AKARI NEAR-INFRARED SPECTROSCOPY OF LUMINOUS INFRARED GALAXIES

Since the Infrared Astronomical Satellite (IRAS; Neugebauer et al. 1984) first opened the all-sky view of the far-infrared universe, a large number of luminous infrared galaxies (LIRGs; 10¹¹≦L_IR(8–1000 μm)  <  10¹² L_☉) and ultraluminous infrared galaxies (ULIRGs; L_IR≧10¹² L_☉) have been identified and studied extensively (see Sanders & Mirabel 1996; Lonsdale et al. 2006; Soifer et al. 2008 for review).

In the local universe, many of them are interacting systems between gas-rich disk galaxies (e.g., Kim et al. 2002; Veilleux et al. 2002; Wang et al. 2006; Kaviraj 2009; Hwang et al. 2010a; but see Elbaz et al. 2007; Lotz et al. 2008; Ideue et al. 2009; Kartaltepe et al. 2010, 2011 for high-z (U)LIRGs). They may evolve into quasars and then into intermediate-mass elliptical galaxies (e.g., Sanders et al. 1988b; Kormendy & Sanders 1992; Genzel et al. 2001; Tacconi et al. 2002; Dasyra et al. 2006; Veilleux et al. 2009a; Rothberg & Fischer 2010; Haan et al. 2011). Their enormous infrared luminosity comes from cool dust primarily heated by young stars (i.e., star formation, SF) and hot dust heated by supermassive black holes rapidly accreting matter (i.e., active galactic nuclei, AGNs). Their contribution to the infrared luminosity density increases with redshift (e.g., Le Floc'h et al. 2005; Magnelli et al. 2009; Goto et al. 2011). Therefore, the study of (U)LIRGs allows us to better understand galaxy–galaxy interactions, starburst–AGN connection, and cosmic SF history.

To identify the energy sources of galaxies (i.e., SF versus AGN), the optical line ratios sensitive to the photoionization source have been used (e.g., Baldwin et al. 1981; Veilleux & Osterbrock 1987; Kewley et al. 2006). Based on this method, the optical spectral types for a large number of (U)LIRGs are also determined (e.g., Veilleux et al. 1995, 1999a; Kewley et al. 2001; Goto 2005; Cao et al. 2006; Hou et al. 2009; Lee et al. 2011). However, the optical spectral classification can be uncertain, in particular for (U)LIRGs, due to the difficulty of detecting dust-enshrouded AGNs. In this case, near-infrared (NIR) spectroscopy, less affected by dust extinction, is a more efficient tool.

There are several ground-based K-band (1.9–2.4 μm) spectroscopic surveys searching for obscured AGN signatures in (U)LIRGs: the presence of broad Paα emission centered at (rest-frame) 1.875 μm or of the high-excitation coronal line [Si vi] at 1.963 μm (e.g., Veilleux et al. 1997, 1999b; Murphy et al. 1999, 2001; Dannerbauer et al. 2005). The ground-based L-band (2.8–4.2 μm) spectroscopy was also used to constrain their energy source (e.g., Imanishi & Dudley 2000; Imanishi & Maloney 2003; Imanishi et al. 2006, 2011; Risaliti et al. 2006, 2010; Sani et al. 2008). The SF-dominated (U)LIRGs show a strong emission line at 3.29 μm attributed to polycyclic aromatic hydrocarbons (PAHs), whereas AGN-dominated (U)LIRGs show a relatively PAH-free continuum attributed to larger-sized hot dust grains (e.g., Moorwood 1986; Imanishi & Dudley 2000). The strong absorption features at 3.05 and 3.4 μm by H₂O ice-covered dust grains and bare carbonaceous dust, respectively, are found in (U)LIRGs with buried AGNs. However, these absorptions are weak or absent in normal SF (U)LIRGs where energy sources and dust are often spatially well mixed (e.g., Imanishi & Dudley 2000; Imanishi & Maloney 2003). For a similar reason, NIR continua of AGN-dominated (U)LIRGs can be much redder than those of SF-dominated (U)LIRGs (e.g., Risaliti et al. 2006).

Thanks to the wide wavelength coverage (2.5–5 μm) and high sensitivity of the Infrared Camera (IRC) on board the AKARI space telescope (Murakami et al. 2007; Onaka et al. 2007), the L-band diagnostic could be applied to more distant and faint galaxies (e.g., Imanishi et al. 2008, 2010a). The AGN detection rate from L-band spectroscopy appears to be larger than that from K-band spectroscopy (roughly 70% versus 20% in ULIRGs) because the K-band diagnostic detects only obvious AGNs. These NIR spectroscopic studies of (U)LIRGs suggest that there are many optically elusive buried AGNs and that the buried AGN fraction increases with increasing infrared luminosity.

