AGN with discordant optical and X-ray classification are not a physical family: Diverse origin in two AGN

Approximately 3-17 percent of Active Galactic Nuclei (AGN) without detected rest-frame UV/optical broad emission lines (type-2 AGN) do not show absorption in X-rays. The physical origin behind the apparently discordant optical/X-ray properties is not fully understood. Our study aims at providing insight into this issue by conducting a detailed analysis of the nuclear dust extinction and X-ray absorption properties of two AGN with low X-ray absorption and with high optical extinction, for which a rich set of high quality spectroscopic data is available from XMM-Newton archive data in X-rays and XSHOOTER proprietary data at UV-to-NIR wavelengths. In order to unveil the apparent mismatch, we have determined the A$_{\rm V}$/N$_{\rm H}$ and both the Super Massive Black Hole (SMBH) and the host galaxy masses. We find that the mismatch is caused in one case by an abnormally high dust-to-gas ratio that makes the UV/optical emission to appear more obscured than in the X-rays. For the other object we find that the dust-to-gas ratio is similar to the Galactic one but the AGN is hosted by a very massive galaxy so that the broad emission lines and the nuclear continuum are swamped by the star-light and difficult to detect.


INTRODUCTION
The standard unified model of AGN (Antonucci 1993, Urry & Padovani 1995 explains the observed differences between optical type-1 and type-2 AGN through orientation effects. If our line-of-sight to the central engine intercepts the nuclear absorber invoked by unified models, the UV/optical continuum emission, the rest-frame UV/optical broad emission lines (line widths ≥1500 km/s in Full Width at Half Maximum, FWHM) and the X-ray emission, originated at sub parsec scales, should be absorbed. In this case, the AGN is classified as type-2. On the contrary, if we have a direct view of the central engine, UV/optical broad emission lines should be detected, while the X-ray emission should have low absorption (NH <4×10 21 cm −2 , the equivalent to AV=2 mag using a Galactic dust-to-gas ratio, Caccianiga et al. 2008). In this case, the AGN is classified as type-1.
The origin of such apparent discrepancies remain unclear, as well as the validity of the unified model for such AGN.
To unveil the nature of such discrepancies, we need detailed studies on these discordant AGN. For X-ray unabsorbed type-2 AGN, one possibility to explain this discrepancy can be the presence of a Comptonthick absorber (intrinsic NH equal or larger that the inverse of the Thomson cross-section: NH > σ −1 T =1.5×10 24 cm −2 ). In this case the direct X-ray emission below 10 keV should be completely suppressed and we would only de-tect scattered nuclear radiation (Braito et al. 2003, Akylas & Georgantopoulos 2009, Braito et al. 2011, Malizia et al. 2012. Since the scattered emission is only 1-3 per cent of the intrinsic AGN emission (Gilli, Salvati & Hasinger 2001, Comastri 2004, Georgantopoulos et al. 2011a, the sources would be identified as low luminosity, unabsorbed type-2 AGN. Another possibility is that the broad UV/optical lines are diluted by the host galaxy emission (Severgnini et al. 2003, Georgantopoulos & Georgakakis 2005, 2008). An alternative explanation could be a high dust-to-gas ratio: normally, AGN show dust-to-gas ratios