1: %%
2: %\documentclass[12pt,referee]{/home/fiore/testi/macros/aastex}
3: %\documentclass[12pt,preprint]{/home/fiore/testi/macros/aastex}
4: %\documentclass[12pt,preprint]{aastex}
5:
6: %% manuscript produces a one-column, double-spaced document:
7:
8: %% \documentclass[manuscript]{aastex}
9:
10: %% preprint2 produces a double-column, single-spaced document:
11:
12: %\documentclass[preprint2]{aastex}
13: %\documentclass{/home/fiore/testi/macros/emulateapj}
14: \documentclass{emulateapj}
15:
16:
17: %\newcommand{\vdag}{(v)^\dagger}
18: %\newcommand{\myemail}{skywalker@galaxy.far.far.away}
19:
20: %% You can insert a short comment on the title page using the command below.
21:
22: %\slugcomment{Not to appear in Nonlearned J., 45.}
23:
24: \shorttitle{Obscured AGNs in the CDFS}
25: \shortauthors{Fiore et al.}
26:
27: \def\ls{{_<\atop^{\sim}}}
28: \def\gs{{_>\atop^{\sim}}}
29: \def\cgs{ ${\rm erg~cm}^{-2}~{\rm s}^{-1}$ }
30:
31:
32: \begin{document}
33:
34: \title{Unveiling obscured accretion in the Chandra Deep Field South}
35: \author{F. Fiore\altaffilmark{1}, A. Grazian\altaffilmark{1},
36: P. Santini\altaffilmark{1,2}, S. Puccetti\altaffilmark{3},
37: M. Brusa\altaffilmark{4}, C. Feruglio\altaffilmark{1},
38: A. Fontana\altaffilmark{1}, E. Giallongo\altaffilmark{1},
39: A. Comastri\altaffilmark{5}, C. Gruppioni\altaffilmark{5},
40: F. Pozzi\altaffilmark{6}, G. Zamorani\altaffilmark{5},
41: C. Vignali\altaffilmark{6}}
42: \altaffiltext{1}{INAF-OAR, via Frascati 33, Monteporzio, I00040, Italy}
43: \altaffiltext{2}{Universita' di Roma La Sapienza, Italy}
44: \altaffiltext{3}{ASI Science data Center, via Galileo Galilei,
45: 00044 Frascati, Italy}
46: \altaffiltext{4}{Max Planck Institut f\"ur extraterrestrische Physik,
47: Giessenbachstrasse 1, D--85748 Garching bei M\"unchen, Germany}
48: \altaffiltext{5}{INAF-OABo, via Ranzani 1, Bologna, Italy}
49: \altaffiltext{6}{Universita' di Bologna, via Ranzani 1, Bologna, Italy}
50:
51: \email{fiore@oa-roma.inaf.it}
52:
53:
54: \begin{abstract}
55: We make use of deep HST, VLT, Spitzer and Chandra data on the Chandra
56: Deep Field South to constrain the number of Compton thick AGN in this
57: field. We show that sources with high 24$\mu$m to optical flux ratios
58: and red colors form a distinct source population, and that their
59: infrared luminosity is dominated by AGN emission. Analysis of the
60: X-ray properties of these extreme sources shows that most of them
61: (80$\pm15\%$) are indeed likely to be highly obscured, Compton thick
62: AGNs. The number of infrared selected, Compton thick AGNs with
63: 5.8$\mu$m luminosity higher than $10^{44.2}$ erg s$^{-1}$ turns out to
64: be similar to that of X-ray selected, unobscured and
65: moderately obscured AGNs with 2-10 keV luminosity higher than
66: $10^{43}$ erg s$^{-1}$ in the redshift bin 1.2-2.6. This ``factor of
67: 2'' source population is exactly what it is needed to solve the
68: discrepancies between model predictions and X-ray AGN selection.
69:
70: \end{abstract}
71:
72: %% Keywords should appear after the \end{abstract} command. The uncommented
73: %% example has been keyed in ApJ style. See the instructions to authors
74: %% for the journal to which you are submitting your paper to determine
75: %% what keyword punctuation is appropriate.
76:
77: \keywords{Active Galactic Nuclei}
78:
79: \section{Introduction}
80:
81: Active Galactic Nuclei (AGN) are not only witnesses of the phases of
82: galaxy formation and/or assembly, but are most likely among leading
83: actors. Indeed, three seminal discoveries indicate tight links and
84: feedbacks between super-massive black holes (SMBH), nuclear activity
85: and galaxy evolution. The first is the discovery of SMBH in the center
86: of most nearby bulge dominated galaxies, and the tight correlation
87: between their masses and galaxy bulge properties (Gebhardt et
88: al. 2000, Ferrarese \& Merritt 2000, Marconi \& Hunt 2003 and
89: references therein). The second is that the growth of SMBH is mostly
90: due to accretion of matter during their active phases, and therefore
91: that most bulge galaxies passed a phase of strong nuclear activity
92: (Soltan 1982, Marconi et al. 2004). The third is that the evolution
93: of AGN is luminosity dependent, with lower luminosity AGN peaking at a
94: redshift lower than luminous QSOs (Hasinger 2003, 2005, Fiore et
95: al. 2003, Ueda et al. 2003, La Franca et al. 2005, Brandt \& Hasinger
96: 2005, Bongiorno et al. 2007), a bimodal behavior recalling the
97: evolution of star-forming galaxies and that of massive spheroids
98: (Cowie et al. 1996, Franceschini et al. 1999, De Lucia et al. 2006).
99: All three discoveries imply that obtaining a complete census of
100: accreting SMBH through the cosmic epochs and constraining accretion
101: efficiency and feedbacks are crucial steps toward the understanding of
102: Galaxy formation and evolution.
103:
104:
105: First attempts to constrain models for the formation and evolution of
106: structure in the Universe using the evolving optical and X-ray AGN
107: luminosity functions have been presented by Granato et al. (2001,
108: 2004), Di Matteo et al. (2005), and Menci et al. (2004, 2005). In
109: particular, the Menci et al. model links the evolution of the galaxies
110: in the hierarchical clustering scenarios with the changing accretion
111: rates of cold gas onto the central SMBH that powers the QSO (Cavaliere
112: \& Vittorini 2000).
113: The results of this model were encouraging, in the sense that it
114: predicts a trend of lower luminosity AGN to peak at increasingly lower
115: redshift, as observed. However, from a quantitative point of view, the
116: model over-predicts by a factor of about 2 the space density of
117: low-to-intermediate luminosity (Seyfert like) AGNs at z=1.5-2.5 with
118: respect to present X-ray observations. Furthermore, Marconi
119: et. al. (2004, 2007 in preparation) derived a SMBH mass function from
120: the X-ray selected AGN luminosity functions (e.g. La Franca et
121: al. 2005) that falls short by a factor of about 2 to the ``relic''
122: SMBH mass function, evaluated using the M$_{BH} - \sigma_V$ / M$_{BH}
123: -$M$_B$ relationships and the local bulge's luminosity function. The
124: most likely explanation for both discrepancies is that present X-ray
125: surveys, although very efficient to probe unobscured and moderately
126: obscured AGN (with column densities up to a few $10^{23}$ cm$^{-2}$,
127: the so-called Compton thin AGNs), miss most of the very highly
128: obscured, but still strongly accreting objects, the so called Compton
129: Thick AGNs, with a column density N$_H\gs10^{24}$ cm$^{-2}$ (see
130: Comastri 2004). Indeed, only a handful of the faintest sources in the
131: Chandra deep fields may be Compton thick (see La Franca et al. 2005
132: and Tozzi et al. 2006). So we still may be viewing just the tip of
133: the iceberg of the highly obscured AGN population. Compton thick
134: objects may well be more common at high redshift, as suggested on
135: theoretical ground by Silk \& Rees (1998) and Fabian (1999) and on
136: observational ground by e.g. Gilli et al. (2001), Worsley et
137: al. (2004, 2006), and La Franca et al. (2005).
