EDP Sciences logo
Subscriber Authentication Point
Search
Menu
Home All issues Volume 646 (February 2021) A&A, 646 (2021) A29 Full HTML
Free Access
Issue
A&A
Volume 646, February 2021
Article Number A29
Number of page(s) 17
Section Extragalactic astronomy
DOI https://doi.org/10.1051/0004-6361/202039401
Published online 02 February 2021

© ESO 2021

1. Introduction

One of the major challenges in current astrophysical research is understanding the physical processes that operate on top of the dark matter distribution to produce luminous structures, such as stars and galaxies. Baryonic physics is highly complex, mainly because it includes electromagnetic interactions; for instance, the interplay between the cooling and heating of baryons (e.g., gas of dust particles) via radiative processes, or the impact of magnetic fields on charged particles. However, such processes operate on spatial scales many orders of magnitude smaller than what can be achieved by state-of-the-art cosmological simulations. Thus, assumptions have to be made on the amplitude and impact of the multitude of possible mechanisms that may affect the formation and evolution of galaxies; for example, gas supply, gas cooling and heating, and the impact of stellar winds on the interstellar medium. These physical processes that govern the births and fates of galaxies are of extreme complexity. However, they are of great interest since galaxies play a pivotal role in the structure of the Universe and are unique tracers of its evolution.

One manifestation of baryonic physics is the formation of supermassive black holes (SMBHs) at the centre of a galaxy. Tight correlations have been found between the mass of the SMBH and the properties of its bulge (e.g., Magorrian et al. 1998; Ferrarese & Merritt 2000). When material is accreted onto these SMBHs, it triggers them, and the galaxy is called an active galactic nucleus (AGN). The energy released during the accretion process is also an important source of heating for both the interstellar (e.g., Morganti 2017) and intergalactic medium (e.g., Kaastra et al. 2014). As a result, it has been hypothesised that SMBHs and their activity play an important role in galaxy evolution (e.g., Brandt & Alexander 2015).

Active galactic nucleus emission can be observed at different wavelengths from X-rays to radio, as different physical mechanisms produce radiation at different wavelengths. In particular, X-ray emission is a trademark of AGN activity. This emission originates from photons produced by the accretion disc that are scattered by the hot corona and emit X-rays through inverse Compton scattering. This process dominates the X-ray emission of the host galaxy and reflects the activity of the central SMBH. Thus, X-rays are often used as a proxy of AGN power (e.g., Lusso et al. 2012; Yang et al. 2019). Important ongoing (e.g., eROSITA) and future (ATHENA) X-ray missions will use the unique window that X-rays offer and provide us a wealth of data to study the tight connection of AGNs with their host galaxies.

The multi-wavelength emission of galaxies can be studied by constructing and modelling their full spectral energy distribution (SED). This method allows us to measure fundamental properties of galaxies, such as their stellar mass M*, star formation rate (SFR), dust mass, and attenuation, while at the same time it breaks degeneracies that plague observations in narrow(er) wavelength ranges. To this end, a number of algorithms have been developed to perform this task via different approaches. A popular approach is based on the energy balance principle, that is, the energy emitted in the infrared (IR; i.e. 5−1000 μm) is equal to the energy absorbed in the UV/optical wavelengths; for example, CIGALE (Code Investigating GALaxy Emission; Burgarella et al. 2005; Noll et al. 2009; Boquien et al. 2019), ProSpect (Robotham et al. 2020), MAGPHYS (da Cunha et al. 2008), Prospector (Leja et al. 2017), and BAGPIPES (Carnall et al. 2018).

As mentioned, AGNs play an important role in galaxy evolution, and their presence affects many parts of the electromagnetic spectrum. Thus, a number of the aforementioned SED fitting algorithms have added an AGN component to the fitting process (e.g., CIGALE, ProSpect) to separate AGNs and galaxy emission. SED algorithms that don’t include an AGN SED component and only account for low-luminosity/obscured AGNs are biased against luminous/unobscured sources and underestimate the contribution of the AGN IR emission to the total IR galaxy emission.

In a recent paper, Yang et al. (2020) presented a new branch of the CIGALE code, X-CIGALE. Compared to CIGALE, X-CIGALE is supplemented with the modelling of AGN X-ray emission and the inclusion of polar dust. Polar dust accounts for extinction of UV and optical radiation, that is commonly found, in particular, in X-ray selected AGNs (Bongiorno et al. 2012). SED fitting algorithms that include X-ray information could be instrumental in the exploitation and interpretation of the large datasets that X-ray surveys will provide.

The goal of this paper is to use the new capabilities of the X-CIGALE code on one of the largest X-ray samples available (XMM-XXL; Pierre et al. 2016). XXL offers a significantly wider luminosity baseline that extends to higher luminosities compared to the fields studied in Yang et al. (2020). Additionally, the size of the database allows us to draw statistically robust conclusions in our tests. The physical properties of the XXL AGNs and their host galaxies, using the CIGALE code, have been the topic of previous studies (e.g., Masoura et al. 2018, 2020, 2021). In this work, we focused on the effect of the new features of X-CIGALE on important SED fitting parameters. Our main purpose is to examine how reliably the algorithm connects the X-ray flux with the UV luminosity and the rest of the other wavelengths, how accurately X-CIGALE can reproduce the X-ray properties of the AGNs, and what improvements the new additions bring in its efficiency on SED decomposition.

2. Data

In this section, we describe the X-ray AGN sample used in our analysis and the methodology we followed to obtain optical and IR identifications.

2.1. The X-ray AGN sample

Throughout our work, we used spectroscopic X-ray AGNs from the XMM-XXL field (Pierre et al. 2016). XXL is an international project based around an XMM Very Large Programme surveying two 25 deg2 extragalactic fields. It has a depth of ∼6 × 10−15 erg cm −2 s−1 in the [0.5−2] keV band for point-like sources, with an exposure time of about 10 ks per XMM pointing. 8445 X-ray sources have been detected in the equatorial subregions (XMM-XXL-N; Liu et al. 2016). 5294 of them have SDSS counterparts. Spectroscopic redshifts are available for 2512 AGNs (Menzel et al. 2016) from SDSS-III/BOSS (Eisenstein et al. 2011; Smee et al. 2013; Dawson et al. 2013). These spectroscopic sources are used in our analysis. Their redshift distribution is presented in Fig. 1.

thumbnail Fig. 1.

Redshift distribution of the X-ray AGN sample.

2.2. IR Photometry

In addition to the optical (SDSS) photometry available for all our sources, we also searched for counterparts in the near-IR, mid-IR, and far-IR part of the spectrum. Mid-IR (allWISE; Wright et al. 2010) and near-IR photometry from the Visible and Infrared Survey Telescope for Astronomy (VISTA; Emerson et al. 2006) were obtained using the likelihood ratio method (e.g., Sutherland & Saunders 1992) as implemented in Georgakakis et al. (2011). The process is described in detail in Georgakakis et al. (2017).

We also used catalogues produced by the HELP1 Collaboration to complement our mid-IR photometry with Spitzer (Werner et al. 2004) observations, and we added far-IR counterparts. HELP provides homogeneous and calibrated multi-wavelength data over the Herschel Multi-tiered Extragalactic Survey (HerMES, Oliver et al. 2012) and the H-ATLAS survey (Eales et al. 2010). The strategy adopted by HELP is to build a master list catalogue of objects as complete as possible for each field (Shirley et al. 2019) and to use the near-IR sources from IRAC surveys as prior information for the IR maps. The XID+ tool (Hurley et al. 2017), developed for this purpose, uses a Bayesian probabilistic framework and works with prior positions. At the end, a flux is measured, in a probabilistic sense, for all the near-IR sources of the master list. The XMM-LSS field was covered by two Spitzer surveys, SpUDS (Spitzer UKIDSS Ultra Deep Survey, Caputi et al. 2011) and SWIRE/SERVS (Lonsdale et al. 2003). The prior positions are defined with the SpUDS and SWIRE/SERVS surveys, and the fluxes are measured for the Spitzer MIPS/24 microns, Herschel PACS and SPIRE bands. In this work, only the MIPS and SPIRE fluxes were considered, given the much lower sensitivity of the PACS observations for this field (Oliver et al. 2012).

The W1 and W2 photometric bands of WISE nearly overlap with IRAC1 and IRAC2 from Spitzer. When a source had been detected by both IR surveys, we only considered the photometry with the highest signal-to-noise ratio (S/N). Similarly, when both W4 and MIPS photometry is available, we only considered the latter due to the higher sensitivity of Spitzer compared to WISE. The number of sources available in each photometric band of our final sample is presented in Table 1. Table 2 shows the sensitivity of each photometric band used in our analysis.

Table 1.

Number of available spectroscopic sources in different photometric bands.

Table 2.

5σ sensitivity (mag) of the photometric bands used in our analysis.

3. Analysis

X-CIGALE requires the intrinsic (i.e. unabsorbed) flux of X-ray AGNs. In this section, we describe how we estimated the X-ray absorption of the sources to infer their intrinsic X-ray flux. We also describe the modules and parameters used for the SED fitting process and present our final sample.

