Multi-wavelength Selected Compton-thick AGNs in Chandra Deep Field-South Survey
Xiaotong Guo, Qiusheng Gu, Nan Ding, Xiaoling Yu, Yongyun Chen
DDraft version December 23, 2020
Typeset using L A TEX twocolumn style in AASTeX63
Multi-wavelength Selected Compton-thick AGNs in
Chandra
Deep Field-South Survey
Xiaotong Guo ( 郭 晓 通 ) ,
1, 2
Qiusheng Gu ( 顾 秋 生 ),
1, 2
Nan Ding ( 丁 楠 ), Xiaoling Yu ( 俞 效 龄 )
1, 2
And Yongyun Chen ( 陈 永 云 ) — School of Astronomy and Space Science, Nanjing University, Nanjing, Jiangsu 210093, China Key Laboratory of Modern Astronomy and Astrophysics (Nanjing University), Ministry of Education, Nanjing 210093, China School of Physical Science and Technology, Kunming University, Kunming 650214, China College of Physics and Electronic Engineering, Qujing Normal University, Qujing 655011, China (Received Aug 10, 2020; Revised Sep 12, 2020; Accepted Dec 2, 2020)
ABSTRACTEven in deep X-ray surveys, Compton-thick active galactic nuclei (CT AGNs, N H (cid:62) . × cm − )are difficult to be identified due to X-ray flux suppression and their complex spectral shape. However,the study of CT AGNs is vital for understanding the rapid growth of black holes and the origin ofcosmic X-ray background. In the local universe, the fraction of CT AGNs accounts for 30% of the wholeAGN population. We may expect a higher fraction of CT AGNs in deep X-ray surveys, however, only10% of AGNs have been identified as CT AGNs in the 7 Ms Chandra
Deep Field-South (CDFS) survey.In this work, we select 51 AGNs with abundant multi-wavelength data. Using the method of the mid-infrared (mid-IR) excess, we select hitherto unknown 8 CT AGN candidates in our sample. Seven ofthese candidates can confirm as CT AGN based on the multi-wavelength identification approach, and anew CT AGN (XID 133) is identified through the mid-IR diagnostics. We also discuss the X-ray originof these eight CT AGNs and the reason why their column densities were underestimated in previousstudies. We find that the multi-wavelength approaches of selecting CT AGNs are highly efficient,provided the high quality of observational data. We also find that CT AGNs have a higher Eddingtonratio than non-CT AGNs, and that both CT AGNs and non-CT AGNs show similar properties of hostgalaxies.
Keywords: galaxies: active — galaxies: nuclei — X-rays: galaxies — infrared: galaxies INTRODUCTIONIt is well known that the X-ray emission has not onlya strong pierce but also exhibits significant property ofthe particle. The X-ray photon is absorbed as it passesthrough the interstellar medium due to Compton scat-tering. If the X-ray obscuring matter has an opticaldepth which is equal to or larger than 1 ( τ = σ T · N H1 (cid:62) N H (cid:62) . × cm − , Comastri 2004). The main charac- Corresponding author: Qiusheng [email protected] σ T is Thomson scattering cross-section ( σ T ≈ . × − cm ),N H is the equivalent neutral hydrogen column density. teristics of the X-ray spectra of CT AGNs are : (1) aflat spectrum with photon index, Γ, less than 1 at ener-gies below 10 keV; (2) an absorption turnoff at energiesabove 10 keV, with the exact cut-off energy dependingon the column density; and (3) a prominent iron K α emission line.CT AGNs are believed to be in a phase of the evo-lutionary scenario of AGNs, which is a rapid growthstate of central supermassive black holes (SMBHs, e.g.,Goulding et al. 2011). Some studies suggest a possibleconnection between CT AGNs and their host environ-ments (e.g., Ricci et al. 2017). For instance, CT AGNsprefer to host in gas-rich environments and galaxy merg-ers (Kocevski et al. 2015). Compared with Compton-thin AGNs, the host galaxies of CT AGNs show higherstar formation rates (SFRs, Goulding et al. 2012). a r X i v : . [ a s t r o - ph . GA ] D ec Guo et al.
Moreover, the cosmic X-ray background synthesis mod-els (e.g., Gilli et al. 2007; Akylas et al. 2012) also requirea significant fraction of CT AGNs. In the local universe,the fraction of CT AGNs is about 30% of the total AGNpopulation (e.g., Ricci et al. 2015). A higher fraction ofCT AGNs is expected at high redshift, where the frac-tion of mass of atomic and/or molecular gas in galaxiesis much higher than in the local universe (e.g., Carilli &Walter 2013). However, CT AGNs are very difficult tobe identified at high redshift, resulting in only a smallfraction ( (cid:46) ∼ .
