1: %\documentclass{emulateapj}
2: \documentclass[11pt,preprint]{aastex}
3:
4: %\usepackage{psfig}
5: %\documentclass[preprint]{aastex}
6:
7: %\documentclass[preprint2]{aastex}
8: %\usepackage{emulateapj5}
9:
10: \newcommand{\vdag}{(v)^\dagger}
11: \newcommand{\myemail}{sani@arcetri.astro.it}
12:
13: %% You can insert a short comment on the title page using the command below.
14:
15: \slugcomment{}
16:
17:
18: \shorttitle{3-5 $\mu$m spectra of ULIRGs}
19: \shortauthors{E. Sani et al.}
20:
21:
22: \begin{document}
23: \title{3-5 $\mu$m spectroscopy of obscured AGNs in ULIRGs \thanks{Based on observations collected at the European Southern Observatory,
24: Chile (proposals ESO 73.B-0574, 79.B-0052).}}
25:
26: \author{E. Sani,\altaffilmark{1} G. Risaliti,\altaffilmark{2,3} M. Salvati,\altaffilmark{2} R. Maiolino,\altaffilmark{4} A. Marconi,\altaffilmark{1} S. Berta,\altaffilmark{5}
27: V. Braito,\altaffilmark{6,7} R. Della Ceca\altaffilmark{8} and A. Franceschini,\altaffilmark{5} }
28: %\\
29: %2pm 21 May 2003}
30:
31: \email{sani@arcetri.astro.it}
32:
33: \altaffiltext{1}{Dipartimento di Astronomia, Universit\`a di Firenze, Largo
34: E. Fermi 2, I-50125 Firenze, Italy}
35: \altaffiltext{2}{INAF - Osservatorio Astrofisico di Arcetri, Largo
36: E. Fermi 5, I-50125 Firenze, Italy}
37: \altaffiltext{3}{Harvard-Smithsonian Center for Astrophysics, 60
38: Garden Street, Cambridge, MA 02138}
39: \altaffiltext{4}{INAF - Osservatorio Astronomico di Roma, Via
40: Frascati 33, I-00040 Monte Porzio Catone, Italy}
41: \altaffiltext{5}{Dipartimento di Astronomia, Universit\`a di Padova,
42: Vicolo dell'Osservatorio 2, I-35122 Padova, Italy}
43: \altaffiltext{6}{Astrophysics Science Division, Code 662, NASA/Goddard Space Flight
44: Center, Greenbelt, MD 20771, USA}
45: \altaffiltext{7}{Department of Physics and Astronomy, Johns Hopkins University,
46: Baltimore, MD 21218, USA}
47: \altaffiltext{8}{INAF - Osservatorio Astronomico di Brera, via Brera 28,
48: I-20121 Milano, Italy}
49:
50: \begin{abstract}
51: We present the results of infrared L-band ($3-4~\mu m$) and M-band ($4-5~\mu m$) VLT-ISAAC spectroscopy of five
52: bright Ultraluminous InfraRed Galaxies (ULIRGs) hosting an AGN. From our analysis we distinguish two
53: types of sources: ULIRGs where the AGN is unobscured (with a flat continuum and no absorption
54: features at $3.4~\mu m$ and $4.6~\mu m$), and those with highly obscured AGNs (with a steep, reddened continuum
55: and absorption features due to hydrocarbons and CO). Starburst activity is also present in all of the sources as inferred
56: from the $3.3~\mu m$ PAH emission line.
57: A strong
58: correlation is found between continuum slope and CO optical depth, which suggests that deep carbon monoxide
59: absorption is a common feature of highly obscured ULIRG AGN. Finally we show that the AGN dominates the $3-4~\mu m$
60: emission, even if its contribution to the bolometric luminosity is small.
61: \end{abstract}
62: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
63:
64: \keywords{galaxies: active - galaxies:starburst - infrared:galaxies}
65:
66: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
67: \section{Introduction}
68:
69: %Ultraluminous InfraRed Galaxies (ULIRGs) are the most powerfull sources in the local Universe. The enormous infrared emission is due to gas and hot dust reprocessing energetic radiation (UV-X) and emitting thermally in IR. The primary energetic source is a combination of two mechanisms: an Active Galactic Nucleus (AGN) and/or a StarBurst activity (SB), but it is not clear which mechanism dominates the power radiated. Moreover if an AGN is obscured by a column density $N_H$ higher than $10^{25}cm^{-2}$ (in X-rays),
70: % The nature and behavior of obscured Active Galactic Nuclei (AGNs) are interesting for different astrophysical phenomena: from the study of local Seyfert galaxies, to the search of quasars into high redshift sources.\\
71: % Relating to this matter is the discovery by IRAS of a class of galaxy, Ultraluminous Infrared Galaxies (ULIRGs), which radiate most of their luminosity in the infrared ($L_{IR}>10^{12}L_\odot$) as dust emission. Two different physical processes are possible as the primary energy source of ULIRGs: nuclear gas accretion onto a supermassive black hole and/or intense starburst activity (SB).\\
72: Determining the role and physical properties of Active Galactic Nuclei (AGNs) in Ultraluminous Infrared Galaxies (ULIRGs) is fundamental for many reasons. ULIRGs are the most luminous sources in the local Universe and dominate the bright end of the infrared counts (Sanders \& Mirabel~1996 and references therein); they represent the local counterpart of the high redshift objects discovered by SCUBA (Barger et al.~1998) that dominate the cosmic sub-millimeter background. Furthermore, \textit{Spitzer} completes the millimeter and sub-~mm surveys with its mid-IR selection function and enhances the sensitivity to warm systems, which shows extreme infrared/optical luminosity ratios (Lonsdale, Farrah \& Smith~2006).
73: Therefore the study of local ULIRGs can unveil the nature of the high redshift sources, which are too faint to be studied in detail with current facilities.
74: Finally, ULIRGs may contain luminous heavily obscured AGNs (by gas with a column density $N_H>10^{24}cm^{-2}$ in X-rays), so it is important to determine the contribution of accretion to their infrared emissivity in order to achieve a complete census of the AGN population in the near Universe.\\
75: %%Although there are substantial differences between AGN and SB, disentangling the two contributions to the infrared (IR) emission is complicated by dust reprocessing which dilute and partially cancel spectral differences between these mechanisms. Several diagnostics at all wavelengths have been proposed to disentangle AGN/SB contributions. The most powerful indicators of the presence of a buried AGN in ULIRGs have been proved to be the hard X-rays and ISO mid-IR spectra (Risaliti et al. 2000, Franceschini et al 2003, Genzel et al. 1998, Lutz et al. 2003).\\
76: %%Theoretically the ratio between X-rays and IR emission in starburst-dominated ULIRGs is expected to be less than $10^{-4}$, and it raises from $10^{-3}$ to $10^{-1}$ in AGN-dominated ULIRGs (Risaliti et al. 2000). However, if the AGN is covered by a column density $N_H$ higher than $10^{25}cm^{-2}$, its presence and total luminosity are hard to determine; only the brighter galaxies can be studied through X-rays astronomy and these are about $10\%$ of classified ULIRGs as shown by \textit{XMM-Newton} (Franceschini et al. 2003) and \textit{Chandra} (Ptak et al. 2003) observations.\\
77: % In the optical observing high ionization narrow emission lines is an indication of the presence of an AGN; nevertheless these measures are affected by high errors due to flux determination and reddened correction. Absence of indicators does not exclude the presence of an AGN absorbed also in optical lines.\\
78: %%In the mid-IR, data from ISO provided powerful diagnostics for the study of bright ULIRGs (Genzel et al. 1998, hereafter G98). High ionization lines in the mid-IR indicate the presence of an AGN, while strong emission features due to polycyclic aromatic hydrocarbon (PAH) molecules indicate starburst activity. However the presence in mid-IR of broad silicates absorptions makes hard to determine the continuum, estimating the PAH relative strength with low precision.\\
79: %%Usually methods described in the above works are able to identify an AGN if they can directly measure its luminosity, but they fail in determining whether an AGN is present or not in case of heavy gas and dust absorption. The ULIRG NGC 6240 represents an extreme example showing SB activity at all wavelengths except in hard X-rays where a flat reflected spectrum and a strong iron $K\alpha$ line point out an AGN (Vignati et al. 1999).
80: %and in the mid-IR ISO data giving a weak sign of the AGN presence through the detection of the [O IV] coronal line (Lutz et al 2003). Recent spectra performed with Spitzer confirm these result (Armus et al. 2006) and they estimate the AGN contribution in terms of $\sim 20\%-24\%$ to the bolometric luminosity of NGC 6240.
81: %%Moreover these methods require good signal-to-noise data for being used to study faint objects: in the X-rays a high S/N is necessary to disentangle the thermal continuum typical of SB from that of the AGN; in the mid-IR is crucial to have an accurate determination of the continuum and, consequently, a reliable measurement of emission/absorption features. Infact, also using space telescopes like \textit{Spitzer} and the high-to-low ionization lines ratio as a tool to estimate the AGN emission, it is crucial to disentangle independently the AGN/SB contribution to the mid-IR continuum to determine the intrinsic dust reddening, despite all this step is complicated by the strong SB emission which dilutes the AGN spectral properties.\\
82: A promising new approach to the study of ULIRGs is infrared L-band ($3-4~\mu m$) spectroscopy.
83: A comparison between the spectral energy distributions of AGN and starburst (hereafter SB) shows that, with the same
84: bolometric luminosity, the L-band emission by an AGN is a factor of $\sim$100 higher
85: than that of a SB (Risaliti et al.~2006, hereafter R06).
86: Therefore, when studying mixed sources, where both an AGN and a SB are present,
87: the emission of the AGN
88: component can be revealed even if the SB is bolometrically dominant and/or the AGN is heavily obscured.
89: Unfortunately, the high background and low atmospheric transmission in the L-band have made
90: high quality spectra of ULIRGs difficult to obtain until recently. Imanishi \& Dudley (2000) studied the physics of
91: emission and absorption between 3 and 4 $\mu m$ with low resolution spectra of the brightest ULIRGs
92: obtained at 4 meter class telescopes.
93: More recently, R06 and Imanishi et al. (2006) obtained L-band spectra of ULIRGs with 8-meter class telescopes (VLT and Subaru).
94: These high quality spectra have shown the great power of L-band diagnostics in disentangling AGN and SB components
95: in ULIRGs. The main results of those studies can be summarized as follows:\\
96: - a large equivalent width of the $3.3~\mu m$ PAH emission feature ($EW_{3.3}\sim100$ nm) is typical
97: of starburst-dominated sources, while the X-ray radiation of an AGN partially or completely destroys
98: PAH molecules (Voit 1992).\\
99: - A strong absorption feature at $3.4~\mu m$ ($\tau_{3.4}>0.2$) due to aliphatic hydrocarbon grains
100: is an indicator of an obscured AGN and, indeed, such a deep absorption requires a point source behind a
101: screen of dusty gas (this has been verified both theoretically and observationally, Imanishi and Maloney 2003).\\
102: - A steep inverted continuum (described by a power law $f_\lambda\propto\lambda^\Gamma$, $\Gamma>1$)
103: suggests the presence of an obscured, reddened AGN. Again, a high value of $\Gamma$ implies strong
104: dust reddening of a compact source.\\
105: %%Two new diagnostic diagrams are proposed combining the above indicators, which separate completely SB and AGN sources; moreover they demonstrate that these diagnostic tools are usefull for classifying weak ULIRGs which can't be studied at other wavelengths.\\
106: %%Besides applying the above diagnostics to Subaru data, Imanishi and collaborators (2006) find out that a significant part of non-Seyfert ULIRGs (LINERs, H$_{II}$ regions, and unclassified) hosts pure buried AGN.
