the results from the optical photometry, the X-ray spectral analysis and even from the optical spectroscopy, are in line with this picture. In fact, while the optical/UV black body radius results ranging between $R_{\rm BB}$=7-13$\times$10¹⁴ cm, from the X-ray spectral fitting we derive a much smaller value for the black body radius, $R_{\rm BB}\sim$1.4$\times$10¹¹ cm, suggesting a different location for the source of the two signals. Moreover, an accretion disk model is able to represent well the X-ray spectrum.

The non-detection of the He ii in the optical spectra taken prior to the X-ray flare may indicate the absence of the X-ray emission at these epochs, as the He ii line is known to be correlated with the soft X-ray photons Pakull & Angebault (1986); Schaerer et al. (2019); Cannizzaro et al. (2021). However we note that in the case of the TDE ASASSN-15oi and AT 2019azh a broad He ii feature is detected well before the onset of a delayed soft X-ray flare Gezari et al. (2017); van Velzen et al. (2021b); Hinkle et al. (2021). Additionally, in the recent work of Nicholl et al. (2022), where the light-curves of 32 optically bright TDEs have been thoroughly analyzed, it has been found that the events without He ii were consistent with a stream-crossing origin for the luminosity, while those with He ii were more consistent with forming accretion disks. The late-time developing of the He ii observed in AT 2017gge could further support this scenario as this TDE seems to transition between these two regimes. However, we caution that there are also alternative explanations connecting both the optical/UV and X-ray emissions to the accretion phenomena, such as the lowering of the optical depth of an outflowing wind or of a reprocessing region, that cannot be excluded from our study. In particular, the delayed X-ray brightening observed in AT 2017gge could still be explained within a reprocessing scenario in which the decreasing of the optical depth of a reprocessing material at later times finally reveals the gas-rich circumnuclear environment from the flare and release the soft X-rays photons, allowing them to power the the narrow He II and iron coronal lines.

In order to understand the impact that a TDE can have on the host galaxy nucleus, we have used the pre-transient SDSS and the Gemini and X-shooter late-time optical spectra (taken after $\sim$408, 572 and 1698 days from the AT 2017gge discovery, respectively) to investigate on the activity status of host galaxy through the BPT diagrams. From this analysis a variation on the host galaxy activity clearly emerges. In particular, while the line ratio measured from the pre-transient SDSS spectrum are consistent with the starforming classification (although its location on the the extreme starburst line suggests the presence of an additional ionising component), the values derived from the Gemini spectra place the galaxy at the border between the composite and the AGN region and well inside the LINER region, suggesting an AGN-like activity between 400 and 600 days from the transient discovery. Finally, the galaxy activity status has come back to the initial values at 1698 days, as derived from the X-shooter spectrum. All this evidence suggests the possibility that the occurrence of the TDE resulted in the enhancement of the activity of the SMBH, which passes from quiescent to an AGN-like, with the formation of a transient stratified photosphere, similar to a BLR, surrounding the source.

In the hypothesis of the formation of a virialized BLR-like photosphere, we test an independent method to derive the SMBH mass by using the SE relation developed by Ricci et al. (2017) and the broad H$\upalpha$ and H$\upbeta$ as measured in the Gemini spectra (which correspond to the phase of the increased activity of the host galaxy in the BPT diagrams). We derive a SMBH mass which is one order of magnitude smaller than what derived by using the canonical $M_{\rm BH}-\sigma_{\star}$ scaling relations developed by McConnell & Ma (2013) and Gültekin et al. (2009) and the stellar velocity dispersion as measured from the X-shooter spectrum ($\log M_{\rm BH}$=5.4-5.7 and $\log M_{\rm BH}$=6.6-6.8, respectively). Given the peak bolometric luminosity of $L_{\rm bol}$=1.4$\times$10⁴⁴ erg s^-1, an Eddington limited accretion would requires a BH mass of $\sim$4$\times$10⁶ M_⊙, consistent with the values obtained from the $M_{\rm BH}-\sigma_{\star}$ relation. Thus, we consider this method more trustworthy and we suggest that the broad lines emitting region is not virialized.