below the Galactic standard or comparable (Maiolino et al. 2001, Vasudevan et al. 2009, Parisi 2011, Marchese et al. 2012, Hao et al. 2013, Burtscher et al. 2016, if this ratio is substantially higher, significant suppression of the broad line emission could take place without strong effects in the X-ray bands. Dust-to-gas ratios well above the Galactic value have been found in some AGN, although such cases are rare (Caccianiga et al. 2004, Trippe et al. 2010, Huang et al. 2012, Malizia et al. 2012, Masetti et al. 2012, Mehdipour, Branduardi-Raymont & Page 2012). In the sample of Maiolino et al. (2001), a sample of AGNs whose X-ray spectrum shows cold absorption and whose optical and/or IR spectrum show at least two broad lines, this is found in 9 per cent of the sources, and in the HBS28 sample (Caccianiga et al. 2004) this is only found in 3 per cent. In other objects optical observations show an intrinsically high Balmer decrement for the Hydrogen broad emission lines, while the X-ray spectra show low absorption (Barcons, Carrera & Ceballos 2003). A dusty-ionized absorber like the one in NGC 7679 (Della Ceca et al. 2001) can produce more relative absorption in the X-rays than in the optical emission. One last possible explanation is a variability scenario since optical and X-ray observations are normally obtained at different epochs. However, we note that, even with simultaneous observations, there are some objects whose optical and X-ray classification do not match (Corral et al. 2005, Bianchi et al. 2008. The objective of this study is to get insight into the physics behind the apparent mismatch between UV/optical and X-ray classification of two low-z AGN selected from the Bright Ultra-hard XMM-Newton Survey (BUXS, Mateos et al. 2012): 2XMMiJ000441.2+000711 (hereafter J00, z=0.1075, Abazajian et al. 2009) and2XMMJ025218.5-011746 (hereafter J02, z=0.0246, Jones et al. 2009). While both sources appear unabsorbed in X-rays, J00 has been optically classified as type-1.9 using the SDSS-DR7 spectrum (Abazajian et al. 2009) and J02 is a Sb edge-on galaxy (de Vaucouleurs et al. 1991) optically classified as a type-2 AGN (6dF spectrum, Jones et al. 2009).
For these sources, we obtained UV-to-NIR XSHOOTER spectra (P.I.: S. Mateos) and by comparing the properties derived in X-rays and in the UV/optical, we have tested three possible scenarios to explain the discordance: a) the presence of a Compton-thick AGN; b) these sources are a normal AGN but in a very massive host galaxy, or weak AGN in a normal host galaxy and their broad UV/optical emission lines are diluted by the host galaxy emission; c) intrinsic non-standard nuclear properties, such as a high dust-to-gas ratio or an intrinsically weak Broad Line Region (BLR). This paper is organized as follows. In Section 2 we ex-plain how our objects are selected. In Section 3 we describe the XMM-Newton data and in Section 4 we describe the XSHOOTER observations. In Section 5 we derive the AGN emission by subtracting the host galaxy emission, as well as the AGN intrinsic reddening, its emission lines and the host galaxy properties. We finally discuss all the possible contributions that can cause a mismatch between the X-ray and UV/optical properties of our objects in Section 6. Throughout this paper errors are 1σ. We assume a ΛCDM cosmology with ΩM=0.3, ΩΛ=0.7 and H0=70 km s −1 Mpc −1 .