107: In this paper we present new L-band spectra of the 5 brightest ULIRGs, which are known from independent observation to host an AGN, and extend our study to the M-band ($4-5~\mu m$), using data obtained with
108: ISAAC at the Very Large Telescope (VLT) in Paranal, Chile. Our main purposes are: (a) to measure the
109: spectral parameters with unprecedented accuracy by determining the continuum over
110: a broader spectral region and with higher signal to noise ratio, (b) to analyze the chemical composition and the physical properties of the absorbing/emitting circumnuclear medium. In particular
111: we are searching for molecular transitions originating only in warm dense gas, and associated with the
112: presence of an AGN, such as the $P$ ($\Delta J=-1$) and $R$ ($\Delta J=+1$) branches of the 1-0
113: rotovibrational CO band near $4.65~\mu m$. In this way we can verify if a deep CO hot gas absorption
114: is a peculiar feature present only in IRAS 00183-7111 (Spoon et al. 2004),
115: or is insted a common signature of heavily obscured AGNs. Even if narrow emission lines are not our main target, in some cases we are able to identify hydrogen
116: atomic and molecular features partially blended with residual sky lines (whose subtraction is incomplete in low resolution spectra, van Dishoeck et
117: al.~2003). Finally, (c) we wish to determine the relative SB/AGN contribution to both the $3-4~\mu m$ and the total
118: IR emission.\\
119: In Section 2 we present our sample and review the known physical properties of each
120: source, which will be compared with our results presented in the following Sections.
121: We present the L and M spectra in Section 3, and analyze them from a physical and chemical point of view in Section 4.
122: %In Section 3, L and M band spectra are presented with reduction technique. The physical and chemical analysis of each new spectrum is presented in Section 4 disentangling the contributions of AGN and SB to L-band emission (we apply the analytic model obtained by R06).\\
123: In Section 5 we discuss our results. More specifically, Section 5.1 discusses the relative AGN/SB contribution to the total luminosity, while the physical properties of the dusty emitter/absorber are illustrated in Section 5.2.
124: Our conclusion are summarized in Section 6.\\
125: Throughout this paper we assume $H_0=70, \Omega_M=0.3, \Omega_\Lambda=0.7$.
126:
127: \section{The sample}
128:
129: The sources presented here have been drawn from the sample of Genzel et al.~(1998) consisting of the 15 ULIRGs in the IRAS Bright Galaxy Sample (Sanders et al.~2003).
130: % This consists of 15 ULIRGs from the IRAS Bright Galaxy Sample (BGS, Sanders et al. 2003). These represent all galaxies in the IRAS all sky survey with galactic latitude $\Vert b\Vert>5^\circ$ and $60~\mu m$ flux density $S_{60}>5.4$ Jy.
131: One of our sources, IRAS 05189-2524 is not included in the original sample
132: but fullfills all its selection criteria (Table \ref{tb:car}).
133: \begin{table*}[!t]
134: \begin{center}
135: % \begin{footnotesize}
136: \begin{tabular}{lcccccccccc}
137: \hline
138: source & z & $S_{12}^a$ & $S_{25}^a$ & $S_{60}^a$ & $S_{100}^a$ & $\log L_{IR}^b$ & J$^c$ & H$^c$ & K$^c$ & optical class\\
139: \hline
140: NGC 6240 & 0.0245 & 0.6 & 3.4 & 22.7 & 27.8 & 11.85 & 10.3 & 9.5 & 9.1 & LINER $^1$ \\
141: IRAS 05189-2524 & 0.0426 & 0.73 & 3.44 & 13.94 & 11.36 & 12.16 & 14.0 & 12.7 &11.7 & Sy2 $^2$ \\
142: IRAS 19254-7245 & 0.0617 & 0.2 & 1.2 & 5.5 & 5.8 & 12.06 & 14.0 & 12.7 & 11.7 & Sy2 $^3$\\
143: IRAS 20551-4250 & 0.0428 & 0.3 & 1.9 & 12.8 & 10.0 & 12.04 & 13.4 & 12.2 & 12.2 & H$_{II}$ $^2$\\
144: IRAS 23060+0505 & 0.173 & 0.2 & 0.4 & 1.2 & 0.8 & 12.43 & 14.1 & 12.9 & 11.6 & Sy2 $^4$\\
145: \hline
146: \end{tabular}
147: % \end{footnotesize}
148: \end{center}
149: % \linespread{1.0}
150: % cercare dati mancanti e scrivere caption
151: \caption{\footnotesize{photometric data for the observed sources. Notes: $^a$: flux density in the IRAS filters
152: in units of Jy. $^b$: Total infrared luminosity, calculated from Eq.(\ref{Esq:1}) and assuming the concordance
153: cosmology ($H_0,\;\Omega_M,\;\Omega_\Lambda$)=(70, 0.3, 0.7), Spergel et al. 2003. $^c$: Infrared Magnitudes
154: from Duc, Mirabel, and Maza (1997). References for the optical classification are: $^1$ Veilleux et al. 1995,
155: $^2$ Veilleux et al. 1999, $^3$ Mirabel et al. 1991, $^4$} Hutchings $\&$ Neff 1988}
156: % \cite{veilleux1,genzel,laurent,im_dud,farrah,rigopoulou,veilleux95}.}}
157: % \linespread{1.6}
158: \label{tb:car}
159: \end{table*}
160:
161: The luminosity limit used to define a ULIRG is $L_{IR}>10^{12}L_\odot$, where $L_{IR}$ is the total
162: 8-1000 $\mu m$ luminosity calculated with the IR flux equation of Sanders and Mirabel (1996):
163: \begin{equation}
164: F_{IR}=1.8\times10^{-11}(13.48~S_{12}+5.12~S_{25}+2.58~S_{60}+S_{100})
165: \label{Esq:1}
166: \end{equation}
167: here $F_{IR}$ is the total IR flux in $erg\;s^{-1}cm^{-2}$, and $S_{12},\;S_{25},\;S_{60},\;S_{100}$
168: are the flux densities in the IRAS filters in units of Jy.\\
169: One source, NGC 6240, has a slightly lower total luminosity. However, it is included in our sample, as in
170: G98, because its flux density at $60~\mu m$ exceeds 5.4 Jy by a factor of 4 (as shown in
171: Table \ref{tb:car}), moreover all of its morphological and physical properties are typical of ULIRGs.\\
172: Our selection among the ULIRGs in G98 has been driven by the following criteria:\\
173: - brightness in the infrared (L \& M) bands;\\
174: - presence of AGNs with different levels of obscuration;\\
175: - observability from the VLT site. \\
176: The final subsample of five ULIRGs have been observed with ISAAC at the VLT with two main motivations:\\
177: 1) We wanted to obtain the first M-band spectra for these ULIRGs, thus covering $2~\mu m$ of spectral range when combined with L-band spectra.
178: The large spectroscopic interval and the high quality of these data allow us to
179: investigate the continuum shape, the physical properties and the chemical composition of the emitting/absorbing medium with unprecedented precision.\\
180: %determine the
181: %continuum level (therefore emission equivalent width, and absorption optical depth) with unprecedent
182: %precision. Moreover we can investigate chemical composition and emitting/absorbing medium properties.\\
183: 2) Our sources are the best studied ULIRGs at other wavelengths. However, in some cases the observations
184: do not provide complete information, as in the case of NGC 6240, which shows a double AGN only
185: in the X-rays. Our aim is to identify both the AGN and the SB components in the IR spectra, and to estimate their relative
186: contribution to the $3-4~\mu m$ and total luminosity.
187:
188: \subsection{Sample properties}
189: In this paragraph we review the morphological and spectral properties of our sample.
190:
191: {\bfseries IRAS 05189-2524}\\
192: IRAS 05189-2524 has a compact structure, typical of a final merging stage.
193: Only a single nucleus is visible in the optical images, and the only evidence for merging is a faint tidal tail (Veilleux, Kim \& Sanders 2002). According to optical spectroscopic data,
194: %showing large lines observed in polarized light,
195: IRAS 05189-2524 has been classified as a type 2 Seyfert galaxy (Veilleux et al. 1999). In the X-rays the absorbing medium is optically thin with a column density $N_H\sim10^{23}cm^{-2}$. Even if the AGN dominates the X-ray emission, the low X-ray to infrared ratio indicates that the AGN is
196: not the dominant energy source (Severgnini et al.~2001).
197: % The L-band spectrum obtained by Imanishi \& Dudley (2000) is characterized by a PAH emission with $EW_{3.3}=4.3\;nm$ and an $H_2O$ ice absorbing featere ($\tau_{3.4}$=0.04), corresponding to a predominance of the AGN which contribute for 90\% to $3-4~\mu m$ emission. Reassuming multiband observations of IRAS 05189-2524 point out the presence of an AGN; howether the SB component is present and probably dominates total luminosity, but not L-band emission.
198:
199: {\bfseries IRAS 19254-7245}\\
200: The {\em Superantennae} galaxy is an interacting system with a separation of 10~kpc between the two nuclei,
201: and prominent tidal tails.
202: It is optically classified as a Seyfert 2 (Mirabel et al. 1991).
203: % moreover imaging performed by HST show that the interacting system could be multiple, being everyone of the two nuclei in its turn double (Borne et al. 1999).
204: The infrared emission originates from the central region with no significant contribution from the tails.
205: Infrared images obtained with the
206: ISO satellite shows that the dominant contribution to the total luminosity comes from the southern nucleus,
207: which also contains most of the
208: molecular gas present in the system (Charmandaris et al. 2002).
209: Polarized spectroscopy reveals that the southern nucleus hosts a heavily absorbed AGN which dominates the L-band emission, contributing for $40-50\%$ to $8-1000~\mu m$ luminosity, while the northern one is in a post starburst phase (Berta, et al. 2003).
210: The X-rays emission arises predominantly from the southern nucleus (Braito et al.~2003). The X-ray continuum
211: is flat and a Fe K$\alpha$ line is present, with EW$\sim 1.4$~keV, typical of a reflection-dominated AGN,
212: whereas the direct continuum is completely obscured by a column density N$_H>10^{24}$~cm$^{-2}$. Assuming a Compton-thick model, the X-rays intrinsic luminosity is L$_{2-10\;keV}>10^{44}$erg~s$^{-1}$, implying that the AGN
213: contribution to the bolometric luminosity is significant. At lower energies (0.5-2 keV) a thermal component
214: is observed (k$_B$T$\sim0.9$~keV), probably related to a compact starburst.
215:
216: {\bfseries IRAS 20551-4250}\\
217: IRAS 20551-4250 is a merging system in an early stage (Franceschini et al. 2003). %(two nuclear components are visible, even if not separate with a weak tail)
218: It is optically classified as a H II region (Veilleux et al. 1999), while in the mid-IR it resembles a SB galaxy
219: (Genzel et al.~1998). The X-rays emission is clearly dominated by an obscured AGN, with
220: luminosity L$_{2-10\;keV}\sim~7.0\times10^{42}$erg~$s^{-1}$ and column density N$_H\sim8\times10^{23}$~cm$^{-2}$
221: (Franceschini et al. 2003).
222: As in the case of IRAS~19254-7245, the relative AGN contribution to the bolometric luminosity is uncertain,
223: but probably significant.
224:
225: {\bfseries IRAS 23060+0505}\\
226: This galaxy has been optically classified as a Seyfert~2.
227: In the infrared its spectral energy distribution and the luminosity (L$_{IR}\sim1.3\times10^{46}$erg~s$^{-1}$)
228: are typical of quasars (Hill et al.~1987). Optical images show a complex morphology with prominent tidal tails
229: (Hutchings \& Neff~1988). Optical and near-infrared spectropolarimetric studies reveal a reflection continuum with an extinction A$_V\sim 2.9$~mag, combined with unpolarized
230: emission extincted by A$_V\sim20$~mag (Young, Hough \& Axon~1996). The X-ray data confirm the presence of a
231: moderately obscured (N$_H\sim8\times10^{22}$~cm$^{-2}$) AGN with L$_{2-10\;keV}\sim~1.5\times10^{44}$erg~s$^{-1}$
232: (Brandt et al.~1997).
233: % In the L-band PAH equivalent width turns out smaller of two orders of magnitude to the expected value for a SB dominated object, so that Imanishi \& Dudley (2000) conclude that $3-4~\mu m$ emission is AGN dominated.