5% in the UKIDSS Ultra Deep Sur-vey field (Masini et al. 2018). Therefore, there shouldexist many CT AGNs that are still missed in deep X-raysurveys.One important thing is how to identify CT AGNs effi-ciently. Currently, to identify CT AGN, the most com-monly used method is X-ray spectroscopy fitting. Forinstance, Lanzuisi et al. (2018) selected 67 CT AGNsfrom 1855 point sources in the
Chandra -COSMOS. How-ever, the major weakness of this method is the low ef-ficiency of selecting CT AGN, and it depends relativelyon the absorption-corrected model of X-ray. Thanks tothe successful operation of NuSTAR, which can pro-vide the X-ray source with a spectrum above 10 keV.The combination of NuSTAR and Chandra or XMM-Newton observations will constitute a broad wavebandX-ray spectrum, which can be more effectively used toidentify CT AGN. Many recent studies have used themethod of broad waveband X-ray spectroscopy fittingto identify CT AGNs (e.g., Kammoun et al. 2019; LaCaria et al. 2019; LaMassa et al. 2019; Toba et al. 2020).Besides, several multi-wavelength techniques have alsobeen developed to pre-select/identify CT AGNs in thepast decade, based on the known CT AGN multi-bandproperties. The method of the mid-infrared (mid-IR)excess, among the multi-wavelength techniques, is usu-ally used to select CT AGN candidates in some studies(e.g., Luo et al. 2011; Lanzuisi et al. 2015b; La Cariaet al. 2019). The multi-wavelength techniques usuallyuse the ratio between X-ray and other bands to identifyCT AGNs, including X-ray to mid-IR luminosity ratio(e.g., Lanzuisi et al. 2015b; Lansbury et al. 2015, 2017)and X-ray to high-ionization optical emission lines (suchas [O III] and [Ne V]) luminosity ratio (e.g., Maiolinoet al. 1998; Cappi et al. 2006; Gilli et al. 2010; Lanzuisiet al. 2015b). The
Chandra
Deep Field-South (CDFS) survey is thedeepest X-ray survey with an exposure time of about7 Ms so far (Luo et al. 2017). There are 1008 sourcesdetected by X-ray, among of which, 711 X-ray sourceswere classified as AGNs. The X-ray sources in the CDFSsurvey contain abundant photometric data from ultra-violet (UV) to infrared (IR), which were collected byStraatman et al. (2016). Some X-ray sources also con-tain optical spectra (e.g., Szokoly et al. 2004; Mignoliet al. 2005; Inami et al. 2017; Herenz et al. 2017) andradio data (e.g., Kellermann et al. 2008; Miller et al.2008, 2013). Since the CDFS survey is currently themost sensitive X-ray deep field, we expect that the frac-tion of CT AGNs can achieve as high as 30% in thissurvey. However, only 71 CT AGNs (10%) are identi-fied by the method of X-ray spectroscopy fitting (Liuet al. 2017; Li et al. 2019; Corral et al. 2019). There-fore, many CT AGNs ( about 20%) have been missedby their methods in the CDFS survey. Lambrides et al.(2020) pointed out that a large population of obscuredAGNs was mis-diagnosed as low-luminosity AGNs in theCDFS survey. The motivation of our work is to find outmissed CT AGNs in the CDFS survey. Moreover, wealso try to understand why these CT AGNs are missedby X-ray spectroscopy fitting.The structure of the paper is as follows. In Section 2,we describe the sample and data. In Section 3, we pre-select 8 CT AGNs via the method of the mid-IR ex-cess. The eight candidates will be identified by a com-bination of multi-wavelength identification approachesin Section 4. In Section 5, we also discuss the X-rayorigin of these eight CT AGNs, the reason why theircolumn densities were underestimated, the efficiency ofthe multi-wavelength approaches of selecting CT AGNs,and comparing the properties of the CT AGNs and thenon-CT AGNs. Finally, we present a brief summary ofthis work in Section 6. We adopt a concordance flat Λ-cosmology with H = 67 . − Mpc − , Ω m = 0 . Λ = 0 .
685 (Planck Collaboration et al. 2020). THE SAMPLE AND DATA2.1.
The AGN sample
The starting point of the analysis is the X-ray sourcesin the CDFS survey. Straatman et al. (2016) providedphotometric catalogs (from UV to IR; a total of 43bands) for the CDFS survey. In order to obtain the pho-tometric data of the X-ray sources in the CDFS survey,Guo et al. (2020) have cross-matched the X-ray catalog(Luo et al. 2017) and the photometric catalog (Straat-man et al. 2016) with a matching radius of 1 arcsec. Thematching results contained 839 X-ray sources. Based on elected CT AGN in CDFS µ m, Her-schel/PACS 100 µ m, and 160 µ m), especially.3) We exclude merger sources identified by their op-tical images.The first criterion ensures that each source is an AGN.The second criterion guarantees that each source hasenough photometric data to run the SED fitting modelreliably. The point spread function in mid-IR and far-IRis large. The merger sources are near each other. Thus,the flux density of the merger sources may be overes-timated at mid-IR and far-IR bands. To be accuratephotometric data, we implement the third criterion. Afinal AGN sample (51 AGNs) is constructed, includingeight known CT AGNs (XID 284, 332, 355, 405, 419,587, 666, and 739), which are identified by the methodof X-ray spectroscopy fitting (Liu et al. 2017; Li et al.2019; Corral et al. 2019).2.2. X-ray data
The X-ray emission of the radio-quiet AGN is believedto arise from the corona above the accretion disk, andit is predominantly produced by the inverse Comptonscattering of photons from the accretion disk. Since theX-ray has strong pierced, the absorption-corrected X-rayluminosities with different correcting models for mostAGNs are consistent. However, for CT AGN, there is asignificant difference with differently correcting models(e.g., Zappacosta et al. 2018). In some studies (e.g.,Luo et al. 2017), the corrected luminosities are used torepresent the intrinsic X-ray luminosities of AGNs.Table 1 shows our sample and the data used in thiswork. Column 5 of Table 1 lists the uncorrected 2–10 keV luminosities, which are converted by apparent0.5–7 keV luminosities from Luo et al. (2017) , wherethey estimated absorption by assuming that the intrin-sic power-law spectrum had a fixed photon index of The 0.5–7 keV luminosities convert to 2–10 keV luminosities:L = L . (ln10 − ln2) / (ln7 − ln0 . , Γ = 2; L =L . (10 − Γ+2 − − Γ+2 ) / (7 − Γ+2 − . − Γ+2 ) , Γ (cid:54) = 2. Where Γ isphoton index. wabs * (zwabs*powerlaw + zgauss + power-law + zwabs*pexrav*constant) model, and provide theabsorption-corrected 2–10 keV luminosities listed in Col-umn 7 of Table 1. To search for a best fit, Li et al. (2019)adopted two models to fit the source spectra. Column8 of Table 1 lists the absorption-corrected 2–10 keV lu-minosities provided by Li et al. (2019). Corral et al.(2019) selected 20 CT AGNs using the automated spec-tral analysis. Column 9 of Table 1 lists the absorption-corrected 2–10 keV luminosities of 3 CT AGNs, whichare obtained by Corral et al. (2019).Moreover, this work also uses the X-ray spectra ofthe eight CT AGN candidates. Their X-ray spectra arethe merged spectra for which the 102 observations arematched to an identically astrometric frame. In thiswork, we use the X-ray spectra produced by Luo et al.(2017). For more detailed data reduction, please seeSection 2.2 of Luo et al. (2017).2.3. Multi-band SED fitting and Mid-IR data
The photometric data from UV to IR can constitutea multi-band SED, which is composed of multiple ther-mal emission components. Thanks to SED fitting tech-nique, we can perform a detailed multi-band SED de-composition, including the stellar radiation, the dust re-radiation, and AGN emission. SED fitting technique canalso constrain some crucially physical quantities of bothAGNs and host galaxies, such as AGN luminosity, AGNfraction, SFR, and stellar mass (M ∗ ). Guo et al. (2020)obtained some physical quantities (including AGN lumi-nosity, AGN fraction, SFR, M ∗ , dust luminosity, AGNtype, and rest-frame 6 µ m luminosity for AGN compo-nent) of all AGNs in our sample through SED fittingusing photometric data from UV to IR. Therefore, somephysical quantities used in this work are from Guo et al.(2020).The mid-IR emission (3–30 µ m) of an AGN is mostdistinct from that of its host galaxy. AGNs are usuallybright in the mid-IR waveband because of the thermalemission from hot dust in the torus, which is heated byabsorbing UV and optical photons from the accretiondisk. The optical depth is low at the mid-IR waveband,and therefore the mid-IR emission of the AGN is notstrongly suppressed. In this paper, we use the 6 µ m Guo et al.