THE SAMPLE
The two AGN analyzed in this work were selected from the wide-angle (44.43 sq. degree) Bright Ultra-hard XMM-Newton Survey ). This is a flux-limited sample of 255 AGN detected in the 4.5-10 keV band with the XMM-Newton observatory. The objects have relatively bright X-ray fluxes f 4.5−10 keV ≥ 6×10 −14 erg s −1 cm −2 . At the time of writing, the optical spectroscopic identification completeness is >98 percent. There are 111 AGN in BUXS with optical spectroscopic classifications 1.8, 1.9 or 2. Of these, the 9 objects showing low X-ray absorption (NH <4×10 21 cm −2 ; Caccianiga et al. 2008) that are visible from Paranal were included in a proposal for followup with XSHOOTER. Only the two objects with the lowest declination were successfully observed with XSHOOTER. We discuss here in detail the properties of these two objects.
For both sources, we have proprietary high resolution UV-to-NIR XSHOOTER spectra, as well as good quality XMM-Newton X-ray spectra (Jansen et al. 2001) in the observed energy range from 0.25 to 10 keV (Table 1 and Fig. 1).

X-RAY PROPERTIES
Source and background spectra were extracted at energies from 0.25 to 10 keV using circular regions. We used the XMM-Newton Science Analysis System (SAS) task eregionanalyze to obtain the circles that maximised the signal-to-noise ratio. For J00 we used radii of 34 and 32 arcsec for the MOS (Strüder et al. 2001) and pn (Turner et al. 2001) cameras, respectively. For J02 we used a radius of 29 arcsec for MOS and 25 arcsec for pn. The background spectra were extracted using circular regions of 50 arcsec radius located in source free regions in the same CCD chip as our objects. The response matrices and effective area curves were obtained using the SAS tasks rmfgen and arfgen, respectively. We combined MOS1 and MOS2 source and background spectra and the corresponding response matrices. The spectra were grouped with a minimum of 15 counts per bin and are shown in Fig. 1.
The X-ray spectroscopic analysis was conducted with the XSPEC package (v12.9.1; Arnaud 1996). We fitted the spectra with a combination of different models to determine the shape of the direct and scattered broad-band continuum components (modeled with power laws), soft excess (modeled with a black body) and rest-frame line-of-sight X-ray absorption. The models take into account the Galactic absorption using column densities taken from Dickey & Lockman (1990). We fitted both the pn and MOS spectra at the   Each component of the X-ray model is plotted with dotted lines. Right: MOS (black) and pn (red) X-ray spectra of J02 and the best fit model (absorbed power law) in solid lines. We also represent the ratio between the data and the best-fitting model.
same time with the parameters of the model tied, except for the continuum normalization to take into account crosscalibration problems between MOS and pn cameras (Mateos et al. 2009).
To accept the detection of a model component, we used the F-test with a significance threshold of 95 per cent. The X-ray luminosities have been computed in the rest-frame 2-10 keV energy band. They are corrected for X-ray absorption both Galactic and intrinsic to the sources. J00: It is X-ray unabsorbed. The 1σ upper limit on the column density is 6.7×10 20 cm −2 . The best-fitting model is a combination of a black body with temperature kT =0.16 +0.01 −0.02 keV and an unabsorbed power law with photon index Γ=1.66±0.09. The X-ray spectra (MOS and pn) are shown in Fig. 1 (left). For a sanity check, we computed the X-ray absorption with a fixed photon index of Γ=1.9, the typical value for type-1 AGN ( . We still classify this object as a low absorption AGN (NH ≤1.3×10 21 cm −2 ). The black body emis-sion is a phenomenological model often used in the literature  to fit the soft X-ray excess. Nevertheless this is not physically correct for J00, as it is well known that the temperature of the black body for a SMBH such as the one for J00 (see Sec. 5.3) is too low to explain the soft X-ray excess. We also tried a more physically motivated model, albeit more complex, replacing the black body model with a hot diffuse gas model (mekal model in XSPEC; Mewe et al. 1985, Mewe et al. 1986, Kaastra et al. 1992, Liedahl et al. 1995. This gives us an upper limit on the column density of 2.4×10 20 cm −2 (with Γ=1.83±0.07, χ 2 /d.o.f=173.7/139). The results, in terms of the X-ray classification as absorbed/unabsorbed, do not change using one model or the other. Therefore we use in this paper the results with the black body since we obtain a more conservative value of NH. The equivalent width of the Fe line at 6.4 keV in rest-frame is formally EW=0.30 +0.16 −0.17 keV, but adding this line is not statistically significant (∆χ 2 =3 for ∆d.o.f.=1, a 1.73σ detection). Looking at the spectrum of J00 ( Fig. 1) we can see a bump at the hard energies, but we believe that is not real, since it is only present in one of the EPIC cameras. This feature is probably associated with residuals in the background subtraction. Nevertheless, since the shape of the continuum is very well determined by the values at lower energies, this is not affecting our best-fit estimates.

