Towards a Complete Census of AGNs in Nearby Galaxies:
A Large Population of Optically Unidentified AGNs
Using Spitzer-IRS spectroscopy, we investigate the ubiquity of Active Galactic Nuclei (AGN) in a complete ( percent), volume-limited sample of the most bolometrically-luminous galaxies ( (0.3–20) ) to Mpc. Our analyses are based on the detection of the high-excitation emission line [NeV] (97.1 eV) to unambiguously identify AGN activity. We find that 17 of the 64 IR-bright galaxies in our sample host AGN activity ( percent), percent of which are not identified as AGNs using optical spectroscopy. The large AGN fraction indicates a tighter connection between AGN activity and IR luminosity for galaxies in the local Universe than previously found, potentially indicating a close association between AGN activity and star formation. The optically unidentified AGNs span a wide range of galaxy type (S0–Ir) and are typically starburst-dominated systems hosting modest-luminosity AGN activity (–). The non-identification of optical AGN signatures in the majority of these galaxies appears to be due to extinction towards the AGN, rather than intrinsically low-luminosity AGN activity; however, we note that the AGN optical signatures are diluted in some galaxies due to strong star-formation activity. Examination of optical images shows that the optically unidentified AGNs with evidence for extinction are hosted in either highly inclined galaxies or galaxies with dust lanes, indicating that obscuration of the AGN is not necessarily due to an obscuring torus. We therefore conclude that optical spectroscopic surveys miss approximately half of the AGN population simply due to extinction through the host galaxy.
keywords:galaxies: active – galaxies: evolution – galaxies: nuclei – infrared: galaxies
The seminal discovery that all massive galaxies in the local Universe harbour super-massive black holes (SMBHs; ) implies that all massive galaxies have hosted Active Galactic Nuclei (AGN) at some time during the last 13 Gyrs (e.g., Kormendy & Richstone 1995; Magorrian et al. 1998; Gebhardt et al. 2000). Sensitive blank-field surveys have traced the evolution of luminous AGN activity out to 5–6, providing a window on the growth of SMBHs across 95% of cosmic time (e.g., Ueda et al. 2003; Croom et al. 2004; Fan et al. 2004; Hasinger et al. 2005; Richards et al. 2006). However, these blank-field surveys typically lack the volume and sensitivity to provide a complete census of AGN activity in the local Universe. Such a census is required to (1) provide a baseline with which to interpret the results obtained for distant AGNs from blank-field surveys, and (2) provide definitive constraints on the growth of SMBHs in the local Universe (e.g., the growth rates of SMBHs; the relative amount of obscured and unobscured SMBH growth; the galaxies and environments where SMBHs are growing).
Arguably, the most complete census of AGN activity in the local Universe is the optical spectroscopic survey of Ho et al.(1997ba,b; hereafter Ho97). Ho97 classified nearly all galaxies with mag in the Northern hemisphere and identified AGNs on the basis of the relative strength of forbidden and permitted emission lines (e.g., Baldwin et al. 1981; Veilleux & Osterbrock 1987; Kewley et al. 2001); objects with a broad permitted line component (FWHM ) were also classified as AGNs. Using this classification scheme, Ho97 found that 10% of galaxies unambiguously host AGN activity (i.e., optically classified as Seyfert galaxies). These AGNs were found to predominantly reside in moderately massive bulge-dominated galaxies (Hubble type E–Sbc). However, since the source selection and classification were performed at optical wavelengths, the Ho97 studies were insensitive to the identification of the most heavily dust-obscured AGNs. For example, the nearby Scd galaxy, NGC 4945 is classified as a starburst galaxy at optical wavelengths and only reveals the presence of AGN activity using mid-IR spectroscopy (Spoon et al., 2000) and X-ray observations (Iwasawa et al., 1993). Various studies have suggested that NGC 4945 is unlikely to be a particularly unusual AGN (e.g., Lutz et al. 2003; Maiolino et al. 2003), indicating that there may be many more optically unidentified AGNs in the local Universe.
X-ray observations have revealed potential AGNs in many galaxies in the Ho97 sample lacking optical AGN signatures (e.g., Ho et al. 2001; Desroches & Ho 2009). However, there is often ambiguity over whether an AGN is producing the X-ray emission in these galaxies (e.g., there can be significant contamination from X-ray binaries). Furthermore, the X-ray emission from Compton-thick AGNs ( cm) can be extremely weak, making it challenging to identify the most heavily obscured AGNs using X-ray data alone (i.e., the keV emission can be a factor 30–1000 times weaker than the intrinsic emission; e.g., Risaliti et al. 1999; Matt et al. 2000). By comparison, due to the extreme conditions required to produce the high ionization emission line [NeV] , (97.1 eV), mid-IR spectroscopy provides an unambiguous indicator of AGN activity in nearby galaxies (e.g., Weedman et al. 2005; Armus et al. 2006).111We note that [OIV] 25.9 m (54.9 eV) is also often used for AGN identification, although energetic starbursts can also produce luminous [OIV] emission; see §3.4. The relative optical depth at mid-IR wavelengths is also considerably lower than at optical wavelengths (/ 50; Li & Draine 2001). Even heavily Compton-thick AGNs that are weak at X-ray energies can be identified using mid-IR spectroscopy (e.g., the well-studied NGC 1068 with has bright [NeV] emission; Sturm et al. 2002). Thus, the identification of high-ionization [NeV] , provides a relatively optically thin means by which to probe the central engine of nearby AGNs. On the basis of the identification of [NeV] emission in a heterogeneous sample of late-type spiral galaxies within the local Universe, Satyapal et al. (2008; hereafter, S08) have suggested that a large number of Sc–Sm galaxies host optically unidentified AGN activity.