In this study, we analyze the AKARI NIR spectra of 36 (U)LIRGs⁶, mainly from the cross-correlation between the IRAS and Sloan Digital Sky Survey (SDSS; York et al. 2000). Combining our new NIR spectroscopic data with those in Imanishi et al. (2008, 2010a), we investigate the NIR properties of a large sample of (U)LIRGs. The structure of this paper is as follows. The target selection is explained in Section 2. Observations and data reduction are described in Section 3, and the method of NIR spectral analysis is given in Section 4. Our findings are discussed and summarized in Sections 5 and 6, respectively. Throughout this paper, we adopt the flat ΛCDM cosmological parameters H₀ = 75 km s⁻¹ Mpc⁻¹, Ω_M = 0.3, and Ω_Λ = 0.7.

We obtained the AKARI NIR spectra of (U)LIRGs in three AKARI open-time programs: the nature of new ULIRGs at intermediate redshift (NULIZ), NIR spectroscopy of composite and LINER LIRGs (CLNSL), and NIR spectroscopy of star-forming infrared galaxies (NISIG). The main targets for the NULIZ program were selected from a catalog of ∼320 ULIRGs in Hwang et al. (2007), which was constructed by cross-correlating the IRAS faint source catalog (Moshir et al. 1992) with the galaxy redshift survey catalogs, including the SDSS Data Release 4 (SDSS DR4; Adelman-McCarthy et al. 2006), 2dF Galaxy Redshift Survey (2dFGRS; Colless et al. 2001), and 6dF Galaxy Survey (6dFGS; Jones et al. 2004, 2005). These ULIRGs are detected up to z ∼ 0.4. Because most ULIRGs observed with the L-band spectrograph are at z < 0.2, we focus on 20 newly discovered ULIRGs at 0.2 <z < 0.4 to extend the NIR spectroscopy to intermediate-z ULIRGs. The upper redshift limit ensures that the 3.3 μm PAH emission falls within the AKARI spectral coverage. Additionally, four nearby ULIRGs were selected, as a control sample, from Soifer et al. (1989), Leech et al. (1994), Stanford et al. (2000), and Sanders et al. (2003).

For the CLNSL and NISIG programs, we first identified ∼14,000 infrared galaxies by the cross-correlation of infrared sources in the IRAS faint source catalog with the spectroscopic sample of galaxies in the SDSS DR7 (Abazajian et al. 2009; see Hwang et al. 2010a for more details). We selected ∼13,000 local infrared galaxies at 0.01 <z < 0.2. We then determined their optical spectral types, star-forming, SF–AGN composite (hereafter, composite), low ionization nuclear emission-line region (LINER), and Seyfert 2, using their emission-line fluxes based on the criteria of Kewley et al. (2006; see Lee et al. 2011 for more details). The line flux measurements were drawn from the Max-Planck-Institute for Astrophysics/Johns Hopkins University value-added galaxy catalog⁷ (MPA/JHU VAGCs; Kauffmann et al. 2003b; Tremonti et al. 2004; Brinchmann et al. 2004). We finally selected 60 K_s-bright⁸ non-Seyfert (i.e., SF, composite, and LINER) galaxies, covering a wide range of infrared luminosities (L_IR = 10¹⁰–10¹³ L_☉). Higher priority was assigned to K_s-bright galaxies in order to obtain high signal-to-noise ratio (S/N) AKARI spectra, and to non-Seyfert galaxies in order to find "buried" AGNs. Most targets are at high ecliptic latitudes due to the limited sky visibility of the AKARI satellite (Sun-synchronous orbit; Murakami et al. 2007), but this does not introduce any bias to the results.

Among 48 observed targets, we use only 36 galaxies with median S/N per pixel for the continuum greater than three for the following analysis. Their basic information, including the IRAS flux densities and optical spectral properties, is summarized in Table 1. We plot the infrared luminosity versus redshift of our sample in Figure 1. The (U)LIRGs in Imanishi et al. (2008, 2010a), hereafter the Imanishi sample,⁹ are also plotted for comparison. The two samples are in a similar range of infrared luminosities overall, while our sample is more numerous than the Imanishi sample at 0.2 <z < 0.4. Imanishi et al. also preferentially selected non-Seyfert (U)LIRGs. In the result, non-Seyfert fractions in these samples are large compared to the fractions among their parent samples (from ∼55% to ∼70% in ULIRGs and from ∼65% to ∼80% in LIRGs).