138:
139: Compton thick AGN at z$\gs$1 can be recovered thanks to the
140: reprocessing of the AGN UV emission in the infrared by selecting
141: sources with AGN luminosities in the mid-infrared and faint
142: near-infrared and optical emission (e.g. Martinez-Sansigre et
143: al. 2005, 2006, Houck et al. 2005, Weedman et al. 2006a, 2006b). We
144: investigate further this issue making use of the multiwavelength data
145: obtained on the Chandra Deep Field South (Giacconi et al. 2002), one
146: of the fields with the deepest coverage at optical, infrared and X-ray
147: wavelengths.
148:
149: A $H_0=70$ km s$^{-1}$ Mpc$^{-1}$, $\Omega_M$=0.3,
150: $\Omega_{\Lambda}=0.7$ cosmology is adopted throughout.
151:
152: \section{Datasets and sample selection}
153:
154: \begin{figure}
155: \includegraphics[height=8.5truecm,width=8.5truecm]{fig1.eps}
156: \caption{$\lambda L_\lambda$ spectral energy distributions of the four
157: additional templates of sources hosting highly obscured AGNs used in this work.
158: From top to bottom: NGC6240; IRAS09104+41091;
159: HELLAS2XMM A2690\_75 and BPM16274\_69}
160: \label{seds}
161: \end{figure}
162:
163: The selection and spectroscopic identification of complete AGN samples
164: from mid-IR surveys is a rather difficult task, because AGNs make only
165: a small fraction of the full mid-IR source population. Using Spitzer
166: IRAC colors results in samples significantly contaminated by
167: star-forming galaxies (e.g. Lacy et al. 2004, Alonso-Herrero et
168: al. 2006, Barmby et al. 2006, Polletta et al. 2006). Comparing the
169: global observed Spectral Energy Distribution (SED) to AGN and galaxy
170: templates proved to be more efficient in selecting AGN samples
171: (Polletta et al. 2006, 2007). However, it is very difficult to assess
172: the completeness of these samples given all complex selection
173: effects. Furthermore, many obscured AGN may still be anyway missed by
174: this technique.
175:
176: We adopt in this paper a somewhat different approach. We do not
177: pretend to select {\it all} AGN through optical and infrared
178: photometry. As explained before X-rays are much more efficient in
179: selecting unobscured (i.e. broad line AGNs) and moderately obscured
180: AGNs. We concentrate our effort on highly obscured AGN only and we
181: limit our analysis to the high infrared luminosity AGN population. The
182: driving consideration is that differences between nuclear and
183: star-formation emission are emphasized comparing the observed SEDs
184: with galaxy templates over a range as broad as possible (also see
185: Martinez-Sansigre et al. 2005,2006, Houck et al. 2005 and Weedman et
186: al. 2006, Magliocchetti et al. 2007). Our primary objective here
187: is to validate our obscured AGN selection criteria and assess the
188: magnitude of the corresponding selection effects using a careful
189: analysis of the deep X-ray data available on the CDFS.
190:
191: \begin{figure*}[t]
192: \begin{tabular}{cc}
193: \includegraphics[height=8.5truecm,width=8.5truecm]{fig2a.eps}
194: \includegraphics[height=8.5truecm,width=8.5truecm]{fig2b.eps}
195: \end{tabular}
196: \caption{Left panel: F($24\mu$m)/F(R) as a function of the 5.8$\mu$m
197: luminosity for three X-ray source samples (GOODS-MUSIC, red symbols,
198: ELAIS-S1, green symbols and HELLAS2XMM, large blue symbols , Pozzi et
199: al. 2007). Open circles = type 1 AGN; filled circles = non type 1
200: AGN; stars = photometric redshifts. Note that F($24\mu$m)/F(R) of non
201: broad line AGN is strongly correlated with the luminosity at
202: 5.8$\mu$m. Right panel: F($24\mu$m)/F(R) as a function of the X-ray to
203: optical flux ratio X/O for the same source samples. All symbols as in left
204: panel; skeleton triangles = sources without optical counterpart.}
205: \label{mirolir}
206: \end{figure*}
207:
208: \subsection{The GOODS-MUSIC catalog}
209:
210: We use in this paper the latest version of the GOODS-MUSIC catalog
211: (Grazian et al. 2006). We limit the analysis to the region fully
212: covered by deep VLT/ISAAC near infrared photometry (143.2 arcmin$^2$)
213: and to the sources with MIPS 24$\mu$m fluxes F(24$\mu$m)$>40\mu$Jy
214: (1729 sources). 24$\mu$m fluxes for all objects
215: in the catalog, were obtained following the procedures described in De
216: Santis et al. (2007).
217:
218: The published version of the catalog contains z band and K band
219: selected sources in the GOODS-south area. The revised version of the
220: catalog used in this work includes 46 objects that are detected only
221: in the 4.5$\mu$m band, i.e. their z and K magnitudes are below the
222: chosen detection threshold. Only 4 of these sources do not have
223: counterparts in the optical and/or in the near infrared images. At
224: the flux limits adopted here, we do not detect any objects at 24$\mu$m
225: that is not detected at shorter wavelengths. In the following we use for each
226: entry of the catalog Vega magnitudes and cgs fluxes.
227:
228: The GOODS-MUSIC catalog includes both monochromatic and total
229: infrared luminosities, for the sources with a reliable spectroscopic
230: or photometric redshifts. Total 8--1000$\mu$m luminosities were
231: computed by integrating the best fit galaxy and AGN templates.
232: Monochromatic luminosities were computed by interpolating the observed
233: SEDs at the rest frame wavelength of interest. A large (factor of
234: 10-30) systematic uncertainty is associated to the total infrared
235: luminosity, being this dominated by the contributions at wavelengths
236: of $\sim100-1000\mu$m, well outside the infrared band used in this
237: paper (1-24 $\mu$m). Different models can produce similar fits
238: below 24$\mu$m but give rise to large differences in the total
239: infrared luminosity. This systematic uncertainty is greatly reduced
240: using the monochromatic luminosity at 5.8 $\mu$m (see e.g. Yan et
241: al. 2007), since this wavelength is within the observed band up to
242: z$\sim$3.2. For this reason in the following we make use of the
243: infrared luminosity at 5.8$\mu$m ($\lambda L_\lambda (5.8\mu$m)) to
244: characterize the infrared power of the CDFS sources.
245:
246: \subsection{Galaxy and AGN templates}
247:
248: A detailed fitting of the observed SEDs using AGN and galaxy templates
249: was also performed. Templates include passive galaxies, star-forming
250: galaxies, unobscured AGNs and highly obscured AGNs. Spectral
251: libraries include both empirical templates (Coleman et al. 1980 datasets,
252: Polletta et al. 2007) and synthetic models (Bruzual \& Charlot 2003,
253: Fioc \& Rocca-Volmerange 1997). We put particular care on the
254: description of highly obscured AGNs. In addition to the templates
255: presented by Polletta et al. (2007) we used four additional templates
256: of sources hosting highly obscured AGNs (figure \ref{seds}). The NGC6240
257: template includes U, B, V, J, H, K, IRAS and ISO
258: photometry. The IRAS09104+41091 includes SDSS u, g, r, i, z
259: photometry, B, V, R, J, H, K photometry, Spitzer IRAC and MIPS
260: photometry, Spitzer IRS spectroscopy and IRAS photometry. For the two
261: HELLAS2XMM sources we use the best fit SED model in Pozzi et
262: al. (2007). These templates span a range of infrared to optical flux
263: ratio significantly broader than the obscured AGN templates used by
264: Polletta el al. (2007).