234:
235: {\bfseries NGC 6240}\\
236: NGC~6240 is an interacting system in an initial state of merging showing two distinct nuclei with gaseous and
237: stellar tails. It is optically classified as a LINER (Veilleux et al.~1995). In the hard X-rays (E$>10$~keV)
238: the high observed luminosity is interpreted as direct AGN emission absorbed by a column density
239: N$_H\sim2\times10^{24}$cm$^{-2}$ (Vignati et al.~1999); in the 2-10~keV range the presence of the AGN is indicated
240: by a Fe $K\alpha$ line with large EW. Recently, observations obtained with \emph{Chandra} revealed that both nuclei host
241: an AGN (Komossa et al.~2003). This has been confirmed by our L-band observations (Risaliti et al.~2006B), which will be summarized in the next Section. The mid-IR continuum of NGC 6240 is typical of a starburst, with an indication
242: of the presence of an AGN coming from high ionization emission lines, such as [OIV] $\lambda25.9~\mu$m (Lutz et al.~2003).
243: A recent study based on observations performed with {\em Spitzer} confirms this result (Armus et al.~2006)
244: and estimates an AGN contribution to the bolometric luminosity of 20-24\%.
245: %PAH emission dominate the L-band spectrum, hydrocarbon absorption is absent while $H_2$ emission is present therefore this source is classified as a SB (Imanishi \& Dudley 2000).\\
246: %Multiband analysis of NGC 6240 gives back an incomplete picture: a) the double AGN is clearly detected only in X-ray observations. b) Optical data and mid-IR continuum indicate the SB process, the AGN is weakly point out by the coronal line detected in mid-IR. c) L-band spectrum is typical of a SB.
247: %It follow that both a double AGN and a SB component contribute to NGC 6240 infrared emission, however: 1) there is not observation giving us a complete information (AGN detected in X-ray, without reveal the SB component). 2) in no case we can disentangle the AGN/SB contributions.
248:
249: \section{Observations and data reduction}
250: The L and M-band spectra were obtained with the instrument ISAAC at the UT1 unit of the Very Large Telescope (VLT) on Cerro Paranal, Chile. All L-band (except for two sources) and M-band data were obtained during two nights on July 28-30, 2004. L-band spectra of IRAS 19254-7245 and IRAS 20551-4250 were obtained on June 2-4, 2002.\\
251: Sky conditions were photometric, with a seeing always of the order of, or better than 1 arcsec in the K-band.
252: The observations were performed with a 90~arcsec long slit, with width of 0.6 or 1.0 arcsec depending on the seeing (see Table~2 for all the relevant observation details).\\
253: The total exposure time varied from $\sim$1 hour for the brightest sources to 2 hours for the faintest one.\\
254: %Sky conditions were photometric, with a seeing always of the order of, or better than 1 arcsec in the K-band.
255: %The observations were performed with a 90x1~arcsec slit for IRAS 05189-2524 IRAS 23060+0505 (L and M-band observations) IRAS 19254-7245 IRAS 20551-4250 (L-band), and a 90x0.6 arcsec slit for NGC 6240 (L and M-band) IRAS 19254-7245 IRAS 20551-4250 (M-band).
256: %The spectral resolution in the L-band (wavelength coverage $\sim 2.9-4.1~\mu m$) was $\lambda/\Delta\lambda\sim360$ for the 1 arcsec slit and $\lambda/\Delta\lambda\sim600$ for the 0.6 arcsec slit, while in the M-band ($\sim4.5-5.1~\mu m$) $\lambda/\Delta\lambda\sim600$ (1 arcsec slit) and $\lambda/\Delta\lambda\sim800$ (0.6 arcsec slit).\\
257: %The total exposure time varied from $\sim$1 hour for the brightest sources to 2 hours for the faintest one.\\
258: %The main data regarding the observations are listed in Table \ref{tb:obs}.\\
259: \begin{table*}[!t]
260: \begin{center}
261: \begin{tabular}{rcccclcccc}
262: \hline
263: Source & Band & T$^a$ & W$^b$ & $\lambda/\Delta\lambda$ $^c$ & Standard Star & Type & Mag$^d$ \\
264: \hline
265: IRAS 05189-2524 & L & 45 & 1.0 & 360 & Hip 115033 & B5V & 4.80 \\
266: & M & 50 & 1.0 & 600 & HR 07221 & B5IV & 4.61 \\
267: \hline
268: NGC 6240 & L & 60 & 0.6 & 600 & HR 05249 & B2IV-V & 4.57 \\
269: & M & 90 & 0.6 & 800 & HR 05249 & B2IV-V & 4.53 \\
270: \hline
271: IRAS 19254-7245/1 & L & 60 & 1.0 & 360 & Hip 077126 & B3V & 5.92 \\
272: /2 & L & 45 & 1.0 & 360 & Hip 000183 & B4V & 5.42 \\
273: & M & 90 & 0.6 & 800 & HR 094703 & B4IV & 4.94 \\
274: \hline
275: IRAS 20551-4250 & L & 60 & 1.0 & 360 & Hip 076126 & B3V & 5.92 \\
276: & M & 60 & 0.6 & 800 & Hip 100881 & B4V & 5.23 \\
277: \hline
278: IRAS 23060+0505 & L & 60 & 1.0 & 360 & Hip 115033 & B5V & 4.80 \\
279: /1 & M & 30 & 1.0 & 600 & Hip 115033 & B5V & 4.82 \\
280: /2 & M & 30 & 1.0 & 600 & HR 07221 & B5IV & 4.61 \\
281: /3 & M & 60 & 1.0 & 600 & HR 07221 & B5IV & 4.61 \\
282: \hline
283: \end{tabular}
284: \end{center}
285: \caption{Observation log for our ISAAC program. IRAS 19254-7245 was observed twice in the L-band, and
286: IRAS 23060+0505 three times in the M-band. Notes: $^a$: observing time in
287: minutes. $^b$: slit width (arcsec). $^c$: spectral resolution. $^d$: L/M magnitude of the standard star.}
288: \label{tb:obs}
289: \end{table*}
290: In order to avoid saturation due to the high background ($\sim3.9\;\mbox{mag/arcsec}^2$ in the L-band,
291: $\sim1.2\;\mbox{mag/arcsec}^2$ in the M-band) the spectra were taken in chopping mode with single exposures
292: of 0.56 s. The spectra were then aligned and merged into a single image. We performed a standard data reduction,
293: consisting of flat-fielding, background subtraction, and spectrum extraction, using the IRAF\footnote{The IRAF package is distributed by the National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation.} 2.11 package.\\
294: For each object we also acquired the spectrum of a standard star immediately before or after the target
295: observation. The star was chosen to have the same airmass as the target, within 0.1, and with L/M magnitude
296: between 4 and 6 (Table~2). This allowed us to obtain high quality spectra with 2 minute long observations. Corrections
297: for sky absorptions and instrumental response were obtained from the standard stars spectra, divided by their
298: intrinsic emission, assumed to be a pure Raleigh-Jeans in the 3-5 $\mu m$ range (all the standards
299: were B stars).\\
300: In order to facilitate the background subtraction, the observations were performed by nodding the source along
301: the slit (with a throw of 15 arcsec). However, the instrumental response is not constant along the slit direction.
302: We compensate for this effect by performing a separate reduction for each of the three spectra relative to the three
303: different positions along the slit. The results before the division by the three standard stars spectra
304: (obtained in the same way) were in general rather different one from the other, both in absolute flux and spectral
305: shape. However, after the division by the stars, the three spectra of all sources overlapped within the errors.\\
306: %This gives us a self-consistency check for our spectral reduction procedure.\\
307: With the aim of obtaining a precise absolute calibration, we took into account aperture effects by analyzing the
308: profiles of both the targets and the standard stars along the slit. We assumed a Gaussian profile and we estimated
309: the fraction of flux inside the slit by assuming a perfect centering. The relative error is
310: $<10\%$ for the L-band, and $<20\%$ for the M-band. The precision of the absolute calibration is confirmed by
311: a) observations of the same source performed in two different nights, with two different standard stars, providing
312: final spectra consistent within the errors; b) the agreement between our spectra and those of
313: Imanishi \& Dudley (2000) for three common sources; and c) the cross calibration between our L-band and M-band spectra,
314: which match within 15\%, as expected from our estimates of the absolute calibration errors.\\
315: For each source we obtained the final spectrum by merging the single spectra relative to the three slit positions,
316: and when two or more observations were present, merging the final calibrated spectra. The errors were estimated
317: from the Poissonian noise in the sky counts, which are by far the dominant source of noise (for ease of comparison,
318: a typical ULIRG has a magnitude $L\sim11-13$, i.e. at least $\sim600$ times fainter than the sky background in the
319: same extraction region).
320: The relative error due to the standard division is negligible when compared to the former.\\
321: The atmospheric transmission is not constant in the $3-5~\mu m$ range. In several narrow spectral intervals in
322: the M band, and one in the L band, the transparency is too low to obtain significant data. We excluded these
323: intervals from our spectral analysis. In other wavelength ranges, such as between 2.9 and 3.2 $\mu m$, and in
324: the M band in general, the atmospheric conditions vary on short timescales. This implies that the sky features
325: are not completely removed with the division by the calibration star. In these intervals of low spectral quality,
326: we rebinned the spectral channels above the instrument resolution, in order to have statistically significant
327: measurements in each spectral point.
328: Finally, for the spectral intervals with good atmospheric transmission and high signal-to-noise ratio,
329: % (such as the brightest nucleus in the 3.4-4.1~$\mu$m interval),
330: we chose the rebinning in order to have $\Delta\lambda<30$ nm, which
331: is a factor of 2 smaller than the spectral resolution.
332:
333: %We fitted the spectra with a power-law continuum and broad Gaussian absorption and emission lines. We left the relative normalizations between the L and M bands free to vary within the uncertainties given above.\\
334: Two out of five sources (IRAS 19254-7245 and NGC~6240) are known mergers.
335: %: in one case (IRAS 19254-7245) two distinct nuclei are resolved in infrared imaging, while in the case of NGC 6240 the nuclei can be distinguished in optical and X-rays observations but not still in infrared.
336: For these sources we chose the slit angle in order to obtain spectra for both nuclei. In the case of NGC~6240
337: we were able to obtain high quality spectra of both nuclei even if their emission wings overlap. A detailed
338: description of the data reduction and analysis for NGC~6240 spectra are reported in Risaliti et al. 2006B. For
339: IRAS 19254-7245 we obtained a high quality spectrum of the brighter nucleus, which accounts for $\sim$95\% of
340: the observed infrared emission, and a low S/N one for the fainter one, sufficient to estimate the flux and
341: continuum slope, but not for a detailed spectral analysis.
342: % None of the spectral features used as indicators to characterize the presence of an AGN are present,
343: Therefore, here we do not discuss the fainter nucleus in the following analysis.\\
344: The final L and M-band spectra of our sources are shown in Figure 1.
345:
346: \section{Data analysis}
347: We present and briefly discuss the spectrum of each source, disentangling the AGN and SB contributions by means of L-band
348: parameters and comparing our results with those reported in Section 2.1. We fitted our spectra with a power-law continuum
349: and broad Gaussian absorption and emission features, using the CIAO 3.3 software package (Fruscione et al.~2006). Due to the small wavelength coverage, to the low S/N ratio, and to the presence of broad absorptions, we left the relative normalization between the L and M bands free to vary within the uncertainties given above, i.e. $15\%$.\\
350:
351:
352: %The reduced spectra relative to our sample are in Figure \ref{fg:spettri}\\
353: In order to estimate the relative contribution of the AGN to the $3-5~\mu m$ ULIRG emission, we use a
354: simple model to disentangle the two components. We assume that the intrinsic spectrum of
355: the AGN, $f_\lambda$(AGN), is a power law ($f_\lambda\propto\lambda^\Gamma$) with spectral index $\Gamma_{int}=-0.5$, in agreement with the L-band
356: spectra of ULIRGs dominated by unabsorbed AGNs (R06), and with L-band spectra of pure type~1 AGNs (Imanishi \& Wada~2004).
357: The observed AGN spectrum is obtained from the intrinsic
358: spectrum, absorbed by a wavelength-dependent optical depth, $\tau(\lambda)\propto\lambda^{1.75}$
359: (Cardelli et al.~1989). The intrinsic starburst component, $f_\lambda$(SB), is modelled by a continuum with $\Gamma_{int}=-0.5$
360: (as for unobscured AGNs) plus a broad emission feature at $3.3~\mu m$ with $EW_{SB}=110$~nm.