Zoom In Zoom Out Reset image size

We observed the target galaxies with the IRC spectrograph (Onaka et al. 2007) on board the AKARI satellite. It is a slitless spectroscopy, but the targets are located inside a 1 × 1 arcmin² window in order to avoid spectral overlap with nearby sources. The 2.5–5 μm wavelength range was covered at a spectral resolution of R ∼120 (∼2500 km s⁻¹) with NIR grism. In the observing modes, the total on-source exposure time per pointing is about six minutes and the 3σ detection limit is roughly 0.6 mJy. Each target was observed one to eight times (mostly twice), depending on its visibility and on K_s-band magnitude. The observation log is given in Table 2. Note that the NULIZ data were obtained in Phase 2 cooled by liquid helium, while the CLNSL and NISIG data were obtained in Phase 3 cooled by a cryocooler.

Data reduction was carried out using the IDL packages prepared by the AKARI team: "IRC Spectroscopy Toolkit Version 20101025" for the NULIZ data and "IRC Spectroscopy Toolkit for Phase 3 data Version 20101025" for the CLNSL and NISIG data. Both packages are available online.¹⁰ This involves dark subtraction, linearity correction, flat-fielding, sub-frame combining, source detection, background subtraction and source extraction, wavelength calibration, and flux calibration. Because one pointing consists of eight (or nine) sub-frames, cosmic rays are efficiently removed. The wavelength calibration accuracy is ∼0.01 μm (∼1 pixel). The absolute flux calibration accuracy is ∼10% at the central wavelength and can be larger than 20% at the edge (Ohyama et al. 2007).

We derived the spatial profile of each source in the two-dimensional spectra obtained with the slitless spectrograph. To extract one-dimensional spectra, we used 3–9 pixels (3–9 × 146) along the spatial direction.

The aperture width (see Column 2 in Table 3) was determined to contain signals larger than the background noise level, which is the root mean square (rms) of fluxes at each side 11–15 pixel away from the center of the source. Since the aperture is large enough to cover the NIR emission even for extended sources, no aperture correction is applied. For the sources observed several times, the spectra were stacked by taking the median. The data collected during the late part of Phase 3 of AKARI are sometimes seriously affected by bad pixels and background noise because of the increase of the temperature in IRC. If there were some data sets substantially inferior to others, then they were not used to obtain the final spectra.

We display the final spectra of 36 (U)LIRGs with the continuum S/N > 3 in Figure 2. These spectra show various continuum shapes, ranging from steeply decreasing ones to nearly flat ones with increasing wavelength. In most spectra, the PAH emission feature at 3.29 μm is clearly seen. Brα (2.63 μm) and Brα (4.05 μm) hydrogen recombination lines and broad absorption features from H₂O ice (3.05 μm) and bare carbonaceous dust (3.4 μm) are detected in some cases.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 3 shows an example of spectral fitting to the NIR spectrum. We fit each emission line of the Brα, Brβ, and 3.3 μm PAH with a single Gaussian function after subtracting the local continuum defined by a linear fit around the line. Especially for the 3.3 μm PAH profile, we choose an asymmetric fitting range (3.15–3.35 μm) in order to avoid contamination of the 3.4 μm sub-peak (see Imanishi et al. 2010a). We compute the equivalent width of the PAH emission using the fitted Gaussian profile on top of the global continuum (see the next paragraph). By using the global continuum instead of the local continuum, we can reduce the effect of an adjacent absorption feature on the equivalent width measurement. We set an upper limit for the non-detection by assuming that the line has a Gaussian profile with a full width at half-maximum (FWHM) and with an amplitude twice the local rms of the continuum-subtracted spectrum. As for the fixed FWHMs, we use the typical values measured from the AKARI spectra in this study: 2500 km s⁻¹ for the Bracket lines and 4000 km s⁻¹ for the PAH feature. The linewidth of the Bracket lines is well matched to the instrumental resolution. The linewidth of the PAH feature is consistent with the PAH profile in other studies (e.g., type-1 sources in Tokunaga et al. 1991).

Zoom In Zoom Out Reset image size

The global continuum slope Γ is determined using a single power-law model (F_λ ∝ λ^Γ) with data points less affected by line features in the observed wavelengths at 2.7–4.8 μm. The data points near the edge of the spectra are excluded because of their large flux uncertainty (Ohyama et al. 2007). The optical depths of the 3.1 and 3.4 μm dust absorption features (τ_3.1 and τ_3.4, respectively) are calculated as the natural logarithmic ratios between the adopted continuum levels and the observed absorption profiles. The continuum levels are adopted from the global continuum, and the absorption profiles are obtained after smoothing the spectrum with a 10 pixel box (∼0.1 μm) to minimize noise effect. We consider that the absorption feature is detected when the maximum difference between the continuum level and the absorption profile is twice as large as the local rms.