UV-TO-NIR OBSERVATIONS
We have obtained UV-to-NIR high resolution spectra for both objects at the Very Large Telescope (VLT) with the VLT/XSHOOTER instrument (Vernet et al. 2011). The instrument divides the light in three paths that lead to three arms: one for the UV light, the second for the visible light and the last one for the near infrared (UVB, VIS and NIR, respectively). Each arm disperses the light with an echelle grating. The observations were taken with a 1.0 ×11 slit for the UVB arm and 0.9 ×11 slits for the VIS and NIR arms, respectively. Table 2 shows some technical details of the configuration setup of the observations.
The spectra were taken in nodding mode. In Table 3 we show some information about the spectra acquisition. We present in Fig. 2 the acquisition images of both objects and indicate the slit projected in the sky, the extraction region of the spectra and the parallactic angle. The observation dates were 2010-09-07 for J00 and 2010-09-04 for J02.
XSHOOTER is equipped with an Atmospheric Dispersion Correction (ADC) that allows the acquisition of the spectrum using any angle at any position in the sky. During our observations, the ADC was functional. For J02 we positioned the slit at an inclination angle close to the minor axis of the host galaxy to allow sky subtraction. Thanks to the ADC we could choose an inclination angle closer to the minor axis of the host galaxy. The observations were taken close to the meridian and with low air mass, meaning that the effect of the atmospheric dispersion is small.
The observations were reduced using the public XSHOOTER pipeline version 2.3.0 with Gasgano, following the instructions described in the XSHOOTER pipeline manual 1 . We used a binning in the wavelength direction 0.02 nm/pix for the UVB and VIS arms and 0.06 nm/pix for the NIR arm, and in the slit direction 0.16 arcsec/pix for the UVB and VIS arms and 0.21 arcsec/pix for the NIR arm, as specified in the XSHOOTER user manual. We used the standard procedures in the pipeline. IRAF Laplacian Cosmic Ray Identification task for spectroscopy (lacos spec, van Dokkum 2001) was applied to the raw images for cosmic ray rejection and then each arm was reduced individually. We used the standard star GD71 (RA=05:52:27.61, DEC=+15:53:13.8) to calibrate the flux of our spectra. There are three different recipes to do the flux calibration: the offset mode, the staring mode and the nodding mode. We used the recipe for the staring mode, as it provides a   better background subtraction. We used the software IRAF to extract the 1D spectra being careful to follow the trace. This was carried out with the routine apall. This is because the spectra have a trace whose centre was not constant in the cross-dispersion direction due to an imperfect order rectification of the pipeline. We joined the spectrum from each XSHOOTER arm following the information about the dichroic crossover region in the XSHOOTER User Manual. To join the UVB and VIS arms, the crossover region is at 5595Å, while between the VIS and NIR it is at 10140Å. The transition regions are 5560-5638Å and 10095-10350Å, respectively. A 0.9-1.1 scaling factor between arms is sometimes needed to match in flux the complete spectrum (López et al. 2016, Nisini et al. 2016. We used the continuum in these regions around the crossover points to compute the flux scaling factors for each arm using the VIS one as reference. We scaled the UVB arm spectra using a factor of 0.9 while for the NIR arm a 1.0 flux scaling factor was acceptable. Errors were propagated through this process. The aperture used to extract the spectra was defined to maximize the signal to noise of the AGN emission. This was carried out using the software IMFIT 3 on the acquisition images of the VLT observations. Both images were taken using the i-band filter. In both targets a bright nuclear source was detected (see contours in Fig. 2). To compute the fraction of AGN light that enters through the slit, for J00 we decomposed the emission in a Gaussian function for the AGN, and a Sersic profile for the host galaxy. J02 is an edge-on galaxy so we used a Gaussian profile for the AGN and two Sersic profiles for the host galaxy, to fit the bulge and the disk. The parameters of the Sersic and Gaussian models (intensity, σ, effective radius, Sersic index), are computed with errors of ∼20 percent or less, indicating reliable fits of the photometric images. In Fig. 2 the contours show that we have enough quality to fit the shape of the different components of our objects. We can trust the reliability of the acquisitions images to measure the slit losses using the width of the Gaussian profiles which are ∼2.5 and ∼3.5 pixels for J00 and J02, respectively.
We estimated that about 75 per cent and 56 per cent of the AGN emission are included in the slit for J00 and J02, respectively.
The reduction of the science spectra is concluded by correcting for Galactic extinction, putting the wavelength and flux in rest frame and converting the air wavelengths into vacuum wavelengths. Information to perform these corrections was obtained from the public data available in NASA/IPAC Extragalactic Database (NED).

ANALYSIS
In this section we describe the steps carried out to isolate the AGN emission in the XSHOOTER spectra and to determine the properties of the AGN and their host galaxies.