265:
266: \subsection {Photometric redshifts and SED fittings}
267:
268: Photometric redshifts were derived by Grazian et al. (2006) fitting
269: only the part of the SED dominated by the integrated stellar
270: population, i.e. $\lambda<5.5 \mu$m (Grazian et al. 2006 and
271: references therein). The synthetic models used are very accurate and
272: complete in the treatment of the star-formation history and evolution
273: of the stellar populations and of the dust content of the galaxy and
274: its evolution. However, while dust extinction is easy to account for
275: because it is a line of sight effect, dust emission and reprocessing
276: is much more complicated to model, since it depends largely on the
277: assumed geometry and covering fraction. Dust emission was therefore
278: not considered in the models, and accordingly the bands above
279: $5.5\mu$m rest frame are ignored in these fits (see Grazian et
280: al. 2006 for further details). This approach produces robust
281: photometric redshifts ($\Delta z/(1+z) <0.05$) for passive galaxies,
282: star-forming galaxies and obscured AGNs, where the nuclear optical and
283: infrared emission is completely blocked or strongly reduced by dust
284: and gas along the line of sight. Unobscured AGNs with power law SEDs
285: are excluded from this analysis (however, most of them have reliable
286: spectroscopic redshift from Cimatti et al. 2002, Szokoly et al. 2004,
287: Le Fevre et al. 2004, Vanzella et al.2005,2006, Mignoli et al. 2005).
288:
289: To characterize each SED we repeated the fit to the observed SED with
290: a library of empirical templates, fixing the redshift to the
291: spectroscopic redshift or, if this is not present, to the photometric
292: redshift obtained as described above. This library includes the 21
293: templates of passive galaxies, star-forming galaxies and AGNs by
294: Polletta et al. (2007) and the four templates of systems hosting a
295: highly obscured AGN in figure \ref{seds}. The best fit template and
296: normalization were used to measure monochromatic and total infrared
297: luminosities.
298:
299: \begin{figure*}[t!]
300: \begin{tabular}{cc}
301: \includegraphics[height=8.5truecm,width=8.5truecm]{fig3a.eps}
302: \includegraphics[height=8.5truecm,width=8.5truecm]{fig3b.eps}
303: \end{tabular}
304: \caption{Left panel: F($24\mu$m)/F(R) as a function of the R-K
305: color for two X-ray source samples (GOODS-MUSIC and HELLAS2XMM, large
306: symbols, from Pozzi et al. 2007). Open circles = type 1 AGN; filled
307: circles = non type 1 AGN; stars = photometric redshifts. Isodensity
308: contours of all GOODS-MUSIC 24$\mu$m sources are overlayed to the
309: plot. The thick continuous lines mark the expectations of three
310: obscured AGN SEDs with redshift increasing from 0 to 4 from left to
311: right. The lower curve represents the colors of a typical low
312: luminosity Seyfert 2 galaxy, the middle curve represents the colors of
313: an obscured AGN from the Pozzi et al sample (A2690\_75), the
314: upper curve represents the colors of IRAS09104+41091. The black dashed
315: lines are the expectations of the SEDs of two star-burst galaxies
316: (M82, lower curve and Arp220, upper curve) for z=0-4. The dotted line
317: is the expectation of a passive elliptical galaxy for z=0-4. Right
318: panel: fraction of GOODS-MUSIC 24$\mu$m sources as a function of the
319: R-K color in three F($24\mu$m)/F(R) bins: solid histogram =
320: F($24\mu$m)/F(R)$<200$; dotted histogram =
321: $200<$F($24\mu$m)/F(R)$<1000$; dashed histogram =
322: F($24\mu$m)/F(R)$>1000$.}
323: \label{mirormk}
324: \end{figure*}
325:
326: \subsection{Optical, near infrared and mid infrared color selection}
327:
328: The longest and shorthest wavelengths at which deep photometry is
329: available on the Chandra Deep Fields are the 24$\mu$m mid-infrared
330: band covered by Spitzer MIPS and the optical bands covered by HST ACS.
331: 24$\mu$m sources with faint optical counterparts must be either
332: luminous AGN whose optical nuclear emission is blocked by dust and
333: gas, or powerful dusty-starburst galaxies. The Mid-infrared to optical
334: flux ratio (\footnote{F(24$\mu$m)/F(R); logF(R)=$-0.4\times
335: R-22.5467$. R magnitudes have been obtained by interpolating the V and
336: I magnitudes in the GOODS-MUSIC catalog provided by HST/ACS}) can
337: therefore be considered a rough estimator of obscured activity (both
338: nuclear and star-formation) in galaxies.
339:
340: Figure \ref{mirolir} shows F(24$\mu$m)/F(R) as a function of $\lambda
341: L_\lambda (5.8\mu$m), for three samples of X-ray sources. Unobscured
342: AGNs (open symbols in figure \ref{mirolir}) have F(24$\mu$m)/F(R) in
343: the range 10-200, uncorrelated with $\lambda L_\lambda (5.8\mu$m), as
344: expected because the nuclear emission dominates both optical and mid
345: infrared wavelengths. Conversely, obscured AGNs (filled symbols) have
346: F(24$\mu$m)/F(R) spanning a broader range, and fairly correlated with
347: $\lambda L_\lambda (5.8\mu$m). This behavior resembles that of
348: moderately obscured AGNs detected in X-rays, for which the X-ray to
349: optical flux ratio (X/O) is strongly correlated with the X-ray
350: luminosity (Fiore et al. 2003, Eckart et al. 2006). The nuclear
351: optical-UV light of these objects is completely blocked, or strongly
352: reduced by dust extinction, and the optical flux is dominated by the
353: host galaxy. On the other hand, the 2-10 keV flux is reduced by only
354: small factors, even for obscuring gas column densities of the order of
355: a few$\times 10^{23}$ cm$^{-2}$. X/O is therefore a good estimator of
356: the ratio between the nuclear flux and the host galaxy starlight
357: flux. While the nuclear AGN X-ray luminosity can span several decades,
358: the host galaxy R band luminosity has a moderate scatter, less than
359: one decade, giving rise to the observed correlation between X/O and
360: the X-ray luminosity. Perola et al. (2004), Mignoli et al. (2004),
361: Brusa et al. (2005) and Cocchia et al. (2007) found that the high
362: X/O sources tend also to be obscured in the X-rays, with column
363: densities of the order of $10^{22-23}$ cm$^{-2}$. It is therefore
364: possible to conclude that a high X/O ratio is a good indicator for
365: both optical and moderate X-ray obscuration in high luminosity
366: sources. Interestingly, the F(24$\mu$m)/F(R) of X-ray selected
367: sources is strongly correlated with X/O (figure \ref{mirolir}, right
368: panel). This suggests that luminous Compton thick AGN, which are faint
369: in X-rays because of the strong photoelectric absorption and Compton
370: scattering, and cannot be selected using their X/O flux ratio, can be
371: recovered using the F(24$\mu$m)/F(R) ratio.