3.1. Estimation of the X-ray properties

To estimate the intrinsic X-ray flux of each source, we need to measure their hydrogen column density, NH, which quantifies the X-ray absorption. To this end, we used the number of photons in the soft (0.5−2.0 keV) and the hard (2.0−8.0 keV) bands that are provided in the Liu et al. (2016) catalogue. We then applied a Bayesian approach to estimate the hardness ratio, HR = , of each source, where H and S are the counts in the soft and hard bands, respectively. Specifically, we utilised the Bayesian Estimation of Hardness Ratios (BEHR; Park et al. 2006) code, which accounts for the Poissonian nature of the observations. These hardness ratio values are then inserted in the Portable, Interactive, Multi-Mission Simulator (PIMMS; Mukai 1993) tool to estimate the NH of each source. In this process, we assume that the power law of the X-ray spectra has a fixed photon index, Γ = 1.8 and the value of the galactic NH is NH = 1020.25 cm−2. The distribution of NH of our AGN sample is presented in Fig. 2. Due to the Bayesian nature of our calculations, some log NH values are below 20.25 (i.e. the galactic absorption).

thumbnail Fig. 2.

Distribution of X-ray absorption, NH, of the X-ray AGN sample.

The X-ray absorption that was estimated through PIMMS, and based on which the intrinsic X-ray flux is inferred, does not necessarily correlate with the dust that the observed mid- and far-IR emission exhibits. Although there are studies that have found a correlation between optical/IR obscuration and X-ray absorption (e.g., Civano et al. 2012), this correlation presents a large scatter (e.g., Jaffarian & Gaskell 2020). This scatter is attributed to, for example, (i) the X-ray column density variability (e.g., Reichert et al. 1985; Yang et al. 2016; ii) absorbing material located at galactic scales (e.g., Malizia et al. 2020) that is mainly heated by star formation rather than AGN; (iii) the fact that different obscuration criteria are sensitive to different amounts of obscuration (e.g., Masoura et al. 2020); and (iv) a dust-to-gas ratio that is different from the Galactic dust-to-gas ratio. Therefore, a source may present X-ray absorption without being optically red (dust-free gas), be heavily X-ray absorbed with broad UV/optical lines (e.g., Li et al. 2019), can be optically red with absorbed AGN emission in its SED without being X-ray absorbed (e.g., Masoura et al. 2020), and can present higher X-ray absorption than what is expected from its optical extinction. Thus, we do not necessarily expect consistency between the X-ray absorption estimated through PIMMS with that from dust, which is estimated by X-CIGALE by modelling the mid- and far-IR emissions.

3.2. SED fitting with X-CIGALE

The fitting capabilities of CIGALE were recently extended to X-rays with the development of X-CIGALE (Yang et al. 2020) in order to improve the characterisation of the AGN component. The X-ray emission is connected to the AGN emission at other wavelengths via the αox−L2500 Å relation of Just et al. (2007), where L2500 Å is the intrinsic (de-reddened) UV luminosity, and αox is the spectral slope between UV (2500 Å) and X-ray (2 keV): αox = −0.3838 log (L2500 Å/L2 ke V). The contribution of X-ray binaries is also considered and modelled as a function of SFR and stellar mass of the host galaxy. The clumpy two-phase torus model, SKIRTOR, based on 3D radiation-transfer (Stalevski et al. 2012, 2016) is used for the UV-to-far-IR emission of the AGN with some modifications keeping the energy balance: the original emission of the accretion disc is updated with the spectral energy distribution of Feltre et al. (2012), and dust extinction and emission in the poles of type 1 AGNs are also considered (e.g., Bongiorno et al. 2012; Tristram et al. 2014; Asmus et al. 2014; Asmus 2019). We refer the reader to Yang et al. (2020) for a full description of X-CIGALE. Here, we describe the main steps followed to build the models and fit our X-ray to far-IR data. The modules and input parameters we used in our analysis are presented in Table 3.

Table 3.

Models and the values for their free parameters used by X-CIGALE for the SED fitting of our galaxy sample.

3.2.1. Galaxy emission

For the sake of simplicity and since we did not study the SFR properties of our sources in detail, we adopted a simple star formation history (SFH). The galaxy component is built using a delayed SFH (SFR ∝ t × exp(−t/τ)). We checked that the addition of a recent burst (e.g., Masoura et al. 2018; Małek et al. 2018) did not change our results. The stellar emission is modelled using the Bruzual & Charlot (2003) template. A Chabrier (2003) initial mass function (IMF) was used, with a metallicity equal to 0.02. The stellar emission was attenuated with the Calzetti et al. (2000) attenuation law. The IR SED of the dust heated by stars was implemented with the Dale et al. (2014) template.

3.2.2. AGN emission

The AGN emission was modelled using the SKIRTOR template (Stalevski et al. 2012, 2016). A polar dust component modelled as a dust screen absorption and a grey-body emission was added. As in Yang et al. (2020), we adopted the Small Magellanic Cloud (SMC; Prevot et al. 1984) extinction curve with E(B − V) as a free input parameter (see Sect. 5 for the effects of E(B − V) on SED fitting). The grey-body dust re-emission is parametrised with a temperature of 100 K and an emissivity index of 1.6. This emission is supposed to be isotropic and thus contributes to the IR emission of both type 1 and type 2 AGNs. We discuss the effect of a modification of the dust temperature and of the extinction curve in Sect. 4.3. The contribution of the AGN to the total SED is quantified by the AGN fraction, fracAGN, defined as the fraction of the total IR emission coming from the AGN. Following Yang et al. (2020), two viewing angles are considered, 30° and 70°, for type 1 and type 2 AGNs, respectively.

The photon index Γ of the AGN X-ray spectrum was fixed to 1.8, for consistency with the value used for the estimation of NH (see Sect. 3.1). We adopted a maximal acceptable value |Δαox|max = 0.2 for the dispersion of the αox−L2500 Å relation (Risaliti & Lusso 2017). This value is also adopted in Yang et al. (2020) and corresponds to ≈2σ scatter in the αox−L2500 Å relation (Just et al. 2007). Nine values of αox were defined (from −1.9 to −1.1 with a step of 0.1), and X-CIGALE added an X-ray flux to each pair (UV to far-IR AGN SED, αox), multiplying the number of AGN models by 9.

3.2.3. SED fitting and parameter estimation

The spectral models described above are normalised to the creation of 1 M⊙. The first step of the fitting process was to scale each model to the observations by minimising its χ2 (Noll et al. 2009; Boquien et al. 2019). After this scaling operation, extensive quantities such as luminosities were defined, including the intrinsic L2500 Å, which is an output of the SKIRTOR module: models that do not satisfy the αox−L2500 Å relation and its maximum dispersion were discarded. The likelihoods of the remaining models were computed and parameter values were estimated from their marginalised probability distribution function (likelihood weighted mean and standard deviation). The best model corresponding to each observed SED is also an output of the code.

3.2.4. Mock catalogues

The validity of a parameter estimation can be assessed through the analysis of a mock catalogue. When this option is chosen, the code considers the best fit of each object and a mock catalogue is built. Each best flux is modified by injecting noise taken from a Gaussian distribution with the same standard deviation as the observed flux. The mock data are then analysed in the same way as the observed data, and the accuracy of the parameter estimation can be tested by comparing input (ground truth) and output (estimated) values. Our tests using the results from the mock catalogues are presented in the appendix.

3.3. Final sample

In our analysis, we compared the estimations of the fracAGN, namely the value of the best model and then mean of the probability density function (PDF) distribution, to select the X-ray AGNs with the most reliable SED measurements. For that purpose, we applied the following criteria: we excluded sources for which the best AGN fraction values are zero, while their Bayesian AGN fraction is greater than 0.4. Such a large difference between the best and the Bayesian values, is a strong indicator that the code failed to accurately fit the SED since the PDF distribution is not consistent with the best model. We also excluded AGNs with reduced χ2, from the SED fitting process. These two criteria exclude ∼4−7% of our sources. Finally, we removed sources with LX <  1042 erg s−1 from our final sample to minimise contamination from inactive galaxies. Table 4 presents the number of sources that satisfy our selection criteria for each configuration of the SED fitting process used in our analysis.

Table 4.

Number of AGNs that satisfy our selection criteria (Sect. 3.3) for different configurations of the SED fitting process.

4. X-CIGALE performance

4.1. X-ray luminosity

First, we compared the intrinsic X-ray luminosities of the AGNs estimated by X-CIGALE with those from the input catalogue. The results are presented in Fig. 3. X-CIGALE estimates are consistent with those from the input catalogue at all luminosities spanned by our sample. A χ2 fit (grey dashed line) gives log LX, X-CIGALE = (0.962 ± 0.023) log LX, data + 1.694 ± 0.085. Green dashed lines present the errors added in quadrature of the X-ray luminosity calculations from X-CIGALE and those quoted in the X-ray catalogue (Menzel et al. 2016). There are 70 outliers, that is, sources for which LX calculations differ by more than 0.5 dex from their input values (shown with error bars). The vast majority of sources for which X-CIGALE underestimates the LX value by more than 0.5 dex are AGNs with high X-ray absorption (NH >  1022.5−23.0 cm−2). 82% of these sources also lack far-IR photometry, and 57% do not have IR coverage above 8 μm. Thus, a possible explanation for the discrepant LX values could be that our NH estimations are not accurate and/or the lack of available IR photometry does not allow X-CIGALE to properly fit these SEDs. Lowering the LX difference threshold to 0.3 dex to characterise a source as an outlier leads to the same conclusions. On the other hand, those sources for which X-CIGALE overestimates their value by more than 0.5 dex (< 0.4% of the total sample) do not present any signs of X-ray absorption (NH <  1021.5 cm−2). The quality of the SED fitting of these systems is good, based on the values (). SED analysis reveals that the AGN emission is obscured in the optical wavelengths. The optical/mid-IR criteria of Yan et al. (2013) do not classify these AGNs as optically red sources. Furthermore, their optical spectra present broad lines. Thus, there are no indications to corroborate with X-CIGALE that these AGNs are absorbed. Therefore, we do not find a plausible explanation for these outliers. However, they only represent < 0.4% of our sample (∼1% if we lower the threshold of the LX difference to 0.3 dex).