Table 1.
Physical properties of our AGN sample.
XID RA DEC z L X App L X Luo L X Liu L X Li L X Cor ν L ν (6 µ m) M BH L bol λ Edd [degree] [degree] [erg s − ] [erg s − ] [erg s − ] [erg s − ] [erg s − ] [erg s − ] [M (cid:12) ] [erg s − ](1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)100 53.006054 -27.694009 1.41 7.427e+43 1.153e+44 1.047e+44 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ∗ · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ∗ · · · · · · · · · · · · · · · · · · · · · · · · ∗ · · · · · · ∗ · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ∗ · · · · · · ∗ · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · Note —(1) is X-ray ID from the Luo et al. (2017) catalog. (2) and (3) are the R.A. and decl. of the source, respectively. (4) is redshift from Guo et al.(2020). (5) and (6) are the apparent rest-frame 2–10 keV luminosity and absorption-corrected 2–10 keV luminosity that is converted by apparent andabsorption-corrected 0.5–7 keV luminosity from Luo et al. (2017). (7), (8) and (9) are the absorption-corrected 2–10 keV luminosity by Liu et al. (2017),Li et al. (2019) and Corral et al. (2019), respectively. (10) is the rest-frame luminosity of the AGN component at 6 µ m. (11), (12) and 13 are BH mass,bolometric luminosity and Eddington ratio. The ∗ represents the BH mass estimated with broad line. elected CT AGN in CDFS µ m luminosities of AGNs, provided by Guoet al. (2020). MID-IR EXCESS PRE-SELECTING CT AGNSThe mid-IR and X-ray luminosities of AGNs areclosely related. Since the X-ray emission of CT AGNsis strongly suppressed, the intrinsic X-ray luminositiesof CT AGNs are usually underestimated. The mid-IRemission of the AGNs is not suppressed because the op-tical depth is low. In order to select CT AGNs, one easyway is to find the AGNs whose intrinsic X-ray luminosi-ties are underestimated.3.1.
The method of Mid-IR excess
The mid-IR and X-ray emission of radio-quiet AGNsare both excellent tracers of supermassive black hole ac-cretion power. Previous studies have investigated thecorrelation between the mid-IR and X-ray emission (e.g.,Lutz et al. 2004; Lanzuisi et al. 2009; Gandhi et al. 2009;Fiore et al. 2009; Chen et al. 2017). For instance, Stern(2015) derived a relation for the radio-quiet AGNs acrossa wide range of luminosity (from local Seyfert galaxieswith L X ∼ erg s − to the most luminous quasarswith L X ∼ erg s − ), which islog L(2–10 keV) = 40 .
981 + 1 . x − . x , (1)where L(2–10 keV) is in unit of erg s − and x ≡ log( ν L ν (6 µ m) / erg s − ). In Stern (2015), the unob-scured and Compton-thin AGNs are in good agreementwith this relation. The relation should also be appro-priate for CT AGNs, while the X-ray luminosities ofCT AGNs mentioned by Stern (2015) are much lowerthan their mid-IR luminosities. The reason is that theabsorption-corrected X-ray luminosities of CT AGNs aremore model-dependent. Thus their intrinsic X-ray lu-minosities are underestimated. Indeed, the method ofmid-IR excess is suitable to pre-select CT AGNs whoseintrinsic X-ray luminosities are strongly underestimated.The above mentioned are all radio-quiet AGNs. Forthe radio-loud AGNs, enhanced X-ray emission is usu-ally observed in radio-loud AGNs (e.g., Miller et al. 2011;Ballo et al. 2012) due to the contribution of the jets tothe total X-ray emission. While the mid-IR emission ofradio-loud AGNs (excluding blazars) is similar to that ofradio-quiet AGNs. The above method of mid-IR excesscan be also used to select CT AGN candidates of radio- loud. There are nine radio-loud AGNs in our sample,including 2 CT AGNs that are identified by this work.The radio-loud AGNs will less impact on the result ofmid-IR excess selection.3.2. CT AGN candidates
The left panel of Figure 1 shows that the mid-IRversus absorption-corrected X-ray luminosities. Theabsorption-corrected X-ray luminosities are estimatedwith a simple model (see Section 4.4 in Luo et al. 2017).The black line represents the relation of Equation 1. Weobtain the 1 σ uncertainty of the Stern (2015) relationwith using the data in Tables of Stern (2015). The grayshaded area is 1 σ dispersion for the Stern (2015) relation(also see Zou et al. 2020). The solid grey circles repre-sent the eight known CT AGNs. The dashed line is 3 σ lower limit. Most of the AGNs locate near the blackline with 1 σ area. However, there are still 12 AGNs be-low the 3 σ lower limit, four of which are the known CTAGNs. The remaining 8 AGNs are also likely to be CTAGNs (hereafter referred to as CT AGN candidates).The right panel of Figure 1 shows that the rest-frame6 µ m versus various absorption-corrected 2–10 keV lu-minosities of 8 known CT AGNs and 8 CT AGN candi-dates. Except for three known CT AGNs, the remainingfive known CT AGNs corrected by Liu et al. (2017), Liet al. (2019), or Corral et al. (2019) are near the blackline with 1 σ area. The absorption-corrected luminosi-ties of the three known CT AGNs are above 3 σ lowerlimit, suggesting that their absorption correction shouldbe reasonable. However, among 8 CT AGN candidates,only three candidates (XID 623, 802, and 844) are cor-rected by Li et al. (2019). Their absorption-corrected lu-minosities are still below the dashed line, suggesting thattheir column densities may be underestimated. Thereare still five candidates that are not corrected by previ-ous works (Liu et al. 2017; Li et al. 2019; Corral et al.2019). To determine whether these eight candidates areCT AGNs, we will perform X-ray diagnostics, opticalspectral diagnostics, and mid-IR diagnostics on theseeight candidates. MULTI-WAVELENGTH DIAGNOSTICS4.1.