Measurements are listed in Table 3. The uncertainties estimated by considering the local rms values are also listed. The median S/Ns of detected features are 6.4, 5.2, 3.6, 3.3, and 3.1 for 3.3 μm PAH, Brα, Brβ fluxes, τ_3.1, and τ_3.4, respectively.

We present the equivalent width of the 3.3 μm PAH emission line (EW_3.3PAH) versus continuum slope (Γ) diagram in Figure 4. It shows that most (U)LIRGs have EW_3.3PAH < 150 nm and −3 < Γ < 0; the distribution of our sample is not different from that of the Imanishi sample.

Zoom In Zoom Out Reset image size

5.1. AGN Diagnostics

5.1.1. AGN Signature in the NIR Spectra

Emission at 3.3 μm is a prominent feature in (U)LIRG NIR spectra. It probably originates from the reprocessing of ultraviolet (UV) radiation by PAH molecules. The contribution of the Pfδ emission line at a similar wavelength is usually negligible. If (U)LIRGs host AGNs, the EW_3.3PAH is suppressed because hot dust emission from the AGN increases the NIR continuum flux level and X-ray photons may destroy PAH molecules (e.g., Smith et al. 2007). A red NIR continuum (i.e., a high value of continuum slope) as well as strong absorption features at 3.1 and 3.4 μm from H₂O ice-covered dust grains inside molecular clouds and bare carbonaceous dust in the diffuse interstellar medium, respectively (see Draine 2003), indicate the presence of highly obscured compact sources. While many of these absorbed sources with red continuum slopes have been shown to harbor buried AGNs (e.g., Risaliti et al. 2006; Sani et al. 2008; Imanishi et al. 2010a), deeply buried starbursts can also produce the same spectral signatures (e.g., Desai et al. 2007; Spoon et al. 2007; Veilleux et al. 2009b) and in general all that is required is a warm, highly obscured heating source. The small PAH equivalent width condition is useful for identifying weakly obscured AGNs, while the large continuum slope and optical depth conditions are efficient in detecting highly obscured AGNs. Therefore, all of these diagnostics are necessary to detect as many obscured AGNs as possible.

Following the criteria in Imanishi et al. (2010a), we regard EW_3.3PAH < 40 nm, Γ(F_λ∝λ^Γ) > −1, τ_3.1 > 0.3, and τ_3.4 > 0.2 as AGN signatures, and classify sources satisfying at least one of these conditions as NIR AGN-detected galaxies (see Column 10 in Table 3). In the results, we find 19 AGNs out of 36 (U)LIRGs based on these criteria. Among these AGNs, there are 13 sources with EW_3.3PAH < 40 nm (68%), five sources with Γ > −1 (26%), five sources with τ_3.1 > 0.3 (26%), and three sources with τ_3.4 > 0.2 (16%). Note that some sources satisfy more than one criterion. Therefore, EW_3.3PAH < 40 nm is the primary criterion to select AGNs. A similar trend is also seen in the Imanishi sample (56%, 43%, 45%, and 10% for EW_3.3PAH, Γ, τ_3.1, and τ_3.4, respectively).

We count the number of sources with AGN signatures among our sample (U)LIRGs in bins of optical spectral type and of infrared luminosity. These results are summarized in Table 4 together with those from the Imanishi sample. The NIR AGN detection rates for our sample and the Imanishi sample are on average 53% and 51%, respectively, and agree well in each bin. Imanishi et al. (2008, 2010a) found that the AGN signature in NIR spectra is more often detected in optical AGN-like and in more luminous galaxies. Our sample appears to follow these trends, as shown in Figure 5. However, it is not conclusive with our data alone because of large uncertainties, in particular, for the NIR AGN detection rate depending on optical spectral type. In the combined sample of 180 (U)LIRGs with L_IR≧10¹¹ L_☉, the NIR AGN detection rate depends on optical spectral type as follows: 36% for SF (U)LIRGs, 55% for composite (U)LIRGs, and 66% for Seyfert 2 (U)LIRGs. Note that most of previously classified LINERs are called composites in this study because we adopted the selection criteria of Kewley et al. (2006) rather than those of Veilleux & Osterbrock (1987) that were used in Imanishi et al. (2008, 2010a). There are two Seyfert 1 (U)LIRGs in the combined sample, and both of them show AGN signatures in their NIR spectra. The total NIR AGN detection rates for LIRGs and ULIRGs are 29% and 65%, respectively. If we select (U)LIRGs without any priority to non-Seyfert galaxies, then the NIR AGN detection rates for (U)LIRGs increase slightly (e.g., 30% for LIRGs and 70% for ULIRGs based on the (U)LIRG sample in Yuan et al. 2010).

Zoom In Zoom Out Reset image size

Table 4. NIR AGN Detection Rate in Bins of Optical Spectral Type and of Infrared Luminosity

Sample	log LIR	Optical Spectral Type				Seyfert 1	Unclassified
	(L☉)	Star-forming	Composite	LINER	Seyfert 2
This study	12.3–13.0	100% (1/1)	100% (3/3)	⋅⋅⋅  (0/0)	50% (1/2)	100% (1/1)	0% (0/1)
	12.0–12.3	50% (2/4)	60% (3/5)	⋅⋅⋅  (0/0)	⋅⋅⋅  (0/0)	⋅⋅⋅  (0/0)	50% (2/4)
	11.0–12.0	0% (0/3)	50% (3/6)	⋅⋅⋅  (0/0)	100% (1/1)	⋅⋅⋅  (0/0)	⋅⋅⋅  (0/0)
	10.0–11.0	0% (0/3)	⋅⋅⋅  (0/0)	100% (2/2)	⋅⋅⋅  (0/0)	⋅⋅⋅  (0/0)	⋅⋅⋅  (0/0)
Imanishi+08,10	12.3–13.0	100% (2/2)	77% (10/13)	⋅⋅⋅  (0/0)	75% (6/8)	⋅⋅⋅  (0/0)	0% (0/2)
	12.0–12.3	67% (4/6)	61% (19/31)	50% (1/2)	70% (14/20)	⋅⋅⋅  (0/0)	57% (4/7)
	11.0–12.0	18% (3/17)	25% (5/20)	⋅⋅⋅  (0/0)	50% (5/10)	100% (1/1)	20% (2/10)

Notes. The values in parentheses mean the number of sources with AGN signature/total number of sources in each bin from our sample and the sample in Imanishi et al. (2008, 2010a). The optical spectral types of the Imanishi sample are taken from Yuan et al. (2010) so that the same criteria of Kewley et al. (2006) are used for the two samples. For sources with multiple nuclei, if AGN signatures are seen at least in one nucleus, then these sources are considered to be AGNs. For an unbiased comparison, the additional interesting sources in the Imanishi sample are not included.

Download table as:  ASCIITypeset image

When we consider non-Seyfert galaxies with AGN signature in the NIR to be optically elusive buried AGNs, the buried AGN fraction (i.e., the number ratio of buried AGNs to non-Seyferts) increases with infrared luminosity: 24% for (U)LIRGs with L_IR = 10¹¹–10¹² L_☉, 60% for (U)LIRGs with L_IR = 10¹²–10^12.3 L_☉, and 84% for (U)LIRGs with L_IR = 10^12.3–10¹³ L_☉. The higher AGN fraction in more luminous galaxies has been reported in previous studies based on the data from not only AKARI but also the Infrared Space Observatory (ISO; e.g., Lutz et al. 1998; Rigopoulou et al. 1999; Tran et al. 2001) and Spitzer Space Telescope (e.g., Desai et al. 2007; Imanishi 2009; Valiante et al. 2009; Veilleux et al. 2009b; Petric et al. 2011), suggesting an important role of AGNs in increasing infrared luminosity.

As expected, we find many non-Seyfert (U)LIRGs with AGN signatures in their NIR spectra (buried AGNs; 49% (55/113)). On the other hand, a substantial number of Seyfert (U)LIRGs do not show the AGN signatures in the NIR (33% (14/43)). The apertures used in the optical spectroscopy (our sample, 3'' diameter fiber; Imanishi sample, 2'' wide slit and 2 kpc extraction) are smaller than those of NIR spectroscopy (covers the entire galaxy size). The line measurements in the optical spectra for our sample are aperture-corrected following the method in Hopkins et al. (2003), but those for the Imanishi sample are not. The small aperture used for the optical spectral classification may miss the extended emission associated with SF, and hence be more sensitive to weak, central AGNs. This helps to understand the existence of Seyferts without AGN signatures in the NIR. We find no difference in the redshift distribution between Seyferts with and without AGN signatures in the NIR spectra. The different aperture size between the optical and NIR spectroscopy seems partially responsible for the disagreement of spectral types.

Regardless of the aperture effect, if the hot dust emission from AGNs is very weak because of a tiny covering of dust around the AGN, then such AGNs are clearly visible in the optical spectrum. However, they would not be distinguishable from SF-dominated galaxies based on the NIR diagnostics (see Nardini et al. 2010). In contrast, when AGNs are really heavily obscured, both optical and NIR diagnostics are less powerful. Then, the observations at other wavelengths are necessary to detect them (e.g., X-ray: Bauer et al. 2010; Teng & Veilleux 2010; mid/far-infrared: Farrah et al. 2007; Spoon et al. 2007; Veilleux et al. 2009b; Hatziminaoglou et al. 2010; Elbaz et al. 2011; but see also Elbaz et al. 2010; Hwang et al. 2010b; radio: Sajina et al. 2008; Imanishi et al. 2010b).

In Figure 6, we present the optical AGN diagnostic diagram based on [O iii]λ5007/Hβ and [N ii]λ6584/Hα line ratios, and NIR AGN diagnostic diagram based on EW_3.3PAH and continuum slope. In panel (a), the optical AGNs (Seyfert+LINER) without the NIR AGN signature have low [O iii]/Hβ ratios compared to those with AGN signatures both in optical and in NIR spectra. On the other hand, in panel (b), the NIR properties of optical AGNs without the NIR AGN signature are not significantly different from those of non-AGNs both in optical and in NIR spectra.

Zoom In Zoom Out Reset image size

5.1.2. AGN Contribution to the Infrared Luminosity

To measure the contribution of buried AGNs to the infrared energy budget of (U)LIRGs, we use the infrared spectral energy distribution (SED) templates and fitting routine of Mullaney et al. (2011), DECOMPIR.¹¹ These templates consist of one AGN and five host-galaxy SEDs. For the host-galaxy SEDs, Spitzer mid-infrared spectra of starburst galaxies are extrapolated to the far-infrared using IRAS photometry. These host-galaxy SEDs are grouped into five categories, referred to as "SB1" through "SB5," in terms of their overall shape and relative strength of PAH features. For the AGN SED, the intrinsic SEDs of AGN-dominated sources are derived after subtracting suitable host-galaxy components from the observed SEDs, and these SEDs are averaged. The infrared SEDs of AGNs show a large spread, mainly dependent on dust distribution around AGNs (i.e., smooth versus clumpy torus structures). However, the different AGN SEDs do not significantly change the resulting AGN contribution to the infrared luminosity in (U)LIRGs (see Mullaney et al. 2011; Pozzi et al. 2012). Based on this SED fitting with sparse photometric data points such as IRAS, Mullaney et al. (2011) found that the intrinsic AGN luminosities measured are actually correlated with those from high-resolution mid-infrared observations of the AGN cores.

We apply this routine to 69 (U)LIRGs with S/Ns > 3 at all four IRAS bands (12, 25, 60, and 100 μm) in the combined sample, and fit their SEDs five times with AGNs and one of host galaxies by allowing renormalization of these two templates. We choose the best-fit solution with the lowest χ² value, computing the AGN contribution to the infrared (8–1000 μm) luminosity from this template set. Figure 7 represents example SEDs with the best-fit AGN and host-galaxy templates. The AGN contribution in (U)LIRGs ranges from 0% to 69%, and is on average 6%–8% in LIRGs and 11%–19% in ULIRGs. Because the combined sample preferentially includes non-Seyferts, the AGN contribution in this study seems to be small compared to other studies (e.g., LIRGs: ∼10% in Petric et al. 2011; ULIRGs: ∼20% in Farrah et al. 2003; 35%–40% in Veilleux et al. 2009b; ∼25% in Nardini et al. 2010; 15%–20% in Risaliti et al. 2010).

Zoom In Zoom Out Reset image size

Figure 8 shows the correlation of the AGN contribution with the presence of AGN signature in the NIR, optical spectral type, and infrared luminosity. Not surprisingly, the AGN contribution is higher in (U)LIRGs with AGN signature in the NIR than in those without AGN signature. The AGN contribution clearly increases with increasing infrared luminosity (see also Donoso et al. 2012), similar to the trend of buried AGN fraction in Figure 5(b). The AGN contribution for Seyfert (U)LIRGs is slightly larger than those for SF and composite (U)LIRGs.

Zoom In Zoom Out Reset image size

5.1.3. AGN Diagnostics with WISE Data

Recently, the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010) opens up the opportunity to probe mid-infrared properties (3.4, 4.6, 12, and 22 μm) for a large sample of galaxies with excellent sensitivity. We use the WISE all-sky survey source catalog¹² to identify the WISE counterparts of the (U)LIRGs observed with AKARI IRC and (U)LIRGs at 0.01 <z < 0.4 in the SDSS DR7 (Hwang et al. 2010a) within 3''.

In Figure 9(a), we compare the continuum slope Γ with WISE [3.4]–[4.6] color (Vega magnitude system; Jarrett et al. 2011) for the AKARI (U)LIRGs. We overplot the expected WISE [3.4]–[4.6] colors from simple power-law continuum models as a function of continuum slope (solid line). The continuum slope Γ and WISE [3.4]–[4.6] color show a tight correlation (Spearman rank correlation coefficient = 0.82; the probability of obtaining the correlation by chance =1.03 × 10⁻³⁰), with some offset and scatter around the expected relation. Most sources have small values of continuum slope compared to the expectation because the presented slopes (fitting range: 2.7–4.8 μm) are affected by the spectrum at <3.4 μm where the stellar population contribution is large (see Lee et al. 2010). The scatter may come from the contamination by the 3.3 μm PAH feature. For the galaxies at z < 0.13, the presence of 3.3 μm PAH emission makes the [3.4]–[4.6] color bluer than the color without the PAH emission. On the other hand, if the galaxies are at 0.26 <z < 0.55, then the presence of 3.3 μm PAH emission makes the [3.4]–[4.6] color redder than the color without the PAH emission. The amount of change in colors depends on redshift and on the strength of PAH emission. From the experiment with the spectra of our sample, we find that the presence of PAH emission can change the [3.4]–[4.6] color by ±0.2 mag, consistent with the scatter in Figure 9(a).

Zoom In Zoom Out Reset image size

We find that the AGN selection criterion based on the continuum slope (i.e., Γ >−1) in this study is roughly equivalent to that of [3.4]–[4.6] >0.8 suggested by Stern et al. (2012; see also Assef et al. 2010; Jarrett et al. 2011). If (U)LIRGs with AGN signatures in the NIR spectra are regarded as genuine AGNs (filled symbols), then the WISE color criterion selects AGNs with 72% completeness (51 out of 71 genuine AGNs satisfy the WISE color criterion) and 76% reliability (51 out of 67 objects which satisfy the WISE color criterion are genuine AGNs).

Figure 9(b) shows the WISE [3.4]–[4.6] colors versus IRAS flux density ratios between 25 and 60 μm (hereafter IRAS 25–60 μm colors) for the SDSS (U)LIRG sample. The IRAS 25–60 μm color is known to be associated with nuclei activity in infrared-luminous galaxies (e.g., de Grijp et al. 1985; Sanders et al. 1988a; Neff & Hutchings 1992; Veilleux et al. 2009b; Lee et al. 2011). AGN-dominated galaxies show warm IRAS 25–60 μm colors (f₂₅/f₆₀≧0.2), while SF-dominated galaxies show cool colors (f₂₅/f₆₀ < 0.2). The composite and SF galaxies show similar distributions in this domain, but the AGNs are significantly different. The median colors of AGNs differ from those of non-AGNs with significance levels of 6.8σ and 7.4σ in the IRAS and WISE, respectively. Interestingly, there are a substantial number of SF (U)LIRGs in the lower-right corner. It seems real even if we consider large uncertainties associated with the IRAS colors. They have warm dust emission without hot dust emission, therefore appearing to be heavily obscured AGNs. As a result, the WISE [3.4]–[4.6] color is a good tracer of AGN-heated hot dust emission, but may not be sufficient to detect heavily obscured AGNs.

5.2. Comparison between Optical and Infrared Properties

5.2.1. Star Formation Rate Indicators

The total infrared continuum, 3.3 μm PAH emission, and recombination lines including Hα and Brα of galaxies are useful indicators of star formation rate (SFR; Kennicutt 1998). In Figures 10(a) and (b), we compare these SFR indicators: infrared luminosity versus (1) Hα and (2) 3.3 μm PAH luminosities. The Hα luminosities are extinction-corrected using the Balmer decrement and Calzetti et al. (2000) extinction curve. The expected relationships are overplotted between these parameters (dotted lines; hereafter SF galaxy sequences): L_Hα/L_IR = 10^−2.25 from the empirical relation in Kennicutt (1998) and L_3.3PAH/L_IR = 10⁻³ from the observations of Mouri et al. (1990) and Imanishi (2002). These panels show that there are large offsets and scatters between the data and the expected relationships, even for SF (U)LIRGs without any AGN signature (pure SF (U)LIRGs; large filled symbols). The offset from the SF galaxy sequence is larger in ULIRGs than in LIRGs, suggesting that Hα and 3.3 μm PAH emissions are more depressed in ULIRGs.

Zoom In Zoom Out Reset image size

There could be several reasons for the strong depression of Hα and 3.3 μm PAH emission in ULIRGs. (1) The amount of dust extinction could be systematically underestimated in ULIRGs. To check this effect, we plot the observed line ratios of Brα/Hα and Hα/Hβ in panel (c). The pure SF (U)LIRGs appear to have larger Brα/Hα ratios by considering the extinction curve of Calzetti et al. (2000) with a typical R_V value of 4.05. This can imply the need of a larger R_V value for the extinction correction in (U)LIRGs (e.g., R_V = 5.01 denoted by the arrow). Some studies also suggest the need of flatter/grayer extinction curves (i.e., large R_V values) in ULIRGs, which may be attributed to supernovae-driven large-size dust grains (e.g., Kawara et al. 2011; Shimizu et al. 2011; see also Boquien et al. 2012). However, because of large scatter in the data and different aperture size between the optical and NIR spectroscopy, it should be checked with a more extensive data set in future studies. (2) Even for ULIRGs without any AGN signature in optical and NIR observations, there could still be hidden AGNs that play a role in the relative depression of line emission. To remove contamination of hidden AGNs, it is necessary to search for AGNs at other wavelengths. However, it is expected that the contribution of buried AGNs is not significant as discussed in Section 5.1.2. (3) The line emission from ULIRGs may be intrinsically weak. Regardless of the presence of AGNs, the SF in ULIRGs produces large infrared continuum emission because a larger fraction of stellar UV photons is absorbed by dust inside star-forming regions under the strong radiation field in ULIRGs (Abel et al. 2009). The PAH emission in ULIRGs may be depressed in the sense that intense radiation fields do not produce photodissociation regions that are necessary for PAH emission, or lead to the destruction of PAH carriers (see Voit 1992; Smith et al. 2007; Veilleux et al. 2009b). It is still unclear whether the ionization state of grains is actually related to infrared luminosity (e.g., Desai et al. 2007; Smith et al. 2007; Veilleux et al. 2009b; Imanishi et al. 2010a; Petric et al. 2011).

5.2.2. Dust Extinction as a Function of Infrared Color

In Figure 11, we present the Brα/Hβ line ratio of (U)LIRGs as a function of IRAS 25–60 μm color. There is an anti-correlation between the two parameters with Spearman rank correlation coefficient = −0.67 and the probability of obtaining the correlation by chance =0.02. This supports that warmer sources are less extinguished than cooler sources (see Veilleux et al. 1999a, 2009b; Keel et al. 2005; Imanishi et al. 2008). When using Balmer decrement values instead, such a correlation becomes weaker since the optical lines do not trace well the buried sources.

Zoom In Zoom Out Reset image size

We obtained AKARI 2.5–5 μm spectra of 36 (U)LIRGs, selected mainly from the IRAS-detected galaxies in the SDSS. We measured the NIR spectral features including continuum slope, 3.3 μm PAH strength, optical depths at 3.1 and 3.4 μm, and Bracket lines in order to find AGN signatures. Together with samples in the literature and ancillary data, we compared optical and infrared properties of (U)LIRGs. Our primary results are summarized below.

• 1.  
  We found that 52% (14/27) of optical non-Seyfert galaxies in our (U)LIRG sample show the AGN signature in their NIR spectra. The NIR AGN detection rate for the combined sample is higher in composite (U)LIRGs than in SF (U)LIRGs, and increases with infrared luminosity.
• 2.  
  We fitted the IRAS photometric data for 69 (U)LIRGs with the AGN/starburst SED templates to compute the AGN contribution to the infrared luminosity. We found that the contribution of buried AGNs to the infrared luminosity is 5%–10%, smaller than the typical AGN contribution in (U)LIRGs including Seyfert galaxies (10%–40%).
• 3.  
  The NIR continuum slopes correlate well with WISE [3.4]–[4.6] colors. Using the WISE [3.4]–[4.6] color versus the IRAS 25–60 μm color domain, we found sources associated with warm dust emission, but without hot dust emission. The WISE color is useful for identifying a large number of AGNs, while it can miss heavily obscured AGNs.

We are grateful to an anonymous referee for comments that helped to improve the manuscript. This work was supported by Mid-career Research Program through NRF grant funded by the MEST (No.2010-0013875). J.C.L., M.K., and J.H.L. are members of the Dedicated Researchers for Extragalactic AstronoMy (DREAM) in the Korea Astronomy and Space Science Institute (KASI). H.S.H. acknowledges the Centre National d'Etudes Spatiales (CNES) and the Smithsonian Institution for the support of his post-doctoral fellowship. This research is based on observations with AKARI, a JAXA project with the participation of ESA. Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory, and the University of Washington. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory, California Institute of Technology, funded by the National Aeronautics and Space Administration.