372:
373: \begin{figure*}
374: \begin{tabular}{cc}
375: \includegraphics[height=8.5truecm,width=8.5truecm]{fig4a.eps}
376: \includegraphics[height=8.5truecm,width=8.5truecm]{fig4b.eps}
377: \end{tabular}
378: \caption{Left panel: the redshift distribution of the
379: F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ sources without direct X-ray
380: detection (107 sources) compared to those of the full GOODS-MUSIC
381: 24$\mu$m sample (1649 sources) and of the X-ray selected AGNs in the
382: GOODS-MUSIC area (150 sources). Right panel: the redshift-infrared
383: luminosity (log$(\lambda L_\lambda (5.8\mu m))$) plane for the full
384: 24$\mu$m GOODS-MUSIC source sample (isodensity contours); the CDFS
385: X-ray AGN sample (open circles = type 1 AGN; filled circles = non type
386: 1 AGN; stars = photometric redshifts); and the sample of
387: F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ sources without direct X-ray
388: detection (filled triangles)}
389: \label{zdist}
390: \end{figure*}
391:
392: Furthermore, since X-ray obscured AGNs tend to have red R-K colors
393: (Brusa et al. 2005 and reference therein), one would also expect that
394: Compton thick AGN have similarly red colors. Indeed figure
395: \ref{mirormk} shows that the F(24$\mu$m)/F(R) of X-ray selected,
396: obscured AGN is correlated with the R-K color, as expected. Figure
397: \ref{mirormk} also shows the iso-density contours of all the
398: GOODS-MUSIC 24$\mu$m sources with F(24$\mu$m)$>40\mu$Jy.
399: Intriguingly, the iso-density contours become narrow in
400: F(24$\mu$m)/F(R) at high R-K values and extend toward the region
401: occupied by obscured X-ray selected AGN at high F(24$\mu$m)/F(R) and
402: high R-K values. The bimodality at high values of F(24$\mu$m)/F(R) of
403: the color distribution of the 24$\mu$m selected sources is evident in
404: the rightmost panel of figure \ref{mirormk}, which shows the fraction
405: of GOODS-MUSIC 24$\mu$m sources as a function of the R-K color in
406: three F(24$\mu$m)/F(R) bins. While at low and intermediate
407: F(24$\mu$m)/F(R) values the distributions are peaked at
408: R-K$\sim2.5-3.5$ and decrease smoothly toward higher R-K values, the
409: distribution of the sources with F(24$\mu$m)/F(R)$>1000$ shows a
410: strong excess at R-K$>4.5$.
411:
412: It is interesting to note that most of the highly obscured AGN
413: selected in the HELLAS2XMM survey on the basis of their high X-ray to
414: optical flux ratio (Pozzi et al. 2007) have F(24$\mu$m)/F(R) higher
415: than a few hundred, and all have R-K$>4.5$. Their SEDs are
416: characterized by a passive early-type galaxy in the optical and in the
417: near infrared, and by an AGN component in the mid infrared. These SEDs
418: redshifted up to z=4 are able to explain the extreme colors of the
419: F(24$\mu$m)/F(R)$>1000$ R-K$>4.5$ sources, unlike the SED of even
420: extreme star-forming galaxies like Arp220 (see figure \ref{mirormk}
421: left panel). This strongly suggests that most of the
422: F(24$\mu$m)/F(R)$>1000$ R-K$>4.5$ sources are powered by an active
423: nucleus. Similar conclusions are found analyzing slightly different
424: color diagrams, like F(24$\mu$m)/F(R) vs. F(24$\mu$m)/F(8$\mu$m) and
425: F(24$\mu$m)/F(R) vs. F(3.6$\mu$m)/F(z).
426:
427: \section{X-ray properties of extreme 24$\mu$m selected sources}
428:
429: Our candidate obscured AGNs are selected from the full 24$\mu$m
430: GOODS-MUSIC sample with the request of having F(24$\mu$m)/F(R)$>1000$
431: and R-K$>4.5$. There are 135 such sources.
432:
433: \subsection{Sources with a direct X-ray detection}
434:
435: Eighteen of the 135 sources have an X-ray detection in Alexander et
436: al. (2003). Four other sources are not formally detected but have
437: more than 4-5 counts (after background subtraction) at the position of
438: the 24$\mu$m source. In summary, 22 of the 135 sources (16\%) have a
439: significant X-ray emission directly visible in the Chandra images.
440:
441: Three of these 22 sources have a spectroscopic redshift, with narrow
442: line optical spectra, the other 19 have a photometric
443: redshift in the GOODS-MUSIC catalog (see Section 2); the median
444: redshift and its interquartile range of these 22 sources are 2.1 (0.5;
445: hereinafter interquartile ranges are indicated in brackets after
446: the median values).
447:
448: The median monochromatic infrared luminosity at 5.8$\mu$m is 44.42
449: (0.37). The X-ray luminosities are in all cases higher than $10^{42}$
450: erg s$^{-1}$, making them bona-fide AGNs. The median logarithmic
451: ratio between the $5.8\mu$m and the 2-10 keV luminosities is 1.07
452: (0.32). As a comparison, the median of the same logarithmic ratio for
453: the full GOODS-MUSIC X-ray sample (150 AGN with measured redshift) is
454: 0.69 (0.47). The probability that the two distributions are drawn from
455: the same parent population is $\ls0.2\%$, using the Kolmogorov-Smirnov
456: test.
457:
458: The hardness ratios indicate in most cases a hard, possibly obscured,
459: X-ray spectrum. Indeed, these sources are among the most obscured ones
460: in the Tozzi et al (2006) analysis, all having column densities higher
461: than a few$\times10^{22}$ cm$^{-2}$ and 2 having column densities as
462: high as $10^{24}$ cm$^{-2}$.
463:
464: Table 1 gives the breakdown of the best fitting templates (see Section
465: 2) to the SEDs of these 22 sources. 15 SEDs are best fitted by one of
466: the templates in figure \ref{seds}.
467:
468: \begin{table}
469: \caption{\bf Table 1: template fits to the SEDs of the sources with
470: F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$}
471: \begin{tabular}{lcc}
472: \hline
473: \hline
474: Template & X-ray det. & Not X-ray det. \\
475: \hline
476: Ellipticals + S0 & - & 2 \\
477: Spirals & - & 1 \\
478: M82+N6090+Arp220 & - & 35 \\
479: I19254 + Mark231 & 4 & 17 \\
480: Seyfert 1.8-2+red QSO & 1 & 5 \\
481: A2690\_75 + BPM16274\_69 & 5 & 34 \\
482: IRAS09104+4109 & 9 & 11 \\
483: N6240 & 1 & 1 \\
484: Seyfert 1 +QSOs & 2 & 1 \\
485: \hline
486: Total & 22 & 107 \\
487: \hline
488: \end{tabular}
489:
490: \end{table}
491:
492:
493: \begin{figure*}
494: \begin{tabular}{cc}
495: \includegraphics[height=8truecm,width=8truecm]{fig5a.eps}
496: \includegraphics[height=8.5truecm,width=8.5truecm]{fig5b.eps}
497: \end{tabular}
498: \caption{Left figure: stacked X-ray images of 24$\mu$m GOODS-MUSIC
499: sources. Images have sides of 18 arcsec, central circles have 2
500: arcsec radii. Upper panels refer to 111 sources with
501: F(24$\mu$m)/F(R)$>1000$ and R-K$>$4.5; lower panels refer to 73
502: sources with F(24$\mu$m)/F(R)$<200$ and R-K$>$4.5. Left panels are
503: stacks in the S 0.3-1.5 keV band, right panels are stacks in the H
504: 1.5-4 keV band. Right figure: the hardness ratio (H-S)/(H+S) of the
505: counts in circles of 2 arcsec radii as a function of the fraction of
506: highly obscured AGN for the sample of 111 24$\mu$m GOODS-MUSIC sources
507: with F(24$\mu$m)/F(R)$>1000$ and R-K$>$4.5. The solid curve is the
508: result of Montecarlo simulations (see text for details); the two thick
509: horizontal lines are the average hardness ratios measured for the
510: F(24$\mu$m)/F(R)$>1000$ and R-K$>$4.5 (upper line) and
511: F(24$\mu$m)/F(R)$<200$ and R-K$>$4.5 sources (lower line). The colored
512: bands mark the hardness ratio statistical uncertainties.}
513: \label{hrstack}
514: \end{figure*}
515:
516: \subsection{Sources without a direct X-ray detection}
517:
518: The total number of sources with F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$
519: and no direct X-ray detection is 111 (we excluded two
520: sources which happen to lie within 5 arcsec from an X-ray source).