Local mid-IR surveys (e.g., Sturm et al. 2002; the Spitzer Infrared Nearby Galaxy Survey [SINGS] Legacy Project [Dale et al. 2006; hereafter D06] and S08) have shown the advantages of using mid-IR spectroscopy as an AGN diagnostic. However, none of these studies have used this diagnostic to provide a complete unambiguous census of AGN activity within the local Universe. Here we use sensitive high-resolution Spitzer-IRS spectroscopy to identify AGNs within a complete ( percent) volume-limited survey of the most bolometrically luminous galaxies () to a distance of Mpc.222 corresponds to the 8–1000 m luminosity, as defined by Sanders & Mirabel (1996). By selecting galaxies at IR wavelengths, our sample will comprise the most active galaxies in the local Universe and will also include the most dust-obscured systems. We use these data to unambiguously identify AGNs using the high-ionization [NeV] emission line to produce the most sensitive census of AGN activity in the local Universe to date. In §2 we outline the construction and data reduction analysis of the sample assembled from the IRAS Revised Bright Galaxy Survey of Sanders et al. (2003; RBGS). In §3 we determine the fraction of local galaxies hosting AGN activity, explore their properties, and compare the results to the previous optical survey of Ho97 to address the key question: Why are a large number of AGNs unidentified at optical wavelengths? In §4 we present our conclusions.
2 The Sample and Data Reduction
2.1 Sample Selection
Using IRAS, the RBGS (Sanders et al. 2003) has provided an accurate census of all IR-bright galaxies (, Jy) in the local Universe. The aim of our study is to identify AGN activity in the most bolometrically luminous galaxies () out to Mpc.333Distances have been calculated using the cosmic attractor model of Mould et al. (2000), which adjusts heliocentric redshifts to the centroid of the local group, taking into account the gravitational attraction towards the Virgo cluster, the Great Attractor, and the Shapley supercluster. The distance constraint of 15 Mpc was placed so as to not include the Virgo cluster at 16 Mpc (i.e., to be representative of field-galaxy populations). The IR luminosity threshold was chosen to be well matched to the flux limit of the RBGS (see Fig. 1) and ensures that we do not include low-luminosity dwarf galaxies and relatively inactive galaxies. In the RBGS there are 68 IRAS detected galaxies to a
|NGC 1068||02h42m41.4s||+00d00m45s||13.7||Sb||11.27||S2 (1.8)||5.29||12.82||0.24||0.76||38235.10||1|
|NGC 4051||12h03m09.8s||+44d31m50s||13.1||Sbc||9.90||S2 (1.5)||3.30||4.47||0.36||0.65||726.25||1|
|NGC 5033||13h13m27.2s||+36d35m40s||13.8||Sc||10.13||S2 (1.9)||4.48||4.68||1.07||2.34||96.28||1|
NOTES: (1) Common galaxy name. (2–3) 2MASS near-IR position of galactic nucleus. (4) Luminosity distance to source in Mpc from the RBGS (Sanders et al. 2003). (5–6) Projected spectral apertures (SH and LH respectively) in kiloparsecs. (7) Morphological classification from RC3 (de Vaucouleurs et al., 1991). (8) Logarithm of IR luminosity (8–1000) from RBGS. (9) Optical spectral class from BPT diagnostics (S2: Seyfert 2; broad-line sub-class in parentheses, HII: Star-forming galaxy, and L: LINER). (10–13) Optical emission line ratios. (14) Observed [OIII] flux in units of . (15) References for published optical data.
REFERENCES: (1) Ho et al. (1997a); (2) Kewley et al. (2001); (3) Veilleux et al. (1995); (4) Veron-Cetty et al. (1986); (5) Moorwood et al. (1996); (6) Moustakas & Kennicutt (2006); (7) Kirhakos & Steiner (1990); (8) Ganda et al. (2006); (9) Barth et al. (2009).
distance of Mpc with , 64 of which have Spitzer-IRS high-resolution spectroscopy publicly available (i.e., percent complete): P3124 (28 objects; PI: D.M. Alexander), P159 (18 objects; PI: R. Kennicutt [SINGS]), P14 (11 objects; PI: J.R. Houck), P59 (4 objects; PI: G. Rieke) and P86 (3 objects; PI: M. Werner).
In Fig. 1 we plot IR luminosity versus luminosity distance for the RBGS and highlight the 64 galaxies with Spitzer-IRS observations in our Mpc sample. The basic properties from the RBGS for the sources are combined with published optical data and listed in Table 1. The objects are all late-type galaxies (Hubble classification of S0 or later), which is unsurprising since early-type galaxies are typically IR faint and undetected by IRAS (e.g., Knapp et al. 1989). The four galaxies that match our selection criteria but lack sufficient high-resolution Spitzer-IRS observations of the central regions are shown in Table 2. Specifically, NGC 3486 has Short-High (SH) and Long-High (LH) observations but the data are noisy and no statistically useful information can be extracted; NGC 4565 and NGC 5457 have only high-resolution observations of extranuclear regions; NGC 5248 has no high-resolution observations. NGC 5248 and NGC 5457 are optically classified as star-forming HII galaxies while NGC 3486 and NGC 4565 are optically classified as Seyfert galaxies (Ho97). Both NGC 3486 and NGC 4565 have published [OIV] fluxes from low-resolution () Spitzer-IRS spectroscopy in Diamond-Stanic et al. (2009); the derived [OIV] luminosities suggest that the AGNs are contributing percent to the total bolometric luminosity of the galaxy.