X-ray diagnostics
The X-ray diagnostics is the most reliable identifica-tion method for CT AGN (e.g., Lanzuisi 2017). Li et al.(2019) have fitted the X-ray spectra of source XID 623,802, and 884 using MYTorus, but did not confirm themas CT AGN. We believe that it is difficult to obtain Radio-loud AGNs: XID 119, 175, 342, 587, 623, 746, 760, 844,and 867.
Guo et al.
43 44 45 46 log L (6 m) [erg s ] l o g L c o r ( - k e V ) [ e r g s ] l o w e r - li m i t Stern15my sampleKnown CT AGN 43 44 45 46 log L (6 m) [erg s ] l o g L c o r ( - k e V ) [ e r g s ] l o w e r - li m i t CT AGN candidateliu2017li2019Corral2019
Figure 1.
Left panel: Rest-frame 6 µ m versus absorption-corrected 2–10 keV luminosities for our AGN sample. The black linerepresents the relation for Stern (2015)(Equation 1). The gray shaded area is 1 σ dispersion, the dashed line is 3 σ lower limit.Open grey circles show our AGN sample, the solid grey circles represent the eight known CT AGNs. Right panel: Rest-frame6 µ m versus different absorption-corrected 2–10 keV X-ray luminosities for 8 known CT AGNs and 8 CT AGN candidates.Grey circles, blue square, green inverted triangles and red triangles represent different X-ray luminosities corrected by Luo et al.(2017), Liu et al. (2017), Li et al. (2019) and Corral et al. (2019). reliable evidence through X-ray spectroscopy fitting toprove that these eight candidates are CT AGNs. There-fore, we only intend to use the characteristics of theirX-ray spectra, in this section, to search for supportingevidence.We fit the X-ray spectra of these eight candidates inXspec v 12.11.0 (Arnaud 1996) using the Cash statis-tic that is more appropriate with the low count regime(Cash 1979). Due to the limited counts of these eightcandidates, we do not bin the spectrum because it mightlose some information. In order to obtain the character-istics of the X-ray spectrum at low-energy (2–10 keV)and high-energy ( >
10 keV), we use phabs * (power-law + zgauss) to fit the rest-frame 2–10 keV spectra,and use phabs * powerlaw to fit the rest-frame above10 keV spectra. The phabs accounts for the Galac-tic absorption, which is fixed at a column density of8 . × − cm − (Stark et al. 1992). The characteris-tics of the X-ray spectra of these eight candidates arederived by fitting their spectra. Their properties andthe characteristics of their X-ray spectra are listed inTable 2.Three candidates (XID 341, 623, and 844) have a flatspectrum with Γ < α emission line with the equivalentwidth of 0 .
234 keV. Therefore, only the source XID 623 is more likely a CT AGN in terms of its X-ray spectrumcharacteristics.4.2.
Optical spectral diagnostics
The torus of an AGN does not obscure the high-ionization narrow optical emission lines from the narrow-line regions (NLRs). As long as the unified schemeholds, narrow emission lines have been considered tobe a useful indicator of the AGN luminosity. There-fore, there should be a correlation between the narrowemission lines and X-ray luminosities, and the relation-ship has been investigated (e.g., Heckman et al. 2005).In this section, we use the high-ionization narrow opti-cal emission lines to identify whether these eight can-didates are CT AGNs. We try to collect the opticalspectra of these eight candidates from the literature(e.g., Szokoly et al. 2004; Mignoli et al. 2005; Inamiet al. 2017; Herenz et al. 2017), and finally we onlyfind two spectra of XID 640 and 889. The spectrumof the source XID 640 originates from Herenz et al.(2017), and XID 889 from Figure 6 (Source 175b) inSzokoly et al. (2004). There is only [O III] λ λ f [O III] = 1 . × − erg s − cm − . In addition, weestimate the flux of the high-ionization optical emissionlines ([O III] λ λ f [O III] = 8 . × − erg s − cm − and f [Ne V] = 1 . × − erg s − cm − , respectively.In this section, the observed X-ray luminosities are fromSection 4.1. elected CT AGN in CDFS Table 2.
The characteristics of X-ray spectra.
XID z L obs (2–10 keV) Γ(2–10 keV)
F e Kα
C-stat/d.o.f.(2–10 keV) Γ( >
10 keV) C-stat/d.o.f.( >
10 keV)[erg s − ](1) (2) (3) (4) (5) (6) (7) (8)341 1.83 8 . +2 . − . × . ± .
81 no 139.7/189 − . ± .
69 150.4/237342 0.76 4 . +2 . − . × . ± .
79 no 227.7/312 · · · · · ·
433 0.70 1 . +0 . − . × . ± .
40 no 142.8/319 · · · · · ·
623 3.88 6 . +1 . − . × − . ± .
83 yes 92.3/99 1 . ± .
77 237.1/340640 0.54 2 . +0 . − . × . ± .
64 no 119.2/353 · · · · · ·
802 2.41 5 . +2 . − . × . ± .
92 no 89.7/161 1 . ± .
35 137.8/278844 1.10 9 . +2 . − . × . ± .
36 no 244.3/264 0 . ± .
56 105.5/148889 0.52 1 . +0 . − . × . ± .
53 no 372.4/360 · · · · · ·
Note —(1) and (2) are XID and redshift. (3) is the apparent rest-frame 2–10 keV luminosity with 1 σ confidence derived by fitting X-rayspectra. (4) is photon index of rest-frame 2–10 keV. (5) represents whether there is a characteristic of the prominent iron K α emission linein the X-ray spectrum. (6) is C-stat/d.o.f. of the best fit for rest-frame 2–10 keV. (7) is photon index above 10 keV. (8) is C-stat/d.o.f.of the best fit above 10 keV. L obs (2-10 keV)/L [O III] 5007 l o g N H [ c m ] L obs (2-10 keV)/L [Ne V] 3426 l o g N H [ c m ]
889 68.3%90.0%
Figure 2.