In the right panel of Fig. 3, we plot the X-ray luminosity calculations of X-CIGALE when the X-ray flux is not included in the SED vs. the LX from the input catalogue. In this case, the scatter is significantly larger compared to the left panel, while the average error of the LX estimates from X-CIGALE, increases by ≈4× (green lines). When there is no fX in the SED, the LX estimates should not be taken at face value. X-CIGALE uses the AGN module SKIRTOR to output plausible L2500 Å values. Since there is no X-ray information to further constrain the L2500 Å parameter by connecting it to the observed X-ray flux, the algorithm provides LX estimations, using the full range of αox (allowed by the Just et al. 2007 relation and the αox dispersion) and weighs over these possible values. Thus, although the code provides LX estimates, these values should be taken with caution.

thumbnail Fig. 3.

Comparison of intrinsic X-ray luminosity in the 2−10 keV band estimated by the SED fitting with that from the input catalogue. Green, solid, and dashed lines present the 1:1 correspondence and the errors added in quadrature of the X-ray luminosity calculations from X-CIGALE and those quoted in the X-ray catalogue (Menzel et al. 2016), respectively. The grey dashed line shows the χ2 fit of the calculations. Symbols are colour-coded based on their NH values. Left panel: results with the X-ray flux included in the SED. X-CIGALE calculations are consistent with those from the input catalogue, at all luminosities spanned by our sample. There are 70 sources for which the LX calculation differs by more than 0.5 dex from their input values (shown with error bars, see text for more details). Right panel: X-ray flux is not included in the SED. In this case, the scatter is significantly larger, while the error of the LX calculations from X-CIGALE, increases by ≈4×.

4.2. The efficiency of X-CIGALE to connect the X-ray-UV luminosity

As mentioned, αox, L2500 Å and L2 keV are the three parameters that are important for X-CIGALE to connect the X-rays with the UV, and thus the other wavelengths during the SED fitting process. In Appendix A.1, we assess the efficiency of X-CIGALE to constrain these three parameters using the mock analysis (see Sect. 3.2.4). In this section, we compare the algorithm’s calculations of L2 keV and L2500 Å.

Figure 4 compares the L2 keV and L2500 Å luminosities with the observed relations of Just et al. (2007) (blue line), used by the code as input (see Sect. 3.2) and Lusso & Risaliti (2016) (green line). The grey line presents the fit on our measurements. There is an overdensity of sources above the Just et al. (2007) relation (at high luminosities) that is due to selection bias. XMM-XXL has a low exposure time and therefore our X-ray sample is biased towards high-luminosity sources.

thumbnail Fig. 4.

L2 keV vs. L2500 Å relation. Blue line presents the Just et al. (2007) relation that X-CIGALE uses to connect the X-ray flux with the UV (2500 Å) luminosity. Green line shows the L2 keV vs. L2500 Å from Lusso & Risaliti (2016). For comparison, we also plot the fit from our calculations (grey line).

4.3. The effect of the extinction law and temperature of polar dust

In Sect. 3.2.2, we mentioned that for the polar dust estimation an SMC extinction curve is adopted and the grey-body dust temperature is set to 100 K. The effect of the addition of polar dust in the SED fitting is discussed in Sect. 5.4. In this section, we examine whether the adoption of different extinction curves and dust temperatures affect the polar dust contribution (through EB − V) and the AGN fraction calculations.

Apart from the SMC extinction curve, X-CIGALE includes the choice of the empirical extinction curves of Calzetti et al. (2000) and Gaskell et al. (2004). Figures 5a and b present the difference of the polar dust estimations between the SMC and the Calzetti et al. (2000), and the SMC and Gaskell et al. (2004) curves, respectively. Both distributions are highly peaked, and therefore the choice of the extinction curve does not affect the polar dust measurements. In Figs. 6a and b, we repeat the same test for the AGN fraction values. Again, the adoption of different extinction curves does not affect the fracAGN.

thumbnail Fig. 5.

Distribution of polar dust measurement differences for different extinction laws (panels a and b) and grey-body dust temperatures (panels c and d). All distributions are highly peaked at zero, which indicates that polar dust calculations are not sensitive to the choice of these parameters.

thumbnail Fig. 6.

Same as in the previous plot, but for the AGN fraction. Different extinction laws (panels a and b) and grey-body dust temperatures (panels c and d) do not affect the fracAGN measurements.

In Figs. 5c, d, and 6c, d, we test whether different temperatures for the grey-body dust re-emission affect the polar dust and AGN fraction calculations, respectively. Specifically, we plot the difference of the aforementioned parameters using values of 100 K (the value used throughout our analysis) as well as 75 K and 50 K. All distributions are highly peaked at zero. We obtain similar results when we increase the temperature to 200 K. We conclude that the choice of the parameters defining the grey-body re-emission does not affect our results.

4.4. The effect of Herschel photometry

Far-IR data combined with mid-IR photometric bands are known to improve the SFR estimations of galaxies hosting an AGN (Hatziminaoglou et al. 2009; Stanley et al. 2018; Masoura et al. 2018) since they constrain the AGN contribution to the IR luminosity of the host galaxy more efficiently. Masoura et al. (2018) used 608 X-ray AGNs in XXL with Herschel detection and found that SFRs estimated without far-IR photometry are systematically underestimated compared to Herschel SFRs (see their Fig. 4). In this section, we examine the effect of Herschel photometry on our estimations. Specifically, we wish to test whether the absence of Herschel photometry affects the AGN fraction and polar dust estimates of X-CIGALE.

For this part of our analysis, we restricted our sample to only those AGNs with reliable (S/N >  3) SPIRE photometry (in addition to the criteria mentioned in Sect. 3.3). This maximises any likely effect on the SED fits. This reduces our X-ray dataset to 328 AGN. We ran X-CIGALE twice: one time using Herschel photometry and a second time without (both runs use the X-ray flux of the sources in the fitting process).

Figures 7a and b present the difference of the polar dust and AGN fraction distributions, from the two runs, respectively. We notice that the addition of Herschel photometry slightly reduces the AGN fraction measurements. Although, the distribution of the difference of the AGN fraction estimations peaks at zero, there is a tail at negative values. Specifically, there are 73 sources (≈22% of this sample) for which the addition of Herschel photometry reduces their AGN fraction value by more than 0.15 (which is equivalent to the error of the two fracAGN estimations added in quadrature, see next sentence). This is also quantified by the mean fracAGN values, fracAGN, Herschel = 0.36 ± 0.08 and fracAGN, no Herschel = 0.43 ± 0.12. We also notice that including far-IR photometry increases the statistical significance of the estimations from 3.6σ to 4.5σ (the significance is defined as the Bayesian value over the error). The addition of Herschel photometry does not affect polar dust measurements. We also examined whether the addition of Herschel photometry affects other parameters estimated by the SED fitting and specifically the X-ray luminosity and the L2500 Å. The distribution of the difference between the LX and L2500 Å parameters with and without far-IR photometry is presented in Figs. 7c and d. Both distributions are highly peaked at zero with small tails at both sides. Thus, the inclusion of Herschel photometry does not seem to affect the estimations of these two parameters.

thumbnail Fig. 7.

Distribution of the difference of polar dust, AGN fraction and X-ray and UV luminosity measurements, with and without Herschel photometry. Polar dust and luminosity calculations are not affected by the existence of far-IR photometry. However, addition of far-IR photometry reduces the fracAGN. This is quantified by the mean values, fracAGN, Herschel = 0.36 ± 0.08 and fracAGN, no Herschel = 0.43 ± 0.12.

5. Advantages of X-ray flux and polar dust components

In this section, we examine the advantages that the introduction of the two new features of X-CIGALE (i.e. the X-ray flux and polar dust) bring to the SED fitting process.

5.1. The effect of the X-ray flux on the AGN fraction measurements

One of the strengths of SED decomposition is that it allows disentangling AGN emission from that of the host galaxy emission. This is a crucial issue since the reliability of any further analysis depends on how reliably the SED fitting code can perform this task. The goal of this part of our analysis is to examine if the introduction of the X-ray information in the SED allows X-CIGALE to improve its efficiency in estimating robust AGN fractions.

In Appendix A.2, we assess the accuracy of X-CIGALE in the estimation of the AGN fraction. Here, we study how the addition of the X-ray flux in the fitting process affects the AGN fraction measurements. In Fig. 8 (left panel), we plot the distribution of the AGN fraction with (black shaded area) and without (blue histogram) the X-ray flux (polar dust is included in both runs). The mean fracAGN value increases from 0.46 ± 0.16 to 0.49 ± 0.14 with the addition of fX. To further examine the effect of the X-ray flux in the AGN fraction estimates, in the right panel we plot the difference of the AGN fraction with and without X-ray flux for different luminosity bins. The distributions peak at zero in all cases. However, at high X-ray luminosities (LX >  1043 erg s−1) a tail starts to appear at positive values, that is, the AGN fraction tends to be higher when the X-ray flux is included in the fitting process. This tail becomes more prominent at the highest luminosity bin. Specifically, for AGN with LX >  1045 erg s−1 the mean AGN fraction increases from 0.58 ± 0.16 without X-ray flux to 0.65 ± 0.13 with X-ray flux. The opposite trend is present for low luminosity AGNs (1042 <  LX <  1043 erg s−1). In this case, the mean fracAGN reduces from 0.28 ± 0.13 to 0.23 ± 0.09 with the X-ray flux.

thumbnail Fig. 8.