NOTES: (1) Common galaxy name. (2) Morphological classification from RC3. (3) Luminosity distance in Mpc from RBGS. (4) Logarithm of IR luminosity (8–1000) from RBGS. (5) Optical spectral class from Ho97 using BPT diagnostics; see Table 1. (6) Status of Spitzer-IRS data: 1. High-resolution observations are publicly available but signal-to-noise is not sufficient for analysis; 2. High-resolution observations available only for extranuclear regions; 3. No high-resolution observations available. (7) Logarithm of [OIV] luminosity from low-resolution () Spitzer-IRS spectroscopy in units of (Diamond-Stanic et al. 2009).
2.2 Data Reduction
Each of the galaxies in our Mpc sample were observed using both the SH (, –) and LH (, –) resolution spectrographs onboard the NASA Spitzer Space Telescope (Houck et al. 2004; Werner et al. 2004). The spectral resolution is for both SH and LH modules. The raw data are compiled from multiple observing programs (see §2.1) and they consist of both spectral mapping and staring observations, with differing exposure times. As an example, in Fig. 2 we project the SH and LH apertures onto a galaxy in our sample (NGC 0278), showing the two differing nod positions.
Many of the objects in our sample do not have dedicated off-source observations. However, as we only require emission-line flux measurements no background subtraction was necessary for the observations (see §184.108.40.206 of the Spitzer-IRS Observers Manual).444The Spitzer-IRS Observers Manual is available at http://ssc.spitzer.caltech.edu/irs/dh/ Furthermore, the targets are bright compared to the background and therefore any background corrections would be small.
For the reduction of the IRS-staring data, Basic Calibrated Data (BCD) images were co-added and mean averaged at each nod position. Few of the observations were pre-processed with the same Spitzer pipeline version. Therefore, to ensure consistency, we extracted the spectra of each galaxy using a custom pipeline based on the Spitzer data reduction packages IRSCLEAN (to apply individual custom bad-pixel masks for each of the BCDs) and SPICE (to extract full slit spectra using the latest flux calibration files: version 17.2). For further information on the high-resolution Spitzer-IRS pre-processing pipeline, see Chapter 7 of the Spitzer Observers Manual.
|()||21.6 eV||97.1 eV||41.0 eV||97.1 eV||54.9 eV||7.9 eV||23.3 eV||8.2 eV|
|()||21.6 eV||97.1 eV||41.0 eV||97.1 eV||54.9 eV||7.9 eV||23.3 eV||8.2 eV|
NOTES: (1) Common galaxy name. (2) Equivalent width of the PAH feature in units of . (3–10) Fluxes and their statistical uncertainties for the measured mid-IR narrow emission lines in units . The mean uncertainty of the fluxes is approximately 10 percent. upper limits are quoted for non-detections. (11) Mid-IR AGN on the basis of [NeV] emission.
For the reduction of the IRS-spectral mapping data, custom bad-pixel masks were again applied to the BCD images. CUBISM (Smith et al., 2007) was then used in conjunction with the latest flux calibration files to construct final data cubes and extract spectra of the central regions matched to the sizes of projected SH and LH apertures.
After extraction of the raw spectra, the ends of each echelle order were trimmed to remove the additional spectral noise caused by the poor response of the grating.555Wavelength trim ranges are given in Table 5.1 of the Spitzer-IRS Observers Manual. Using the redshifts given in the RBGS, each extracted spectrum was shifted to rest-wavelength for further spectral analysis.
Solely for presentation purposes, single continuous spectra of each slit were produced. Echelle orders were matched by fitting each spectral continuum from a given order with either a first or second order polynomial. Each echelle order continuum was then combined by matching and calibrating to the 1st echelle order of the relevant slit to construct the final SH and LH continua. Due to the different aperture sizes (and hence continuum fluxes), we have not attempted to match the SH and LH spectra. Fig. 12 shows the reduced spectra for each of the sources using the SH and LH slits (left and right panels respectively).
2.3 Measuring Emission-line Properties
Second-order polynomials were used to model the continuum and gaussian profiles were simultaneously fit to spectral features to determine emission-line fluxes. These were calculated using the IDL-based spectral analysis tool SMART (Higdon et al., 2004).666SMART was developed by the IRS Team at Cornell University and is available through the Spitzer Science Center at Caltech. Fluxes or upper-limits are given in Table 3 for the following emission lines: [NeII] , [NeV] , [NeIII] , [NeV] , [OIV] , [FeII] , [SIII] and [SiII] ). Polycyclic Aromatic Hydrocarbon (PAH; e.g., Draine 2003) features are detected in many of the galaxies; we use the PAH equivalent width in our analyses and report these values in Table 3. Due to their broad profiles, the strength of the PAH features are measured using multiple gaussians. The results obtained from our continuum and emission-line fitting procedure are shown in Fig. 3 for all [NeV] detected galaxies.