Left panel: Rest-frame 2–10 keV to [O III] luminosity ratio as a function of N H . The cyan shaded regions correspondto 68 .
3% around the < X/[O III] > ratio (Maiolino et al. 1998). The solid red stars are for CT AGN candidates. Right panel:Rest-frame 2–10 keV to [Ne V] luminosity ratio as a function of N H . The cyan and pink shaded regions correspond to 68 . < X/[Ne V] > ratio (Gilli et al. 2010). Maiolino et al. (1998) provided a diagnostic based onthe ratio between the rest-frame 2–10 keV and the [O III]luminosities. The left panel of Figure 2 shows that theL obsX / L [O III] ratio as a function of N H . The cyan shadedregions correspond to ± σ around the < X/[O III] > ra-tio (Maiolino et al. 1998). Based on the informationin the left panel of Figure 2, we can derive the columndensities of these two candidates. There is a 68.3% prob-ability that the column density of the source XID 640is larger than 4 . × cm − . Similarly, the columndensity of the source XID 889 is larger than 10 cm − .Besides, Gilli et al. (2010) provided another diagnos- tic based on the < X/[Ne V] > ratio. The right panelof Figure 2 shows that the L obsX / L [NeV] ratio as a func-tion of N H . Based on the information in the right panelof Figure 2, we can derive that the column density ofthe source XID 889 is N H > × cm − with a 90%probability. Through the above optical diagnosis, wecan identify that the source XID 889 is a CT AGN.4.3. Mid-IR diagnostics
The column densities of AGNs can be estimatedby comparing the X-ray to mid-IR luminosities (e.g.,Lanzuisi et al. 2015b; Lansbury et al. 2015, 2017). Fig-
Guo et al.
43 44 45 46 log L (6 m) [erg s ] l o g L o b s ( - k e V ) [ e r g s ] N H = . × c m Fiore2009 relationmy sampleKnown CT AGNCT-AGN Candidates
Figure 3.
Rest-frame 2–10 keV observed X-ray versusrest-frame 6 µ m luminosity for our sample. The blackline represents the relation for Fiore et al. (2009). Thedashed line indicates the same relationship but where theX-ray luminosities are absorbed by a column density ofN H = 1 . × cm − . The open grey circles are our AGNsample and the solid grey circles represent CT AGNs identi-fied by X-ray. The solid red stars are the candidates of CTAGN. ure 3 shows the distribution of observed X-ray versus6 µ m luminosities for our sample. Except that the X-rayluminosities of eight candidates (solid red stars) are fromTable 2, the X-ray luminosities of the remaining AGNsare from Table 1. The black line represents the relationfrom Fiore et al. (2009). Assuming that the intrinsicpower-law spectrum has a photon index of 1.8 in the 2–10 keV energies range, the dashed line indicates the rela-tionship after being absorbed by N H = 1 . × cm − gas , which means that the sources below the dashedline should be CT AGNs. We find that the 7 knownCT AGNs are below the dashed line, indeed. Amongeight CT AGN candidates, seven of which are belowthe dashed line. More interesting thing is that theseseven candidates seem to have a higher column den-sity than the eight known CT AGNs. One exceptionis XID 844 sitting above the dashed line, suggestingthat this candidate may be not a CT AGN, but it isstill a heavily obscured AGN. The column density ofthis candidate is 1 . +0 . − . × cm − obtained by X-rayspectroscopy fitting (Li et al. 2019), and it may be un-derestimated compared to the result of Figure 3 (thecolumn density of XID 844 is about 10 cm − ). Be-sides, we find a source (the blue star in Figure 3, XID133) that is not selected as a CT AGN candidate, but Absorbed model : phabs. it is below the dashed line. The column density of thissource is 1 . +0 . − . × cm − (Li et al. 2019), its up-per limit is larger than 1 . × cm − . The columndensity of this source, which is estimated by comparingits X-ray to mid-IR luminosity, should be larger than1 . × cm − . Combining these two cases, we caninfer that the source XID 133 should be a CT AGN.Luckily, we identify hitherto unknown 8 CT AGNs bythe multi-wavelength approaches. The summary of themulti-wavelength diagnostics is listed in Table 3. Table 3.
The summary of the multi-wavelength diagnos-tics.
XID z DiagnosticsX-ray Optical Mid-IR(1) (2) (3) (4) (5)133 3.47 CT341 1.83 · · · · · ·
CT342 0.76 · · · · · ·
CT433 0.70 · · · · · ·
CT623 3.88 CT · · ·
CT640 0.54 · · · > × cm − (68.3%) CT802 2.41 · · · · · · CT844 1.10 · · · · · · non-CT889 0.52 · · ·
CT CT
Note —(1) and (2) are XID and redshift. (3), (4), and (5) areX-ray, optical spectrum, and mid-IR diagnostics. DISCUSSION5.1.
X-ray origin for CT AGNs
For the galaxies that host AGNs, the majority of theX-ray emission originates from the center AGNs, anda small amount of X-ray emission can be contributedby the X-ray binaries (XRBs) in their host galaxies. Ifthe central AGN is a CT AGN, then the X-ray emissionfrom the central AGN may be completely absorbed. Inorder to diagnose whether the X-ray emission of 8 newCT AGNs originates from central AGNs or the XRBpopulations, we compare the observed X-ray and theX-ray luminosities expected from XRB populations.The X-ray luminosity from XRB populations is com-posed of low-mass X-ray binaries (LMXBs) and high-mass X-ray binaries (HMXBs). Previous studies haveshown that the total X-ray luminosity from the LMXBand HMXB population is proportional to M ∗ and SFR ofthe host galaxy, respectively (e.g., Gilfanov et al. 2004;Lehmer et al. 2016). Lehmer et al. (2016) provided the elected CT AGN in CDFS XRB = L
LMXB + L
HMXB = α (1 + z) γ M ∗ + β (1 + z) δ SFR , where log α = 29 .