Left: distribution of the AGN fraction, with X-ray flux (black shaded area) and without X-ray flux (blue histogram). Right: distribution of the difference of the AGN fraction, with and without X-ray flux, for different luminosity bins. All distributions are normalised to unity. A tail appears at positive values, i.e., the AGN fractions are higher when X-ray flux is included in the SED fitting, for AGNs with LX <  1043 erg s−1 that becomes more prominent for the most luminous sources (LX <  1045 erg s−1). On the other hand, there is a tail at negative values, i.e. the AGN fractions are lower with X-ray flux, for the fainter AGNs (LX <  1043 erg s−1, blue line).

The statistical significance of the AGN fraction measurements improves with the inclusion of the X-ray flux, as can be seen by the numbers presented above. This is also illustrated in Fig. 9, where we plot the difference of the significance, Δσ, of the AGN fraction estimations with and without X-ray information, for sources with LX >  1045 erg s−1 and difference in fracAGN >  0.2 (shaded histogram) and sources with LX <  1043 erg s−1 and difference in fracAGN <  −0.2. Therefore, we conclude that the addition of the X-ray flux in the SED fitting results in more robust AGN fraction measurements, while its effect on fracAGN is small and depends on the AGN power.

thumbnail Fig. 9.

Difference of the significance, Δσ, of the AGN fraction estimations with and without X-ray information, for sources with LX >  1045 erg s−1 and difference in fracAGN >  0.2 (shaded histogram) and sources with LX <  1043 erg s−1 and difference in fracAGN <  −0.2. The statistical significance of the AGN fraction measurements improves significantly with the inclusion of the X-ray flux, especially for the most luminous AGNs.

5.2. The effect of X-ray flux on the estimation of the L2500 Å and AGN type

Another important parameter of the SED fitting is the intrinsic L2500 Å luminosity. In addition to the reasons mentioned in Sect. 3.2, L2500 Å is also a proxy of the AGN power. Here, we study the effect of the X-ray flux on the estimation of L2500 Å. Figure 10 (left panel) compares the L2500 Å estimations with and without X-ray flux in the SED for type 1 and 2 sources. The classification is based on their inclination angle values, as calculated by X-CIGALE. 2264 sources satisfy our selection criteria, described in Sect. 3.3, in both runs (with and without X-ray flux). In Fig. 10, we only consider the 1941 sources (86%) for which the inclination angle from the best fit has the same value in the two runs. For both types, we observe a scatter, but no systematic effect between the two estimations.

thumbnail Fig. 10.

Comparison of L2500 Å values with and without X-ray flux in the fitting process. Left panel: comparison of sources that are classified as the same type, with and without X-ray flux in the SED fitting process. Blue symbols present the results for type 1 sources (i = 30°) and red symbols for type 2 sources (i = 70°). Right panel: comparison of L2500 Å values for sources with different types from the two runs (with and without X-ray flux). X-CIGALE decreases/increases the intrinsic accretion power (and thus the intrinsic L2500 Å) depending on whether the source is type 1/2.

Of 2264 AGNs, 1810 are classified as type 1 with the X-ray flux, but 47 (∼2.5%) become type 2 without the X-ray flux. On the other hand, from the 454 sources classified as type 2 with fX, 276 (∼60%) change to type 1 without X-ray flux. The 323 (47+276) sources with different classifications in the two runs are shown in the right panel of Fig. 10. X-CIGALE increases L2500 Å when the type changes from type 1 to type 2 with the inclusion of the X-ray flux and lowers L2500 Å when the type changes from type 2 to type 1 with fX. This result is expected, considering the way that X-CIGALE works. (mid-) IR emission is considered anisotropic, that is, it depends on the viewing angle. The same source will have lower IR flux when viewed edge on (type 2) compared to face on (see Fig. 4 in Stalevski et al. 2012). On the other hand, X-ray flux is considered isotropic. Therefore, a type 2 source will have a higher (and ) than a type 1 source. For a given observed IR emission, X-CIGALE will decrease/increase the intrinsic accretion power (and thus the intrinsic L2500 Å) depending on whether the source is type 1/2.

Our analysis revealed that the inclusion of the X-ray information does not significantly affect the classification of sources that are type 1 based on the run with fX, but it does change the characterisation of the majority of the AGNs classified as type 2 with X-ray flux. To investigate further, the different classification for the 276 AGNs classified as type 2 when the X-ray information is included in the SED, but classified as type 1 without fX, we compare their X-CIGALE classification with that from optical (SDSS) spectra (Menzel et al. 2016). Among those sources that lie at z <  1, ∼85% are classified as narrow-line AGNs (NLAGN2; see Sect. 3.3.2 in Menzel et al. 2016), meaning the optical spectrum agrees with the classification of X-CIGALE when using the X-ray flux. On the other hand, at z >  1.5, ∼85% of the AGNs present broad lines in their optical spectra (BLAGN1), meaning their classification does not agree with that of X-CIGALE when the X-ray flux is included. We conclude that, at z <  1, the addition of the X-ray information increases the reliability of the SED fitting regarding the classification type of a source. At high redshifts, however, X-CIGALE seems to misclassify some sources: among the 703 AGNs at z >  1.5 in our total sample, 103 (≈15%) are classified as type 2 with X-ray flux and as type 1 without.

We further investigated plausible reasons for the misclassification of sources at high redshift when fX is added to the fitting process. The vast majority of these sources lack photometry above 5 microns. In the absence of IR photometry and without considering the X-ray flux, the introduction of a type 1 component gives a flexibility to fit the UV/optical data, while a type 2 component does not contribute to this emission. Thus, without fX, X-CIGALE is more likely to classify high redshift sources as type 1, with the risk of over-fitting the UV/optical data. The inclusion of X-ray flux provides an additional constraint for a type 1 template (via the Just et al. 2007 relation), whereas the code can freely scale a type 2 template to match any given X-ray flux. Thus, with fX, the code will preferentially classify sources as type 2. In this configuration (SED with only UV/optical data), the difference between a type 1 and a type 2 component is not meaningful.

We conclude that the misclassification of AGNs at high redshifts is mostly due to lack of available IR photometry, which makes the type assignation almost unconstrained. Future surveys (JWST; Gardner et al. 2006) could provide mid-IR photometry for these distant AGNs.

5.3. Ability to quantify the AGN contribution to the total galaxy X-ray emission

Star-forming galaxies may emit X-rays that originate from X-ray binaries, supernovae remnants, and hot gas (e.g., Fabbiano 1989). Previous studies have shown that X-ray luminosity can be used as a proxy of SFR (e.g., Ranalli et al. 2003; Laird et al. 2005). However, in X-ray-luminous AGNs, like those used in this study, the X-ray emission predominately originates from the supermassive black hole, and therefore these systems are usually identified and excluded by studies that use the X-ray emission to trace star formation (e.g., Persic & Rephaeli 2002; Laird et al. 2005).

As already mentioned, X-CIGALE can also model the X-ray emission that originates from low- and high-mass X-ray binaries and hot gas. Thus, in a similar manner to the fracAGN that quantifies the contribution of the AGN IR emission to the total galaxy IR emission, we define the X-ray AGN fraction as the ratio of the X-ray AGN emission to the total X-ray emission of the galaxy (AGN + binaries + hot gas). In Fig. 11, we plot the X-ray AGN fraction as a function of the X-ray luminosity, estimated by X-CIGALE. As expected, for the vast majority of sources the AGN X-ray emission contributes more than 90% of the total X-ray emission of the galaxy. Therefore, X-CIGALE models the X-ray emission and its various components to offer the capability of estimating the SFR of galaxies not only from IR indicators but also from X-rays, in those systems that are not AGN dominated.

thumbnail Fig. 11.

X-ray AGN fraction vs. X-ray luminosity estimated by X-CIGALE. The X-ray AGN fraction is defined as the ratio of the X-ray AGN emission to the total X-ray emission of the galaxy (AGN + binaries + hot gas). As expected, for the vast majority of sources the AGN X-ray emission contributes more than 90% of the total X-ray emission of the galaxy.

5.4. Polar dust and AGN fraction

Another important feature of X-CIGALE is polar dust, quantified with EB − V, as a free parameter to account for dust extinction in all AGNs, regardless of their classification into type 1 and 2 (see Sect. 3.2.2). In Appendix A.3, we test whether X-CIGALE can successfully constrain EB − V. Here, we study the effect of adding polar dust as a free parameter to the SED fitting results on the AGN fraction measurements.

Figure 12 presents the distribution of the AGN fraction with (black shaded area) and without (blue histogram) polar dust. In both runs, X-ray flux is included. The addition of polar dust to the fitting process increases the AGN fraction. The mean fracAGN is 0.37 ± 0.12 without polar dust and 0.49 ± 0.14 with the addition of polar dust. A similar increase is observed when the two runs do not include fX. This trend is also similar regardless of the AGN type, classified based on the inclination angle. The increase of the AGN fraction is expected since the introduction of polar dust allows X-CIGALE to account for obscuring material in the poles of AGNs.

thumbnail Fig. 12.

Distribution of AGN fraction, with (black shaded area) and without (blue histogram) polar dust. The addition of polar dust to the fitting process increases the AGN fraction.

5.5. Polar dust and AGN type

Repeating the same exercise as for the inclusion of the X-ray flux, we examine whether there are sources for which the classification, based on the estimated inclination angle, changes with the introduction of polar dust (fX is not included in these runs). Our analysis shows that only 0.01% of the AGNs classified as type 2 with polar dust, change classification without it. However, 14% of the sources characterised as type 1 when polar dust is included change to type 2 without it. The addition of attenuation by polar dust provides X-CIGALE with the flexibility to attribute part of the absorption to polar dust.