AGN and host galaxy continuum decomposition
To decompose the extracted spectra into AGN and host galaxy emission at UV-to-NIR wavelengths we used the software STARLIGHT (Cid Fernandes et al. 2005, Mateus et al. 2006.The best-fitting model was obtained by minimizing the χ 2 statistic. We modeled the spectra with a composite model including a host galaxy spectrum plus an absorbed (by nuclear and host galaxy extinction) AGN spectrum. To reduce noise and to follow the recommendations of the STARLIGHT manual we rebinned the spectrum to 2 A bins. SED decomposition is an approach commonly used to compute the relative contributions from the AGN and their hosts. Nevertheless, both the good spectral resolution and wide wavelength coverage of the X-SHOOTER spectra allow us to calculate the AGN and stellar components without resorting to a full SED decomposition.
We need to assume an spectral shape for the rest-frame UV to near-IR AGN continuum. There is substantial scatter in the continuum shapes of individual AGN which also depend on the SMBH mass and can vary with time (Koratkar & Blaes 1999, Schmidt et al. 2012, Baron et al. 2016. Since our aim is to reproduce the intrinsic AGN continuum by finding the model that best fits each source we have adopted the broken power law models (F λ ∝ λ α ) from Polletta et al. (2007). In these models the spectral index ranges from α=-1.9 to -1.4 for λ <10000Å and α=-0.8 to -0.6 redwards, and the break is located at 10000Å. This break is likely associated with the change in the AGN continuum slope between the IR bump and the Big blue bump (Koratkar & Blaes 1999). We used this information to create a set of broken power laws with the previously mentioned index range to reproduce the intrinsic AGN continuum emission of our objects. The steps in the power law index to create the grid of broken power law models are ∆α=0.125 for the blue region and ∆α=0.05 for the red.
The next parameter needed in our fit is the obscuration of the nuclear region of the AGN. We used the extinction model of the Small Magellanic Cloud (SMC; Gordon et al. 2003) as it is the one that fits better the AGN spectra (Hopkins al. 2004). We tried different extinction models but the SMC is the one that minimized the χ 2 . In particular we checked the Large Magellanic Cloud Super-Shell and Average models from Gordon et al. (2003), the Calzetti et al. (2000) law, the Milky Way model form Allen (1976), and the model from Cardelli, Clayton & Mathis (1989). In addition, using the SMC model we obtained the most conservative AV values. We constrained the nuclear extinction to be between AV=10 mag and AV=0 mag.
The host galaxy contribution is modeled using the Single Stellar Population (SSP) templates from the Bruzual & Charlot (2003) library. We used 45 models with metallicities Z=0.05, 0.02 and 0.004 in units of the solar value and ages ranging from 1 Myr to 13 Gy. We obtained SSP in the 0.8-13 Gyr range for our objects.
In order to fit the AGN and host galaxy continuum emission only, we masked out the spectral ranges where the AGN emission lines and the telluric absorption lines are expected (see grey and yellow bands in Fig. 3). We excluded the regions between 6850-6950Å, 7165-7210Å and 7550-7725Å, where some residuals of telluric features in the Bruzual & Charlot library are present.
The final spectral range used in the fit is between 3700 and 16000Å (rest-frame). Fig. 3 shows the results of the spectral decomposition. We see that in both objects, the total emission is dominated by the host galaxy. The preferred models are α blue =-1.90 and α red =-0.6 for J00 and for J02 α blue =-1.78 and α red =-0.6. The resulting values of intrinsic absorption associated to the AGN emission are AV=2.04±0.30 mag and AV=2.19±0.33 mag for J00 and for J02, respectively.
We computed the errors in the extinction by varying the power-law index and then fitting again and calculating the χ 2 statistics. The errors obtained are very small. We added an additional 15 percent error contribution to account for the overall uncertainty of the SMC model used (Gordon et al. 2003). This contribution to the error is the one that dominates.
We believe that the results from the AGN and host galaxy decomposition reported in this section are robust. The XSHOOTER spectra have a sufficiently large wavelength range to constrain in a robust way the host galaxy contribution, which is the major contribution at optical wavelengths. It is clear that in the spectra there are several stellar features that help constraining the emission from the hosts (see Fig. 3). If more relative contribution of the nuclei is present in the spectra, this would in fact, flatten the stellar features to a more featureless contribution. The STARLIGHT software fits this stellar features and the results in this paper are the best fitting ones.

Narrow line and broad line Balmer decrements
After removing the host galaxy component and after having taken into account the slit losses, we estimated the Balmer decrement by using the broad and narrow components of the Hα and H β emission lines. We fitted the nuclear spectra from rest-frame 6270 to 6800Å for Hα. For the H β region we fitted from rest-frame 4600 to 5050Å for J00 and from 4800 to 5050Å for J02. We used a wider wavelength range in the first object to ensure fitting the whole broad H β emission line. To model the AGN narrow emission lines we used Gaussian functions, assuming that they share the same width in velocity space. Hydrogen broad emission lines are fitted with Gaussian profiles as well. We included also a parameter for the broad line that is a shift of the centre of the Gaussian with respect to the vacuum values. This is because some AGN show broad lines with an offset, sometimes of a thousand of km/s due to strong winds (Sulentic et al. 2000, Steinhardt et al. 2012, Gaskell & Goosmann 2013. For J00 the shift is ∼30Å (∼1350km/s in velocity) while for J02 is ∼3Å (∼150km/s), not unusual with respect to observed shifts (Sulentic et al. 2000). We used a power law to fit the continuum around the lines. The line parameters were obtained with the CIAO's SHERPA fitting tool (Freeman et al. 2001).
Since the broad H β component was not detected in any of the spectra, only an upper limit could be computed for the Balmer decrement from the BLR. Line parameters and Balmer decrements are reported in Table 4. The narrow line Balmer decrement will be compared in Sec. 6.3 with the absorption in the X-rays and the extinction of the UV-to-NIR continuum.
In Fig. 4 we show our results while best-fitting values are indicated in Table 4.