361: This is in agreement with the average observed L-band spectra of starburst-dominated ULIRGs (Imanishi \& Dudley~2000, R06)\footnote{We note that our pure starburst spectra
362: differ from the average spectra of lower-luminosity
363: normal galaxies (which have $\Gamma\sim-2$, Lu et al. 2003), probably due
364: to a larger contribution by hot dust in powerful SB.}. The ULIRG emission, $f_\lambda$, is obtained as the combination of the two components:
365: \begin{equation}
366: f_\lambda=\alpha f_\lambda\mbox{(AGN)}e^{-\tau(\lambda)}+(1-\alpha)f_\lambda\mbox{(SB)},
367: \label{fm:dec}
368: \end{equation}
369: where $\alpha$ is the fraction of the L-band luminosity due to the AGN. We express the optical
370: depth as $\tau(\lambda)=\tau_L\times (\lambda/3.3~\mu m)^{-1.75}$, where $\tau_L$ is the optical depth
371: at $3.3~\mu m$. Therefore, the model has two free parameters, $\alpha$ and $\tau_L$. These parameters and
372: their errors can be computed from the observed $\Gamma$ and $EW_{3.3}$ in the composite spectrum as follows:
373: \begin{equation}
374: \tau_L=\frac{EW_{SB}(\Gamma-\Gamma_{int})}{\beta(EW_{SB}-EW_{3.3})}
375: \end{equation}
376: and
377: \begin{equation}
378: \alpha=\frac{EW_{SB}-EW_{3.3}}{EW_{SB}-EW_{3.3}(1-\mbox{e}^{-\tau_L})},
379: \end{equation}
380: where $\beta=1.75$ is the slope of extinction curve in the L-band.
381:
382: % From the composite spectrum, the equivalent width of the $3.3~\mu m$ feature and the continuum slope can be computed as a function of the above parameters.
383: %Therefore, we can find the unique solution for $\tau_L$ and $\alpha$ which reproduces the observed values of $EW_{3.3}$ and $\Gamma$.
384: %We estimate the errors from measured $EW_{3.3}$ and $\Gamma$ ones.
385: An additional source of systematic errors is the uncertainty on the intrinsic AGN and SB spectral
386: slope, and, more important, on the intrinsic EW$_{SB}$ in SBs, which is found to span from $\sim80$~nm to $\sim150$~nm in a pure starburst sample (see R06 for further details). If these are taken into account, the true errors could be larger than the final one by
387: a factor of $\sim2$ (for further details on this model and on errors estimate, we refer to R06 where the model is also applied to a larger sample of ULIRGs).\\
388: %, in which the model is also applied to a larger sample of ULIRGs.}.
389: In order to give a quantitative estimate of this uncertainty, we have listed in Table~4 the model parameters of Eq.~3 and Eq.~4 by assuming different values of the intrinsic PAH equivalent width ($EW_{SB}$). As expected from the Equations, by increasing $EW_{SB}$, $\tau_L$ decreases and $\alpha$ increases. However the parameters are consistent within the errors for all the sources; IRAS~05189-2524 and NGC~6240~S present discrepancies of the order of $3-5~\%$ due to our simple model, which does not reproduce details observed in these sources such as a complex continuum (IRAS~05189-2524), or a broad emission line (the Br$\alpha$
390: in NGC~6240~S). The L-band absorption, $\tau_L$, varies a lot (50\%) only in IRAS~20551-4250, discussed in detail below.
391:
392: The results of our spectral analysis are summarized in Table~\ref{tb:l}.\\
393:
394: The above analysis is based on the continuum and emission features.
395: The broad absorption features present in our L-band and M-band spectra (Fig.~2) were analyzed separately.
396:
397: The optical depth of the absorption features were estimated with a Gaussian fit, and correcting for the starburst
398: continuum dilution, as estimated using the model described above, in order to calculate the optical depth
399: with respect to the AGN component only. Specifically, we adopted the following
400: estimate:
401: \begin{equation}
402: \tau_\lambda = -\ln \frac{(F_\lambda(\mbox{\scriptsize{TOT}})-A_\lambda)-F_\lambda(\mbox{\scriptsize{SB}})}{F_\lambda(\mbox{\scriptsize{AGN}})}=-\ln\frac{F_\lambda(\mbox{\scriptsize{AGN}})-A_\lambda}{F_\lambda(\mbox{\scriptsize{AGN}})}
403: \end{equation}
404:
405: where $A_\lambda$ is the Gaussian amplitude, and $F_\lambda(\mbox{\scriptsize{TOT}})$,$F_\lambda(\mbox{\scriptsize{SB}})$ and $F_\lambda(\mbox{\scriptsize{AGN}})$ are
406: the total, SB and AGN continuum fluxes. All the values refer to the central wavelength of the
407: absorption feature (3.4~$\mu$m or 4.6~$\mu$m).
408:
409: In order to better investigate the properties of the CO absorption features, we compared our M-band spectra with the
410: low
411: resolution Spitzer spectrum of IRAS00183-7111 (Spoon et al.~2004) and high resolution ISAAC data of NGC4945
412: (Spoon et al.~2003). Due to the low resolution and the high atmospheric noise of the M-band spectra,
413: we cannot use a complex analytical model like in the case of IRAS 00183-7111 (Spoon et al.~2004), nor can we try and reproduce the
414: M-band features by varying the contributions of different CO ice and gas components as for NGC~4945 (Spoon et al.~2003 and
415: references therein). However, we compared the position
416: and the full width at half maximum of the absorption features with the ones of those high signal-to-noise spectra.
417: The results for each single source are discussed below.\\
418:
419:
420: \begin{figure*}[!h]
421: \begin{center}
422: % \vspace{-.5cm}
423: \includegraphics[scale=0.40]{f1.eps}
424: \end{center}
425: % \linespread{1.0}
426: % \vspace{-.8cm}
427: \caption{$\lambda-f_\lambda$ plot of our sources; spectral points and continuum slope are reported
428: and the principal features, with measured physical parameters (see Table~3) are indicated. Bad atmospheric regions are in yellow.}
429: % \linespread{1.6}
430: \label{fg:spettri}
431: \end{figure*}
432:
433:
434: {\bfseries IRAS 05189-2524}\\
435: $\bullet$ The spectral index of the $\lambda-f_\lambda$ continuum is $\Gamma=-0.546\pm0.004$, typical of starburst
436: and unobscured AGN (R06). The continuum steepness below $2.9~\mu m$ is interpreted as due to the host galaxy
437: emission in the K-band, therefore we have neglected these points in determining the slope.
438: %In any case our choice does not affect the parameter determination thanks to the large wavelengths range and good statistic.
439: The PAH emission at $3.3~\mu m$ has $EW_{3.3}=4.0\pm0.2\;nm$, almost two orders of magnitude smaller than in a
440: pure starburst (Imanishi \& Dudley 2000, R06), thus indicating the presence of an AGN. As shown in
441: Figure \ref{fg:spettri}, a simple power law appears inadequate to reproduce the details of the continuum emission.
442: In particular, the slope $\Gamma$ increases at wavelengths $\lambda > 3.8~\mu m$.
443: %This is due to the high quality of this spectrum (the highest S/N in our sample) which make it possible to reveal local deviations from a pure power law continuum.
444: We will not discuss any further these small deviations only visible in the highest S/N cases, and will use
445: the average spectral slope in order to analyze in a homogeneous way all the sources in the sample.\\
446: $\bullet$
447: The M-band spectrum of IRAS~05189-2524 shows possible absorption features due to $\mbox{CO}_2$ (centered at $4.26~\mu m$ rest frame) and CO (Fig.~1 and Fig.~2).
448: This interpretation is
449: not unique due to the uncertainty in the absolute calibration. It could be possible, within the calibration errors, to
450: interpret the spectral points at 4.7-5.0~$\mu$m as the continuum level, and the higher flux points at $\sim4.6~\mu$m as due to an emission feature.
451: However, no such a feature is known at the observed wavelengths. Moreover,
452: comparing our spectra with literature ones, the positions and FWHM of the two Gaussians reproducing the $4.8~\mu m$
453: absorption profiles are consistent within the errors with Spitzer data (Spoon et al.~2004).
454: Based on these considerations,
455: we interpret the observed spectral shape as due to CO absorption features, and therefore we choose the normalization of the M-band spectrum in order to obtain a good match between the extrapolated L-band
456: continuum and the highest flux points in the $4.5-5~\mu m$ interval. With this choice, the cross-calibration factor
457: between L-band and M-band is 0.98.
458: IRAS 05189-2524 shows a spectral profile analogous to IRAS 00183-7111, with an
459: optical depth of CO absorption $\tau_{4.6}\sim0.1$. $\mbox{CO}_2$ absorption is also partially observed at the
460: lower end of the M-band spectral range.
461: Beside broad absorption features, a series of H$_2$ and other atomic emission lines are visible in the M-band
462: spectrum. However, it is impossible to estimate their properties for two reasons, both due to the low resolution of our spectra:
463: molecular sky lines are not well subtracted, and a forest of fundamental rotovibrational absorption lines of gaseous
464: CO (as detected in NGC 4945 by Spoon et al.~2003) are blended with the emission lines.\\
465: If the measured spectral parameters are interpreted according to the model described above, the 3-5~$\mu m$ emission of
466: IRAS 05189-2524 is dominated by an AGN ($\alpha>0.9$) with little or no absorption ($\tau_L<0.03$). A significant
467: starburst contribution is also present, as revealed by the PAH emission line.
468: % On the whole our interpretation is
469: On the whole our interpretation is consistent with published data (Imanishi \& Dudley~2000).
470: %, but the $3.4~\mu m$ absorption revealed by Imanishi \& Dudley (2000) is not real, depending on a low S/N and less accuracy in the continuum determination.
471:
472: {\bfseries IRAS 23060+0505}\\
473: This is the most luminous and highest redshift source in our sample. The observed spectrum closely resembles that
474: of IRAS 05189-2524 (Fig. \ref{fg:spettri}), where the emission of an unobscured AGN dominates the 3-5~$\mu$m
475: spectrum.\\
476: $\bullet$ Despite the high flux of the source, the spectrum is one of the noisiest in our sample, due to bad
477: atmospheric conditions during the observation. The only feature present in the L-band spectrum is a weak
478: PAH emission with
479: $EW_{3.3}=1.8\pm0.7\;nm$. Therefore, an AGN is present and dominates the L-band emission, producing a strong
480: dilution of the PAH line. The continuum spectral slope is
481: $\Gamma=-0.36\pm0.02$ consistent with a SB and an unobscured AGN.\\
482: $\bullet$ No narrow features are detected in the M-band.\\
483: % spectrum due to the high background noise. %At $3.95~\mu m$ (rest frame) we identify the [Si IX] coronal line. With a ionization potential of 330 eV it clearly indicates the presence of an AGN, having a flux ($\sim1.2\times10^{-13}erg\;s^{-1}cm^{-2}$) consistent with Sy2 values (Lutz, Maiolino et al. 2002).
484: The normalization factor for the M-band is 1.09, within calibration uncertainties.\\
485: Applying the analytical model described above we find that the L-band emission is dominated by the AGN emission
486: ($\alpha>0.9$) with no observed absorption ($\tau_L<0.09$). Our analysis is consistent with the previous
487: ones (Section 2.1 and Imanishi \& Dudley 2000).\\
488: %IRAS 05189-2524 and IRAS 23060+0505 represent ULIRG with a $3-5~\mu m$ emission dominated by unobscured AGN plus a small contribution of SB. Therefore hot dust, illuminated by the direct emission of the accretion disc, produces the observed spectrum. Lack of hydrocarbon and H$_2$O ice absorption points out the absence of a compact absorber. The L-band starburst contribution is strongly diluted, even if clearly detected. As we will see quantitatively in Par. such a contribution can be relevant, if not dominant, for the total luminosity of these two sources.
489:
490: {\bf IRAS 19254-7245}\\
491: IRAS~19254-7245 is a merging system with two nuclei separated by 10~kpc (Mirabel Lutz \& Maza 1991). In our ISAAC observation the two nuclei are clearly distinguishable.
492: %\footnote{For further details on this extraction technique see Risaliti et al.~2006B.}.
493: However, as reported in Section 2.1, the northern nucleus is too faint to obtain a spectrum with a signal-to-noise ratio good enough for a detailed analysis. Therefore, we will focus on the southern nucleus, accounting for $\sim$95\% of the observed 3-5~$\mu$m emission.