We compare the sources that change from type 1 to type 2 when we ignore polar dust with the classification from optical SDSS spectra. ∼70% of the AGNs present broad lines in the optical continuum. We conclude that the introduction of polar dust as a free parameter in the SED fitting process improves the accuracy of X-CIGALE in the source type classification.

5.6. Polar dust and AGN detection efficiency

X-ray detection provides the most reliable method to identify AGNs. Although X-rays have proven efficient in penetrating the intervening absorbing material (e.g., Luo et al. 2008), even hard X-rays are absorbed by huge amounts of gas and dust. Thus, X-ray-selected AGNs are biased against the most heavily absorbed sources. On the other hand, mid-IR surveys are less affected by extinction. The material that absorbs AGN radiation even at hard X-ray energies, is heated by the AGN and re-emits at infrared wavelengths. Therefore, mid-IR-selected AGNs can detect AGNs that are missed by X-ray surveys (e.g., Georgantopoulos et al. 2008; Fiore et al. 2009). Spitzer was the first infrared survey used to select AGNs via colour selection criteria (e.g., Stern et al. 2005; Donley et al. 2012). Later, these techniques were adapted and used for WISE. For instance, Mateos et al. (2012) used an AGN selection method based on three WISE colours. Assef et al. (2018) used the W1 and W2 criteria of Stern et al. (2012) and extended them to fainter magnitudes. However, IR selection techniques are biased against low-luminosity AGNs. In a recent study, Pouliasis et al. (2020) showed that SED decomposition can provide a complementary tool to the X-ray and IR selection techniques, since it makes use of large wavelength ranges to efficiently disentangle accretion from star formation (Ciesla et al. 2015; Dietrich et al. 2018; Małek et al. 2018).

Our AGN sample consists of X-ray selected sources with high X-ray luminosity (LX >  1042 erg s−1), therefore the contamination from non-AGN systems should be minimal. Thus, we wish to examine whether the addition of polar dust in the fitting process affects the efficiency of the SED decomposition to find a strong AGN component (fracAGN) and compare it with the efficiency of IR colour-selection criteria to detect AGN candidates among X-ray sources.

In Fig. 13, we plot the W1 − W2 vs. W2 for the X-ray AGNs in our sample. With magenta and grey, we mark the sources that are selected as IR AGN candidates with a confidence level of 75% and 90%, respectively, using the Assef et al. (2018) criteria. Circles present all the X-ray AGNs in our dataset, which were detected by WISE and colour-coded based on the AGN fraction measurements of X-CIGALE. To estimate them, we included the polar dust as a free parameter, but their X-ray flux was ignored. There are 1956 sources with WISE counterparts that also fulfill our selection criteria (Sect. 3.3). Of 1956, 689 (∼35%) were detected by Assef et al. (2018) with 90% confidence (889/1956 ≈ 45%, with 75% confidence). 671 out of the 689 sources (∼97%) have fracAGN >  0.2 (631/689 ≈ 92% with fracAGN >  0.3), meaning the SED decomposition reveals strong AGN contribution in the IR emission of the system. In Appendix A.2, we examine how many of the sources with measured fracAGN >  0.2 (and fracAGN >  0.3) have true fracAGN <  0.2, that is, the AGN fraction has been miscalculated and the AGN contribution to the IR emission of the galaxy is weaker than assumed. Our analysis shows < 10% contamination from such systems (Fig. A.3). Among IR-selected AGNs with 75% confidence, the numbers are 851/889 ≈ 96% with fracAGN >  0.2 (795/889 ≈ 90% with fracAGN >  0.3). Therefore, X-CIGALE finds a strong AGN component in more than 90% of the IR-selected AGNs. However, as we notice in Fig. 13, there is a large number of fainter AGNs that are missed by IR colour-selection criteria, but the SED fitting finds strong AGN IR emission. Specifically, 85% of the sources have fracAGN >  0.2 and 74% have fracAGN >  0.3. Thus, even considering the more strict criterion of fracAGN >  0.3, SED decomposition manages to detect a larger fraction of X-ray selected AGNs than IR colour criteria (74% vs. 45%), while at the same time it also uncovers > 90% of the IR colour-selected AGNs. Repeating the same exercise without the inclusion of polar dust in the SED fitting process yields 71% sources with fracAGN >  0.2 and 55% with fracAGN >  0.3 among the non-IR detected AGNs. Thus, the addition of polar dust allows SED decomposition to uncover a larger fraction of X-ray AGNs. The results are summarised in Table 5. On the other hand, the addition of the X-ray flux only marginally improves the AGN detection efficiency. Specifically, among the non-IR-selected AGN, modelling the SEDs with X-ray flux and polar dust yields 87% of the sources with fracAGN >  0.2 and 75% with fracAGN >  0.3, compared to 85% and 74%, respectively, using only the polar dust in the fitting process.

thumbnail Fig. 13.

W1 − W2 vs. W2 for X-ray AGNs in our sample (circles). Sources selected as IR AGNs with 75% and 90% confidence, using the criteria of Assef et al. (2018), are in magenta and grey, respectively. The remaining sources of our X-ray AGN sample are presented with circles, colour coded based on the AGN fraction measurements of X-CIGALE. ∼97% of IR selected AGN have fracAGN >  0.2 (∼92% with fracAGN >  0.3). This percentage drops to 71% (55% with fracAGN >  0.3) without polar dust.

Table 5.

Polar dust and AGN detection efficiency of the SED decomposition.

In Fig. 14, we show a repetition of the same exercise using the Spitzer colour-selection criteria of Donley et al. (2012). 968 AGNs in our sample were detected by Spitzer and follow our selection criteria (Sect. 3.3). 409/968 (42%) are IR-selected AGNs, based on the criteria presented by Donley et al. (2012). 98% of these IR-selected AGNs have fracAGN >  0.2 (92% with fracAGN >  0.3). Among the 968 sources with Spitzer counterparts, ∼80% have fracAGN >  0.2 and ∼67% have fracAGN >  0.3. These numbers drop to ∼65% and ∼49%, when polar dust is not included in the SED fitting process.

thumbnail Fig. 14.

Colour-colour distribution of X-ray AGNs in our sample (circles) using Spitzer photometry. Circles are colour-coded based on the AGN fraction estimations of X-CIGALE. Sources selected as IR AGN using the colour criteria of Donley et al. (2012), are marked with purple squares.

From this part of the analysis, we conclude that the addition of polar dust improves the ability of the algorithm to disentangle the AGN and host galaxy IR emission, and thus increases the efficiency of the SED decomposition to detect a strong AGN component.

6. Summary

In this work, we used the SED fitting code X-CIGALE to model the SEDs of ∼2500 X-ray AGNs in the XMM-XXL field. X-CIGALE has some important features to model the AGN emission. It accounts for polar-dust extinction that is commonly found, especially in the case of X-ray-selected AGNs and includes X-ray data in the SED fitting process.

Our analysis shows that the estimated L2500 Å and L2 keV follow the adopted Just et al. (2007) relation, as well as other similar relations in the literature (Lusso & Risaliti 2016). The SED fitting results are not sensitive to the choice of the extinction law used and the temperature of polar dust. This result holds for our X-ray dataset and the available photometry. The addition of far-IR (Herschel) photometry does not statistically affect the polar dust and AGN fraction measurements; however, it slightly increases the statistical significance of the latter (from 3.6σ to 4.5σ).

About half of the sources identified as type 2 with the addition of X-ray flux are found to be type 1 in the absence of X-ray information. Visual inspection of randomly selected optical (SDSS) spectra, revealed that for ∼85% of the AGNs that lie at z <  1, the optical spectrum agrees with X-CIGALE’s classification when using fX. However, at higher redshifts (z >  1.5) lack of available IR photometry does not allow the algorithm to classify sources in a robust manner. Inclusion of polar dust also improves the agreement of the AGN type classification between X-CIGALE and optical spectra. We conclude that the addition of the X-ray flux and polar dust in the fitting process improves the accuracy of the classification of AGNs as type 1 or 2.

Compared to IR selection techniques (Donley et al. 2012; Assef et al. 2018), X-CIGALE recovered > 90% of the IR-selected AGNs. Among the X-ray-detected AGNs that are not IR selected, SED decomposition attributes a large AGN component to the IR emission of the system (fracAGN >  0.2) for ∼85% of them. This number drops to ∼70% when polar dust is not included in the SED modelling. Thus, the addition of polar dust improves the efficiency of the SED decomposition to detect AGN.

One of the most important strengths of SED fitting is that it allows the disentangling of the IR emission of AGNs from that of the host galaxy, quantified by fracAGN in X-CIGALE. The new features of X-CIGALE improve the AGN fraction measurements. Specifically, the addition of the X-ray flux improves the statistical significance of the AGN fraction measurements, in particular for luminous (LX >  1045 erg s−1) sources (Fig. 9). The addition of polar dust increases the fracAGN estimations, since the AGN contributes more to the IR emission of the system.

The conclusions of our analysis hold under the condition that (mid-) IR photometry is available in the SED fitting process. A lack of IR photometry may result in unreliable X-ray luminosity calculations, lower and less robust AGN fraction estimates, and AGN type misclassifications.

1

The Herschel Extragalactic Legacy Project (HELP; http://herschel.sussex.ac.uk/) is a European-funded project to analyse all the cosmological fields observed with the Herschel satellite. All the HELP data products can be accessed on HeDaM (http://hedam.lam.fr/HELP/).

Acknowledgments

The authors thank the anonymous referee for their comments. The authors thank Raphael Shirley and Yannick Roehlly for their help to retrieve IR fluxes on the XMM-LSS field. GM acknowledges support by the Agencia Estatal de Investigación, Unidad de Excelencia María de Maeztu, ref. MDM-2017-0765. MB acknowledges FONDECYT regular grant 1170618 XXL is an international project based around an XMM Very Large Programme surveying two 25 deg2 extragalactic fields at a depth of ∼6 × 10−15 erg cm −2 s−1 in the [0.5−2] keV band for point-like sources. The XXL website is http://irfu.cea.fr/xxl/. Multi-band information and spectroscopic follow-up of the X-ray sources are obtained through a number of survey programmes, summarised at http://xxlmultiwave.pbworks.com/. This research has made use of data obtained from the 3XMM XMM-Newton serendipitous source catalogue compiled by the 10 institutes of the XMM-Newton Survey Science Centre selected by ESA. This work is based on observations made with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and NASA. Funding for the Sloan Digital Sky Survey IV has been provided by the Alfred P. Sloan Foundation, the U.S. Department of Energy Office of Science, and the Participating Institutions. SDSS-IV acknowledges support and resources from the Center for High-Performance Computing at the University of Utah. The SDSS website is www.sdss.org. SDSS-IV is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS Collaboration including the Brazilian Participation Group, the Carnegie Institution for Science, Carnegie Mellon University, the Chilean Participation Group, the French Participation Group, Harvard-Smithsonian Center for Astrophysics, Instituto de Astrofísica de Canarias, The Johns Hopkins University, Kavli Institute for the Physics and Mathematics of the Universe (IPMU)/University of Tokyo, Lawrence Berkeley National Laboratory, Leibniz Institut für Astrophysik Potsdam (AIP), Max-Planck-Institut für Astronomie (MPIA Heidelberg), Max-Planck-Institut für Astrophysik (MPA Garching), Max-Planck-Institut für Extraterrestrische Physik (MPE), National Astronomical Observatories of China, New Mexico State University, New York University, University of Notre Dame, Observatório Nacional/MCTI, The Ohio State University, Pennsylvania State University, Shanghai Astronomical Observatory, United Kingdom Participation Group, Universidad Nacional Autónoma de México, University of Arizona, University of Colorado Boulder, University of Oxford, University of Portsmouth, University of Utah, University of Virginia, University of Washington, University of Wisconsin, Vanderbilt University, and Yale University. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. The VISTA Data Flow System pipeline processing and science archive are described in Irwin et al. (2004), Hambly et al. (2008) and Cross et al. (2012). Based on observations obtained as part of the VISTA Hemisphere Survey, ESO Program, 179.A-2010 (PI: McMahon). We have used data from the 3rd data release. This work is based [in part] on observations made with the Spitzer Space Telescope, which was operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA The project has received funding from Excellence Initiative of Aix-Marseille University – AMIDEX, a French ‘Investissements d’Avenir’ programme.

References

  1. Asmus, D. 2019, MNRAS, 489, 2177 [CrossRef] [Google Scholar]
  2. Asmus, D., Hönig, S. F., Gandhi, P., Smette, A., & Duschl, W. J. 2014, MNRAS, 439, 1648 [NASA ADS] [CrossRef] [Google Scholar]
  3. Assef, R. J., Stern, D., Noirot, G., et al. 2018, ApJS, 234, 23 [NASA ADS] [CrossRef] [Google Scholar]
  4. Bongiorno, A., Merloni, A., Brusa, M., et al. 2012, MNRAS, 427, 3103 [NASA ADS] [CrossRef] [Google Scholar]
  5. Boquien, M., Burgarella, D., Roehlly, Y., et al. 2019, A&A, 622, A103 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  6. Brandt, W. N., & Alexander, D. M. 2015, A&ARv, 23, 1 [Google Scholar]
  7. Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000 [NASA ADS] [CrossRef] [Google Scholar]
  8. Burgarella, D., Buat, V., & Iglesias-Páramo, J. 2005, MNRAS, 360, 1413 [NASA ADS] [CrossRef] [Google Scholar]
  9. Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682 [NASA ADS] [CrossRef] [Google Scholar]
  10. Caputi, K. I., Cirasuolo, M., Dunlop, J. S., et al. 2011, MNRAS, 413, 162 [NASA ADS] [CrossRef] [Google Scholar]
  11. Carnall, A. C., McLure, R. J., Dunlop, J. S., & Davé, R. 2018, MNRAS, 480, 4379 [NASA ADS] [CrossRef] [Google Scholar]
  12. Chabrier, G. 2003, PASP, 115, 763 [Google Scholar]
  13. Ciesla, L., Charmandaris, V., Georgakakis, A., et al. 2015, A&A, 576, A10 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  14. Civano, F., Elvis, M., Brusa, M., et al. 2012, ApJS, 201, 30 [NASA ADS] [CrossRef] [Google Scholar]
  15. Cross, N. J. G., Collins, R. S., Mann, R. G., et al. 2012, A&A, 548, A119 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  16. da Cunha, E., Charlot, S., & Elbaz, D. 2008, MNRAS, 388, 1595 [Google Scholar]
  17. Dale, D. A., Helou, G., Magdis, G. E., et al. 2014, ApJ, 784, 83 [NASA ADS] [CrossRef] [Google Scholar]
  18. Dawson, K. S., Schlegel, D. J., Ahn, C. P., et al. 2013, AJ, 145, A10 [NASA ADS] [CrossRef] [Google Scholar]
  19. Dietrich, J., Weiner, A. S., Ashby, M. L. N., et al. 2018, MNRAS, 480, 3562 [CrossRef] [Google Scholar]
  20. Donley, J. L., Koekemoer, A. M., Brusa, M., et al. 2012, ApJ, 748, 142 [Google Scholar]
  21. Eales, S., Dunne, L., Clements, D., et al. 2010, PASP, 122, 499 [NASA ADS] [CrossRef] [Google Scholar]
  22. Eisenstein, D. J., Weinberg, D. H., Agol, E., et al. 2011, AJ, 142, A72 [NASA ADS] [CrossRef] [Google Scholar]
  23. Emerson, J., McPherson, A., & Sutherland, W. 2006, The Messenger, 126, 41 [NASA ADS] [Google Scholar]
  24. Fabbiano, G. 1989, ARA&A, 27, 87 [NASA ADS] [CrossRef] [Google Scholar]
  25. Feltre, A., Hatziminaoglou, E., Fritz, J., & Franceschini, A. 2012, MNRAS, 426, 120 [NASA ADS] [CrossRef] [Google Scholar]
  26. Ferrarese, L., & Merritt, D. 2000, ApJ, 539, 9 [NASA ADS] [CrossRef] [Google Scholar]
  27. Fiore, F., Puccetti, S., Brusa, M., et al. 2009, ApJ, 693, 447 [NASA ADS] [CrossRef] [Google Scholar]
  28. Gardner, J. P., Mather, J. C., Clampin, M., et al. 2006, Space Sci. Rev., 123, 485 [NASA ADS] [CrossRef] [Google Scholar]
  29. Gaskell, C. M., Goosmann, R. W., Antonucci, R. R. J., & Whysong, D. H. 2004, ApJ, 616, 147 [NASA ADS] [CrossRef] [Google Scholar]
  30. Georgakakis, A., Coil, A. L., Willmer, C. N. A., et al. 2011, MNRAS, 418, 2590 [NASA ADS] [CrossRef] [Google Scholar]
  31. Georgakakis, A., Salvato, M., Liu, Z., et al. 2017, MNRAS, 469, 3232 [NASA ADS] [CrossRef] [Google Scholar]
  32. Georgantopoulos, I., Georgakakis, A., Rowan-Robinson, M., & Rovilos, E. 2008, A&A, 484, 671 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  33. Hambly, N. C., Collins, R. S., Cross, N. J. G., et al. 2008, MNRAS, 384, 637 [NASA ADS] [CrossRef] [Google Scholar]
  34. Hatziminaoglou, E., Fritz, J., & Jarrett, T. H. 2009, MNRAS, 399, 1206 [NASA ADS] [CrossRef] [Google Scholar]
  35. Hurley, P. D., Oliver, S., Betancourt, M., et al. 2017, MNRAS, 464, 885 [NASA ADS] [CrossRef] [Google Scholar]
  36. Irwin, M. J., Lewis, J., Hodgkin, S., et al. 2004, Proc. SPIE, 5493, 411 [NASA ADS] [CrossRef] [Google Scholar]
  37. Jaffarian, G. W., & Gaskell, C. M. 2020, MNRAS, 493, 930 [NASA ADS] [CrossRef] [Google Scholar]
  38. Just, D. W., Brandt, W. N., Shemmer, O., et al. 2007, ApJ, 685, 1004 [NASA ADS] [CrossRef] [Google Scholar]
  39. Kaastra, J. S., Kriss, G. A., Cappi, M., et al. 2014, Science, 345, 64 [NASA ADS] [CrossRef] [Google Scholar]
  40. Laird, E. S., Nandra, K., Adelberger, K. L., Steidel, C. C., & Reddy, N. A. 2005, MNRAS, 359, 47 [NASA ADS] [CrossRef] [Google Scholar]
  41. Leja, J., Johnson, B. D., Conroy, C., van Dokkum, P. G., & Byler, N. 2017, ApJ, 837, 170 [NASA ADS] [CrossRef] [Google Scholar]
  42. Li, J., Xue, Y., Sun, M., et al. 2019, ApJ, 877, 5 [NASA ADS] [CrossRef] [Google Scholar]
  43. Liu, Z., Merloni, A., Georgakakis, A., et al. 2016, MNRAS, 459, 1602 [NASA ADS] [CrossRef] [Google Scholar]
  44. Lonsdale, C. J., Smith, H. E., Rowan-Robinson, M., et al. 2003, PASP, 115, 897 [NASA ADS] [CrossRef] [Google Scholar]
  45. Luo, B., Bauer, F. E., Brandt, W. N., et al. 2008, ApJS, 179, 19 [NASA ADS] [CrossRef] [Google Scholar]
  46. Lusso, E., & Risaliti, G. 2016, ApJ, 819, 154 [NASA ADS] [CrossRef] [Google Scholar]
  47. Lusso, E., Comastri, A., Simmons, B. D., et al. 2012, MNRAS, 425, 623 [NASA ADS] [CrossRef] [Google Scholar]
  48. Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285 [Google Scholar]
  49. Małek, K., Buat, V., Roehlly, Y., et al. 2018, A&A, 620, A50 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  50. Malizia, A., Bassani, L., Stephen, J. B., Bazzano, A., & Ubertini, P. 2020, A&A, 639, A5 [CrossRef] [EDP Sciences] [Google Scholar]
  51. Masoura, V. A., Mountrichas, G., Georgantopoulos, I., et al. 2018, A&A, 618, A31 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  52. Masoura, V. A., Georgantopoulos, I., Mountrichas, G., et al. 2020, A&A, 638, A45 [CrossRef] [EDP Sciences] [Google Scholar]
  53. Masoura, V. A., Mountrichas, G., Georgantopoulos, I., & Plionis, M. 2021, A&A, in press, https://doi.org/10.1051/0004-6361/202039238 [Google Scholar]
  54. Mateos, S., Alonso-Herrero, A., Carrera, F. J., et al. 2012, MNRAS, 426, 3271 [NASA ADS] [CrossRef] [Google Scholar]
  55. Menzel, M.-L., Merloni, A., Georgakakis, A., et al. 2016, MNRAS, 457, 110 [NASA ADS] [CrossRef] [Google Scholar]
  56. Morganti, R. 2017, Nat. Astron., 1, 39 [NASA ADS] [CrossRef] [Google Scholar]
  57. Mukai, K. 1993, Legacy, 3, 21 [Google Scholar]
  58. Noll, S., Burgarella, D., Giovannoli, E., et al. 2009, A&A, 507, 1793 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  59. Oliver, S. J., Bock, J., Altieri, B., et al. 2012, MNRAS, 424, 1614 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  60. Park, T., Kashyap, V. L., Siemiginowska, A., et al. 2006, ApJ, 652, 610 [NASA ADS] [CrossRef] [Google Scholar]
  61. Persic, M., & Rephaeli, Y. 2002, A&A, 382, 843 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  62. Pierre, M., Pacaud, F., Adami, C., et al. 2016, A&A, 592, A1 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  63. Pouliasis, E., Mountrichas, G., Georgantopoulos, I., et al. 2020, MNRAS, 495, 1853 [CrossRef] [Google Scholar]
  64. Prevot, M., Lequeux, J., Maurice, E., Prevot, L., & Rocca-Volmerange, B. 1984, A&A, 132, 389 [Google Scholar]
  65. Ranalli, P., Comastri, A., & Setti, G. 2003, A&A, 399, 39 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  66. Reichert, G. A., Mushotzky, R. F., Holt, S. S., & Petre, R. 1985, ApJ, 296, 69 [NASA ADS] [CrossRef] [Google Scholar]
  67. Risaliti, G., & Lusso, E. 2017, Astron. Nachr., 338, 329 [CrossRef] [Google Scholar]
  68. Robotham, A. S. G., Bellstedt, S., Lagos, C. D. P., et al. 2020, MNRAS, 495, 905 [CrossRef] [Google Scholar]
  69. Shirley, R., Roehlly, Y., Hurley, P. D., et al. 2019, MNRAS, 490, 634 [CrossRef] [Google Scholar]
  70. Smee, S., Gunn, J. E., Uomoto, A., et al. 2013, AJ, 146, A32 [NASA ADS] [CrossRef] [Google Scholar]
  71. Stalevski, M., Fritz, J., Baes, M., Nakos, T., & Popović, L. Č. 2012, MNRAS, 420, 2756 [NASA ADS] [CrossRef] [Google Scholar]
  72. Stalevski, M., Ricci, C., Ueda, Y., et al. 2016, MNRAS, 458, 2288 [NASA ADS] [CrossRef] [Google Scholar]
  73. Stanley, F., Harrison, C. M., Alexander, D. M., et al. 2018, MNRAS, 478, 3721 [NASA ADS] [CrossRef] [Google Scholar]
  74. Stern, D., Eisenhardt, P., Gorjian, V., et al. 2005, ApJ, 631, 163 [NASA ADS] [CrossRef] [Google Scholar]
  75. Stern, D., Assef, R. J., Benford, D. J., et al. 2012, ApJ, 753, 30 [NASA ADS] [CrossRef] [Google Scholar]
  76. Sutherland, W., & Saunders, W. 1992, MNRAS, 259, 413 [NASA ADS] [CrossRef] [Google Scholar]
  77. Tristram, K. R. W., Burtscher, L., Jaffe, W., et al. 2014, A&A, 563, A82 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  78. Werner, M. W., Roellig, T. L., Low, F. J., et al. 2004, ApJS, 154, 1 [NASA ADS] [CrossRef] [Google Scholar]
  79. Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010, AJ, 140, 1868 [Google Scholar]
  80. Yan, L., Donoso, E., Tsai, C.-W., et al. 2013, AJ, 145, 55 [NASA ADS] [CrossRef] [Google Scholar]
  81. Yang, G., Brandt, W. N., Luo, B., et al. 2016, ApJ, 831, 145 [NASA ADS] [CrossRef] [Google Scholar]
  82. Yang, G., Brandt, W. N., Alexander, D. M., et al. 2019, MNRAS, 485, 3721 [CrossRef] [Google Scholar]
  83. Yang, G., Boquien, M., Buat, V., et al. 2020, MNRAS, 491, 740 [NASA ADS] [CrossRef] [Google Scholar]

Appendix A: Mock catalogue analysis

As mentioned in Sect. 3.2.4, X-CIGALE offers the option to create and analyse mock catalogues based on the best-fit model of each source of the dataset. Here, we use this to assess the efficiency of X-CIGALE in estimating important parameters. Throughout this section, the Bayesian estimates of the mock values are presented.

A.1. The efficiency of X-CIGALE in estimating αox and L2500 Å

In Fig. A.1, we present a test of X-CIGALE’s reliability in estimating the αox parameter using the results of the mock catalogue. Specifically, we compare the true values of the parameters from the best fit of the dataset to the Bayesian values of the parameter obtained from the fit of the mock catalogue. The distribution of the difference of the two values clearly peaks at zero and ∼95% of the sources are within ±0.1. This demonstrates that the algorithm can reliably recover this parameter.

thumbnail Fig. A.1.

Efficiency of X-CIGALE to constrain the αox parameter using the mock catalogue (see Sect. 3.2.4). The distribution peaks at zero, with 95% of the sources lying within ±0.1. This demonstrates that the algorithm can successfully estimate αox.

In the same fashion, Fig. A.2 presents the efficiency of X-CIGALE to constrain L2500 Å with (left panel) and without (right panel) the X-ray flux. When the X-ray flux is included in the SED, L2500 Å is well constrained. When there is no X-ray information in the SED, the scatter is larger. This is expected based on how X-CIGALE calculates this parameter in the absence of the X-ray flux (see Sect. 3.2.2).

thumbnail Fig. A.2.

Comparison of L2500 Å values estimated by X-CIGALE for the mock sources with the true values (data). Left panel: when X-ray flux is included in the SED, L2500 Å is well constained. Right panel: X-ray flux is not included in the data. Although X-CIGALE recovers the parameter successfully, the scatter is larger compared to the results with the X-ray flux.

A.2. The efficiency of X-CIGALE in estimating the AGN fraction

Here, we examine the accuracy of X-CIGALE in the estimation of the AGN fraction. In Fig. A.3, we plot the AGN fraction values estimated by the fit of the mock catalogue vs. the true values (i.e. the values from the best fit of the data). Circles present the mean measurements, and their standard deviation is also plotted. Median values are shown by triangles. There is a good agreement between the estimated and true values.

thumbnail Fig. A.3.

AGN fraction values estimated by the fit of the mock catalogue vs. the true values (from the fit of the data). Circles present the mean measurements and their standard deviation is also plotted. Median values are shown by triangles.

In Fig. A.4, we plot the difference of the AGN fraction values estimated by fitting the mock catalogue from the true values. We present the results when the X-ray flux has been included in the SED (shaded histograms) and without including the X-ray flux (green lines). The results are split into X-ray luminosity bins. Based on our analysis, X-CIGALE can successfully recover the AGN fractions, regardless of whether the X-ray flux is included in the SED fitting or not, at all X-ray luminosities.

thumbnail Fig. A.4.

Difference of the AGN fraction values estimated by fitting the mock catalogue from the true values. We present the results when the X-ray flux has been included in the SED (shaded histograms) and without including the X-ray flux (green histograms). The results are split into X-ray luminosity bins. X-CIGALE can successfully recover the AGN fractions, regardless of whether the X-ray flux is included in the SED fitting or not, at all X-ray luminosities.

In Sect. 5.6, we examine the fraction of X-ray AGNs for which a strong AGN component is measured (i.e. fracAGN >  0.2). In Fig. A.5, we test the level of contamination in these calculations, that is, the percentage of sources for which the true AGN fraction is < 0.2, albeit the measured fracAGN >  0.2. Using a threshold of measured (mock) fracAGN >  0.2, the contamination is ≈10%, which drops to ∼6.5% if we use a higher threshold of fracAGN >  0.3. For this calculation, we ran X-CIGALE with polar dust and without X-ray flux, in accordance with the configuration used in Sect. 5.6.

thumbnail Fig. A.5.