521:
522: Four of these sources have a spectroscopic redshift, 99 have
523: photometric redshift in the GOODS-MUSIC catalog. For four sources we
524: could only compute a lower limit to the redshift. In conclusion, we
525: have redshifts or limits for 107 sources. Both median redshift and
526: infrared luminosity are similar to those of the 22 sources with a
527: direct X-ray detection.
528:
529: The redshift and infrared luminosities distributions of the
530: F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ sources are compared in figure
531: \ref{zdist} to those of the full GOODS-MUSIC 24$\mu$m and X-ray
532: selected samples. The F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ sources
533: have a redshift distribution with median redshift 1.91 (0.30).
534: Excluding the lower limits $<$z$>$=1.9 (0.28). This distribution is
535: shifted toward higher redshifts than both the full GOODS-MUSIC
536: 24$\mu$m source sample and the X-ray selected AGN sample (figure
537: \ref{zdist} left panel). The right panel of figure \ref{zdist} shows
538: that moderately obscured, X-ray selected sources are concentrated
539: below z=1.5 and span a range of infrared luminosities log$(\lambda
540: L_\lambda (5.8\mu m))\sim42-45.3$ Their median 2-10 keV and infrared
541: luminosities are 43.16 (0.56) and 43.84 (0.57) respectively. On the
542: other hand, the F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ sources are
543: concentrated between z=1.2 and z=2.6 and have infrared luminosities
544: from log$(\lambda L_\lambda (5.8\mu m))\sim43.5$ to log$(\lambda
545: L_\lambda (5.8\mu m))\sim45.5$, with a median 44.34 (0.30),
546: luminosities similar to that of the X-ray selected AGNs at the same
547: redshift.
548:
549: Table 1 gives the results of the template fits to the observed SEDs of
550: the 107 sources with a redshift. In 65\% of the cases AGN templates
551: provide the best fit. The majority of these SEDs are best fitted by
552: one of the templates in figure \ref{seds}, the others are best fitted
553: by one of the AGN templates of Polletta et al. (2007). Only one SED
554: is best fitted by the template of a type 1 AGN. 36 SEDs are best
555: fitted by star-forming galaxy templates (in 30 cases the best fit is
556: obtained using powerful star-forming galaxy templates, like those of
557: Arp220 and NGC6090). In conclusion, the results of the SED fitting
558: analysis confirm that the majority of the sources with
559: F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ may be highly obscured
560: AGNs. However, we remark that the results of the template fitting
561: should not be always considered a quantitative determination of the
562: nature of each single source, since in many cases different templates
563: can produce fits with similar $\chi^2$ in the observed optical to
564: 24$\mu$m band, producing a degeneracy difficult to account for.
565: Therefore, the results of the template fittings should only be taken
566: as a qualitative indication of the population properties of the
567: samples.
568:
569:
570: \subsection{X-ray stacking analysis}
571:
572: To validate our highly obscured AGN selection sample and assess
573: quantitatively the fraction of bona-fide AGNs in the
574: F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ sample, we performed a detailed
575: ``stacking'' analysis of the X-ray data of the 24$\mu$m selected
576: sources. Indeed, thanks to the low Chandra background, it is possible
577: to increase the effective exposure time and derive average properties
578: of undetected objects, using these stacking techniques: counts at the
579: positions of known sources are co-added in order to probe X-ray
580: emission substantially below the single source sensitivity limit.
581:
582: We performed a stacking analysis of the 111 sources not directly
583: detected in the Chandra 1 Msec X-ray image. As control samples we used
584: the 22 sources with F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ and an X-ray
585: counterpart, a sample of sources with F(24$\mu$m)/F(R)$>1000$ and
586: R-K$<3.5$ (51 sources after the exclusion of the sources directly
587: detected in the X-ray image) and a sample of sources with
588: F(24$\mu$m)/F(R)$<200$ and R-K$>$4.5 (73 sources after the exclusion
589: of the sources directly detected in the X-ray image). The total
590: exposure times for the four source samples are 94 Msec, 19.5Msec, 43
591: Msec and 61 Msec respectively.
592:
593: The top panels of figure \ref{hrstack}a) show the X-ray stack of the
594: 111 24$\mu$m sources with F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ in two
595: energy bands. This is compared to the stack of 73 sources with
596: F(24$\mu$m)/F(R)$<200$ and R-K$>4.5$ and no direct X-ray detections
597: (lower panels). We choose the bands 0.3-1.5 keV (soft band, S) and
598: 1.5-4 keV (hard band, H) to keep the level of the internal background
599: as low as possible and similar in the two bands. At a typical
600: redshift of 2 these bands correspond to rest frame energies of 0.9-4.5
601: keV and 4.5-12 keV respectively. The stack of the high
602: F(24$\mu$m)/F(R) and high R-K sources produces a detection in both
603: soft and hard bands, with count rates $(0.98\pm0.20)\times10^{-6}$
604: counts/s and $(1.25\pm0.26)\times10^{-6}$ counts/s in the two bands
605: respectively (using a 2 arcsec radius extraction region). The
606: hardness ratio H-S/H+S measured for this sample is therefore
607: 0.12$\pm$0.15. Conversely, no detection in either bands is obtained
608: for the stack of 51 sources with high F(24$\mu$m)/F(R) and low R-K,
609: with 1$\sigma$ upper limits of $\sim2\times10^{-7}$ counts/s. The
610: stack of the low F(24$\mu$m)/F(R) and high R-K sources produces a
611: significant detection only in the soft band, with count rate
612: $(1.89\pm0.32)\times10^{-6}$ counts/s. The count rate measured in the
613: hard band is significant at less than 3 $\sigma$ and corresponds to
614: $(4.9\pm1.9)\times10^{-7}$ counts/s. The
615: corresponding hardness ratio is H-S/H+S=-0.58$\pm$0.18. The stack of
616: the 22 sources with F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ and an X-ray
617: counterpart produces a hardness ratio H-S/H+S$=-0.06\pm$0.01. In
618: conclusion, the stack of the 111 sources with F(24$\mu$m)/F(R)$>1000$
619: and R-K$>4.5$ and without a direct X-ray detection produces a
620: significant signal in both soft and hard X-ray bands. It is
621: interesting to note that its hardness ratio suggests an average
622: spectrum harder than even the average spectrum of the 22 sources with
623: similar IR and optical colors but with a direct X-ray detection.
624:
625: \subsection{Simulations to assess the fraction of obscured AGN in the
626: 24$\mu$m source samples}
627:
628: We used the observed flux in the stacked images, together with the
629: hardness ratio H-S/H+S, to constrain the fraction of highly obscured
630: AGN in the samples. To this purpose, we generated simulated X-ray
631: count rates and hardness ratios as a function of the fraction of the
632: AGN, assuming that the F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ source
633: sample is made by obscured AGNs and star-forming galaxies. We started
634: from the observed redshift and infrared luminosities, and for the
635: obscured AGN we assumed a log($\lambda L_\lambda
636: (5.8\mu$m)/L(2-10 keV)) luminosity ratio chosen randomly in the range
637: 0.4-1.2. The lower value is the typical ratio found by Silva et
638: al. (2003) for low luminosity Seyfert 2 galaxies with column density
639: log$N_H\ls 24$. It is also similar to the ratio found for the two
640: HELLAS2XMM sources A2690\_75 and BPM16274\_69. The upper value is the
641: ratio found for the powerful obscured QSO IRAS09104+4109 (Piconcelli
642: et al. 2007).