30, log β = 39 . γ = 2 .
19, and δ = 1 .
02 for 2–10 keV, respectively.
40 41 42 43 log L obs (2-10 keV) [erg s ] l o g L o b s / L X R B Figure 4.
Observed 2–10 keV X-ray luminosity over 2–10keV X-ray luminosity from XRB populations (L obs / L XRB )vs. observed 2–10 keV X-ray luminosity (L obs ) for the 8 newCT AGNs. The gray shaded area shows the expected 1 σ dispersion of the derived L obs / L XRB (0.17 dex, Lehmer et al.2016). The dashed line is L obs / L XRB = 2.
Figure 4 presents the observed X-ray over XRB X-ray luminosities as a function of observed X-ray lumi-nosities. The gray shaded area shows the expected 1 σ dispersion of the derived L obs / L XRB (0.17 dex, Lehmeret al. 2016). There are two CT AGNs (XID 341 and 433)that locate in the gray shaded area, indicating that theirobserved X-ray emission may be entirely contributed bythe XRBs in their host galaxies. The dashed line isL obs / L XRB = 2, indicating that the X-ray emission fromthe center AGNs is equal to that from the XRBs. TheCT AGNs XID 640 and 802 are near the dashed line.However, their lower limits are lower than or equal tothose expected from XRB populations. We can’t ex-clude the possibility that their X-ray emission is fromthe XRB populations. There are still four remainingCT AGNs (XID 133, 342, 623, and 889) that are abovethe dashed line, suggesting that the majority of theirX-ray emission still originates from the center AGNs.However, the lower limits of XID 342 and 889 are belowthe dashed line. We cannot rule out the possibility thatthe emission of the XRB populations dominates the X-ray emission of these two AGNs. Moreover, the X-ray spectrum analysis in Section 4.1 seems to assess whetherthe observed X-ray emssion is produced by the AGNs.The measured photon indices of four CT AGNs (XID342, 433, 640, and 802) are Γ >
3. However, the gen-erally measured photon index for AGN is between 1.4and 2.6, suggesting that their observed X-ray emissionshould not be dominated by (or come from) the AGNs.Combining the above information, we can infer the X-ray origin of these eight CT AGNs, and the conclusionsare as follows: (1) The observed X-ray emission of twoCT AGNs (XID 133 and 623) should be dominated bythe center AGNs. (2) The observed X-ray emission oftwo CT AGNs (XID 341 and 433) may come from theXRB populations. (3) The observed X-ray emission ofthree CT AGNs (XID 342, 640, and 802) should not bedominated by the AGNs. (4) There is still a CT AGN(XID 889) whose X-ray emission is dominated by cen-tral AGNs with a strong possibility. However, it cannotexclude the possibility that the emission of the XRBpopulations dominates its X-ray emission.5.2.
Why the N H is underestimated? In Section 5.1, we have demonstrated that the X-rayemission of the two CT AGNs (XID 133 and 623) mainlyoriginates from the central AGN. The column density ofXID 133 obtained by the X-ray spectrum (Li et al. 2019)is reasonable within the error range (see Section 4.3 fordetailed). While the XID 623 is a high-redshift source(z = 3.88), its rest-frame energies are mainly above 10keV. One reason may be that the X-ray spectra above10 keV are not significantly absorbed. Moreover, theXID 623 is a radio-loud AGN, so its jet will provide apart of the X-ray emission. Previous studies (Luo et al.2017; Li et al. 2019) did not consider the contributionof the jet, thus underestimated its column density.The X-ray emission of two CT AGN (XID 341 and433) might originate from the XRBs in the host galax-ies and suggested that their X-ray emission originatedfrom AGNs was entirely absorbed. Therefore, the col-umn densities of the center AGNs cannot be obtained byfitting their X-ray spectra. Similarly, the column densityof the three AGNs (XID 342, 640, and 802) cannot bealso obtained, because their X-ray emission is not dom-inated by AGNs. However, Luo et al. (2017) found thatthe X-ray spectra of those five CT AGNs presented weakabsorbed or non-absorbed. Thus the column densitiesof these five CT AGNs are severely underestimated.The X-ray emission of XID 889 is uncertain whetherit is dominated by the AGN. If the AGN does not dom-inate the X-ray emission, then the reason for its under-estimation of column density should be similar to theabove five CT AGNs. If the AGN dominates, its X-ray0
Guo et al. emission may be the soft scattered component of thepolar region, while the X-ray emission from the nucleusis hidden. However, Luo et al. (2017) misunderstoodthis source as a weakly absorbing AGN. Thus it columndensity is severely underestimated.5.3.
Selection efficiency of CT AGNs
In the deep field survey, only about 10% of AGNscan be identified as CT AGN by X-ray spectroscopyfitting, showing the efficiency is low. In our sample,eight (15.6%) CT AGNs have been confirmed by X-rayspectroscopy fitting. The seven known CT AGNs arealso confirmed in this work (see Section 4.3 for details).Besides, we identify newly 8 (15.6%) CT AGNs by themulti-wavelength approaches. The multi-wavelength ap-proaches of selecting CT AGNs have a higher efficiencythan the X-ray spectroscopy fitting. However, the multi-wavelength approaches require that a source has multi-wavelength data and that its AGN component decom-posed by the SED is reliable, so our sample is only asmall part of the AGNs in the CDFS survey. In otherwords, the multi-wavelength approach cannot be appliedto all AGNs in the CDFS survey. In our sample, 16sources can be identified as CT AGN, i.e. the fractionof CT AGNs is 31.2%, which agrees well with the the-oretical expectation (e.g., Ricci et al. 2015; Lansburyet al. 2017).5.4.
Accretion of central BH of CT AGNs
CT AGNs usually have a higher accretion rate thannon-CT AGNs (e.g., Draper & Ballantyne 2010). Here,we compare the Eddington ratio λ Edd = L bol / L Edd ofCT AGNs and non-CT AGNs in our sample. WhereL bol is the bolometric luminosity, L
Edd is Edding-ton luminosity which is related to BH mass (i.e.,L
Edd = 1 . × M BH / M (cid:12) erg s − ).The bolometric luminosity is the total luminosityemitted at all wavelengths by the AGN (e.g., Krawczyket al. 2013; Duras et al. 2020). The AGN luminosityobtained by Guo et al. (2020) is the integrated lumi-nosity from UV to IR. The luminosity from UV to IRfar exceeds the luminosity of other wavelengths. There-fore, for Type 1 AGNs, we use the AGN luminositiesinstead of their bolometric luminosities. Due to Type 2AGN emission in the UV/optical bands is known to beobscured, we estimate their bolometric luminosities byintegrating the AGN emission component in the range1–1000 µ m and simply rescaling the result by a factor of1.7 (Pozzi et al. 2007; Zappacosta et al. 2018). There are6 AGNs with broad optical emission lines in their spec-tra. Therefore, we estimated the BH masses of these sixAGNs with the broad emission lines (H β or Mg II, Net-zer & Marziani 2010; Trakhtenbrot & Netzer 2012). The BH masses of other AGNs are estimated to use M BH –M ∗ relation with a scaling of 0.003 (Suh et al. 2020). z l o g E dd CT AGNnon-CT AGN 2 4 6 8CT AGNnon-CT AGN
Figure 5.
Left panel: Eddington ratio vs. redshift for oursample. The solid red stars indicate the 16 CT AGNs, andthe open grey circles represent non-CT AGNs. The dottedline is λ Edd = 1. Right panel: The distribution of Eddingtonratio for our sample. The red filled histogram is for theEddington ratio of CT AGNs, and the gray histogram is forthe Eddington ratio of non-CT AGNs.
The left panel of Figure 5 presents the Eddington ratioversus redshift for our sample. Except for 3 CT AGNs(XIDs 133, 623, and 802), the Eddington ratios of theremaining AGNs are lower than 1. Interestingly, thesethree CT AGNs of super-Eddington accretion are allnewly identified in this work. The multi-wavelength ap-proaches seem to prefer the selection of CT AGN witha high Eddington ratio. The right panel of Figure 5shows the distributions of the Eddington ratio for CTAGNs and non-CT AGNs. We use the Kolmogorov–Smirnov test (KS-test) to examine their distributions ofEddington ratio: the p-value is 0.010, which suggeststhat their distributions are different. The Eddingtonratios of CT AGNs are significantly higher than thoseof non-CT AGNs. Our results support that the BHmasses of CT AGNs are more rapid growth than thoseof non-CT AGNs and that CT AGNs may be a rapidgrowth phase of BH masses in the evolutionary scenarioof AGNs (Goulding et al. 2011).5.5.
The properties of CT AGN host galaxies
Some studies suggested that the occurrence of CTAGNs is related to the properties of their host galaxies(e.g., Ricci et al. 2017). Goulding et al. (2012) pointedout that the host galaxies of CT AGNs in the nearbyuniverse show a high level of star formation. Therefore,we compared the SFRs and stellar masses of the hostgalaxies of CT AGNs and non-CT AGNs in our sample.Figure 6 presents the relation between SFRs and stellar elected CT AGN in CDFS log M * [M ] l o g S F R [ M y r ] z=0.5z=2non-CT AGNCT AGN0.51.0 CT AGNnon-CT AGN 0.5 1.0 1.5 Figure 6.
SFR vs. M ∗ distribution for our sample. Forcomparison, the data points of 16 CT AGNs (solid red stars)and non-CT AGNs (open gray circles) are plotted. The yel-low and blue lines show the main sequences of star formationat z ∼ . z ∼
2, respectively (Schreiber et al. 2015). masses for the host galaxies. We also use the KS-testto examine their stellar mass distributions, and the p-value is 0.87. We repeat the KS-test to examine theirSFR distributions, and the p-value is 0.97. These sug-gest that the properties of CT AGN host galaxies aresimilar to those of non-CT AGN host galaxies in oursample. Our results are against the claim of Gouldinget al. (2012), and the main reason may be that mostof the sources in our sample are high redshift AGNs.The environment of AGN host galaxies at high-redshiftmay be different from the local universe. In addition,our sample contains fewer CT AGNs, we cannot ruleout our results are biased in sample selection. In sub-sequent work, we will expand the CT AGN sample andfurther study whether there are differences in the hostgalaxies of both CT AGN and non-CT AGN. SUMMARYIn this work, we find out missed CT AGNs fromthe CDFS survey using the multi-wavelength techniques and discussed their properties. Firstly, we constructa sample containing 51 AGNs with abundant multi-wavelength data. We select hitherto unknown 8 CTAGN candidates using the method of the mid-IR ex-cess. Diagnosis based on the characteristics of their X-ray spectra, only the X-ray spectrum of source XID 623shows all characteristics of CT AGNs. Through the opti-cal spectrum diagnosis, the source XID 889 can be iden-tified as CT AGN, while the source XID 640 can onlyobtain a lower limit of the column density. Except forthe source XID 844, the remaining seven candidates canbe identified as CT AGN in the mid-infrared diagnosis.Besides, a new CT AGN (XID 133) is also found in sec-tion 4.3. Subsequently, we discuss the X-ray origin ofthese eight CT AGNs. Except that the X-ray emissionof the two sources (XID 133 and 623) is still from thecentral AGNs, most or all of the X-ray emission of the re-maining six sources originates from the XRB populationin their host galaxies. We also discuss the reason whytheir column densities were underestimated in previousstudies. We find that the multi-wavelength approachesof selecting CT AGNs are highly efficient, provided thehigh quality of observation data. Finally, we compareCT AGNs and non-CT AGNs in our sample. We findthat CT AGNs have a higher Eddington ratio than non-CT AGNs, and that both CT AGNs and non-CT AGNsshow similar properties of host galaxies.ACKNOWLEDGMENTSWe sincerely thank the anonymous referee for usefulsuggestions. We thank Bin Luo for helpful discussions.We acknowledge financial support from the NationalKey R&D Program of China grant 2017YFA0402703(Q.S.Gu) and National Natural Science Foundation ofChina grant 11733002 (Q.S.Gu). This research makesuse of data from
Chandra
Deep Field-South Survey. Weacknowledge the extensive use of the following Pythonpackages:
Software : numpy, pandas (Wes McKinney 2010), mat-plotlib (Hunter 2007), astropy(Astropy Collaborationet al. 2013), Code Investigating GALaxy Evolution (Bo-quien et al. 2019).REFERENCES
Akylas, A., Georgakakis, A., Georgantopoulos, I.,Brightman, M., & Nandra, K. 2012, A&A, 546, A98Arnaud, K. A. 1996, Astronomical Society of the PacificConference Series, Vol. 101, XSPEC: The First TenYears, ed. G. H. Jacoby & J. Barnes, 17 Astropy Collaboration, Robitaille, T. P., Tollerud, E. J.,et al. 2013, A&A, 558, A33Ballo, L., Heras, F. J. H., Barcons, X., & Carrera, F. J.2012, A&A, 545, A66 Guo et al.
Boquien, M., Burgarella, D., Roehlly, Y., et al. 2019, A&A,622, A103Cappi, M., Panessa, F., Bassani, L., et al. 2006, A&A, 446,459Carilli, C. L., & Walter, F. 2013, ARA&A, 51, 105Cash, W. 1979, ApJ, 228, 939Chen, C.-T. J., Hickox, R. C., Goulding, A. D., et al. 2017,ApJ, 837, 145Comastri, A. 2004, Astrophysics and Space Science Library,Vol. 308, Compton-Thick AGN: The Dark Side of theX-Ray Background, ed. A. J. Barger, 245Corral, A., Georgantopoulos, I., Akylas, A., & Ranalli, P.2019, A&A, 629, A133Draper, A. R., & Ballantyne, D. R. 2010, ApJL, 715, L99Duras, F., Bongiorno, A., Ricci, F., et al. 2020, A&A, 636,A73Fiore, F., Puccetti, S., Brusa, M., et al. 2009, ApJ, 693, 447Gandhi, P., Horst, H., Smette, A., et al. 2009, A&A, 502,457Gilfanov, M., Grimm, H. J., & Sunyaev, R. 2004, MNRAS,347, L57Gilli, R., Comastri, A., & Hasinger, G. 2007, A&A, 463, 79Gilli, R., Vignali, C., Mignoli, M., et al. 2010, A&A, 519,A92Goulding, A. D., Alexander, D. M., Bauer, F. E., et al.2012, ApJ, 755, 5Goulding, A. D., Alexander, D. M., Mullaney, J. R., et al.2011, MNRAS, 411, 1231Guo, X., Gu, Q., Ding, N., Contini, E., & Chen, Y. 2020,MNRAS, 492, 1887Harrison, F. A., Craig, W. W., Christensen, F. E., et al.2013, ApJ, 770, 103Heckman, T. M., Ptak, A., Hornschemeier, A., &Kauffmann, G. 2005, ApJ, 634, 161Herenz, E. C., Urrutia, T., Wisotzki, L., et al. 2017, A&A,606, A12Hunter, J. D. 2007, Computing in Science and Engineering,9, 90Inami, H., Bacon, R., Brinchmann, J., et al. 2017, A&A,608, A2Kammoun, E. S., Miller, J. M., Zoghbi, A., et al. 2019,ApJ, 877, 102Kellermann, K. I., Fomalont, E. B., Mainieri, V., et al.2008, ApJS, 179, 71Kocevski, D. D., Brightman, M., Nandra, K., et al. 2015,ApJ, 814, 104Krawczyk, C. M., Richards, G. T., Mehta, S. S., et al. 2013,ApJS, 206, 4La Caria, M. M., Vignali, C., Lanzuisi, G., Gruppioni, C.,& Pozzi, F. 2019, MNRAS, 487, 1662 LaMassa, S. M., Yaqoob, T., Boorman, P. G., et al. 2019,ApJ, 887, 173Lambrides, E. L., Chiaberge, M., Heckman, T., et al. 2020,ApJ, 897, 160Lansbury, G. B., Gandhi, P., Alexander, D. M., et al. 2015,ApJ, 809, 115Lansbury, G. B., Alexander, D. M., Aird, J., et al. 2017,ApJ, 846, 20Lanzuisi, G. 2017, Astronomische Nachrichten, 338, 316Lanzuisi, G., Piconcelli, E., Fiore, F., et al. 2009, A&A,498, 67Lanzuisi, G., Ranalli, P., Georgantopoulos, I., et al. 2015a,A&A, 573, A137Lanzuisi, G., Perna, M., Delvecchio, I., et al. 2015b, A&A,578, A120Lanzuisi, G., Civano, F., Marchesi, S., et al. 2018, MNRAS,480, 2578Lehmer, B. D., Basu-Zych, A. R., Mineo, S., et al. 2016,ApJ, 825, 7Li, J., Xue, Y., Sun, M., et al. 2019, ApJ, 877, 5Liu, T., Tozzi, P., Wang, J.-X., et al. 2017, ApJS, 232, 8Luo, B., Brandt, W. N., Xue, Y. Q., et al. 2011, ApJ, 740,37—. 2017, ApJS, 228, 2Lutz, D., Maiolino, R., Spoon, H. W. W., & Moorwood,A. F. M. 2004, A&A, 418, 465Maiolino, R., Salvati, M., Bassani, L., et al. 1998, A&A,338, 781Marchesi, S., Civano, F., Elvis, M., et al. 2016, ApJ, 817, 34Masini, A., Civano, F., Comastri, A., et al. 2018, ApJS,235, 17Mignoli, M., Cimatti, A., Zamorani, G., et al. 2005, A&A,437, 883Miller, B. P., Brandt, W. N., Schneider, D. P., et al. 2011,ApJ, 726, 20Miller, N. A., Fomalont, E. B., Kellermann, K. I., et al.2008, ApJS, 179, 114Miller, N. A., Bonzini, M., Fomalont, E. B., et al. 2013,ApJS, 205, 13Netzer, H., & Marziani, P. 2010, ApJ, 724, 318Planck Collaboration, Aghanim, N., Akrami, Y., et al.2020, A&A, 641, A6Pouliasis, E., Mountrichas, G., Georgantopoulos, I., et al.2020, MNRAS, 495, 1853Pozzi, F., Vignali, C., Comastri, A., et al. 2007, A&A, 468,603Ricci, C., Ueda, Y., Koss, M. J., et al. 2015, ApJL, 815, L13Ricci, C., Bauer, F. E., Treister, E., et al. 2017, MNRAS,468, 1273 elected CT AGN in CDFS13