494: The high quality spectrum of the southern nucleus can be used to analyze the substructures of the absorption features at 3.4~$\mu$m and in the 4.5-5~$\mu$m range. An accurate analysis of the L-band spectrum is presented in Risaliti et al. (2003). Here we report new measurements of the spectral properties obtained by adding the M-band data.\\
495: $\bullet$ The L-band spectrum shows both PAH emission, with an equivalent width $EW_{3.3}=24.2\pm1.8\;nm$,
496: and two clear absorbed AGN features: a steep continuum ($\Gamma=2.22\pm0.03$), and a strong aliphatic hydrocarbons absorption
497: ($\tau_{3.4}=0.82\pm0.05$), three times higher than the maximum value expected if the dust is spatially mixed with the
498: emitting source ($\tau\sim0.2$, Imanishi \& Maloney 2003).\\
499: $\bullet$ As in the case of IRAS 05189-2524, the observed $4-5~\mu m$ absorption is ascribed to vibrational
500: modes of the CO molecule. We assume that the high flux points at rest frame wavelength $\lambda\sim4.4 \mu$m are at
501: the continuum level, and therefore lay on the extrapolation of the L-band continuum (Fig.~1 and Fig.~2). With this choice the
502: normalization factor
503: between the L-band and the M-band is 1.13.\\
504: Our low resolution spectrum does not allow a detailed analysis of the single substructures. However, it is
505: possible to compare the overall absorption profile with that due to gaseous CO
506: (such as in IRAS~00183-7111, Spoon et al.~2004) and that due to solid CO (as NGC~4945,
507: Spoon et al.~2003). This comparison is further
508: complicated by the redshift of our sources, which moves the end of the absorption
509: profile out of the M-band (Fig.~1). Despite these limitations, we can qualitatively assess that the
510: absorption profile seems to be due to gaseous CO with $\tau_{4.6}=0.44\pm0.04$ (Fig.~2), and we reproduce the absorption profile with two Gaussians
511: centered within the errors at the expected position for the P and R-branches of the vibration mode of CO gas.
512:
513:
514: \begin{figure*}[!h]
515: \begin{center}
516: % \vspace{-.5cm}
517: \includegraphics[scale=0.45]{f2.eps}
518: \end{center}
519: % \linespread{1.0}
520: % \vspace{-.8cm}
521: \caption{Absorption features in the spectra of IRAS~05189-2524, IRAS~19254-7245, IRAS~20551-4250 and NGC~6240: the normalized flux profiles are shown in red. The broad double-branched features centered at $4.67~\mu$m (rest frame) are identified
522: with the P- and R- branches of the fundamental vibration mode of CO gas; the red wing of CO$_2$ feature ($\lambda_{rest}=4.26~\mu$m) is detected only in IRAS~05189-2524.
523: We obtain these profiles by fitting the M-band data with the proper continuum powerlaw and a set of Gaussian curves.}
524: % \linespread{1.6}
525: \label{fg:assorbimenti}
526: \end{figure*}
527:
528:
529: This is also suggested by the lack of H$_2$ emission lines $\lambda4.69~\mu$m, indicating a
530: temperature higher than $\sim$2000~K.\\
531: % is not revealed in our spectrum, probably due to a so hot medium that H$_2$ sublimes. We attribute the observed absorption to CO gas of which we can see only the R-branch of the fundamental vibrational mode (the P-branch is invisible because of the redshift). For this reason the M-band points with an higher flux density are considered part of the continuum and the data above $4.7~\mu m$ (observer frame) belong to the CO absorption, and the normalization factor is 1.13. We estimate optical depth $\tau_{4.6}\sim0.4$ and ascribe CO absorption to the presence of an absorbed AGN consistently with the interpretation for IRAS 00183-7111.\\
532: The analytical model confirms these results: an AGN dominates the L-band emission ($\alpha>0.9$), and it is
533: absorbed by a large amount of dust which reddens the spectrum
534: with an absorption of $\tau_L\sim2$. This source has been observed in the $3-5~\mu m$ range by our group for the
535: first time, and the results are consistent with X-rays observations,which also indicates the presence of a powerful, obscured
536: AGN (Braito et al.~2003).
537:
538: {\bfseries IRAS 20551-4250}\\
539: The spectrum of IRAS 20551-4250 (Fig.~\ref{fg:spettri}) is similar to that of IRAS 19254-7345, with a steep inverted
540: continuum and deep absorption features. The L-band spectrum has been discussed in R06. Here we add the M-band
541: spectrum to the analysis.\\
542: % In this case we include also the spectral points relative to the bad $3.18-3.31~\mu m$ atmospheric band (in yellow in Figure \ref{fg:spettri}) thanks to the special good observing conditions, which in the over observation are so bad that the error bars are as large as the entire plotted flux range.
543: %Our new measure considering also M-band data are as follows.\\
544: $\bullet$ The L-band spectrum shows both a $3.3~\mu m$ PAH emission with a large equivalent width $EW_{3.3}=90\pm7\;nm$,
545: pointing to a powerful SB, and two clear absorbed AGN features: a steep inverted continuum with $\Gamma=3.10\pm0.03$,
546: and a significant aliphatic hydrocarbons absorption $\tau_{3.4}=1.5\pm1.1$.
547: We note that this is the case with the largest correction of $\tau_{3.4}$ due to SB dilution with respect to the observed value
548: (observed $\tau_{3.4}=0.3\pm0.1$, R06). Also, the wavelength of the absorption peak is shifted to $\sim3.47~\mu$m. This is
549: probably due to a high abundance of electronegative groups such as -OH and -N0$_2$ in the aliphatic hydrocarbon dust molecules (see R06 for a detailed discussion).
550: \\
551: % and chemical analysis of the ices absorptions).\\
552: $\bullet$ The M-band features consist of a deep CO absorption and a Pf$\beta$ ($4.65~\mu m$) emission line. Thanks
553: to the low redshift ($z=0.0428$) we can observe the P- and R-branches of gaseous CO, in which the absorptions forest
554: identified in NGC4945 can not be resolved, however, because of our low spectral resolution combined with the high atmospheric
555: variability (Fig.~1 and Fig.~2).
556: % generate the huge observed noise.
557: %Even if for we distinguish two CO vibrational modes,
558: If the whole absorption feature is modeled with one Gaussian broad line, the peak position and FWHM are consistent with gaseous CO and we estimate an optical depth
559: $\tau_{4.6}=2.2\pm0.6$.
560: Atmospheric turbulence and CO absorption bands prevent us from measuring the profile and equivalent width of the
561: Pf$\beta$ line.
562: The cross correlation factor between L and M bands is 0.83.\\
563: %Likewise IRAS 19254-7245, a strong dust absorbed AGN (reddened continuum, hydrocarbons and molecular gas absorptions) combined with a significant circumnuclear starburst (UV excess gives PAH transitions), generate IRAS 20551-4250 spectrum.\\
564: Our decomposition model confirms that the L-band spectrum is dominated by an AGN ($\alpha>0.9$) with the
565: highest dust reddening ($\tau_L\sim11$) in our sample.
566: This high value of $\tau$ is necessary in order to reconcile two apparently contradictory observational results: the high
567: equivalent width of the 3.3~$\mu$m feature, implying a dominance of the starburst at $\lambda_{rest}=3.3$~$\mu$m,
568: and the steep continuum, implying a dominant contribution of an absorbed AGN at $\sim4 \mu$m.
569: Anyway, we note that the estimated L-band optical depth is affected by a large relative error, of the order of $40\%$. Furthermore, we note that the estimate of $\tau$ can be strongly affected by our assumption of a fixed value of
570: EW$_{3.3}$ with respect to the pure starburst continuum. If we allow for larger values of this parameter we
571: increase the fraction of continuum due to the AGN. As a consequence, the observed continuum slope can be
572: reproduced with a less absorbed AGN component. For example, if we assume an intrinsic EW$_{3.3}=130$~nm instead
573: of the standard value
574: of 110~nm (well within the spread observed in pure starbursts, R06) we obtain $\tau=6.7\pm1.3$; varying $EW_{3.3}$ in the range $110-150$~nm we obtain a $\tau_L$ ranging from 11 to 5.
575:
576: {\bfseries NGC 6240}\\
577: VLT data pertaining to NGC~6240 have been discussed in a separate paper (Risaliti et al.~2006B); in the following
578: we report a summary of our result for completeness and comparison with the other four bright ULIRGs.
579: NGC 6240 is the only source in our sample in which we are able to study both nuclei in detail.
580: Our results confirm the early detection of the double AGN obtained in the hard X-rays with Chandra (Komossa et
581: al. 2003): in the brightest nucleus (see Figure \ref{fg:spettri}) a broad Br$\alpha$ emission line is detected
582: (with a FWHM corresponding to a velocity of $1800\pm200$~km~s$^{-1}$), the only broad line region evidence
583: identified in the spectrum of this source.
584: The continuum is flat ($\Gamma=-0.17\pm0.03$) and a PAH emission is clearly detected, with $EW_{3.3}=48\pm2$ nm,
585: a factor $\sim2$ smaller than in a pure starburst.\\
586: The faintest nucleus (see Figure~\ref{fg:spettri}, and Figure~2) shows a steep $3-5~\mu m$ emission ($\Gamma=1.28\pm0.17$),
587: typical of a reddened AGN (NGC~6240N $\tau_L\sim2$), as in the cases of IRAS~20551-4250 and IRAS~19254-7245, with a strong
588: CO absorption feature in the M band ($\tau_{4.6}=1.5\pm1.4$). The only difference between NGC~6240N and the above IRAS sources consists in a small hydrocarbon absorption feature
589: (with an upper limit for the optical depth $\sim0.05$),which indicates a possible different dust composition (see discussion in Section 5.2).\\
590: %We note that thanks to the low redshift it would be possible to detect CO$_2$ absorption in this source; however
591: %the brightest nucleus (NGC~6240~S) is unobscured, while the NGC~6240~N spectrum has a low S/N ratio, making impossible any detection of
592: %such features below
593: %4.4~$\mu$m (detector frame).
594:
595: Summarizing our results (see Table~3), we can interpret the data as follows: the starburst process is always present and an AGN
596: is clearly detected
597: in all sources. A different amount of dust along our line of sight generates the observed spectral differences.
598: In particular the large amount of gas and dust producing the $3.4~\mu m$ and $4.6~\mu m$ absorption bands together
599: with a reddened continuum can be related only with a point source behind a screen of dusty gas, such as an AGN. Our M-band spectra show that
600: the presence of deep CO absorption features in the 4.6-5~$\mu$m range is always associated with a reddened AGN continuum.
601: %, being improbable that such material can be spread covering the SB distance scales as suggested by either quantitative consideration on hydrocarbons absorption (Imanishi \& Maloney 2003) and comparative observation between type 2 Seyfert galaxies and ULIRGs (Imanishi \& Dudley 2000, R06).
602: % We can explain these four ULIRGs spectra in a simple manner: an AGN is always present with the typical starburst emission, than the spectral differences can be ascribed to a different amount of dust along our line of sight.
603: %Moreover this interpretation, through which is possible to distinguish two extreme cases (ULIRGs hosting an unobscured AGN and sources with obscured AGNs), is useful to understand the NGC 6240 ambiguous ULIRG.
604: % IRAS 19254-7245 and IRAS 20551-4250 are representative of ULIRGs with $3-5~\mu m$ emission dominated by obscured AGNs
605:
606: %In conclusion, comparing NGC 6240 spectra with the extreme cases illustrated, we individuate an AGN in both nuclei: an unobscured ones in southern (brightest) nucleus, and highly obscured AGN in northern (faintest) nucleus, confirming Komossa et al. (2003) X-ray analysis. Moreover we can quantitatively estimate the relative AGN/SB contribution to the L-band emission and the correspondent dust absorption for both nuclei: for the Southern one the lowest $\alpha$ coefficient of 0.7\\
607: %In Table~\ref{tb:l} we summarize the results of the spectral analysis of our five sources.
608: \begin{table*}
609: \begin{center}
610: \begin{footnotesize}
611: \begin{tabular}{lcccccccc}
612: \hline
613: Source & $\Gamma$ & $EW_{3.3}$ & $\tau_{3.1}$ & $\tau_{3.4}$ & $\tau_{4.6}$ & $\tau_L$ & $\alpha$ (\%) & $\alpha_{\tiny{BOL}}$ (\%)\\
614: \hline
615: NGC 6240 S & $-0.17\pm0.03$ & $48\pm2$ & $0.2\pm0.05$ & -- & -- & $0.34\pm0.03$ & $64\pm4$ & $<1$ \\
616: NGC 6240 N & $1.28\pm0.17$ & $50\pm12$ & -- & $<0.05$ & $1.5\pm1.4$ & $1.4\pm0.1$ & $80\pm20$ & $2\pm1$\\
617: I 05189-2524 & $-0.55\pm0.01$ & $4.0\pm0.2$ & -- & -- & $0.118\pm0.007$ & $<0.03$ & $>95$ & $11.3\pm0.5$ \\
618: I 19254-7245 & $2.21\pm0.03$ & $24\pm2$ & -- & $0.82\pm0.05$ & $0.44\pm0.04$ & $1.82\pm0.02$ & $>95$ & $10\pm1$\\
619: I 20551-4250 & $3.10\pm0.03$ & $90\pm7$ & -- & $1.5\pm1.1$ & $2.2\pm0.6$ & $5-11^a$ & $>95$ & $>90$ \\
620: I 23060+0505 & $-0.36\pm0.02$ & $1.8\pm0.7$ & -- & -- & -- & $<0.09$ & $>95$ & $25\pm8$\\
621: \hline
622: \end{tabular}
623: \end{footnotesize}
624: \end{center}
625: % \linespread{1.0}
626: \caption{{\footnotesize Columns 2-5: best fit estimates for the continuum slope
627: $\Gamma$ ($f_\lambda\propto\lambda^\Gamma$), the
628: equivalent width of the 3.3~$\mu$m PAH feature, $EW_{3.3}$ in units of nm, and the optical depths of the absorption features
629: due to water ice ($\lambda\sim3.1~\mu$m), aliphatic hydrocarbons ($\lambda\sim3.4~\mu$m) and CO ($\lambda\sim3.4~\mu$m). The optical depth values are toward the AGN emission, after subtracting the starburst contribution.
630: Columns~7-9: estimated values of the continuum extinction $\tau_L$, the contribution of the AGN component to the L-band luminosity ($\alpha$) and to the bolometric luminosity
631: ($\alpha_{\tiny{BOL}}$), obtained, respectively, from Eq.~3, Eq~4 and Eq.~5.}}
632: % \linespread{1.6}
633: \label{tb:l}
634: \end{table*}
635: %
636: \begin{table*}
637: \begin{center}
638: \begin{tabular}{l|cc|cc|cc}
639: & \multicolumn{2}{c}{EW$_{SB}=80$~nm} & \multicolumn{2}{c}{EW$_{SB}=110$~nm} & \multicolumn{2}{c}{EW$_{SB}=150$~nm} \\
640: \hline
641: Source & $\tau_L$ & $\alpha$ & $\tau_L$ &$\alpha$ & $\tau_L$ & $\alpha$ \\
642: \hline
643: NGC 6240 S & $0.5\pm0.1$ & $51\pm4$ & $0.34\pm0.03$ &$64\pm4$ & $0.3\pm0.03$ & $74\pm4$ \\
644: NGC 6240 N & $2.7\pm1.3$ & $>90$ & $1.4\pm0.1$ &$80\pm20$ & $1.5\pm0.3$ & $>90$ \\
645: IRAS 05189-2524 & $<0.03$ & $>95$ & $<0.03$ &$>95$ & $<0.01$ & $>95$ \\
646: IRAS 19254-7245 & $2.2\pm0.1$ & $>95$ & $1.82\pm0.02$ &$>95$ & $1.85\pm0.05$ & $>95$ \\
647: IRAS 20551-4250 & - & - & $8\pm3$ &$>95$ & $5.1\pm0.7$ & $>95$ \\
648: IRAS 23060+0505 & $<0.08$ & $>95$ & $<0.09$ &$>95$ & $<0.08$ & $>95$ \\
649: \hline
650: \end{tabular}
651: \end{center}
652: \caption{{\footnotesize Comparison between the best fit values of $\alpha$ and $\tau$ obtained assuming three different values of $EW_{SB}$
653: (see text for details).
654: }}
655: \label{tb:conf}
656: \end{table*}
657:
658: \section{Discussion}
659: The analysis of our 3-5~$\mu$m spectra leads to the following important results: \\
660: $\bullet$ Assuming fixed AGN and starburst intrinsic templates, and allowing for a free relative contribution of the two components and a free absorption of the AGN component we can evaluate quantitatively the AGN and starburst contribution to the $3-5~\mu$m flux in all sources.\\
661: % This allows a quantitative estimate of the relative L-band contribution of the two components.\\
662: $\bullet$ In all sources with clear indication of an obscured AGN in the continuum spectrum, we find strong gaseous CO absorption features in the M-band.
663: %This confirms that these features are common among heavily obscured AGN
664:
665: %In the following we discuss in more detail these two issues.
666: \subsection{Estimates of the AGN and SB contributions}
667:
668: The estimates of the relative contributions of the AGN and starburst components to the L-band flux, presented in
669: Section~4 and listed in Table~3, can be complemented with analogous estimates for the contributions to the
670: total luminosity.
671:
672: In order to achieve this, we need to take into account
673: the L-band to bolometric ratios of pure starbursts, $R_{SB}$, and pure AGNs, $R_{AGN}$.
674: From the analysis of a large sample of nearby ULIRGs (R06,
675: Risaliti, Imanishi \& Sani submitted) we estimate $R_{SB}\sim 5\times10^{-3}$ and $R_{AGN}\sim1$. The uncertainties
676: on these values, estimated by the spread of the single objects with respect to the average ratios, are of the
677: order of 30\%.
678: While there are no other ways to determine $R_{SB}$, it is possible to compare our estimate of $R_{AGN}$ with
679: the value obtained from quasar SEDs (e.g. Elvis et al.~1994). From these average SEDs, based on local PG quasars,
680: we obtain $R_{AGN}\sim0.5$\footnote{Assuming that in ULIRGs all the intrinsic emission is
681: reprocessed in the IR, we estimated the bolometric luminosity in QSOs from their total intrinsic emission, from near-IR to X-rays.}, which is consistent with our estimate, considering (a) the large spread in
682: quasar SEDs (Elvis et al~1994) and that recent works based on Spitzer data show that the SED of PG quasars
683: is also heavily contaminated by a starburst component (Netzer et al.~2007).
684:
685: % e qui dovrei dire di quanto?
686: The correction factor for the AGN contribution to bolometric luminosity is:
687: \begin{equation}
688: \alpha_{\mbox{\tiny{BOL}}}=\alpha/(\alpha+K(1-\alpha)).
689: \label{fm:abol}
690: \end{equation}
691: where $K=R_{AGN}/R_{SB}\sim200$.
692: The errors on this estimate are larger than those on $\alpha$, because of the high uncertainty on K (a factor of $\sim2$).
693: %ampliare un po'!!!!!!!!!!!!!!! quanto grande st'incertezza...
694: However,
695: it is easily seen from Eq.~\ref{fm:abol} that this does not affect our main conclusion, which
696: is $\alpha_{\mbox{\tiny{BOL}}}<1$, provided that K$\gg1$. The results are summarized in Table~\ref{tb:l} and Table~4.\\
697: To check the consistency of our model, and of Eq.~6, we estimate the expected bolometric flux and compare it
698: with the IRAS one. The expected flux is given by:
699: \begin{equation}
700: F_{\mbox{\tiny{BOL}}}=\frac{R_{\mbox{\tiny{AGN}}}\alpha+R_{\mbox{\tiny{SB}}}(1-\alpha)}{\alpha~e^{-\tau}+(1-\alpha)}F(3.3),
701: \label{eq:exp}
702: \end{equation}
703: where $F(3.3)$ is the observed continuum flux at 3.3~$\mu$m.
704: Fig.\ref{fg:flussi} show that the two values are in good agreement.
705:
706: \begin{figure*}[!h]
707: \begin{center}
708: \includegraphics[scale=0.70]{f3.eps}
709: \end{center}
710: \caption{Comparison between the total IR luminosity inferred with our model
711: from the observed L-band emission, assuming bolometric correction (see text for details), and the luminosity measured
712: by IRAS. Red lines mark the $\pm30\%$ relative error.}
713: \label{fg:flussi}
714: \end{figure*}
715:
716: %\begin{table*}
717: %\begin{center}
718: % \begin{footnotesize}
719: % \begin{tabular}{lcc}
720: % \hline
721: % \hline
722: % Source & $F_{\mbox{\tiny{BOL}}}$ & $F_{\mbox{\tiny{IRAS}}}/F_{\mbox{\tiny{BOL}}}$\\
723: % NGC 6240 & 6.0 & 0.34 \\
724: % I 05189-2524 & 2.5 & 0.55 \\
725: % I 19254-7245 & 2.5 & 0.21 \\
726: % I 20551-4250 & 3.6 & 0.28 \\
727: % I 23060+0505 & 0.59 & 0.26\\
728: % \hline
729: % \end{tabular}
730: % \end{footnotesize}
731: % \end{center}
732: % \linespread{1.0}
733: %\caption{{\footnotesize Comparison between the expected bolometric fluxes and the IRAS ones. Column~2:
734: %bolometric fluxes in unit of $10^{-9}$~erg~s$^{-1}$cm$^{-2}$ obtained through Eq.~\ref{eq:exp}. Column~3:
735: %Ratio between the bolometric flux and IRAS flux (as obtained using Eq.~\ref{Esq:1})}}
736: %\end{table*}
737:
738: %We note that even if the AGN dominates the $3-5~\mu m$ ULIRGs emission, its contribution to the bolometric luminosity can be small. This confirms the effectiveness of our method: it is possible to detect clearly an AGN in the 3-5~$\mu$m band (and quantitatively estimate its power) even when it is heavily obscured and/or its contribution to the total luminosity is not dominant.
739:
740: %As mentioned above, the estimates obtained from our model are affected not only by the measurement uncertainties,
741: %but also by the assumptions about the AGN and SB templates. Two cases where this limits could be particularly relevant are IRAS~20551-4250 and NGC~6240, which are briefly discussed below:\\
742:
743: %IRAS 20551-4250: in this source a strong starburst contribution is apparent in the L-band spectrum ($EW_{3.3}\sim90$~nm); however its bolometric emission appears to be dominated by the AGN component ($\alpha_{BOL}\sim99\%$).
744: %This high estimate is related to the high estimated optical depth probably due to the large amount of hot dust that obscure the AGN preventing PAH molecules from sublimation; this is confirmed by the $EW_{3.3=90\;nm}$ and the highest L-band optical depth ($\tau_L\sim11$) which inverts the continuum slope. Infact if we express Equation \ref{fm:abol} in function of $\tau_L$ and $EW_{3.3}$ using equation A2 of R06 we obtain:
745: %\begin{equation}
746: % \alpha_{BOL}=\frac{EW_{SB}-EW_{3.3}}{EW_{SB}+(1-ke^{-\tau_L})EW_{3.3}},
747: %\label{fm:abol_int}
748: %\end{equation}
749: %where $EW_{SB}$ is the value for a pure starburst mechanism. Considering the case of IRAS 20551-4250 with $EW_{3.3}\longrightarrow EW_{SB}$ and huge value of $\tau_L$ it results $\alpha_{BOL}\longrightarrow 1$. Infact to reproduce a strong molecular emission with reddening the model requires a huge bolometric AGN contribution. Howether the model is a first order approximation of the observed flux and it appears more realistic an AGN bolometric emission of the same order of the starburst one. \\
750: %The interacting system NGC 6240 has the smaller contributions due to AGNs emission, and this explains the difficulty to detect the double AGN except in X-rays: both the active nuclei are intrinsically faint ($\alpha_{BOL}=2\&\pm0.2\%$ for the Southern nucleus, $\alpha_{BOL}=7\%\pm4\%$ for the Northern one), and can be neglected to account for the bolometric luminosity with respect the starburst process.\\
751: %These examples justifies our choice of studying ULIRGs properties with L and M-band spectra (as described in the Introduction), and points out the power of high signal-to-noise ratio as tools to individuate AGN and SB presence and their relative contribution to the observed emission even when the AGN is negligible for the IR-emission because intrinsically faint and/or obscured.
752:
753:
754: \subsection{Physical properties of the dusty absorber}
755:
756: The presence of a dusty absorber covering the AGN component is revealed by three different observational
757: features:\\
758: $\bullet$ The slope of the AGN component, indicating the reddening of the continuum emission.\\
759: $\bullet$ The absorption feature at 3.4~$\mu$m due to aliphatic hydrocarbon dust grains.\\
760: $\bullet$ The absorption features in the 4.5-5.0~$\mu$m range due to CO gas.\\
761: Considering the L and M bands absorption features in our five sources (Table~3) we note that the presence of deep CO
762: absorption ($\tau_{4.6}>0.15$) is always connected with the $3.4~\mu m$ feature (with the exception of the fainter nucleus of NGC 6240, for which we can estimate only the $\tau_{3.4}$ upper limit), consistent with our physical interpretation.
763: % the only possible exception being the fainter nucleus of NGC 6240 ().
764: Moreover, these absorption structures are always associated with a steep continuum, which again is a signature of
765: an obscured AGN.
766:
767: Lutz et al. (2004) found no signatures of strong CO absorption both in type 1 and type 2 AGNs, and suggested
768: that the deep CO absorption features detected in two ULIRGs (Spoon et al.~2003, 2004) were not directly related to
769: the obscuring torus responsible for the type~1/type~2 optical classification.
770:
771: Here we confirm that such features are signatures of the circumnuclear region of highly obscured ULIRGs.
772: An exception is IRAS 05189-2524, which is not heavily absorbed in the continuum but shows a weak
773: CO absorption feature
774: ($\tau_{4.6}<0.1$). However, this optical depth is below the detection limit for most
775: of the AGN analyzed by Lutz et al.~(2004) with ISO.
776: As a consequence, it is possible that CO absorption is common among type~2~AGN, but with an optical depth
777: significantly smaller than the typical values found in ULIRGs.
778:
779: A more detailed analysis of the relation between continuum and line/band absorption can be performed by
780: checking for possible quantitative correlations between CO and hydrocarbon absorption features and continuum reddening of the AGN component.
781:
782:
783: \begin{figure*}[!h]
784: \begin{center}
785: \includegraphics[scale=0.70]{f4.eps}
786: \end{center}
787: \caption{Continuum reddening versus absorption features optical depth. The hydrocarbon data are shown in red, in blue are the CO measurements; the isolated blue point corresponds to IRAS~05189-2524,which show a low L-band absorption and a relatively low $\tau_{4.6}$ (see Table~3 and Section~5.2). Spectra with no continuum reddening nor absorption lines lie in the yellow band.}
788: \label{fg:tau}
789: \end{figure*}
790:
791:
792: Figure~4 relates the continuum reddening, $\tau_L$, versus the CO and $3.4~\mu$m hydrocarbon
793: optical depths, for all the objects with significant
794: detection of at least one of these localized absorption features.
795: The data are from our sample, from R06 and from a re-analysis of the Imanishi et al.~2006 sample using
796: our model (Risaliti, Imanishi \& Sani submitted).
797: The figure confirms the correlation discussed above: all the objects with detected absorption features
798: have reddened continua, while all objects without absorption features are not heavily reddened.
799: However, we note that no correlation is present within the heavily obscured sources between continuum reddening
800: and absorption features, nor between the two absorption features themselves.
801:
802: This suggests that the dust composition is not constant among ULIRGs: a fixed dust composition would
803: imply a linear correlation between all absorption effects. We plan to investigate the variations in dust
804: composition complementing our data with Spitzer IRS observations (Sani et al.~2007, in prep.).
805:
806: %\begin{equation}
807: %N_H/A_V=1.9\times10^{21}mag^{-1}cm^{-2},
808: %\label{fm:nh}
809: %\end{equation}
810: %\begin{equation}
811: %A_L \sim 0.04 \times A_V
812: %\label{fm:algal}
813: %\end{equation}
814: %(assuming a galactic dust/gas ratio, Bohlin, Savage \& Drake 1978). We can obtain a second assessment using
815: %hydrocarbons $3.4~\mu m$ optical depth as follows (Pendleton et al. 1994):
816: %\begin{equation}
817: %A_L=(12\pm4)\tau_{3.4}.
818: %\label{fm:tauh}
819: %\end{equation}
820: %%Now we compare the above estimation with that implied by $\tau_L$ obtained thanks to our model, and connected to L-band extintion by its definition:
821: %\begin{equation}
822: %A_L=1.086\times \tau_L.
823: %\end{equation}
824:
825: %\begin{table*}[!t]
826: %\begin{center}
827: % \begin{footnotesize}
828: % \begin{tabular}{lccccc}
829: % \hline
830: % \hline
831: % Source (IRAS) & $N_H(X)^a\;(10^{23}cm^{-2})$ & $A_L(X)^b$ & $\tau_{3.4}^c$ & $A_L(\tau_{3.4})^d$ & $A_L(obs)^e$ \\
832: % \hline
833: % NGC 6240 S & $>20$ & 20 & $<0.03$ & $<0.3$ & $>0.5$ \\
834: % \hline
835: % NGC 6240 N & $>20$ & 20 & $<0.05$ & $<0.5$ & 2.2 \\
836: % \hline
837: % 05189-2524 & 4 & 9 & $<0.02$ & $<0.2$ & $<0.3$ \\
838: % \hline
839: % \hline
840: % 19254-7245 & $>10$ & 20 & 0.59 & 7 & 1.6 \\
841: % \hline
842: % 20551-4250 & 8 & 17 & 0.23 & 3 & $>7$ \\
843: % 23060+0505 & 0.8 & 2 & $<0.03$ & $<0.3$ & $<0.2$ \\
844: % \hline
845: % \hline
846: % \hline
847: % \end{tabular}
848: % \end{footnotesize}
849: %\end{center}
850: %% \linespread{1.0}
851: % \vspace{-.5cm}
852: %\caption{{\footnotesize Comparison between our model and X-rays and L-band data. $^a$and $^b$ dust to gas ratio
853: %(see references in Section 2.1) and relative extintion obtained using (\ref{fm:algal}). $^c$ $\tau_{3.4}$ by our
854: %data, and $^d$ the relative extintion from (\ref{fm:tauh}). $^c$ L-band extintion from our model.}}
855: %% \linespread{1.6}
856: %\label{tb:estmia}
857: %\end{table*}
858: %e note that we can not estimate realistically the CO column density because of the high dust temperature, with a
859: %orresponding significant $3-5~\mu m$ thermal emission which can cause shallow absorption line even in case of
860: %arge columns of warm molecular gas.
861: %Our observed extintion is one order of magnitude smaller than ones obtained with X-rays observations (assuming a
862: %galactic parameters). Supposing correct our model, and fixing the ULIRGs amount of gas, our result implies a lower
863: %dust/gas ratio for our sources than for the Milky Way. We are confident that our result is robust for two reasons:\\
864: %1) If $\tau_\lambda$ were such that $A_L(X)\sim A_L$, theoretical spectra, obtained though the model, would have
865: %a strong steepness which is not observed.\\
866: %2) Using Equation~\ref{fm:tauh} (valid for galactic extintion curve) to deabsorb the observed flux, we would
867: %obtain an AGN intrinsic luminosity greater than source bolometric luminosity.\\
868: %We can explain this discrepancy in two ways:\\
869: %a) A lower dust-to-gas ratio than Galactic is present. This is usually observed in Seyfert Galaxies (Maccararo
870: %et al. 1982, Maiolino at al. 2001b), and is probably common also in higher redshift quasar (Risaliti \& Elvis 2005).
871: %The low ratio can be due either to the presence of a dust-free region of gas (i.e. the part of the absorbing
872: %gas lies within the dust sublimation radius) or to dust grains with, on average, larger sizes than Galaxy ones
873: %(Maiolino et al. 2001a).\\
874: %b) The amount of absorbing gas/dust toward the X-rays emitting region can be higher than in the case of more
875: %extended L-band emitting region. The X-rays are supposed to be produced within a few tens of Schwartzschild radii
876: %of the central AGN black hole (for example $10^{-2}$ pc for a $10^8\;\mbox{M}_\odot$), while the hot dust region
877: %must be at a distance R larger than the sublimation radius (for an AGN with an intrinsic optical/UV luminosity of
878: %$10^{46}erg\;s^{-1}$, R$>1$ pc).
879:
880: \section{Conclusions}
881: We have presented 3-5 $\mu m$ high quality, low resolution spectra of the five brightest ULIRGs containing an AGN,
882: and we have studied the physical properties of the circumnuclear medium which affects the relative AGN/SB observed
883: contributions. The high statistic of the data, together with the large wavelength range, allows us to measure spectral diagnostics
884: ($EW_{3.3}$, $\Gamma$, $\tau_{3.4}$ and $\tau_{4.6}$) with unprecedented precision. In our sample we can distinguish
885: nuclei with an unobscured AGN (low $EW_{3.3}$, flat continuum and absence of strong absorption) in IRAS 05189-2524,
886: IRAS 23060+0505 and the brighter nucleus of NGC 6240, and sources with a heavily absorbed AGN (significant $EW_{3.3}$,
887: steep inverted continuum with strong dust and gas absorptions) in IRAS 19254-7245, IRAS 20551-4250 and the fainter
888: nucleus of NGC 6240. The detection of PAH emission line points to the presence of starburst activity in all sources.\\
889: To quantify our findings we used a simple analytical model for the deconvolution of the AGN
890: and starburst contributions to the bolometric luminosity. All galaxies are
891: dominated by the AGN component in the L-band,
892: while the bolometric emission is dominated by starburst emission in all cases, with the possible exception of IRAS 20551-4250.\\
893: For the first time we detect strong CO absorption in sources with heavily reddened continua.\\
894: The lack of correlation between continuum reddening and CO/hydrocarbon absorption features within
895: heavily obscured sources (see Fig.~4) suggests that
896: the dust composition is highly variable from source to source.\\
897:
898: \textit{Acknowledgments}. We are grateful to the anonymous referee for very useful
899: comments which significantly improved this work.
900: This publication makes use of data products from the
901: 2MASS, which is a joint project of the University of Massachusetts
902: and the Infrared Processing and Analysis Center/California Institute
903: of Technology, funded by the National Aeronautics and Space Administration
904: and the National Science Foundation. Also, we made
905: use of theNASA/IPAC Extragalactic Database (NED) which is operated
906: by the Jet Propulsion Laboratory, California Institute of Technology,
907: under contract with the National Aeronautics and Space
908: Administration.
909:
910: \appendix
911: \section{Derivation of $\tau_L$ and $\alpha$}
912: %Relation between observed parameters and model ones}\\
913:
914: In the following we derive Eq.~(3) and Eq.~(4) which express the model parameters $\alpha$ and $\tau_L$ as a function of the observed EW$_{3.3}$ and $\Gamma$. The intrinsic EW$_{SB}$ is defined as:
915: \begin{equation}
916: EW_{SB}=\frac{\int_\lambda f_{PAH}~d\lambda}{(1-\alpha)\lambda^{\Gamma_{int}}},
917: \end{equation}
918: while, considering that the AGN continuum dilutes the PAH EW, the observed values is:
919: \begin{equation}
920: EW_{3.3}=\frac{\int_\lambda f_{PAH}~d\lambda}{\alpha\lambda^{\Gamma_{int}}+(1-\alpha)\lambda^{\Gamma_{int}}},
921: \end{equation}
922: where $\Gamma_{int}=-0.5$ as shown in Section~4.
923: From Eq.~(A1) and Eq.~(A2), we obtain:
924: \begin{equation}
925: EW_{3.3}=\frac{EW_{SB}~(1-\alpha)}{\alpha e^{-\tau_L}+(1-\alpha)}.
926: \end{equation}
927: The observed spectral index $\Gamma$ is obtained from the continuum between $\lambda_1=3.3~\mu$m and $\lambda_2=4~\mu$m (corresponding to the wavelength range with the lowest background noise):
928: \begin{equation}
929: \left\{
930: \begin{array}{ll}
931: \Gamma\simeq\frac{\ln (f_{\lambda_1}/f_{\lambda_2})}{\ln \lambda_1/\lambda_2} \\
932: \frac{f_{\lambda_2}}{f_{\lambda_1}}=\left(\frac{\lambda_2}{\lambda_1}\right)^{\Gamma_{int}}\frac{\alpha e^{-\tau(\lambda_2)}+(1-\alpha)}{\alpha e^{-\tau(\lambda_1)}+(1-\alpha)}.
933: \end{array}
934: \right.
935: \end{equation}
936: A first order
937: linearization of the above equations,
938: based on the the approximation:\\
939: $\tau(\lambda)=\tau_L\left( \frac{\lambda}{\lambda_L}\right) ^{-\beta}\simeq\tau_L\left( 1-\beta\frac{\lambda_2-\lambda_1}{\lambda_1}\right) $,
940: ($\beta=1.75$) leads to:
941: \begin{equation}
942: \Gamma=\Gamma_{int}+\frac{\beta~\alpha\tau_L e^{-\tau_L}}{\alpha e^{-\tau_L}+(1-\alpha)}.
943: \end{equation}
944: We obtain Eq.~(3) and Eq.~(4) by inverting the system formed by Eq.~(A3) and Eq.~(A5).
945: A numerical solution of the system without the linear approximation of the extinction term
946: leads to values very similar (and always consistent within the errors) to those estimated
947: with Equations 3 and 4.
948:
949:
950:
951: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
952: \begin{thebibliography}{}
953: %\bibitem[1]{860} Abbott, D.C. 1982, ApJ, 259, 282
954: %\bibitem[Arav et al.(2001)]{2001ApJ...561..118A} Arav, N.~et al.\ 2001,
955: %\apj, 561, 118
956:
957: \bibitem[Armus et al.(2006)]{2006ApJ...640..204A} Armus, L., et al.\ 2006, \apj, 640, 204
958:
959: \bibitem[Barger et al.(1998)]{1998Natur.394..248B} Barger, A.~J., Cowie, L.~L., Sanders, D.~B., Fulton, E., Taniguchi, Y., Sato, Y., Kawara, K., \& Okuda, H.\ 1998, \nat, 394, 248
960:
961: \bibitem[Berta et al.(2003)]{2003A&A...403..119B} Berta, S., Fritz, J., Franceschini, A., Bressan, A., \& Pernechele, C.\ 2003, \aap, 403, 119
962:
963: \bibitem[Bohlin et al.(1978)]{1978ApJ...224..132B} Bohlin, R.~C., Savage, B.~D., \& Drake, J.~F.\ 1978, \apj, 224, 132
964:
965: %\bibitem[Borne et al.(1999)]{1999Ap&SS.266..137B} Borne, K.~D., et al.\ 1999, \apss, 266, 137
966:
967: \bibitem[Braito et al.(2003)]{2003A&A...398..107B} Braito, V., et al.\ 2003, \aap, 398, 107
968:
969: \bibitem[Brandt et al.(1997)]{1997MNRAS.290..617B} Brandt, W.~N., Fabian, A.~C., Takahashi, K., Fujimoto, R., Yamashita, A., Inoue, H., \& Ogasaka, Y.\ 1997, \mnras, 290, 617
970:
971: \bibitem[Cardelli et al.(1989)]{1989ApJ...345..245C} Cardelli, J.~A., Clayton, G.~C., \& Mathis, J.~S.\ 1989, \apj, 345, 245
972:
973: \bibitem[Charmandaris et al.(2002)]{2002A&A...391..429C} Charmandaris, V., et al.\ 2002, \aap, 391, 429
974:
975: \bibitem[Duc et al.(1997)]{1997A&AS..124..533D} Duc, P.-A., Mirabel, I.~F., \& Maza, J.\ 1997, \aaps, 124, 533
976:
977: \bibitem[Elvis et al.(1994)]{1994ApJS...95....1E} Elvis, M., et al.\ 1994, \apjs, 95, 1
978:
979: \bibitem[Franceschini et al.(2003)]{2003MNRAS.343.1181F} Franceschini, A., et al.\ 2003, \mnras, 343, 1181
980:
981: \bibitem[Fruscione et al.(2006)]{2006SPIE.6270E..60F} Fruscione, A., et
982: al.\ 2006, \procspie, 6270
983:
984: \bibitem[Genzel et al.(1998)]{1998ApJ...498..579G} Genzel, R., et al.\ 1998, \apj, 498, 579
985:
986: \bibitem[Hill et al.(1987)]{1987ApJ...316L..11H} Hill, G.~J., Wynn-Williams, C.~G., \& Becklin, E.~E.\ 1987, \apjl, 316, L11
987:
988: \bibitem[Hutchings \& Neff(1988)]{1988AJ.....96.1575H} Hutchings, J.~B., \& Neff, S.~G.\ 1988, \aj, 96, 1575
989:
990: \bibitem[Imanishi et al.(2006)]{2006ApJ...637..114I} Imanishi, M., Dudley, C.~C., \& Maloney, P.~R.\ 2006, \apj, 637, 114
991:
992: \bibitem[Imanishi \& Dudley(2000)]{2000ApJ...545..701I} Imanishi, M., \& Dudley, C.~C.\ 2000, \apj, 545, 701
993:
994: \bibitem[Imanishi \& Maloney(2003)]{2003ApJ...588..165I} Imanishi, M., \& Maloney, P.~R.\ 2003, \apj, 588, 165
995:
996: \bibitem[Imanishi \& Wada(2004)]{2004ApJ...617..214I} Imanishi, M., \&
997: Wada, K.\ 2004, \apj, 617, 214
998:
999: \bibitem[Komossa et al.(2003)]{2003ApJ...582L..15K} Komossa, S., Burwitz, V., Hasinger, G., Predehl, P., Kaastra, J.~S., \& Ikebe, Y.\ 2003, \apjl, 582, L15
1000:
1001: \bibitem[Lonsdale et al.(2006)]{2006astro.ph..3031L} Lonsdale, C., Farrah,
1002: D., \& Smith, H.\ 2006, ArXiv Astrophysics e-prints, arXiv:astro-ph/0603031
1003:
1004: \bibitem[Lu et al.(2003)]{2003ApJ...588..199L} Lu, N., et al.\ 2003, \apj,
1005: 588, 199
1006:
1007: \bibitem[Lutz et al.(2004)]{2004A&A...426L...5L} Lutz, D., Sturm, E., Genzel, R., Spoon, H.~W.~W., \& Stacey, G.~J.\ 2004, \aap, 426, L5
1008:
1009: \bibitem[Lutz et al.(2003)]{2003A&A...409..867L} Lutz, D., Sturm, E., Genzel, R., Spoon, H.~W.~W., Moorwood, A.~F.~M., Netzer, H., \& Sternberg, A.\ 2003, \aap, 409, 867
1010:
1011: \bibitem[Lutz et al.(2002)]{2002A&A...396..439L} Lutz, D., Maiolino, R., Moorwood, A.~F.~M., Netzer, H., Wagner, S.~J., Sturm, E., \& Genzel, R.\ 2002, \aap, 396, 439
1012:
1013: \bibitem[Lutz et al.(2000)]{2000ApJ...530..733L} Lutz, D., et al.\ 2000, \apj, 530, 733
1014:
1015: \bibitem[Maccacaro et al.(1982)]{1982ApJ...257...47M} Maccacaro, T., Perola, G.~C., \& Elvis, M.\ 1982, \apj, 257, 47
1016:
1017: \bibitem[Maiolino et al.(2001)]{2001A&A...365...37M} Maiolino, R., Marconi, A., \& Oliva, E.\ 2001, \aap, 365, 37
1018:
1019: \bibitem[Maiolino et al.(2001)]{2001A&A...365...28M} Maiolino, R., Marconi, A., Salvati, M., Risaliti, G., Severgnini, P., Oliva, E., La Franca, F., \& Vanzi, L.\ 2001, \aap, 365, 28
1020:
1021: \bibitem[Mirabel et al.(1991)]{1991A&A...243..367M} Mirabel, I.~F., Lutz, D., \& Maza, J.\ 1991, \aap, 243, 367
1022:
1023: \bibitem[Pendleton et al.(1994)]{1994ApJ...437..683P} Pendleton, Y.~J., Sandford, S.~A., Allamandola, L.~J., Tielens, A.~G.~G.~M., \& Sellgren, K.\ 1994, \apj, 437, 683
1024:
1025: \bibitem[Ptak et al.(2003)]{2003ApJ...592..782P} Ptak, A., Heckman, T., Levenson, N.~A., Weaver, K., \& Strickland, D.\ 2003, \apj, 592, 782
1026:
1027: \bibitem[Risaliti et al.(2000)]{2000A&A...356...33R} Risaliti, G., Maiolino, R., \& Bassani, L.\ 2000, \aap, 356, 33
1028:
1029: \bibitem[Risaliti et al.(2003)]{2003ApJ...595L..17R} Risaliti, G., et al.\ 2003, \apjl, 595, L17
1030:
1031: \bibitem[Risaliti \& Elvis(2005)]{2005ApJ...629L..17R} Risaliti, G., \& Elvis, M.\ 2005, \apjl, 629, L17
1032:
1033: \bibitem[Risaliti et al.(2006)]{2006MNRAS.365..303R} Risaliti, G., et al.\ 2006, \mnras, 365, 303
1034:
1035: \bibitem[Risaliti et al.(2006)]{2006ApJ...637L..17R} Risaliti, G., et al.\ 2006B, \apjl, 637, L17
1036:
1037: \bibitem[Sanders \& Mirabel(1996)]{1996ARA&A..34..749S} Sanders, D.~B., \& Mirabel, I.~F.\ 1996, \araa, 34, 749
1038:
1039: \bibitem[Sanders et al.(2003)]{2003AJ....126.1607S} Sanders, D.~B., Mazzarella, J.~M., Kim, D.-C., Surace, J.~A., \& Soifer, B.~T.\ 2003, \aj, 126, 1607
1040:
1041: \bibitem[Severgnini et al.(2001)]{2001A&A...368...44S} Severgnini, P., Risaliti, G., Marconi, A., Maiolino, R., \& Salvati, M.\ 2001, \aap, 368, 44
1042:
1043: \bibitem[Spergel et al.(2003)]{2003ApJS..148..175S} Spergel, D.~N., et al.\ 2003, \apjs, 148, 175
1044:
1045: \bibitem[Spoon et al.(2003)]{2003A&A...402..499S} Spoon, H.~W.~W., Moorwood, A.~F.~M., Pontoppidan, K.~M., Cami, J., Kregel, M., Lutz, D., \& Tielens, A.~G.~G.~M.\ 2003, \aap, 402, 499
1046:
1047: \bibitem[Spoon et al.(2004)]{2004ApJS..154..184S} Spoon, H.~W.~W., et al.\ 2004, \apjs, 154, 184
1048:
1049: \bibitem[van Dishoeck et al.(2003)]{2003Msngr.113...49V} van Dishoeck, E.~F., et al.\ 2003, The Messenger, 113, 49
1050:
1051: \bibitem[Veilleux et al.(2002)]{2002ApJS..143..315V} Veilleux, S., Kim, D.-C., \& Sanders, D.~B.\ 2002, \apjs, 143, 315
1052:
1053: \bibitem[Veilleux et al.(1999)]{1999ApJ...522..113V} Veilleux, S., Kim, D.-C., \& Sanders, D.~B.\ 1999, \apj, 522, 113
1054:
1055: \bibitem[Veilleux et al.(1995)]{1995ApJS...98..171V} Veilleux, S., Kim, D.-C., Sanders, D.~B., Mazzarella, J.~M., \& Soifer, B.~T.\ 1995, \apjs, 98, 171
1056:
1057: \bibitem[Vignati et al.(1999)]{1999A&A...349L..57V} Vignati, P., et al.\ 1999, \aap, 349, L57
1058:
1059: \bibitem[Young et al.(1996)]{1996MNRAS.280..291Y} Young, S., Hough, J.~H., Axon, D.~J., Ward, M.~J., \& Bailey, J.~A.\ 1996, \mnras, 280, 291
1060:
1061: \bibitem[Voit(1992)]{1992MNRAS.258..841V} Voit, G.~M.\ 1992, \mnras, 258,
1062: 841
1063: \end{thebibliography}
1064: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1065:
1066:
1067: \end{document}
1068:
1069: