Spitzer/IRS 5-35 um Low-Resolution Spectroscopy of the 12 um Seyfert Sample

Abstract

We present low-resolution 5.5-35 um spectra for 103 galaxies from the 12 um Seyfert sample, a complete unbiased 12 um flux limited sample of local Seyfert galaxies selected from the IRAS Faint Source Catalog, obtained with the Infrared Spectrograph (IRS) on-board Spitzer Space Telescope. For 70 of the sources observed in the IRS mapping mode, uniformly extracted nuclear spectra are presented for the first time. We performed an analysis of the continuum emission, the strength of the Polycyclic Aromatic Hydrocarbon (PAH) and astronomical silicate features of the sources. We find that on average, the 15-30 um slope of the continuum is alpha_{15-30}=-0.85+-0.61 for Seyfert 1s and -1.53+-0.84 for Seyfert 2s, and there is substantial scatter in each type. Moreover, nearly 32% of Seyfert 1s, and 9% of Seyfert 2s, display a peak in the mid-infrared spectrum at 20 um, which is attributed to an additional hot dust component. The PAH equivalent width decreases with increasing dust temperature, asindicated by the global infrared color of the host galaxies. However, no statistical difference in PAH equivalent width is detected between the two Seyfert types, 1 and 2, of the same bolometric luminosity. The silicate features at 9.7 and 18um in Seyfert 1 galaxies are rather weak, while Seyfert 2s are more likely to display strong silicate absorption. Those Seyfert 2s with the highest silicate absorption also have high infrared luminosity and high absorption (hydrogen column density N_H>10^23 cm^-2 as measured from the X-rays. Finally, we propose a new method to estimate the AGN contribution to the integrated 12 um galaxy emission, by subtracting the "star formation" component in the Seyfert galaxies, making use of the tight correlation between PAH 11.2 um luminosity and 12 um luminosity for star forming galaxies. Comment: accepted for publication in ApJ

Full text

Page 1

arXiv:0906.2004v1 [astro-ph.CO] 10 Jun 2009
Spitzer/IRS 5-35µm Low-Resolution Spectroscopy of the 12µm
Seyfert Sample
Yanling Wu1, Vassilis Charmandaris2,3, Jiasheng Huang4, Luigi Spinoglio5, Silvia
Tommasin5,6
yanling@ipac.caltech.edu, vassilis@physics.uoc.gr, jhuang@cfa.harvard.edu,
luigi.spinoglio@ifsi-roma.inaf.it, Silvia.Tommasin@ifsi-roma.inaf.it
ABSTRACT
We present low-resolution 5.5-35µm spectra for 103 galaxies from the 12µm
Seyfert sample, a complete unbiased 12µm flux limited sample of local Seyfert
galaxies selected from the IRAS Faint Source Catalog, obtained with the Infrared
Spectrograph (IRS) on-board Spitzer Space Telescope. For 70 of the sources
observed in the IRS mapping mode, uniformly extracted nuclear spectra are
presented for the first time. We performed an analysis of the continuum emission,
the strength of the Polycyclic Aromatic Hydrocarbon (PAH) and astronomical
silicate features of the sources. We find that on average, the 15-30µm slope
of the continuum is < α15−30 >=-0.85±0.61 for Seyfert 1s and -1.53±0.84 for
Seyfert 2s, and there is substantial scatter in each type. Moreover, nearly 32%
of Seyfert 1s, and 9% of Seyfert 2s, display a peak in the mid-infrared spectrum
at 20µm, which is attributed to an additional hot dust component. The PAH
equivalent width decreases with increasing dust temperature, as indicated by
the global infrared color of the host galaxies. However, no statistical difference
in PAH equivalent width is detected between the two Seyfert types, 1 and 2,
of the same bolometric luminosity. The silicate features at 9.7 and 18µm in
Seyfert 1 galaxies are rather weak, while Seyfert 2s are more likely to display
1Infrared Processing and Analysis Center, California Institute of Technology, 1200 E. California Blvd,
MC 314-6, Pasadena, CA 91125
2University of Crete, Department of Physics, GR-71003, Heraklion, Greece
3IESL/Foundation for Research and Technology - Hellas, GR-71110, Heraklion, Greece, and Chercheur
Associ´ e, Observatoire de Paris, F-75014, Paris, France
4Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA, 02138
5Istituto di Fisica dello Spazio Interplanetario, INAF, Via Fosso del Cavaliere 100, I-00133 Rome, Italy
6Dipartimento di Fisica, Universit di Roma, La Sapienza, Roma, Italy

Page 2

– 2 –
strong silicate absorption. Those Seyfert 2s with the highest silicate absorption
also have high infrared luminosity and high absorption (hydrogen column density
NH>1023cm−2) as measured from the X-rays. Finally, we propose a new method
to estimate the AGN contribution to the integrated 12µm galaxy emission, by
subtracting the ”star formation” component in the Seyfert galaxies, making use
of the tight correlation between PAH 11.2µm luminosity and 12µm luminosity
for star forming galaxies.
1.Introduction
for which most of their nuclear and often bolometric luminosities is produced in
Active galaxies are galaxies in which one detects radiation from their nucleus which
is due to accretion onto a super-massive black hole (SMBH) located at the center. The
spectrum of an Active Galactic Nucleus (AGN) is typically flat in νfν. The fraction of the
energy emitted from the AGN compared with the total bolometric emission of the host
can range from a few percent in moderated luminosity systems (Lbol < 1011L⊙) to more
than 90% in quasars (Lbol> 1012L⊙) (see Ho 2008, and references therein). As a subclass,
Seyfert galaxies are the nearest and brightest AGNs, with 2-10keV X-ray luminosities less
than ∼ 1044ergs−1and their observed spectral line emission originates principally from highly
ionized gas. Seyferts have been studied at many wavelengths, from X-rays, ultraviolet,optical,
to infrared (IR) and radio. The analysis of their optical spectra has lead to the identification
of two types, Seyfert 1s (Sy 1s) and Seyfert 2s (Sy 2s), with the type 1s displaying features
of both broad (FWHM>2000 km s−1) and narrow emission lines, while the type 2s only
narrow-line emission.
The differences between the two Seyfert types have been an intense field of study for
many years. Are they due to intrinsic differences in their physical properties, or are they
simply a result of dust obscuration that hides the broad-line region in Sy 2s? A so-called
unified model has been proposed (see Antonucci 1993; Urry & Padovani 1995), suggesting
that Sy 1s and Sy 2s are essentially the same objects viewed at different angles. A dust
torus surrounding the central engine blocks the optical light when viewed edge on (Sy 2s)
and allows the nucleus to be seen when viewed face on (Sy 1s). Optical spectra in polarized
light (Antonucci & Miller 1985) have indeed demonstrated for several Sy 2s the presence of
broad lines, confirming for these objects the validity of the unified model. However, the exact
nature of this orientation-dependent obscuration is not clear yet. Recently, more elaborate
models, notably the ones of Elitzur (2008), Nenkova et al. (2008), and Thompson et al.
(2009) suggest that the same observational constraints can also be explained with discrete

Page 3

– 3 –
dense molecular clouds, without the need of a torus geometry.
The study of Seyfert galaxies is interesting also from a cosmological perspective, as
they trace the build up of SMBHs at the centers of galaxies. Observations up to 10keV have
established that the cosmic X-ray background (CXB) is mostly due to Seyferts with a peak in
their redshift distribution at z∼0.7 (Hasinger et al. 2005). Furthermore, theoretical modeling
of the observed number counts suggests that CXB at 30keV is also dominated by obscured
Seyferts at z∼0.7 (Gilli et al. 2007; Worsley et al. 2005). Given the strong ionization field
produced by the accretion disk surrounding a SMBH, the dust present can be heated to
near sublimation temperatures, making an AGN appear very luminous in the mid-infrared
(mid-IR).
Mid-IR spectroscopy is a powerful tool to examine the nature of the emission from
AGNs, as well as the nuclear star-formation activity. Since IR observations are much less
affected by dust extinction than those at shorter wavelengths, they have been instrumental
in the study of obscured emission from optically thick regions in AGNs. This is crucial to
understand the physical process of galaxy evolution. With the advent of the Infrared Space
Observatory (ISO), local Seyferts have been studied by several groups (see Verma et al. 2005,
for a review). Mid-IR diagnostic diagrams to quantitatively disentangle the emission from
AGNs, starbursts and quiescent star-forming (SF) regions have been proposed, using both
spectroscopy and broad-band photometry (i.e. Genzel et al. 1998; Laurent et al. 2000). The
recent launch of the Spitzer Space Telescope (Werner et al. 2004) has enabled the study of
AGN with substantially better sensitivity and spatial resolution. In particular, using the
Infrared Spectrograph (IRS1) (Houck et al. 2004a) on board Spitzer, Weedman et al. (2005)
demonstrated early into the mission the variety in the morphology displayed by the mid-IR
spectra of eight classical AGNs. Since then, large samples of AGNs have been studied in
detail, in an effort to quantify their mid-IR properties (Buchanan et al. 2006; Sturm et al.
2006; Deo et al. 2007; Gorjian et al. 2007; Hao et al. 2007; Tommasin et al. 2008). In ad-
dition, new mid-IR diagnostics have been developed to probe the physics of more complex
sources, such as luminous and ultra luminous infrared galaxies (LIRGs/ULIRGs), which may
also harbor AGNs. These were based on correlating the strength of the Polycyclic Aromatic
Hydrocarbons (PAHs), high excitation fine-structure lines, as well as silicate features (i.e.
Armus et al. 2007; Spoon et al. 2007; Charmandaris 2008; Nardini et al. 2008).
The extended 12µm galaxy sample is a flux-limited (down to 0.22Jy at 12µm) sample
of 893 galaxies selected from the IRAS Faint Source Catalog 2 (Rush et al. 1993). As dis-
1The IRS was a collaborative venture between Cornell University and Ball Aerospace Corporation funded
by NASA through the Jet Propulsion Laboratory and the Ames Research Center.

Page 4

– 4 –
cussed by Spinoglio & Malkan (1989), all galaxies emit a nearly constant fraction (∼7%) of
their bolometric luminosity at 12µm. As a result, selecting active galaxies based on their
rest frame 12µm fluxes is the best approach to reduce selection bias due to the variations in
their intrinsic spectral energy distributions (SED). A total of 116 objects from this sample
have been optically classified as Seyfert galaxies (53 Sy 1s and 63 Sy 2s), providing one of the
largest IR selected unbiased AGN sample. This sample also has ancillary data in virtually
all wavelengths, thus making it the most complete data set for systematically studying the
fundamental issues of AGNs in the infrared. Low-resolution 5.5-35µm Spitzer/IRS spectra
of 51 Seyferts from the 12µm sample have been published by Buchanan et al. (2006), who
focused on the study of the Seyfert types and the shape of mid-IR SED using principal com-
ponent analysis. Based on this analysis and comparing with radio data, where available, they
estimate the starburst contribution to the observed spectrum and find it to appear stronger
in Sy 2s, in contrast to the unified model. However, high resolution Spitzer spectroscopy
on 29 objects by Tommasin et al. (2008) does not find a clear indication of stronger star
formation in Sy 2s than Sy 1s.
In this paper, we extend earlier work and study the mid-IR properties and nature of
the dust enshrouded emission from 103 Seyferts of the 12µm Seyferts, nearly 90% of the
whole sample. We use low resolution Spitzer/IRS spectra, focusing mainly on their broad
emission and absorption features and provide for the first time our measurements of the PAH
emission and strength of Si absorption features. Our observations and data reduction are
presented in §2. In §3, we show our analysis on the mid-IR continuum shape, PAH emission
and silicate strength of Seyfert galaxies. A new method to separate the star-formation and
AGN contribution in the 12µm continuum is proposed in §4. Finally, we summarize our
conclusions in §5.
2.Observations and Data Reduction
Since the launch of Spitzer in August 2003, a large fraction of the 12µm Seyfert galaxies
have been observed by various programs using the low-resolution (R∼64-128) and high-
resolution (R∼600) modules of Spitzer/IRS. These observations are publicly available in the
Spitzer archive. A total of 84 galaxies have been observed with the low-resolution spectral
mapping mode of the IRS by the GO program “Infrared SEDs of Seyfert Galaxies: Starbursts
and the Nature of the Obscuring Medium” (PID: 3269), and mid-IR spectra extracted from
the central regions of the maps for 51 galaxies were published by Buchanan et al. (2006).
Spectra for the remaining sources have been taken as part of a number of Guaranteed, Open
Time, as well as Legacy programs with program identifications (PID) 14, 61, 86, 96, 105, 159,

Page 5

– 5 –
3237, 3624, 30291, and 30572. For the purpose of this work, which is mainly to use the PAH
emission features to diagnose starburst and AGN contribution, we are focusing on the low-
resolution IRS spectra (Short Low, SL: 5-15µm; Long Low, LL: 15-37µm). We performed a
complete search of the Spitzer Science Center (SSC) data archive and retrieved a total of 103
sources with at least Short Low observations. Among these objects, 47 are optically classified
as Sy 1s and 56 as Sy 2s2. We adopt the spectral classification of Rush et al. (1993) for the
Seyfert types. Even though in a few cases, the classification may be ambiguous or may
have changed, we do not expect our results, which are of statistical nature, to be affected.
The complete list of the objects analyzed in this paper including their coordinates, IRAS
fluxes, IR luminosity, redshift, Seyfert type and Spitzer program identification number are
presented in Table 1. The redshift and luminosity distribution of the 12µm Seyfert sample
and the galaxies with IRS data studied in this paper is displayed in Figure 1.
With the exception of the data from the SINGS Legacy program (PID 159), of which
we directly used spectra available at the SSC3, all other raw datasets are retrieved from the
Spitzer archive and reduced in the following manner.
All except three4of the 12µm Seyferts with cz<10,000kms−1have been observed with
the IRS spectral mapping mode (PID 3269). This enabled us to also study the circumnuclear
activity and examine the contribution of the host galaxy emission to the nuclear spectrum.
Using datasets from this program, mid-IR spectra obtained from the central slit placement of
each map with point-source extraction were published by Buchanan et al. (2006). However,
as these authors have noted, since the observations had been designed as a spectral map,
blind telescope pointing was used. As a result, no effort was made to accurately acquire
each target and to ensure that the central slit of each map was indeed well centered on the
source. Moreover, for sources where the mid-IR emission is extended, using a point-source
extraction method would likely result in an overestimate of the flux densities, due to the slit
loss correction function applied. This would consequently affect the SED of the galaxy.
For most of the data reduction in this paper, we used The CUbe Builder for IRS Spectra
Maps (Smith et al. 2007b, CUBISM), in combination with an image convolution method.
This method is developed explicitly for IRS mapping mode observations where the map
2The 12µm Seyferts not included in this study due to lack of Spitzer/IRS SL spectra are 6 Sy 1s (Mrk1034,
M-3-7-11, Mrk618, F05563-3820, F15091-2107, E141-G55) and 7 Sy 2s (F00198-7926, F00521-7054, E541-
IG12, NGC1068, F03362-1642, E253-G3, F22017+0319).
3The SINGS data products are available at: http://data.spitzer.caltech.edu/popular/sings/. The nuclear
spectra were extracted over a 50′′×33′′region on the nucleus.
4The three exceptions are NGC1097, NGC1566, NGC5033, and they were also observed by SINGS.

Page 6

– 6 –
size is small and involves the following steps. Spectral cubes were built using CUBISM
from the IRS maps. Sky subtraction was performed by differencing the on and off source
observations of the same order in each module (SL and LL). The observations from PID
3269 were designed to minimize redundancy and maximize the number of sources that can
be observed using a limited amount of telescope time. As a result these maps consisted of
only 13 steps in SL and 5 steps in LL with no repetition and for a source with extended
mid-IR emission, the small map may not encompass the entire source, e.g. NGC1365. To
obtain an accurate SED from the extracted region of the galaxy, one needs to ensure that the
same fraction of source fluxes at all wavelengths is sampled. Since the point spread function
(PSF) changes from 5 to 35µm, we adopted an image convolution method to account for the
change in the full width half maximum (FWHM) of the PSF: 2-dimensional images at each
wavelength were convolved to the resolution at the longest wavelength (35µm). Then low-
resolution spectra were extracted with matched apertures, chosen to encompass the whole
nuclear emission. Even though the image convolution method dilutes the fluxes included in
the extraction aperture, it does ensure an accurate SED shape, especially for small maps that
cannot encompass the extended emission for the source. A comparison of the LL spectra
before and after image convolution for NGC1365 can be found in Figure 2.
A number of tests with varying sizes for the spectral extraction aperture were performed
in order to select the optimum size. An aperture of 4×3 pixels in LL was adopted, which
corresponds to an angular size of 20.4′′×15.3′′. Then a complete 5–35µm spectrum of all
galaxies centered on their nucleus was extracted5. The main motivation behind this choice
was to ensure that we could produce an accurate overall SED of the extracted region, even
in cases where the emission was extended. This was essential so that we have a reliable
measurement of the continuum emission, in order to calculate the mid-IR slope, and to
estimate the strength of the silicate features. Further tests were performed by extracting
just the SL spectra of the sample using the smallest aperture possible (2×2 pixels in SL
or 3.6′′× 3.6′′). It was found that, in general, the measured fluxes and EWs of the PAH
features did not differ more than 20% from the integrated spectra over the larger apertures,
which indicates that these objects are likely to be less than 20% more extended than the
point-spread-function of point sources. This suggests that most of the sources were not very
extended in the mid-IR and more than 80% of their flux originates from a point source
unresolved to Spitzer. Scaling the spectra between different orders and modules was not
needed for most of the cases, and when an order mismatch was detected, the SL spectra
were scaled to match LL. In Table 2, where we present our measurements, the sizes of the
5For two Sy 2s NGC1143/4 and NGC4922, only SL data were available, thus only 5–15µm spectra were
obtained.

Page 7

– 7 –
extraction apertures as well as their corresponding projected linear sizes on the sky are
also listed. A histogram of the physical size of extraction aperture for the whole sample
is presented in Figure 3 with the dotted line, while in the same figure, we also show the
distribution for sources that are observed in spectral mapping mode with the solid line. All
the sources that are extracted from a projected size of more than 20 kpc are observed using
the staring mode and reduced with point-source extraction method.
For data obtained with the IRS staring mode, the reduction was done in the following
manner. We started from intermediate pipeline products, the “droop” files, which only
lacked stray light removal and flat-field correction. Individual pointings to each nod position
of the slit were co-added using median averaging. Then on and off source images were
differenced to remove the contribution of the sky emission.
images were extracted with the Spectral Modeling, Analysis, and Reduction Tool (SMART,
Higdon et al. 2004) in a point source extraction mode, which scaled the extraction aperture
with wavelength to recover the same fraction of the diffraction limited instrumental PSF.
Note that since the width of both the SL and LL slits is 2 pixels (3.6′′and 10.2′′respectively),
no information could be retrieved along this direction from areas of the galaxy which were
further away. The spectra were flux calibrated using the IRS standard star α Lac, for which
accurate template was available (Cohen et al. 2003). Finally, using the first order of LL
(LL1, 20–36µm) spectrum to define the absolute value of the continuum, the flux calibrated
spectra of all other low-resolution orders were scaled to it.
Spectra from the final 2-D
For nine sources in the 12µm Seyfert sample, both spectral mapping and staring mode
observations had been obtained and were available in the SSC data archive. For these galax-
ies, the spectra were extracted following the above mentioned recipes for all observations.
In order to ascertain again how extended the emission from these source was, we compared
the resulting spectra of the same source. With two exceptions, NGC4151 and NGC7213,
all other seven spectra obtained in staring mode required a scaling factor larger than 1.15
between the SL and LL modules. This further suggests that the nuclear emission in those
sources is indeed extended, and point-source extraction may not be appropriate. For galax-
ies with only staring mode observations, we also checked the difference between the overlap
region (∼ 15µm) in the SL and LL modules. None of them require a scaling factor more
than 1.15, suggesting that those objects are point-like at least along the direction of the IRS
slits.

Page 8

– 8 –
3. Results
3.1.Global Mid-IR spectra of Seyfert Galaxies
It has been well established that the mid-IR spectra of Seyfert galaxies display a variety
of features (see Clavel et al. 2000; Verma et al. 2005; Weedman et al. 2005; Buchanan et al.
2006; Hao et al. 2007, and references therein). This is understood since, despite the optical
classification of their nuclear activity, emission from the circumnuclear region, as well as of
the host galaxy, also influences the integrated mid-IR spectrum of the source. Our complete
12µm selected Seyfert sample provides an unbiased framework to study the statistics on their
mid-IR properties. Earlier work by Buchanan et al. (2006) on just 51 AGN from this sample
presented a grouping of them based on their continuum shapes and spectral features. In this
section, we explore the global shape of the mid-IR spectra for the complete Spitzer/IRS set
of data on the 12µm Seyferts.
We examine if the average mid-IR spectrum of a Sy 1 galaxy is systematically different
from that of a Sy 2. The IRS spectra for 47 Sy 1s and 54 Sy 2s with full 5.5-35µm spectral
coverage, normalized at the wavelength of 22µm, are averaged and plotted in Figure 4. For
comparison, we over-plot the average starburst template from Brandl et al. (2006). It is
clear that the mid-IR continuum slope of the average Sy 1 spectrum is shallower than that
of Sy 2, while the starburst template has the steepest spectral slope, indicating a different
mixture of hot/cold dust component in these galaxies (also see Hao et al. 2007). This would
be consistent with the interpretation that our mid-IR spectra of Sy 2s display a strong
starburst contribution, possibly due to circumnuclear star formation activity included in the
aperture we used to extract the spectra from, consistent with the findings of Buchanan et al.
(2006) that star formation is extended and it is not a purely nuclear phenomenon. PAH
emission, which is a good tracer of star formation activity (F¨ orster Schreiber et al. 2004),
can be detected in the average spectra of both Seyfert types, while it is most prominent in the
average starburst spectrum. PAH emission originates from photo-dissociation region (PDR)
and can easily be destroyed by the UV/X-ray photons in strong a radiation field produced
near massive stars and/or an accretion disk surrounding a SMBH (Voit 1992; Laurent et al.
2000; Clavel et al. 2000; Sturm et al. 2002; Verma et al. 2005; Weedman et al. 2005). In the
12µm Seyfert sample, we detect PAH emission in 37 Sy 1s and 53 Sy 2s, that is 78% and
93% for each type respectively. This is expected since the apertures we used to extract the
mid-IR spectra for the 12µm sample correspond in most cases to areas of more than 1 kpc
in linear dimensions (see Figure 3). As a result, emission from the PDRs associated with the
extended circumnuclear region and the disk of the host galaxy is also encompassed within the
observed spectrum. High ionization fine-structure lines, such as [NeV]14.32µm/24.32µm,
are clearly detected even in the low-resolution average spectrum of Sy 1. This signature

Page 9

– 9 –
is also visible, though rather weak, in the average spectrum of Sy 2, while it is absent in
the average starburst template. Due to photons of excitation energy higher than 97eV and
typically originating from the accretion disk of an AGN, [NeV] serves as an unambiguous
indicator of an AGN. Even though the low-resolution module of IRS was not designed for
studying fine-structure lines, we are still able to detect [NeV] emission in 29 Sy 1s and
32 Sy 2s, roughly 60% for both types. Another high ionization line, [OIV]25.89µm, with
ionization potential of 54eV, also appears in both Seyfert types (42 Sy 1s and 41 Sy 2s),
and is stronger in the average spectrum of Sy 1. The [OIV] emission line can be powered by
shocks in intense star forming regions or AGNs (see Lutz et al. 1998; Schaerer & Stasi´ nska
1999; Bernard-Salas et al. 2009; Hao et al. 2009). In our sample it is probably powered
by both, given the large aperture we adopted for spectral extraction. More details and a
complete analysis of mid-IR fine-structure lines for 29 galaxies from the 12µm Seyfert sample
are presented in Tommasin et al. (2008), while the work for the entire sample is in progress
(Tommasin et al. 2009).
An atlas with mid-IR low-resolution spectra of the 12µm Seyfert sample is included at
the end of this paper6. We find that a fraction of our sources show a flattening or a local
maximum in the mid-IR continuum at ∼20µm, which had also been noted as a “broken
power-law” in some Seyfert galaxies by Buchanan et al. (2006). Another more extreme such
case was the metal-poor blue compact dwarf galaxy SBS0335-052E (Houck et al. 2004b),
and it was interpreted with the possible absence of larger cooler dust grains in the galaxy.
Since the change of continuum slope appears at ∼20µm, we use the flux ratio at 20 and
30µm, F20/F30, to identify these objects. After further examination of the spectra, we find
15 Sy 1s and 4 Sy 2s, which have a F20/F30≥0.95. We call these objects “20µm peakers”.
To analyse the properties of “20µm peakers”, we plot the average IRS spectra of the
19 sources in Figure 5. As most of these sources are type 1 Seyferts (15 out of 19), we
also overplot the average Sy 1 spectrum for comparison. In addition to their characteristic
continuum shape, a number of other differences between the “20µm peakers” and Sy 1s are
also evident. PAH emission, which is clearly detected in the average Sy 1 spectrum, appears
to be rather weak in the average “20µm peaker” spectrum. The high-ionization lines of
[NeV] and [OIV] are seen in both spectra with similar strength, while low-ionization lines,
especially [NeII] and [SIII], are much weaker in the average spectrum of “20µm peakers”. If
we calculate the infrared color of a galaxy using the ratio of F25/F60(see section 3.3 for more
detailed discussion on this issue), we find an average value of 0.75 for the “20µm peakers”,
while it is 0.30 for the other “non-20µm peaker” Sy 1s in the 12µm sample. Finally, the
6All spectra are also available in electronic format.

Page 10

– 10 –
average IR luminosities of the “20µm peakers” and Sy 1s do not show significant difference,
with log(LIR/L⊙)=10.96 for the former and log(LIR/L⊙)=10.86 for the latter. These results
are consistent with the “20µm peakers” being AGNs with a dominant hot dust emission
from a small grain population heated to effective temperatures of ∼ 150K and a possible
contribution due to the distinct emissivity of astronomical silicates at 18µm. Their radiation
field must also be stronger than a typical Sy 1, since it destroys the PAH molecules around
the nuclear region more efficiently. We should stress that unlike SBS0335-052E, whose global
IR SED peaks at ∼30µm, and likely has limited quantities of cold dust, the mid-IR peak
we observe in these objects at ∼20µm is likely a local one7. This becomes more evident in
Figure 6, where we also include the scaled average 60 and 100µm flux densities for the “20µm
peakers”. It is clear that their far-IR emission increases with wavelength, thus confirming
the presence of ample quantities cold dust in these objects. To contrast the global SED of
these objects with the whole Seyfert sample we include in Figures 7 and 8 the average SED
of the Sy 1s and Sy 2s of the sample. All SEDs have been normalized as in Figure 4 at
22µm. Unlike the “20µm peakers” one can easily observe the regular increase of the flux
from ∼15 to 60µm in the average Sy 1 and Sy 2 SEDs. A more detailed analysis of the
dust properties of the “20µm peakers” in comparison with typical active galaxies will be
presented in a future paper.
3.2. The PAH emission in the 12µm Seyferts
In this section, we explore some of the properties of the PAH emission in our sample
and contrast our findings to the previous work. To quantify the strength of PAH emission,
we follow the usual approach and measure the fluxes and equivalent widths (EWs) of the 6.2
and 11.2µm PAH features from their mid-IR spectra. Even though the 7.7µm PAH is the
strongest of all PAHs (Smith et al. 2007a), we prefer not to include it in our analysis. This
is due to the fact that its measurement is affected by absorption and emission features next
to it and depends strongly both on the assumptions of the various measurement methods
(spline or Drude profile fitting) as well as the underlying continuum. Furthermore, it often
spans between the two SL orders, which could also affect its flux estimate. The 6.2 and
11.2µm PAH EWs are derived by integrating the flux of the features above an adopted
continuum and then divide by the continuum fluxes in the integration range. The baseline
is determined by fitting a spline function to the selected points (5.95-6.55µm for the 6.2µm
PAH and 10.80-11.80µm for the 11.2µm PAH). The PAH EWs as well as the integrated
fluxes are listed in Table 2. The errors including both flux calibration and measurement are
7However, some objects, such as Mrk335, Mrk704 and 3C234, do have IRAS ratios of F25/F60>1.

Page 11

– 11 –
estimated to be less than ∼15% on average.
The first study on the PAH properties of a large sample of Seyferts was presented by
Clavel et al. (2000), using ISO/PHOT-S 2.5–11µm spectra and ISO/CAM broad band mid-
IR imaging. The authors suggested that there was a statistical difference in the strength of
the PAH emission and the underlying hot continuum (∼ 7µm) emission between type 1 and
type 2 objects. They also found that Sy 2s had stronger PAHs than Sy 1s, while Sy 1s had
higher hot continuum associated with emission from the AGN torus. This was consistent
with an orientation depended depression of the continuum in Sy 2s. The interpretation of
these results was challenged by Lutz et al. (2004), who attributed it to the large (24′′×
24′′) aperture of ISO/PHOT-S, and the possible contamination from the host galaxy. More
recently, Deo et al. (2007) have found a relation between the 6.2µm PAH EWs and the
20-30µm spectral index8, with a steeper spectral slope seen in galaxies with a stronger
starburst contribution. This is understood since galaxies hosting an AGN are “warmer” and
have an IR SED peaking at shorter wavelengths thus appearing flatter in the mid-IR (see
also next section). Given the global correlations between star formation activity and PAH
strength (F¨ orster Schreiber et al. 2004; Peeters et al. 2004; Wu et al. 2005; Calzetti et al.
2005, 2007, i.e.), star-forming galaxies are expected to also have stronger PAH features. In
Figure 9, we plot the 15-30µm9spectral index for the 12µm Seyfert sample as a function of
their 11.2µm PAH EWs. The diamonds indicate the starburst galaxies from Brandl et al.
(2006)10. A general trend of the PAH EWs decreasing as a function of 15-30µm spectral index
is observed in Figure 9, even though it is much weaker than the anti-correlation presented by
Deo et al. (2007). Starburst galaxies are located on the upper left corner of the plot, having
very steep spectral slopes, with < α15−30>=-3.02±0.50, and large PAH EWs, nearly 0.7µm.
Seyfert galaxies spread over a considerably larger range in spectral slopes as well as PAH
EWs. Sy 1s and Sy 2s are mixed on the plot. On average, the 15-30µm spectral index is
< α15−30>=-0.85±0.61 for Sy 1s and < α15−30>=-1.53±0.84 for Sy 2s. Note that although
the mean spectral slope is slightly steeper for Sy 2s, there is substantial scatter, as is evident
by the uncertainties of the mean for each types (see also Figure 7 and 8).
It is well known that the flux ratio of different PAH emission bands is a strong function
of PAH size and their ionization state (Draine & Li 2001). The 6.2µm PAH emission is due
8The spectral index α is defined as log(F1/F2)/log(ν1/ν2).
9We choose to use α15−30 so that we can directly make use of the spectral index measurement for the
starburst galaxies in the Brandl et al. (2006) sample.
10We have excluded the 6 galaxies that have AGN signatures from the Brandl et al. (2006) starburst
sample.

Page 12

– 12 –
to C-C stretching mode and the 11.2µm feature is produced by C-H out-of-plane bending
mode (Draine 2003). In Figure 10, we display a histogram of the 11.2µm to 6.2µm PAH flux
ratios for the 12µm Seyferts. Given the relatively small number of starburst galaxies in the
Brandl et al. (2006) sample (16 sources), we also included 20 HII galaxies from the SINGS
sample of Smith et al. (2007a), thus increasing the number of SF galaxies to 36 sources
and making its comparison with the Seyferts more statistically meaningful. However, since
Smith et al. (2007a) adopted multiple Drude profile fitting for the measurement of PAH
features, for reasons of consistency, we re-measured the PAH fluxes of the 20 SINGS galaxies
with our spline fitting method. From Figure 10, we can see that both the Seyferts and SF
galaxies, indicated by the solid and dashed line respectively, appear to have very similar
distribution of PAH 11.2µm/6.2µm band ratios. The average PAH flux ratios for the Seyfert
sample is 0.94±0.37 while for the SF galaxies is 0.87±0.24, which agree within 1-σ. This
is also consistent with the findings of Shi et al. (2007), who reported similar 11.2µm/7.7µm
flux ratios between a sample of higher redshift AGNs (PG, 2MASS and 3CR objects) and the
SINGS HII galaxies. This implies that even though the harsh radiation field in AGNs may
destroy a substantial amount of the circumnuclear PAH molecules, and does so preferentially
– the smaller PAHs being destroyed first (Draine & Li 2001; Smith et al. 2007a) – it likely
does not do so over a large volume. Enough molecules in the circumnuclear regions do remain
intact and as a result, the aromatic features that we observe from Seyferts are essentially
identical to those in SF galaxies.
The relative strength of PAH emission can also be used to examine the validity of the
unified AGN model. As mentioned earlier, this model attributes the variation in AGN types
as the result of dust obscuration and relative orientation of the line of sight to the nucleus
(Antonucci 1993). Sy 1s and Sy 2s are intrinsically the same but appear to be different in the
optical, mainly because of the much larger extinction towards the nuclear continuum of Sy
2s when viewed edge on. The latest analysis of the IRS high-resolution spectra of 87 galaxies
from the 12µm Seyfert sample shows that the average 11.2µm PAH EW is 0.29±0.38µm
for Sy 1s and 0.37±0.35µm for Sy 2s (Tommasin et al. 2009, in preparation). As we show
in Table 3, the 11.2µm PAH EWs of the whole 12µm Seyfert sample (90 objects excluding
upper limits) is 0.21±0.22µm for the Sy 1s and 0.38±0.30µm for the Sy 2s. The difference
we observe in the PAH EWs between the two Seyfert types is somewhat larger than the one
reported by Tommasin et al. (2009, in preparation), though still consistent within 1σ. This
further suggests that there is little discernible difference between Sy 1s and Sy 2s at this
wavelength.
If the observed AGN emission in the infrared does not depend on the line of sight of
the observer, one can compare the circumnuclear PAH emission of Sy 1s and Sy 2s of similar
bolometric luminosities to test the unified model. If we bin the sources according to their

Page 13

– 13 –
IR luminosity, we find an average PAH EW of 0.22±0.23µm for Sy 1s and 0.40±0.30µm
for Sy 2s with LIR<1011L⊙; while the average PAH EWs are 0.19±0.19µm for Sy 1s and
0.37±0.30µm for Sy 2s with LIR≥1011L⊙(see Table 3 for a summary of these results). The
difference between the two Seyfert types is still less than 1σ. This result is not in agreement
with the findings of Buchanan et al. (2006) who based on a principle component analysis find
that in their subset of 51 galaxies the Sy 2s show a stronger starburst eigenvector/template
contribution than the Sy 1s. As the authors suggest this might be due to a bias of selection
effects. Our result can be interpreted as indicating that there is some, but not substantial
obscuration in the mid-IR. As a consequence we are able to probe deep into the nuclear region,
sampling most of the volume responsible for the mid-IR emission. This result is consistent
with a similar finding of Buchanan et al. (2006) who compared the mid-IR to radio ratio
for their sample. They concluded that the observed factor of ∼2 difference between the two
types would imply either a smooth torus which is optically thin in the mid-IR or a clumpy
one containing a steep radial distribution of optically thick dense clumps (Nenkova et al.
2008).
3.3.Cold/Warm AGN diagnostics
The IRAS 25 and 60µm flux ratio has long been used to define the infrared color
(“warm” or “cold”) of a galaxy, with “warm” galaxies having a ratio of F25/F60 >0.2
(Sanders et al. 1988). In Figure 11, we plot the 11.2µm PAH EW as a function of the
flux ratio between F25and F60for the 12µm Seyfert sample. The IRAS 25 and 60µm fluxes
were compiled from Rush et al. (1993) and Sanders et al. (2003) and are listed in Table 1.
The aperture of the IRAS broad band filters is on the order of a few arcminutes on the sky11
and typically encompass the whole galaxy, while the PAH EW is measured from a spectrum
of a smaller region centered on the nucleus of the galaxy (see Table 2). Nevertheless, we
observe a clear trend of the 11.2µm PAH EW decreasing with F25/F60ratio in Figure 11. On
this plot, we also include the SF galaxies from Brandl et al. (2006) and Smith et al. (2007a).
All SF galaxies appear to be clustered on the top left corner of the plot, having high PAH
EWs and low F25/F60 values, suggesting strong star formation and cooler dust tempera-
tures. The observed suppression of PAH emission seen in the warm AGNs implies that the
soft X-ray and UV radiation of the accretion disk, which destroys the PAH molecules, is
also reprocessed by the dust and dominates the mid- and far-IR colors. More specifically,
11According to the IRAS Explanatory Supplement Document for unenhanced coadded IRAS images the
resolution is approximately 1’× 5’, 1’× 5’, 2’× 5’ and 4’× 5’ at 12, 25, 60 and 100µm, respectively ( see
http://lambda.gsfc.nasa.gov/product/iras/docs/exp.sup ).

Page 14

– 14 –
warm Sy 1s have an average 11.2µm PAH EW of 0.10±0.12µm, while for Sy 2s the value
is 0.18±0.24µm. Similarly for the cold sources, the average PAH EW is 0.40±0.23µm for
Sy 1s and 0.59±0.19µm for Sy 2s. We observe a ∼3σ difference in the PAH EWs between
the cold and warm sources, independent of their Seyfert type. This indicates that as the
emission from the accretion disk surrounding the SMBH of the active nucleus contributes
progressively more to the IR luminosity, its radiation field also destroys more of the PAH
molecules and thus diminishes their mid-IR emission.
The trend of PAH EWs decreasing with F25/F60has also been detected in a large sample
of ULIRGs studied by Desai et al. (2007). The luminosities of the 12µm Seyfert sample are
more comparable with LIRGs, thus our work extends the correlation of Desai et al. (2007)
to a lower luminosity range. This is rather interesting since deep photometric surveys with
Spitzer can now probe normal galaxies as well as LIRGs at z∼1 (Le Floc’h et al. 2005), a
fraction of which are known to host an AGN based on optical spectra and mid-IR colors
(Fadda et al. 2002; Stern et al. 2005; Brand et al. 2006).
We have also investigated the dependence of the 6.2µm PAH EW on the F25/F60ratio for
our sample. A similar trend of the 6.2µm EWs decreasing with F25/F60is observed as well,
though there is larger scatter than the one seen with the 11.2µm feature. This is probably
due to the fact that the 6.2µm PAH EW is intrinsically fainter and only upper limits could
be measured for a number of source (see Table 2) . Despite the scatter, this trend is still
a rather important finding, because for high redshift galaxies (z>2.5), the 6.2 and 7.7µm
PAHs might be the only features available in the wavelength range covered by the IRS, thus
measuring them can reveal essential information on the star-formation luminosity and dust
composition of high-redshift galaxies (see Houck et al. 2005; Yan et al. 2005; Teplitz et al.
2007; Weedman & Houck 2008).
3.4. The Silicate Strength of the 12µm Seyferts
In the mid-IR regime, one can examine not only the structure of complex organic
molecules and determine their aromatic or aliphatic nature, but can also probe the chem-
istry of dust grains (see van Dishoeck 2004). One of the most prominent continuum fea-
tures in the 5-35µm range is the one associated with the presence of astronomical silicates
in the grains, which are characterized by two peaks in the emissivity at 9.7 and 18µm (see
Dudley & Wynn-Williams 1997). The silicate features had been detected in absorption in SF
galaxies, protostars and AGN for over 30 years (ie Gillett et al. 1975; Rieke & Low 1975), but
it was the advent of space observatories such as ISO and Spitzer, which allowed for the first
time of study over a wide range of astronomical objects and physical conditions. According

Page 15

– 15 –
to the unified model, an edge-on view through the cool dust in the torus will cause the silicate
feature to be seen in absorption, while with a face-on view, the hot dust at the inner surface
of the torus will cause an emission feature for the silicate (Efstathiou & Rowan-Robinson
1995). Even though silicate in emission at 9.7µm had already been observed in the SF region
N66 (Contursi et al. 2000), emission at both 9.7µm and 18µm in AGNs and Quasars was de-
tected with Spitzer (Siebenmorgen et al. 2005; Hao et al. 2005; Sturm et al. 2005), providing
strong support for the unified model (Antonucci 1993). Using Spitzer/IRS data, Hao et al.
(2007) compiled a large, though inhomogeneous sample of AGNs and ULIRGs, and uni-
formly studied the silicate features in these galaxies. Using the same sample, Spoon et al.
(2007) proposed a new diagnostic of mid-IR galaxy classification based on the strength of
silicate and PAH features. To put in the same context the properties of the silicate feature
in the 12µm Seyfert sample, we also measured the strength of the silicate at 9.7µm, using
the definition and approach of Spoon et al. (2007) as:
Ssil= lnfobs(9.7µm)
fcont(9.7µm)
(1)
where fcont(9.7µm) is the flux density of a local mid-IR continuum, defined from the
5-35µm IRS spectrum. Sources with silicate in emission have positive strength and those
in absorption negative. Buchanan et al. (2006) did identify two AGN with silicate emission
and two more with broad silicate absorption out of the 51 sources they studied. In this
paper, we provide for the first time measurements on the silicate features for this complete
sample. We follow the prescription of Spoon et al. (2007) and Sirocky et al. (2008) for the
continuum definition and identify the sources to be PAH-dominated, continuum-dominated
and absorption-dominated. The values of Ssilmeasured from the 12µm Seyferts can be found
in Table 2.
In Figure 12a, we plot the 11.2µm PAH EWs as a function of the 9.7µm silicate
strength12. We observe that most Sy 1s are located close to Ssil=0 and the range of their
silicate strength is rather narrow, with the exception of one galaxy, UGC5101, which is also
one of the ULIRGs of the Bright Galaxy Sample (Armus et al. 2004, 2007). This is in agree-
ment with the results of Hao et al. (2007), who found that Sy 1s are equally likely to display
the 9.7µm silicates feature in emission and in absorption. The Sy 2s have a larger scatter in
the value of the silicate strength, with most of them showing the feature in absorption. The
12A similar plot using the 6.2µm PAH EWs was proposed by Spoon et al. (2007) as a mid-IR galaxy
classification method