Binned measurements of true AGN fraction values (data) for fracAGN, mock >  0.2 (green histogram) and fracAGN, mock >  0.3 (black histogram) from the mock analysis. For sources with fracAGN, mock >  0.2, there is ≈10% contamination (i.e. sources for which the AGN component is low: fracAGN, data <  0.2). For sources with fracAGN, mock >  0.3, the contamination is ∼6.5%.

A.3. The efficiency of X-CIGALE in the estimation of polar dust contribution

To check whether X-CIGALE can successfully constrain the effect of dust extinction by polar dust, in Fig. A.6 we plot the Bayesian estimations of the EB − V parameter for the polar dust vs. its exact value. The scatter (1σ variations) is substantial. Furthermore, for EB − V <  0.5 the fit is not sensitive to an incremental increase in the reddening parameter. The sensitivity is better at higher values of EB − V, but the parameter is still not well determined. The picture remains the same regardless of whether the X-ray flux is taken into account in the SED fitting. These results suggest that it is redundant to use in the SED configuration process, values of EB − V within small intervals, and that the estimated values of the parameter are not well constrained.

thumbnail Fig. A.6.

Polar dust estimations from mock sources vs. input (data) values. Although the 1σ deviation is large, mock values increase with as the data values do. However, the algorithm is not sensitive to incremental parameter increases. The trends are similar regardless of whether the X-ray flux is included (left panel) or not (right panel) in the fitting process.

All Tables

Table 1.

Number of available spectroscopic sources in different photometric bands.

In the text
Table 2.

5σ sensitivity (mag) of the photometric bands used in our analysis.

In the text
Table 3.

Models and the values for their free parameters used by X-CIGALE for the SED fitting of our galaxy sample.

In the text
Table 4.

Number of AGNs that satisfy our selection criteria (Sect. 3.3) for different configurations of the SED fitting process.

In the text
Table 5.

Polar dust and AGN detection efficiency of the SED decomposition.

In the text

All Figures

thumbnail Fig. 1.

Redshift distribution of the X-ray AGN sample.
In the text
thumbnail Fig. 2.

Distribution of X-ray absorption, NH, of the X-ray AGN sample.
In the text
thumbnail Fig. 3.

Comparison of intrinsic X-ray luminosity in the 2−10 keV band estimated by the SED fitting with that from the input catalogue. Green, solid, and dashed lines present the 1:1 correspondence and the errors added in quadrature of the X-ray luminosity calculations from X-CIGALE and those quoted in the X-ray catalogue (Menzel et al. 2016), respectively. The grey dashed line shows the χ2 fit of the calculations. Symbols are colour-coded based on their NH values. Left panel: results with the X-ray flux included in the SED. X-CIGALE calculations are consistent with those from the input catalogue, at all luminosities spanned by our sample. There are 70 sources for which the LX calculation differs by more than 0.5 dex from their input values (shown with error bars, see text for more details). Right panel: X-ray flux is not included in the SED. In this case, the scatter is significantly larger, while the error of the LX calculations from X-CIGALE, increases by ≈4×.
In the text
thumbnail Fig. 4.

L2 keV vs. L2500 Å relation. Blue line presents the Just et al. (2007) relation that X-CIGALE uses to connect the X-ray flux with the UV (2500 Å) luminosity. Green line shows the L2 keV vs. L2500 Å from Lusso & Risaliti (2016). For comparison, we also plot the fit from our calculations (grey line).
In the text
thumbnail Fig. 5.

Distribution of polar dust measurement differences for different extinction laws (panels a and b) and grey-body dust temperatures (panels c and d). All distributions are highly peaked at zero, which indicates that polar dust calculations are not sensitive to the choice of these parameters.
In the text
thumbnail Fig. 6.

Same as in the previous plot, but for the AGN fraction. Different extinction laws (panels a and b) and grey-body dust temperatures (panels c and d) do not affect the fracAGN measurements.
In the text
thumbnail Fig. 7.

Distribution of the difference of polar dust, AGN fraction and X-ray and UV luminosity measurements, with and without Herschel photometry. Polar dust and luminosity calculations are not affected by the existence of far-IR photometry. However, addition of far-IR photometry reduces the fracAGN. This is quantified by the mean values, fracAGN, Herschel = 0.36 ± 0.08 and fracAGN, no Herschel = 0.43 ± 0.12.
In the text
thumbnail Fig. 8.

Left: distribution of the AGN fraction, with X-ray flux (black shaded area) and without X-ray flux (blue histogram). Right: distribution of the difference of the AGN fraction, with and without X-ray flux, for different luminosity bins. All distributions are normalised to unity. A tail appears at positive values, i.e., the AGN fractions are higher when X-ray flux is included in the SED fitting, for AGNs with LX <  1043 erg s−1 that becomes more prominent for the most luminous sources (LX <  1045 erg s−1). On the other hand, there is a tail at negative values, i.e. the AGN fractions are lower with X-ray flux, for the fainter AGNs (LX <  1043 erg s−1, blue line).
In the text
thumbnail Fig. 9.

Difference of the significance, Δσ, of the AGN fraction estimations with and without X-ray information, for sources with LX >  1045 erg s−1 and difference in fracAGN >  0.2 (shaded histogram) and sources with LX <  1043 erg s−1 and difference in fracAGN <  −0.2. The statistical significance of the AGN fraction measurements improves significantly with the inclusion of the X-ray flux, especially for the most luminous AGNs.
In the text
thumbnail Fig. 10.

Comparison of L2500 Å values with and without X-ray flux in the fitting process. Left panel: comparison of sources that are classified as the same type, with and without X-ray flux in the SED fitting process. Blue symbols present the results for type 1 sources (i = 30°) and red symbols for type 2 sources (i = 70°). Right panel: comparison of L2500 Å values for sources with different types from the two runs (with and without X-ray flux). X-CIGALE decreases/increases the intrinsic accretion power (and thus the intrinsic L2500 Å) depending on whether the source is type 1/2.
In the text
thumbnail Fig. 11.

X-ray AGN fraction vs. X-ray luminosity estimated by X-CIGALE. The X-ray AGN fraction is defined as the ratio of the X-ray AGN emission to the total X-ray emission of the galaxy (AGN + binaries + hot gas). As expected, for the vast majority of sources the AGN X-ray emission contributes more than 90% of the total X-ray emission of the galaxy.
In the text
thumbnail Fig. 12.

Distribution of AGN fraction, with (black shaded area) and without (blue histogram) polar dust. The addition of polar dust to the fitting process increases the AGN fraction.
In the text
thumbnail Fig. 13.

W1 − W2 vs. W2 for X-ray AGNs in our sample (circles). Sources selected as IR AGNs with 75% and 90% confidence, using the criteria of Assef et al. (2018), are in magenta and grey, respectively. The remaining sources of our X-ray AGN sample are presented with circles, colour coded based on the AGN fraction measurements of X-CIGALE. ∼97% of IR selected AGN have fracAGN >  0.2 (∼92% with fracAGN >  0.3). This percentage drops to 71% (55% with fracAGN >  0.3) without polar dust.
In the text
thumbnail Fig. 14.

Colour-colour distribution of X-ray AGNs in our sample (circles) using Spitzer photometry. Circles are colour-coded based on the AGN fraction estimations of X-CIGALE. Sources selected as IR AGN using the colour criteria of Donley et al. (2012), are marked with purple squares.
In the text
thumbnail Fig. A.1.

Efficiency of X-CIGALE to constrain the αox parameter using the mock catalogue (see Sect. 3.2.4). The distribution peaks at zero, with 95% of the sources lying within ±0.1. This demonstrates that the algorithm can successfully estimate αox.
In the text
thumbnail Fig. A.2.

Comparison of L2500 Å values estimated by X-CIGALE for the mock sources with the true values (data). Left panel: when X-ray flux is included in the SED, L2500 Å is well constained. Right panel: X-ray flux is not included in the data. Although X-CIGALE recovers the parameter successfully, the scatter is larger compared to the results with the X-ray flux.
In the text
thumbnail Fig. A.3.

AGN fraction values estimated by the fit of the mock catalogue vs. the true values (from the fit of the data). Circles present the mean measurements and their standard deviation is also plotted. Median values are shown by triangles.
In the text
thumbnail Fig. A.4.

Difference of the AGN fraction values estimated by fitting the mock catalogue from the true values. We present the results when the X-ray flux has been included in the SED (shaded histograms) and without including the X-ray flux (green histograms). The results are split into X-ray luminosity bins. X-CIGALE can successfully recover the AGN fractions, regardless of whether the X-ray flux is included in the SED fitting or not, at all X-ray luminosities.
In the text
thumbnail Fig. A.5.

Binned measurements of true AGN fraction values (data) for fracAGN, mock >  0.2 (green histogram) and fracAGN, mock >  0.3 (black histogram) from the mock analysis. For sources with fracAGN, mock >  0.2, there is ≈10% contamination (i.e. sources for which the AGN component is low: fracAGN, data <  0.2). For sources with fracAGN, mock >  0.3, the contamination is ∼6.5%.
In the text
thumbnail Fig. A.6.

Polar dust estimations from mock sources vs. input (data) values. Although the 1σ deviation is large, mock values increase with as the data values do. However, the algorithm is not sensitive to incremental parameter increases. The trends are similar regardless of whether the X-ray flux is included (left panel) or not (right panel) in the fitting process.
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.