Of the 64 galaxies in the Mpc sample, 16 have published Spitzer-IRS data in S08 and 18 have published Spitzer-IRS data in Dale et al. (2009; hereafter D09). However, due to differing data-reduction routines and approaches in the detection of emission lines (e.g., we measured emission-line properties in apertures centred on the near-IR nucleus, S08 searched for emission lines in small apertures across the circumnuclear region of each galaxy, and D09 constrained emission-line properties in large apertures across the extent of each galaxy), we have re-analysed all of the galaxies to provide self-consistent results. However, despite these differing approaches, we find average emission-line flux variances of only 10–30% between our study and that of S08 and D09 for those galaxies with with detected [NeII] , [NeIII] , and [OIV] emission; when comparing to D09 we scaled our emission-line fluxes by the difference in aperture size between D09 and our study.
3.1 The Discovery of a Significant Population of Optically Unidentified AGNs
Of the 64 objects in our Mpc sample (presented in Tables 1 and 3), 17 have detections of the [NeV] emission line, and therefore have unambiguous evidence for AGN activity.777Theoretical modeling has shown that [NeV] can also be produced in galaxies containing a large population of Wolf-Rayet (WR) stars (Schaerer & Stasińska, 1999). Two WR galaxies are present in our volume-limited sample (IIZw40 and NGC 1569). However, [NeV] emission is not detected in either source, and their mid-IR continuum and spectral features are clearly distinct from those of AGNs. We therefore find an overall AGN fraction of percent in the most bolometrically luminous galaxies to Mpc ().888Small-number Poisson statistical errors are calculated for upper and lower limits based on the confidence levels given in Gehrels (1986).
Of the 16 galaxies in common between S08 and our sample, S08 identified [NeV] (14.32 or 24.32 ) in four systems (NGC 3556; NGC 3938; NGC 4536; NGC 5055). We do not detect significant [NeV] in any of these four galaxies, which may be due to differences in the positions and sizes of the apertures used to extract the spectra within CUBISM. Of the 18 galaxies in common between D09 and our sample, D09 identified [NeV] in three systems (NGC 3621; NGC 5033; NGC 5194), all of which we also identify here; however, we also identified [NeV] emission in two systems where D09 quote [NeV] upper limits (NGC 3627; NGC 5195), which could be due to dilution of the [NeV] emission by the host galaxy in the large apertures used by D09. We note that differences in the identification of weak [NeV] emission can also be due to the adopted emission-line detection procedure and signal-to-noise ratio threshold.
Seven ( percent) of the 64 galaxies in our sample are unambiguously identified as AGNs using classical optical emission-line diagnostics (i.e., optically classified as Seyfert galaxies; e.g., Baldwin et al. 1981; Veilleux & Osterbrock 1987; Kauffmann et al. 2003); see Table 1. All of these optically identified AGNs are classified as AGNs at mid-IR wavelengths in our analysis. In Fig. 5, we plot the optical emission-line ratios of the 53 galaxies with good-quality optical spectra; of the other 11 galaxies in the sample, three do not have a sufficient number of detected emission lines to classify at optical wavelengths using emission-line diagnostics, and eight do not have published optical spectroscopy. Ten of our mid-IR classified AGNs are not unambiguously identified as AGNs at optical wavelengths: five are classified as HII galaxies (NGC 0613, NGC 1448, NGC 1792, NGC 4945, and NGC 5128), four are classified as LINERS (NGC 0660, NGC 3627, NGC 3628, and NGC 5195), and one does not have good-quality optical spectroscopic data (ESO121-G006). Although ESO121-G006 does not have a good-quality optical spectrum, since it is hosted in a highly inclined galaxy we would expect it to be an optically unidentified AGN; see §3.3 and Fig. 9. Large optical spectroscopic studies (e.g., Veilleux et al. 1995; Ho97) have speculated that the central regions of many LINERs are likely to be powered by AGN activity. Indeed, using Chandra X-ray observations, Ho et al. (2001) suggest percent of all LINERs host AGNs. By contrast, using our mid-IR diagnostics, we find only percent of IR-bright LINERs appear to be AGNs, which could be due to a number of factors (e.g., different sample selection and ambiguous evidence for AGN activity at X-ray energies; i.e., X-ray binaries).
The optical spectroscopy for the galaxies in our sample comes from a variety of different studies with a range in emission-line sensitivities. However, the majority of the galaxies are in Ho97, which arguably comprises the most sensitive optical spectroscopy of a large number of nearby galaxies. Of the 38 galaxies present in both our sample and Ho97, we find that eight are mid-IR identified AGNs (NGC 0660, NGC 1068, NGC 3627, NGC 3628, NGC 4051, NGC 5033, NGC 5194 and NGC 5195), only four of which are unambiguously identified as AGN at optical wavelengths. One of the optically unidentified AGNs is NGC 5195, which may be a binary AGN system with NGC 5194 in the Whirlpool galaxy. Potential AGN activity has also been found in NGC 5195 using Chandra observations (Terashima & Wilson, 2004).
Optical spectroscopic surveys have found that AGNs typically reside in moderately massive bulge-dominated galaxies (Hubble-type: E–Sbc; (0.1–3) ; e.g., Ho97; Heckman et al. 2004). In Fig. 6 we show the histogram of galaxy morphology for our Mpc sample. We find that the host galaxies of our mid-IR identified AGNs cover a wide range of galaxy type (S0–Ir). However, in contrast to Ho97 and Heckman et al. (2004), we find that a large fraction of Sc–Sd-type galaxies host AGN activity at mid-IR wavelengths ( percent; i.e., a comparable AGN fraction to that found in Sab–Sbc galaxies). This shows that late-type galaxies typically assumed to host pseudo bulges (Sc–Sd) can harbour AGN activity, and therefore must host a SMBH. As found in previous studies, this indicates that galaxies without classical bulges can host SMBHs (e.g., Greene et al. 2008; Barth et al. 2009).
In Fig. 6 we also show the incidence of AGN activity as a function of IR luminosity. In the moderate-luminosity IR bin ( (1–3) ), we find a large AGN fraction of percent, suggesting that the overall AGN fraction for our sample may be a lower limit. The smaller AGN fractions found in the lower (; percent) and higher (; percent) IR luminosity bins could be due to relatively weaker AGN sensitivity limits (i.e., a higher / emission-line ratio) and large uncertainties due to small-number statistics, respectively. Indeed, we find similar AGN fractions (of order 10 percent) in both the lower IR luminosity and moderate IR luminosity bins if we only consider AGNs identified with /, suggesting that further AGNs remain to be detected in the lower IR luminosity; see Fig. 7a. This may indicate that the overall AGN fraction in our IR-bright sample may be closer to 40 percent.
The large AGN fraction found in our study indicates a tighter connection between AGN activity and IR luminosity for galaxies in the local Universe than previously found, exceeding the AGN fraction obtained with optical spectroscopy by up-to an order of magnitude (e.g., compared to the results for galaxies in Veilleux et al. 1999). This may indicate a close association between AGN activity and star formation, as is typically expected given the tight relationship between SMBH and spheroid mass in the local Universe (e.g., Magorrian et al. 1998; Gebhardt et al. 2000). There are probably two reasons why we identify a significantly larger AGN fraction than previously found: (1) mid-IR spectroscopy provides a more sensitive probe of AGN activity than optical spectroscopy, and (2) our galaxies are very nearby, allowing us to identify faint AGN signatures.
3.2 Why are AGN signatures often absent at optical wavelengths?
Of the seventeen galaxies in our Mpc sample that unambiguously host AGN activity, ten ( percent) lack AGN signatures at optical wavelengths. Here we explore the three most likely reasons why the AGN signatures are absent in the optical spectra of these galaxies: (1) the optically unidentified AGNs are intrinsically lower luminosity systems, (2) the optically unidentified AGNs have a larger fraction of star formation/stellar light that dilutes the optical AGN signatures, or (3) the optically unidentified AGNs are more heavily obscured at optical wavelengths.
3.2.1 Are optically unidentified AGNs intrinsically low luminosity?
The non detection of optical AGN signatures in the optically unidentified AGNs may be due to the AGNs being lower luminosity systems. In Fig. 8a, we show the [NeV] luminosities of the mid-IR identified AGNs and a sample of well-studied local AGNs from Panessa et al. (2006) as a function of their [OIII] Å luminosity. We characterise the [NeV]–[OIII] luminosity relationship using a regressional fit.999The fit to the data was obtained using the IDL-based robust bi-sector linefit algorithm ROBUST_LINEFIT.
The tightness in the [NeV]–[OIII] luminosity relationship indicates that the [NeV] luminosity provides a reliable measurement of the intrinsic luminosity of the AGN; the optically identified AGNs lie on this relationship if we correct the [OIII] luminosity for extinction as measured using the Balmer decrement (see §3.2.3).
The optically identified AGNs in our sample cover a broader range of [NeV] luminosities than the optically unidentified AGNs (– and –, respectively). However, since there are optically identified AGNs with similar luminosities to the optically unidentified AGNs, this indicates that the dominant reason for the non detection of the optically unidentified AGNs cannot be due to them hosting intrinsically lower-luminosity AGN activity. Fig. 8a also shows that the non identification of optical AGN signatures in the optically unidentified AGNs is not due to low-sensitivity optical spectroscopy since the [OIII]/[NeV] luminosity ratios are lower than that given in Equation 1.
3.2.2 Are optically unidentified AGNs star-formation dominated?
The non-identification of AGNs at optical wavelengths could be due to dilution from star-formation signatures, which we can test with our data. The mid-IR continua of starburst galaxies are typically characterised by strong PAH features at , 6.2, 7.7, 8.6, 11.3, 12.7, 14.2 and combined with a steep spectral slope at far-IR () wavelengths (Brandl et al., 2006). By contrast, PAH features tend to be weak or absent in AGN-dominated systems (e.g., NGC 1068; Rigopoulou et al. 2002). It is apparent from the Spitzer-IRS spectroscopy presented in Fig. 3 and 12, that most of the optically unidentified AGNs exhibit star-formation signatures at mid-IR wavelengths, indicating that they host joint AGN–starburst activity. However, to test whether the lack of AGN signatures at optical wavelengths is due to dilution from star-formation, we need to compare the relative AGN–star-formation contributions for both the optically identified and unidentified AGNs.
As the sources in the sample are well resolved, we can use the Spitzer-IRS spectra to quantify the relative strengths of the star-formation and AGN activity in the circumnuclear regions of the mid-IR identified AGNs. Previous studies have shown that the equivalent width of the IR-detected PAH features at 6.2 and are well correlated with the AGN-starburst activity occurring within a galaxy (e.g., Genzel et al. 1998; Laurent et al. 2000; Peeters et al. 2004; Dale et al. 2006). However, given the spectral coverage of the Spitzer-IRS spectroscopy for the Mpc sample ( 9.9–37.2 m), here we calibrate and use the equivalent width of the PAH feature (EW, which is detected in every galaxy) to indicate the relative AGN–star-formation contribution of each galaxy. Martín-Hernández et al. (2006) find that for a starburst-dominated circumnuclear region EW (100 percent SF) m. Comparing this to an AGN-dominated system (i.e., NGC 1068) we find EW (100 percent AGN) m. The mid-IR emission line ratios can also be used to constrain the relative contributions of AGN and stellar emission (e.g., Dale et al. 2006). Since [NeII] emission is primarily produced by star-formation activity within a galaxy, and [NeV] is solely attributed to AGN activity, the ratio of these emission lines is also a strong tracer of the relative strengths of these two processes (e.g., Sturm et al. 2002).
In Fig. 7b we present the predicted EW and [NeII] – [NeV] flux ratios for different AGN–star-formation contributions; this is similar in principal to Genzel et al. (1998). Both of the AGN–star-formation ratio estimates are in good agremment for the majority of the mid-IR AGNs, with a mean dispersion of percent. We note that the predicted AGN–star-formation contributions for four ( percent) of the galaxies (NGC 1068, NGC 4051, NGC 5128 and NGC 5643) differ by more than a factor of two from the mixing model, whilst the other mid-IR identified AGNs lie within a factor of 40 percent. One of these such outliers is the FR-1 radio galaxy, NGC 5128 (EWEW), which could be due to dilution of the PAH feature by an underlying synchrotron component (related to the radio emission) that is emitting at mid-IR wavelengths. On the basis of Fig. 7b, the IR emission for five ( percent) of the mid-IR identified AGNs (NGC 1068, NGC 6300, NGC 4051, NGC 5643 and NGC 1448) has a significant contribution from AGN activity ( percent), only one of which is an optically unidentified AGN (NGC 1448). As may be expected, these five galaxies also host the most luminous AGNs, as shown in Fig. 8a. The IR emission for the other twelve ( percent) mid-IR identified AGNs appears to be star-formation dominated (AGN contribution percent), nine of which are optically unidentified AGNs. These analyses have been performed using the SH module of Spitzer-IRS, which traces only the circumnuclear region of these galaxies. However, we get qualitatively similar results if we consider the / ratios, which should provide a measure of the contribution of the AGN to the total IR luminosity of the galaxy (i.e., as measured by IRAS); see Fig. 7a.
On the basis of these analyses we therefore derive qualitatively similar conclusions to those in §3.2.1. Clearly, there is a difference in the distribution of relative AGN–star formation strengths, with the optically unidentified AGNs being typically more star-formation dominated than the optically identified AGNs. However, since approximately half of the optically identified AGNs are also star-formation dominated, dilution from star-formation signatures is unlikely to be the dominant cause for the lack of optical AGN signatures in all of the optically unidentified AGNs. Indeed, as we show in §3.2.3 and §3.3, strong H emission produced by young stars is likely to be the primary reason for the lack of AGN optical signatures (i.e., a low [OIII]–H ratio) in only three of the optically unidentified AGNs.
3.2.3 Are optically unidentified AGNs heavily dust obscured?
The dominant reason for the lack of AGN optical signatures could be due to dust obscuration. A good measure of dust obscuration within a host galaxy is the so-called Balmer decrement (the flux ratio; e.g., Ward et al. 1987). In Table 1 we show that both the optically identified and optically unidentified AGNs cover similar ranges in Balmer decrements, apparently indicating no difference in optical extinction between the two populations. However, many late-type galaxies are found to be very dust/gas rich, and for galaxies such as these, the Balmer decrement may be a poor measure for high-levels of extinction. For example, NGC 4945 is an optically unidentified AGN (optically classified as an HII galaxy), and using analyses at near-IR wavelengths it is estimated to have a V-band extinction of –20 magnitudes (Moorwood et al., 1996); however, the optical Balmer decrement of the galaxy would suggest that it is relatively unobscured ( mags). To remove this ambiguity of additional attenuation to the optical emission, we require a measure of dust-obscuration that is capable of probing greater optical depths than optical spectroscopy alone.
Here we estimate the extinction towards the AGNs using [OIV] at mid-IR wavelengths and [OIII] Å at optical wavelengths.101010We note that it would also be useful to determine the amount of extinction using the [NeV] emission lines at and Å. Unfortunately, this is not currently possible due to the lack of available data at ultra-violet wavelengths for the majority of our sample. Also, this ratio is additionally dependent on temperature variations of the photoionised gas within the narrow line region of the AGN. [OIV] has often been used as a relatively dust-obscuration independent indicator of the AGN luminosity (e.g., Genzel et al. 1998; D06), and we show in §3.4 that it is reliable for [NeV]–identified AGNs; however, see §3.4 for caveats in using [OIV] to directly identify AGN activity. By contrast, the optically detected [OIII] Å emission lines produced in the narrow-line regions of AGNs can be subject to strong reddening by dust/gas, and therefore the [OIV]–[OIII] emission-line ratio should provide a good indicator for the presence of heavy obscuration. A similar approach using the [OIV]–[OIII] ratio has been taken by Diamond-Stanic et al. (2009) for assessing the obscuration in a sample of X-ray luminous AGNs. There could also be some dependence of the [OIV]–[OIII] ratio on the hardness of the radiation field (e.g., as measured using the [NeIII]–[NeII] ratio; Brandl et al. 2006); however, we find no strong dependence in our sample. We calibrate the extinction correction factor in magnitudes by assuming , as found for Seyfert 2 galaxies (e.g., Meléndez et al. 2008), combined with the expected dust-reddening at Å and (e.g., Osterbrock & Ferland 2006). Although the [OIII] emission-line constraints are typically obtained in a narrower slit than the [OIV], since the majority of the [OIII] and [OIV] emission is likely to be produced close to the AGN (i.e., the ionising source), aperture effects are probably not significant.
In Fig. 8b we relate the two different measures of dust-extinction within the host galaxies obtained using the Balmer decrement and our [OIV]–[OIII] flux ratio estimate. Similar levels of optical extinction are derived for the optically identified AGNs on the basis of both the Balmer decrement and the [OIV]–[OIII] flux ratios, implying moderate levels of dust obscuration (–3 mags). However, by contrast, although optically unidentified AGNs have similar Balmer decrements to the optically identified AGNs, many have significantly higher [OIV]–[OIII] flux ratios. This suggests that the optically unidentified AGNs are so heavily extincted at optical wavelengths that the Balmer decrement no longer provides a reliable estimate of the amount of obscuration, suggesting mags. Indeed, on the basis of the [OIV]–[OIII] flux ratio for NGC 4945 we estimate that the AGN is obscured behind a screen of magnitudes, which is in good agreement with the detailed constraints of Moorwood et al. (1996). Additionally, Alexander et al. (1999) find using near-IR spectroscopy magnitudes for NGC 5128, which is again consistent with our extinction estimation derived from the [OIV]–[OIII] flux ratio ( magnitudes). We predict that if the [OIII] emission was adjusted for an additional absorption within the narrow line region (NLR) as found by our diagnostic, these heavily extinguished optically unidentified AGNs would be moved into the Seyfert region of an optical BPT diagram. For example, assuming there is no further reddening to the H emission, an additional NLR extinction of is consistent with an increase by a factor of 6 in [OIII] flux (i.e., the correction required for NGC 5128 to be optically classified as a Seyfert galaxy; see Fig. 5).
In Fig. 8a, we show that three of the optically unidentified AGNs (NGC 0613, NGC 3627 and NGC 5195) are within the intrinsic scatter of the [OIII]-[NeV] relation (shaded region), and consequently do not appear to be deficient in [OIII] flux. Indeed, in Fig. 8b we find that these particular objects do not appear to be heavily extinguished at optical wavelengths ( mags). It is therefore likely that these are not classified as AGNs in optical surveys due to enhanced H flux from young stars, which lowers the [OIII]/H flux ratio in a BPT diagram (see Fig. 5). This is in agreement with the AGN–star-formation ratios estimated for these objects with Spitzer-IRS; see Fig. 6b. However, it is clear from Fig. 8a,b that the majority of the optically unidentified AGNs are highly extinguished in [OIII] Å flux.
3.3 First-order constraints on the source of the optical extinction
The obscuration towards some of the AGNs implied by Fig. 8a,b may be due to either a dusty torus (as predicted by the unified model for AGNs; e.g., Antonucci 1993) or extinction through the host galaxy (e.g., Malkan et al. 1998; Matt 2000). Here we can test the latter hypothesis by looking for evidence of host-galaxy extinction (i.e., a highly inclined galaxy or obscuring dust lanes) using available optical imaging; we defer exploration of the former hypothesis to Goulding et al. (in prep.) where we search for evidence of nuclear obscuration in our sample using sensitive X-ray observations. In Fig. 9 we compare optical images of the optically identified and unidentified AGNs. The optically identified AGNs in our sample are hosted by relatively face-on galaxies while, by contrast, many of the optically unidentified AGNs are hosted in either highly inclined galaxies or have dust lanes that obscure the central regions. Indeed, the four optically unidentified AGNs with the highest [OIV]–[OIII] flux ratios, which imply mags, are all hosted by galaxies that are either highly inclined or have obscuring dust lanes; ESO121-G006 does not have good-quality optical spectroscopy, however, since it resides in a highly inclined host galaxy, it is likely to be an optically unidentified AGN with a high [OIV]–[OIII] flux ratio. Furthermore, three of the optically unidentified AGNs are found to be hosted in relatively face-on galaxies (NGC 0613, NGC 3627 and NGC 5195), similar to that of the optically identified AGNs, suggesting there is little extinction from the host galaxy towards these AGNs. This is consistent with our findings in the previous section that these galaxies do not appear to have heavily extinguished [OIII] emission (see Fig. 8a).
Our findings extend the analyses of Malkan et al. (1998), who found that dust lanes can dictate the observed optical AGN type, by now showing that the host galaxy can have a large effect on even the identification of AGN activity. We find that four ( percent) of the ten optically unidentified AGNs reside in highly inclined galaxies, compared to 4 percent (Seyfert 1) and 7 percent (Seyfert 2) of the optically identified AGNs in Malkan et al. (1998), respectively. Four ( percent) of the optically unidentified AGNs also appear to have obscuring dust lanes, compared to 10 and 20 percent of the Seyfert 1 and Seyfert 2 galaxies in Malkan et al. (1998), respectively.111111We note that high-resolution HST observations are required for the optically unidentified AGNs to provide a consistent comparison to Malkan et al. (1998). Whilst these findings are empirical and may be subject to small number statistics, there appears to be strong evidence to suggest that the non-identification of optical AGN signatures in the majority of these galaxies is due to extinction through the host galaxy, indicating that obscuration of the nucleus is not necessarily due to an obscuring torus. We also note that similar conclusions have been proposed for the non-identification of AGN signatures in 0.5–1 X-ray identified AGNs (e.g., Rigby et al. 2006).
3.4 Can further AGNs be identified using other mid-IR emission-line diagnostics?
In our analyses thus far we have conservatively identified AGNs using the [NeV] emission line. However, there may be further AGNs in our sample that can be identified using different mid-IR emission lines. Other studies have used the high-excitation emission line [NeV] , as well as intermediate-excitation emission lines such as [OIV] (54.9 eV) and [NeIII] (41.0 eV). To explore whether we can identify further AGNs in our sample, we investigate these other potential AGN indicators here.
[NeV] has the same ionisation potential as [NeV] but it has a higher critical electron density ( versus ; Sturm et al. 2002). [NeV] is detected in 11 of the 64 galaxies in our sample, all of which are also detected at [NeV] . The smaller AGN identification fraction found using [NeV] may be due to the lower relative sensitivity of data in the LH module in some cases. We also note that a large number of pixels around were damaged by solar flares early in the Spitzer mission, and are flagged by the Spitzer pre-processing pipeline as either damaged or unreliable (‘hot’ pixels). In our custom reduction pipeline we removed these pixels from the analysis and fitting routines to reduce the potential for spurious detections. For example, from analysis of the BCD images for NGC 3938 before and after cleaning using a custom version of IRSCLEAN, we find that the possible [NeV] emission reported by S08 falls below our detection threshold of . Furthermore, an additional ‘hot’ pixel at (in the observed frame) is not removed by default in IRSCLEAN due to the high incidence of adjacent cleaned pixels. During the reduction process, care must be taken to remove such pixels to ensure reliable detections. Additionally, we do not find any further AGNs in our Mpc sample on the basis of the [NeV] emission line and urge caution when identifying AGNs at low redshifts solely on the detection of weak [NeV] with Spitzer-IRS observations. Certainly, at higher redshifts (i.e., where the observed [NeV] emission is not affected by the spurious pixels at ) detections of weak [NeV] without complimentary [NeV] will be possible for systems with high levels of relative dust extinction; however, this is likely to be rare as (Chiar & Tielens, 2006).
[OIV] is often attributed to AGN activity. However, since it is an intermediate excitation emission line it can also be detected in galaxies experiencing heightened starburst activity (D06). [OIV] is detected in 41 of the 64 galaxies in our sample, including all of the 17 AGNs identified using [NeV]. For the mid-IR identified AGNs, we find that [OIV] and [NeV] emission are well correlated, suggesting that [OIV] is a good proxy for the intrinsic AGN luminosity in [NeV]–identified AGNs; see Fig. 10. The regressional fit is characterised by the equation given below, with a spread in the data of only 0.24 dex (see also Footnote 9). We highlight the caveat that this relation may only hold for the emission-line and IR luminosity range explored here. To rigorously assess the validity of this relation for all galaxies a further study probing a large luminosity range must be undertaken, which is beyond the scope of these analyses presented here.
Our finding of a tight relationship between [OIV] and [NeV] suggests that any AGN activity with bright [NeV] emission should also have bright [OIV] emission. Contrary to this, S08 identified two AGNs (NGC 4321 and NGC 4536) using [NeV] that had undetected [OIV] emission. NGC 4536 is present in our sample but we do not identify [NeV] using our conservative reduction and analysis techniques. Four of the galaxies (IC2056; NGC 3175; NGC 3184; NGC 3368) in our sample with [OIV] detections lie above the regressional fit in Fig. 10 and more sensitive mid-IR spectroscopy may identify AGN activity with [NeV] in these systems. A further 20 galaxies undetected in [NeV] lie below the regressional fit and, therefore, unless AGNs can have weak [NeV] and bright [OIV] emission, these do not appear to host AGN activity. However, we note that the 13 galaxies in our sample that are undetected in both [OIV] and [NeV] may host AGN activity substantially below our detection limits. Furthermore, for galaxies with L, the source of the [OIV] emission becomes uncertain as a non-negligible fraction may be produced by star-formation (e.g., the starburst-AGN, NGC 4945). Therefore, the [OIV] emission may not be a strong tracer of the intrinsic luminosity of the AGN (i.e., the [NeV] luminosity) for galaxies with L. Further probes of the AGN luminosity (e.g., hard X-ray luminosities) are required to test this further and we defer this analysis to Goulding et al. (in prep.).
[NeIII] is also an intermediate excitation emission line and could be produced by AGN or star-formation activity. Previous surveys (e.g., Farrah et al. 2007; Groves et al. 2008; Meléndez et al. 2008) have selected potential AGNs on the basis of high [NeIII]–[NeII] flux ratios ([NeIII]/[NeII]; Meléndez et al. 2008). In Fig. 11 we show that while this diagnostic identifies many of the AGN-dominated systems in our sample, it does not find all of the AGNs. The production of [NeIII] appears to be complex and may have strong dependences on star-formation, gas density, temperature, metallicity and AGN activity (Brandl et al. 2006); for example, there is a similar range of [NeIII]–[NeII] ratios for both the optically unidentified AGNs and the galaxies without clear signatures of AGN activity. Furthermore, the two WR galaxies in our sample also have high [NeIII]–[NeII] flux ratios, suggesting that further criteria are required to unambiguously identify AGN activity. However, three of our galaxies not identified as AGNs (NGC 4490; NGC 4736; NGC 6744) have high [NeIII]–[NeII] flux ratios and may host AGN activity below our [NeV] sensitivity limit.
We have presented the initial results of a sensitive volume-limited Spitzer-IRS spectral survey of all ( percent) bolometrically luminous () galaxies to Mpc. We place direct constraints on the ubiquity of AGN activity in the local universe. Our main findings are the following:
By conservatively assuming that the detection of the high-excitation [NeV]