SMBH masses
From the XSHOOTER nuclear spectra, we derived an estimate of the SMBH masses for our targets. There are many different ways to compute SMBH masses (Trippe 2015). We used the luminosity and FWHM of the Hα broad emission line and the expression from Greene & Ho (2005) (1) where MSMBH/M is the mass of the SMBH in solar units. LHα is the intrinsic luminosity of the Hα emission line in erg s −1 , this is, corrected for both the nuclear and the host galaxy extinction, computed from the spectral fits. Finally FWHMHα is the FWHM of the broad Hα emission line in km s −1 . We also have corrected the FWHM of the lines by the instrumental spectral dispersion by subtracting the instrumental broadening in quadrature. We show the SMBH masses computed using Eq. 1, as well as the relevant luminosities, in Table 4. In addition we show the bolometric and Eddington luminosities for our sources, as well as the Eddington ratio (L bol /L Edd. ). The bolometric luminosities are calculated using the expression L bol =9×λL5100 (Kaspi et al. 2000), being L5100 the monochromatic luminosity of the unreddened nuclear emission at rest-frame 5100Å. The Eddington ratio is within the expected values for nearby AGN with similar bolometric luminosities and SMBH masses (Panessa et al. 2006).

Host galaxy masses
It is well known that the SMBH mass and the spheroidal mass of the host galaxies follow a linear relation (Merritt & Ferrarese 2001, Park et al. 2012. In this study we calculate both the dynamical mass and the stellar mass from the spheroidal component of the host galaxy.

Stellar masses
STARLIGHT provides stellar masses for its best fit models. We have corrected those for slit losses using the spheroidal components obtained by IMFIT (see Sec. 4): the fractions of the spheroid light that went through the slit are 9.0 and 6.2 percent for J00 and J02, respectively.
The stellar mass derived using STARLIGHT is computed using mass-to-light relations, converting each SSP contribution to stellar mass. STARLIGHT does not compute  errors in the best-fitting values. In Bell & de Jong (2001) it is discussed the stellar mass-to-light ratio and the uncertainties associated to this ratio. This uncertainties are all of order 0.1-0.2 dex. We use the largest value as the error for our values to be conservative. Our results are shown in Table 5.

Dynamical masses
As a sanity check, we also computed the dynamical mass of the spheroidal component of the host galaxies. We used the relation between the line of sight velocity dispersion (σe) and the dynamical mass.
The observed width of the NaI doublet at rest-frame λλ5896, 5890Å (Na ID) is a stellar absorption feature that can be used to calculate σe (Spiniello et al. 2012). The NaID feature is a convolution of the resolution of the instrument, the intrinsic width of the stellar population and σe. To compute σe, we used the instrumental resolution from Table 2. For the stellar population, we used the Single Stellar Population (SSP) template from the Bruzual and Charlot 2003 library (Bruzual & Charlot 2003) that based on the results of STARLIGHT, is the most dominant SSP for each host galaxy: the SSP of 0.9 Gyr and Z=0.05 for J00, and 11 Gyr and Z=0.02 for J02. The standard dispersion of the Gaussian function gives the desired σe.
In Fig. 5 we show the NaID and the results of the fits. We plot in this figure the template used in each object with the spectral resolution at the region of the doublet. The spectral resolution is computed using the FWHM of the closest Figure 5. Spectrum (in black) of the Na I Doublet of J00 (left) and J02 (right) and its fit (in red). In grey we show the error of each spectrum. In blue we plot the template used to measure σe in the region of NaID. The template has a spectral resolution of 1.28Å.  arc line of the XSHOOTER observation. To calculate the dynamical mass of our AGN hosts we used the Virial relations from Cappellari et al. (2006) and Taylor et al. (2010).
where re is the effective radius of the spheroidal mass of the host galaxy, σe is the line-of-sight velocity dispersion, the factor k(n) that depends on the Sersic index and G is the universal gravitational constant. Using the software IMFIT we obtained re and the Sersic index by fitting the acquisition images from the VLT. To compute the error in the spheroidal mass we propagate errors. We summarize in Table 5 the properties of our AGN hosts. We clearly see that our dynamical mass estimates are compatible with the stellar masses calculated before. The studied galaxies are massive but not atypical (Vitale et al. 2013).

DISCUSSION
In the following subsections we discuss one by one the possible causes of the apparent discordant properties of our AGN in the UV/optical range and in X-rays.

Compton-thick or Compton-thin obscuration
Based on the observed X-ray properties, we can rule out that our objects are Compton-thick. First of all the broad band continuum shape is too steep to be produced mainly by reflected emission as demonstrated in Sec. 3. Comptonthick sources usually have Γ ∼1.0 (Brightman et al. 1999, Winter et al. 2008, Georgantopoulos et al. 2011b, Del Moro et al. 2016. Compton-thick AGN are expected and known to display large EW Fe K emission lines at 6.4 keV (≤ 1 keV; Gandhi et al. 2014), due to the highly suppressed underlying continuum. No Fe line is detected by the fits with high significance in any of our sources.
The Compton-thin nature of the sources is supported by the high L 2−10 keV /L [O III] ratio, where L [O III] is the unreddened luminosity of the [O III] emission line at restframe λ5007Å. This is because if we only detect in Xrays the soft scattered component (which is only a few percent of the intrinsic AGN power), the X-ray luminosity can be largely underestimated. Therefore we can use L [O III] as a proxy of the bolometric luminosity, and compare it to the L 2−10 keV . Compton-thick sources have L 2−10 keV /L [O III] <0.1-1.0 (Bassani et al. 1999, Akylas & Georgantopoulos 2009). This ratio is 69 for J02 and 22 for J00. This effectively excludes the Compton-thick character of both of our sources.
Alternatively, we could use the mid-IR to LX ratio to identify Compton-thick obscuration (Gandhi et al. 2009, Asmus et al. 2011, Mateos et al. 2015, Stern 2015. If we use the three shortest λ-bands of the Wide-Field Infrared Survey Explorer (WISE, 3.4, 4.6, 6, 12 µm;Wright et al. 2010), we find that our sources have mid-IR colours that fall outside of the region occupied by AGN , because their catalogued fluxes have significant contamination from the AGN hosts.

Host-SMBH relations
We estimated the SMBH-to-host galaxy mass ratio for our two sources (see Table 6) and we compared them with the value reported by Merritt & Ferrarese (2001), log(MSMBH/M bulge ) =-2.9 with a rms of 0.5 dex. This has been derived by using the samples of bulges and elliptical galaxies of Ferrarese et al. (2001) and Gebhardt et al. (2000). For J00 we obtain log(MSMBH/M bulge ) =-2.77, which is consistent with the Meritt & Ferrarese relation. This means that the galaxy is as massive as expected by its SMBH. J02 has log(MSMBH/M bulge ) =-4.15, that is more than 2 times the rms below the standard relation. For J02 host galaxy dilution could explain, at least in part, the lack of broad emission line detection in the 6dF spectrum. This is because we expect more impact of the star-light dilution on the AGN emission compared to AGN with less massive host galaxies. There are many examples in the literature of apparently normal galaxies (eg. XBONGs) in which, after the host galaxy contamination is removed, AGN emission is revealed (Severgnini et al. 2003, Georgantopoulos & Georgakakis 2005). In the next sections we will investigate if this is the only factor that contributes to the observed optical/X-ray discrepancy.
6.3 Dust-to-gas ratio of the obscuring medium and Balmer decrement Typically AGN have dust-to-gas ratios lower than or compatible with the Galactic value (Maiolino et al. 2001, Vasudevan et al. 2009, Parisi 2011, Marchese et al. 2012, Hao et al. 2013, Burtscher et al. 2016). This result does not appear only in X-ray selected studies, but also on optical and IR selected AGN (Wilkes et al. 2002, Young, Elvis & Risaliti 2008. We have compared AV (see Table 7) and NH (see Table 1) for our sources. For J00 AV/NH ≥2.61×10 −21 mag cm −2 , while for J02 AV/NH=1.30 +1.8 −1.1 ×10 −21 mag cm −2 . The Galactic relation is AV/NH=5.3×10 −22 mag cm −2 . J00 shows an AV/NH more than 5 times the Galactic value. For J02 the value is consistent with the Galactic. As mentioned in Sec. 5.1, the SMC extinction model is one of the most conservative measurements on the AV of all models taken into account. The results provided by the other extinction models do not change the results in terms of the dust-to-gas ratio.
A dust-to-gas ratio higher than the Galactic value explains the observed properties of J00. It is a scenario offered in Panessa & Bassani (2002), to explain the different optical and X-ray classification of unabsorbed Seyfert 2. In this case a strong contribution of dust has less effect in the X-ray emission than the optical one. There are a few examples in the literature of higher dust-to-gas than the Galactic value (Trippe et al (Bassani et al. 1999, Pappa et al. 2001, Carrera, Page & Mittaz 2004) through the expression E(B-V)=2.07×log((Hα/H β )/3) and R V =3.1. The A V,X−ray is the optical extinction corresponding to the N H column density using the SMC model of Gordon et al. (2003).

Intrinsically weak BLR emission
To investigate whether our AGN have a BLR with nonstandard properties (e.g. underluminous) we have determined the luminosity ratio LNLR/LBLR from the broad and narrow components of the Hα emission line (see Table 4). Then we compare our values with the relation between LNLR/LBLR and LBLR found for AGN of similar z and luminosities to ours from Stern & Laor (2012). Our objects have LNLR/LBLR ratios within the observed 0.4 dex scatter hence, none of our sources appears to have intrinsically weak BLR.

Variability
AGN are highly variable sources across the full electromagnetic spectrum at both long and short time scales (Ulrich, Maraschi & Urry 1997, Mateos et al. 2007, Krumpe et al. 2010, Hernández-García et al. 2015, LaMassa et al. 2015. This is originated by variability in the accretion rate and by extinction variability in the line-of-sight material (Markowitz et al. 2014, Miniutti et al. 2014. As the X-ray and UV/optical observations have not been taken simultaneously we cannot rule out that this might also contribute to the observed mismatch between the optical and X-ray properties of our AGN. As a check, we carried out the UV-to-optical spectral decomposition into AGN and host galaxy emission with the public SDSS-DR7 spectrum of J00 (taken in 2000-09-05), using the same broken power law for the AGN emission and the same host galaxy emission from Sec. 5.1. In this test we compare the intrinsic flux of the AGN, so we have taken into account that the SDSS spectra were taken with a 3 fiber and hence a higher fraction of AGN enters through the fiber and the host galaxy contribution is higher than in the XSHOOTER spectra. Analyzing the results there is some variation, that is best fitted by an extinction change (AV,SDSS=0.69±0.10 mag versus the one obtained by XSHOOTER, that is AV,XSH=2.04±0.29 mag) instead of a change in the emission of the AGN. The computed AV gives a higher dust-to-gas ratio than the Galactic (AV/NH >8.8×10 −22 mag cm −2 ), as with the XSHOOTER data. The 6dF optical spectrum of J02 is not of sufficient quality for conducting a spectral decomposition analysis.
We conclude that extinction variability could be present in at least one of the sources (the one with the largest AV/NH ratio). Even so, a higher than Galactic dust-to-gas ratio is also needed in that source. For this source we only observe a marginal variation of the intrinsic flux of 1.6σ. Simultaneous X-ray and optical observations are needed to assess the actual importance of variability.

CONCLUSIONS
In this work we have investigated the origin of the apparent mismatch of the optical and X-ray classifications of two AGN with high optical extinction but low X-ray absorption. In our two selected objects there is a clear broad line in Hα using the data from VLT/XSHOOTER after a careful removal of the host galaxy contribution.
We discussed several scenarios that could explain the discordance of our observations. We ruled out a Compton-thick nature of our sources, on the basis of the L 2−10 keV /L [OIII] ratio. We also discarded that the total or partial absence of broad lines in the spectrum is caused by an intrinsically weak BLR emission.
The origin of the mismatch for each object is found to be different. An intrinsically different AV/NH is the best explanation for the properties of the object J00 without the need to invoke variability. This obscuring material has a dustto-gas ratio really different than the majority of the AGN population. The other object, J02, has a massive host galaxy in comparison with its SMBH, so that the broad emission lines and the nuclear continuum are swamped by the host galaxy star-light which makes them very difficult to detect.