643:
644: For the star-forming galaxies we used a log($\lambda L_\lambda
645: (5.8\mu$m)/L(2-10 keV)) luminosity ratio between 2 and 2.8. These two
646: values are obtained assuming a total infrared to 2-10 keV logarithmic
647: luminosity ratio of 3.6 (Ranalli et al. 2003) and correcting this for
648: the ratio between the total infrared luminosity and the 5.8$\mu$m
649: luminosity of powerful star-forming galaxies and spirals templates.
650:
651: We further assumed that the star-forming galaxies are not obscured in
652: X-rays, while the AGNs are highly obscured. For the latter
653: objects, we adopted a flat logN$_H$ distribution from 23 to 26
654: cm$^{-2}$.
655:
656: For a given fraction of AGN in the sample, we first decide, using a
657: random generator, whether each source is an AGN or a star-forming
658: galaxy. Then, if the source turned out to be an AGN, we choose an
659: absorbing column density using again a random generator and the
660: assumed logN$_H$ distribution. We computed unobscured 2-10 keV
661: luminosities from infrared luminosities, and then X-ray fluxes by
662: folding a power law spectrum with spectral energy index of 0.8 reduced
663: at low energy by photoelectric absorption with a Chandra response
664: matrix. For column densities $\gs10^{25}$ cm$^{-2}$ we assumed that
665: the direct emission is completely blocked by photoelectric absorption
666: and Compton scattering. For these sources we assumed a reflection
667: component with normalization 1/100 of the direct component and same
668: spectral index.
669:
670: A small fraction of the simulated count rates (20-30\%) is higher than
671: the Chandra detection limit. This was expected, given the fact
672: that 22 out of 135 sources with F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$
673: do have X-ray counterparts. Since we are interested in the fraction
674: of AGN in the sources not directly detected in the Chandra images,
675: these count rates have been excluded from the following analysis.
676:
677: The result of the simulations is shown in figure \ref{hrstack}, right
678: panel. According to our simulations, the average hardness ratio of the
679: F(24$\mu$m)/F(R)$>1000$ and R-K$>4.5$ source sample without X-ray
680: detection is reproduced if 80$\pm$15\% of the sources in the sample
681: are highly obscured AGNs. The hardness ratio of the sources with
682: F(24$\mu$m)/F(R)$<200$ and R-K$>4.5$ is reproduced if the fraction of
683: the obscured AGN is in the range 0-20\%. Changing the assumptions
684: made to produce the simulations within reasonable ranges changes only
685: slightly this result.
686: %To account for these systematic uncertainties we
687: %increase by an additional 5\% the confidence intervals of the AGN
688: %fraction.
689:
690: The result of the simulations concerning the F(24$\mu$m)/F(R)$>1000$
691: and R-K$>4.5$ source sample is well consistent with the
692: indication coming from the 1-24$\mu$m SED fitting with galaxy and AGN
693: templates presented in the previous section.
694:
695: \section{Discussion and Conclusions}
696:
697: We selected a source sample with extreme mid-infrared to optical flux
698: ratio (F(24$\mu$m)/F(R)$>$1000) and red optical colors (R-K$>$4.5) in
699: the GOODS CDFS area. These sources are among the most luminous sources
700: in the GOODS 24$\mu$m sample, because of the correlation of
701: F(24$\mu$m)/F(R) with the infrared luminosity (see figures
702: \ref{mirolir} and \ref{zdist}, right panel). The fraction of these
703: sources not directly detected in the Chandra images produces a
704: significant stacked signal in both Chandra soft and hard X-rays bands.
705: A detailed analysis based on Montecarlo simulation shows that the
706: stacked count rates and hardness ratios can be reproduced if this
707: source population is dominated by highly obscured (N$_{\rm H}>{\rm
708: a~few}\times 10^{23}$ cm$^{-2}$) AGNs.
709:
710: This conclusion is further confirmed by the following
711: consideration. If we assume that this source population is dominated
712: by star-forming galaxies, we can use the observed infrared luminosity
713: and the dust-corrected UV luminosities to derive two different
714: estimates of star-formation rates. On one hand total infrared
715: luminosities have been converted to a star-formation rate using the
716: following formula: $SFR=L_{IR} erg s^{-1}/2.24\times 10^{43}
717: M_{\odot}/yr$, (Kennicutt 1998). On the other hand the UV
718: star-formation rate is an output of the fits of the observed SEDs with
719: synthetic models. These models have been obtained using the
720: Bruzual \& Charlot (2003) models, parametrizing the star-formation
721: histories with a variety of exponentially declining laws (with
722: timescales ranging from 0.1 to 15 Gyrs) and metallicities (from
723: $Z=0.02Z_{\odot}$ to $Z=2.5Z_{\odot}$, see Grazian et al. 2006 and
724: references therein for further detail). Dust extinction is added to
725: all star-forming models, with $0<E(B-V)<1.1$ and a Calzetti
726: attenuation curve. Given the very red colors of our objects in the
727: optical bands, they are typically fit with star-forming templates with
728: a large dust extinction ($E(B-V)>0.5$ ), resulting in a star-formation
729: rate much larger than what derived by a simple, dust free UV-to-SFR
730: conversion.
731:
732: The median logarithmic star-formation rate from infrared luminosity is
733: 2.31 (0.59) $M_{\odot}/$yr. The median logarithmic UV star-formation
734: rate is 0.83 (0.80)$M_{\odot}/$yr, a factor 30 lower. The large
735: mismatch between these two estimates strongly suggests that the
736: infrared luminosity of this source sample is not dominated by
737: star-formation but rather by accretion. As a comparison, the median
738: infrared and UV star-formation rates of the sources with
739: F(24$\mu$m)/F(R)$<200$ and R-K$>$4.5 in our control sample are 1.26
740: (0.38) and 1.10 (0.30), consistent with each other, suggesting that
741: the infrared luminosity of these sources is not dominated by nuclear
742: accretion (in agreement with their X-ray hardness ratio, see previous
743: section).
744:
745: We can now compare the number of highly obscured AGNs selected at
746: 24$\mu$m and not detected in X-rays, with the number of unobscured and
747: moderately obscured AGNs directly seen in the X-ray images. To reduce
748: the importance of complex selection effects we limit this analysis to
749: the sources in the redshift bin 1.2-2.6. At z=2.6 the 5.8$\mu$m
750: luminosity corresponding to the limit of 40$\mu$Jy at 24$\mu$m is
751: log($\lambda L_\lambda(5.8\mu$m))=44.3-44.2 for Seyfert
752: 2,1.8 galaxy SEDs, and 44.4-44.5 for the four SEDs in figure
753: \ref{seds}. Assuming a log($\lambda L_\lambda (5.8\mu$m)/L(2-10 keV))
754: luminosity ratio in the range 0.4-1.2 implies unobscured 2-10 keV
755: luminosities in the range 43.3-43.9. This suggests to limit the
756: comparison to the sources with logL(2-10 keV)$>43$ and
757: log($\lambda L_\lambda(5.8\mu$m))$\gs$44.2.
758:
759: In the 1.2-2.6 redshift bin there are at least 46$\pm$10 sources with
760: F(24$\mu$m)/F(R)$>$1000, R-K$>4.5$, log($\lambda
761: L_\lambda(5.8\mu$m))$\gs$44.2 and no direct X-ray detection (80$\pm15\%$
762: of 57 sources), which are probably highly obscured AGNs, and 44 X-ray
763: selected AGNs, 7 of which with broad lines in their optical
764: spectra. The number of 24$\mu$m selected, presumably highly obscured
765: AGNs, missed by the CDFS X-ray survey is therefore similar to
766: the number of X-ray selected AGNs.
767:
768: The median 5.8$\mu$m luminosity of the 57 F(24$\mu$m)/F(R)$>$1000
769: and R-K$>4.5$ sources without direct X-ray detection in the redshift
770: and luminosity bins defined above is log($\lambda L_\lambda
771: (5.8\mu$m))=44.48 (0.30), implying a median unobscured 2-10 keV
772: luminosity in the range 43.3-44.1, adopting our assumption on the
773: infrared to X-ray flux ratio. At z=2.6 these luminosities translate to
774: X-ray fluxes between a few times $10^{-16}$ and $10^{-14}$ \cgs,
775: meaning that these sources would have been easily detected if they
776: were not highly obscured.
777:
778: We have compared our findings with the predictions of the La Franca et
779: al. (2005) and Gilli et al. (2007) models. In the redshift bin
780: 1.2--2.6 and in the area covered by our 24$\mu$m sample the La Franca
781: et al. (2005) luminosity function predicts about 75 AGNs with
782: logL(2-10 keV)$>43$, 20 of which with a column density $\gs10^{24}$
783: cm$^{-2}$. The absorption distribution adopted by the Gilli et
784: al. (2007) AGN synthesis model for the Cosmic X-ray Background
785: predicts about 40 Compton Thick AGNs with the same limits for redshift
786: and luminosity. We find 46$\pm$10 24$\mu$m selected, highly obscured
787: AGNs, a number about twice the La Franca et al. (2005) prediction but
788: well consistent with the Gilli et al. (2007) prediction.
789:
790: Marconi et. al. (2004, 2007 in preparation) derived a SMBH mass
791: function from the X-ray selected AGN luminosity functions that falls
792: short by a factor of about 2 to the ``relic'' SMBH mass function. This
793: difference can be greatly alleviated adding the population of infrared
794: selected, highly obscured AGNs to the AGN selected in the X-ray band
795: below 10 keV.
796:
797: Finally, we note that the population of infrared selected, highly
798: obscured AGNs, can help in reconciling the predictions of models of
799: galaxy evolution with the observed AGN number densities at
800: z$\gs1-1.5$. Indeed, the Menci et al (2004) model predicts a number of
801: low-to-intermediate luminosity AGNs at z=1.5-2.5 about twice with
802: respect to that measured through X-ray 2-10 keV surveys (La Franca et
803: al. 2005, Fiore 2006). We found above that in the redshift bin 1.2-2.6
804: the number of X-ray selected AGN with logL(2-10keV)$>43$ is comparable
805: to that of the infrared selected, highly obscured AGNs. Therefore,
806: adding together the two populations would give a total number density
807: of AGNs similar to that predicted by the Menci et al. (2004)
808: model. More detailed, quantitative, AGN number density calculations,
809: spanning wider redshift and luminosity ranges, are beyond the purpose
810: of this paper and will be presented elsewhere (Fiore et al. in
811: preparation).
812:
813: Most of the GOODS-MUSIC 24$\mu$m selected sources with high
814: F(24$\mu$m)/F(R), high R-K and no X-ray detection have a very faint
815: optical and near infrared counterpart, only 14\% of the sample having
816: R$\ls26$. Furthermore, only a handful of objects have a mid infrared
817: flux high enough (F(24$\mu$m$\gs0.5$mJy) to allow Spitzer IRS
818: spectroscopy. The spectroscopic identification of $\sim$90\% of the
819: sample must therefore await the advent of ELTs or the launch of JWST.
820:
821: Our analysis is limited to AGNs of intermediate luminosity at
822: z$\sim2$. To extend the coverage of the luminosity-redshift plane
823: requires a complementary observation strategy. Particularly useful to
824: this purpose is the SWIRE survey, which has both Spitzer and optical
825: medium-deep coverage on $\sim 50$ deg$^2$ of sky. As an example, the
826: SWIRE survey contains hundreds of sources with extreme
827: F(24$\mu$m)/F(R) flux ratios and red optical-near infrared colors.
828: Interestingly, several dozens of these extreme sources have 24$\mu$m
829: flux higher than $\sim 1$ mJy, allowing Spitzer IRS spectroscopy,
830: and/or optical magnitude brighter than R$\sim25$, allowing
831: spectroscopy with 8m class telescopes. Furthermore, all these sources
832: are well within the reach of Herschel instruments between 75 and 500
833: $\mu$m. Such long wavelengths observations can greatly help in
834: separating nuclear activity from star formation when the two
835: components have comparable integrated luminosities, e.g. in low
836: luminosity, Seyfert like, AGNs.
837:
838:
839: \acknowledgments
840:
841: We are grateful to Fabio La Franca and Roberto Maiolino for useful
842: discussions. We thank an anonymous referee for comments that helped
843: improving the presentation. After submission of this paper to ApJ we
844: became aware of a work by a different group reaching similar
845: conclusions (Daddi et al. 2007). We thank Emanuele Daddi for
846: providing a copy of his manuscript before submission and for useful
847: discussions. Part of this work was supported by ASI/INAF contracts
848: I/023/05/0 and I/024/05/0 and by PRIN/MUR grant 2006-02-5203. Part of
849: this work was supported by the Deutsches Zentrum f\"ur Luft-- und
850: Raumfahrt, DLR project numbers 50 OR 0207 and 50 OR 0405.
851:
852: \begin{thebibliography}{}
853: \bibitem[]{} Alexander, D. M., Bauer, F. E., Brandt, W. N. et al.
854: 2003, \aj, 126, 539
855: \bibitem[]{} Alonso-Herrero, A. et al. 2006, \apj, 640, 167
856: \bibitem[]{} Barmby, P., Alonso-Herrero, A., Donley, J.L.
857: et al. 2006, \apj, 642, 126
858: \bibitem[]{} Bongiorno, A. et al. 2007, \aap submitted, astro-ph/0704.1660,
859: \bibitem[]{} Brandt, W.N., Hasinger, G. 2005, ARA\&A, 43, 1056
860: \bibitem[]{} Brusa, M., Comastri, A., Daddi, E. et al. 2005 \aap, 432, 69
861: \bibitem[Bruzual \& Charlot 2003]{bc03} Bruzual, G., \& Charlot, S. 2003,
862: MNRAS, 344, 1000
863: \bibitem[]{} Cavaliere, A. \& Vittorini, V. 2000, \apj, 543, 599
864: \bibitem[Cimatti et al. 2002]{cimatti02} Cimatti, A. et al. 2002,
865: \aap, 381, L68
866: \bibitem[]{} Cocchia, F., Fiore, F., Vignali, C. et al. 2007, \aap, 466, 31
867: \bibitem [Coleman, Wu \& Weedman 1980]{coleman}{Coleman, G. D., Wu, C-C,\&
868: Weedman, D. W. 1980, ApJS, 43, 393}
869: \bibitem[]{} Comastri, A. 2004, in "Supermassive Black Holes in the
870: Distant Universe", Ed. A. J. Barger, Kluwer Academic, vol. 308, p.245
871: \bibitem[]{} Cowie L.L., Songaila, A., Hu, E.M., Cohen, J.G., 1996, AJ,
872: 112, 839
873: \bibitem[]{} Daddi, E. et al. 2007, ApJ submitted, astro-ph/0705.2832
874: \bibitem[]{} De Lucia, G., Springel, V., White, S.D.M., Croton, D., Kauffmann,
875: G., 2006, MNRAS, 366, 499
876: \bibitem[]{} De Santis, C., Grazian, A., Fontana, A., Santini, P. 2007,
877: New Astronomy, 12, 271
878: \bibitem[]{} Di Matteo, T., Springel, V., Hernquist, L. 2005, Nature, 433, 604
879: \bibitem[]{} Eckart, M.E., Stern, D., Helfand, D. et al. 2006, ApJS, 165, 19
880: \bibitem[]{} Fabian, A.C. 1999 \mnras, 308, L39
881: \bibitem[]{} Ferrarese, L. \& Merritt, D. 2000, \apj, 539, L9
882: \bibitem[]{} Fioc, M. \& Rocca-Volmerange, B. 1997, \aap, 326, 950
883: \bibitem[]{} Fiore, F., Brusa, M., Cocchia, F. et al. 2003, A\&A, 409, 79
884: \bibitem[]{} Fiore, F. 2006, in ``AGN and galaxy evolution'',
885: Mem. SAIt, v.77, p.694, astro-ph/0603823
886: \bibitem[]{} Franceschini, A., Hasinger, G., Miyaji, T.,
887: Malguori, D. 1999, \mnras, 310, L5
888: \bibitem[]{} Gebhardt, K., Kormendy, J., Ho, L. et al. 2000, \apj, 543, L5
889: \bibitem[]{} Giacconi, R., Zirm, A., Wang, J., et al. 2002, \apjs, 139, 369
890: \bibitem[]{} Gilli R., Salvati, M., Hasinger G. 2001, A\&A, 366, 407
891: \bibitem[]{} Gilli R., Comastri A., Hasinger G. 2007, A\&A, 463, 79
892: \bibitem[]{} Granato, G.L., Silva, L., Monaco, P., Panuzzo, P.,
893: Salucci, P., De Zotti, G., \& Danese, L. 2001, \mnras, 324, 757
894: \bibitem[]{} Granato, G.L., De Zotti, G., Silva, L., Bressan, A., Danese, L.
895: 2004, \apj, 600, 580
896: \bibitem[]{} Grazian, A., Fontana, A., de Santis, C. et al. 2006, \aap,
897: 449, 951
898: \bibitem[]{} Hasinger, G. 2003, AIP Conf. Proc. 666, 227
899: \bibitem[]{} Hasinger, G., Miyaji, T., Schmidt, M. 2005 \aap,
900: 441, 417
901: \bibitem[]{} Houck, J.R., et al. 2005, \apjl, 622, L105
902: \bibitem[]{} Kennicutt, R.C., Jr. 1998, ARAA, 36, 189
903: \bibitem[]{} Lacy, M. et al. 2004, \apjs, 154, 166
904: \bibitem[]{} La Franca, F., Fiore, F., Comastri A. et al. 2005, ApJ, 635, 864
905: \bibitem[Le Fevre et al. 2004]{vvds} Le Fevre, O., Vettolani, G., Paltani, S.,
906: Tresse, L., Zamorani, G., Le Brun, V., Moreau, C., Bottini, D., Maccagni, D.,
907: Picat, J. P. et al. 2004, \aap, 428, 1043
908: \bibitem[]{} Magliocchetti, M, Silva, L, Lapi, A. de Zotti, G., Granato, G.L.,
909: Fadda, D., Danese, L. 2007, \mnras, 375, 1121
910: \bibitem[]{} Marconi, A., Hunt, L. 2003, \apjl, 589, L21
911: \bibitem[]{} Marconi, A., Risaliti, G., Gilli, R., Hunt, L. K.,
912: Maiolino, R., Salvati, M. 2004, \mnras, 351, 169
913: \bibitem[]{} Matt. G. 2000, \aap, 355, L31
914: \bibitem[]{} Martinez-Sansigre, A., Rawlings, S., Lacy, M. et al. 2005,
915: Nature, 436, 666
916: \bibitem[]{} Martinez-Sansigre, A., Rawlings, S., Lacy, M. et al. 2006,
917: \mnras, 370, 1479
918: \bibitem[]{} Menci N., Fiore, F., Perola, G.C, Cavaliere, A. 2004, \apj,
919: 606. 58
920: \bibitem[]{} Menci, N., Fontana, A., Giallongo, E., Salimbeni, S. 2005 \apj,
921: 632, 49
922: \bibitem[]{} Mignoli, M., Pozzetti, L., Comastri, A. et al. 2004,
923: \aap, 418, 827
924: \bibitem[Mignoli et al. 2005]{k20} Mignoli, M., Cimatti, A., Zamorani, G.,
925: Pozzetti, L., Daddi, E., Renzini, A., Broadhurst, T., Cristiani, S.,
926: D'Odorico, S., Fontana, A. et al. 2005, \aap, 437, 883
927: \bibitem[]{} Piconcelli, E., Fiore, F., Nicastro, F., Mathur, S., Brusa, M.,
928: Comastri, A., Puccetti, S. 2007, \aap in press, astro-ph/0707.2465
929: \bibitem[]{} Perola, G. C, Puccetti, S., Fiore, F. et al. 2004,\aap, 421, 491
930: \bibitem[]{} Polletta, M., Wilkes, B., Siana, B. et al. 2006, \apj, 642, 673
931: \bibitem[]{} Polletta, M., Tajer, M., Maraschi, L. et al. 2007, \apj, 663, 81
932: \bibitem[]{} Pozzi, F., Vignali, C., Comastri, A. et al. 2007, \aap, 468, 603
933: \bibitem[]{} Ranalli, P., Comastri, A., Setti, G. 2003, \aap, 399, 99
934: %\bibitem[]{} Sanders, D.B., Mirabel, I.F. 1996, ARA\&A, 34, 749
935: \bibitem[]{} Silk, J., Rees, M.J. 1998, \aap 311, L1
936: \bibitem[]{} Soltan, A. 1982, \mnras, 200, 115
937: \bibitem[Szokoly et al. 2004]{cxo} Szokoly, G. P., Bergeron, J., Hasinger, G.,
938: Lehmann, I., Kewley, L., Mainieri, V., Nonino, M., Rosati, P., Giacconi, R.,
939: Gilli, R. et al. 2004, ApJS, 155, 271
940: \bibitem[]{} Tozzi, P. , Gilli R. Mainieri V. et al. 2006, \aap, 451, 457
941: \bibitem[]{} Ueda Y., Akiyama M., Ohta K., Miyaji T.
942: 2003, \apj, 598, 886
943: \bibitem[Vanzella et al. 2005]{vanzella05} Vanzella, E., Cristiani, S.,
944: Dickinson, M., Kuntschner, H., Moustakas, L. A., Nonino, M., Rosati, P.,
945: Stern, D., Cesarsky, C., Ettori, S., et al. 2005, \aap, 434, 53
946: \bibitem[Vanzella et al. 2006]{vanzella06} Vanzella, E., Cristiani, S.,
947: Dickinson, M., et al. 2006, \aap, 454, 423
948: \bibitem[]{} Weedman, D.W., Le Floc'h, E., Higdon, S.J.U., Higdon, J.L.,
949: Houck, J.R. 2006a
950: \apj, 638, 613
951: \bibitem[]{} Weedman, D.W., Soifer, B.T., Hao, L. et al. 2006b, \apj, 651, 101
952: \bibitem[]{} Worsley, M.A., Fabian, A. C., Barcons, X. et al. 2004, MNRAS,
953: 352, L28
954: \bibitem[]{} Worsley, M.A., Fabian, A. C., Bauer, F.E., Alexander, D.M.,
955: Brandt, W.N., Lehmer, B.D. 2006, \mnras, 368, 1735
956: \bibitem[]{} Yan, L., Sajina, A., Fadda, D., Choi, P., Armus, L., Helou G.,
957: Teplitz, H., Frayer, D., Surace, J. 2007, \apj, 658, 778
958:
959: \end{thebibliography}
960:
961: \end{document}
962: