AKARI IRC INFRARED 2.5–5 μm SPECTROSCOPY OF A LARGE SAMPLE OF LUMINOUS INFRARED GALAXIES

Based on the recently estimated dust extinction curve (Nishiyama et al. 2008, 2009), the amount of dust extinction at 3.3 μm is as small as ∼1/30 of that in the optical V band (λ = 0.55 μm). This indicates that the flux attenuation of the 3.3 μm PAH emission is not significant (a factor of <2.5) if the dust extinction A_V is less than 30 mag. For this reason, the observed 3.3 μm PAH emission luminosities roughly map out the intrinsic luminosities of modestly obscured (A_V < 30 mag) PAH-emitting normal starburst activity.

The relationship L_3.3 PAH/L_IR ∼ 10⁻³ has been used for this purpose (Mouri et al. 1990; Imanishi 2002). Mouri et al. (1990) observed a limited number of starburst galaxies, and in most of these, not all of the starburst regions were covered by their apertures. The proportion L_3.3 PAH/L_IR ∼ 10⁻³ was derived after aperture correction. The aperture correction was based on the ratio of the infrared aperture size to the optical extent of each galaxy, which could entail an uncertainty. Imanishi (2002) supported the proportion L_3.3 PAH/L_IR∼ 10⁻³ for starbursts on the basis of the rough agreement of 3.3 μm PAH-derived starburst luminosities (without a dust extinction correction) and UV-derived starburst luminosities (with a dust extinction correction) in three Seyfert galaxies. For the prototypical starburst galaxy M82, Satyapal et al. (1995) argued that the proportion L_3.3 PAH/L_IR ∼ 0.8 × 10⁻³ holds, once the effects of dust extinction have been corrected. We now have AKARI IRC slit-less spectra for a large number of LIRGs, many of which may be starburst-dominated, without strong AGN signatures in the infrared range. Because the 3.3 μm PAH emission from spatially extended starburst regions of LIRGs should be covered in the AKARI IRC slit-less spectra, we can use observational data to directly test the validity of the proportion L_3.3 PAH/L_IR ∼ 10⁻³ for estimating intrinsic starburst luminosity, without introducing a potentially uncertain aperture correction.

Figure 3 compares the 3.3 μm PAH emission luminosities measured with ground-based slit spectroscopy and AKARI IRC slit-less spectroscopy. For ULIRGs (Figure 3(a)), the two sets of measurements roughly agree, with some scatter (partly originating from measurement errors for faint sources), as expected from the physically compact nature of the infrared emission of ULIRGs in general (<300 pc; Soifer et al. 2000), and their relatively large distances. The largest discrepancy among the 3.3 μm PAH luminosity measurements is found in Mrk 231 (z = 0.042), one of the nearest ULIRGs in the sample, as indicated in the figure. Mrk 231 exhibits spatially extended starburst activity (Krabbe et al. 1997), the bulk of which can be missed with ground-based narrow-slit (<1''–2''wide) spectroscopy, because of the large apparent extension resulting from its proximity. However, except for the nearest ULIRG, Figure 3(a) suggests that the bulk of the 3.3 μm PAH emission from ULIRGs can be recovered effectively from ground-based narrow-slit spectra.

Zoom In Zoom Out Reset image size

Figure 3(b) compares the 3.3 μm PAH luminosity measurements obtained from ground-based narrow-slit spectroscopy and AKARI IRC slit-less spectroscopy for LIRGs. The AKARI IRC measurements are systematically larger, which is reasonable because the infrared emission from LIRGs is, in most cases, spatially extended (up to ∼ kpc; Soifer et al. 2001), and LIRGs are generally closer than ULIRGs, so that their apparent spatial extensions are also larger.

Figure 4 compares the rest-frame equivalent widths of the 3.3 μm PAH emission features (EW_3.3 PAH) measured by ground-based slit spectroscopy and AKARI IRC slit-less spectroscopy. If the 3.3 μm PAH emission properties do not exhibit substantial spatial variation, the EW_3.3 PAH values are less susceptible to aperture effects. For ULIRGs, the two sets of measurements roughly agree, with some scatter, part of which may originate from measurement errors for faint sources (Figure 4(a)). The LIRGs plotted in Figure 4(b) are generally bright, and so we do not expect large measurement errors. In Figure 4(b), many LIRGs display smaller EW_3.3 PAH values in the AKARI IRC data. A possible explanation is that ground-based slit spectroscopy preferentially probes strongly PAH-emitting, very active nuclear starburst regions, while stellar photospheric continuum emission from inactive old stars (which lack PAH-exciting UV photons) in the outer regions of galaxies may contribute to the AKARI IRC slit-less spectra and reduces the EW_3.3 PAH values.

Zoom In Zoom Out Reset image size

In Figure 5(a), the observed ratios of 3.3 μm PAH to infrared luminosity (L_3.3 PAH/L_IR) for starburst-classified LIRGs (i.e., no AGN signatures in the AKARI IRC 2.5–5 μm spectra; see next subsection) are represented by open circles. The upper bound of the L_3.3 PAH/L_IR distribution for these starburst-classified LIRGs is indeed ∼10⁻³. We have now obtained direct evidence from AKARI IRC slit-less spectroscopic observations that the relationship L_3.3 PAH/L_IR = 10⁻³ is appropriate for quantitatively estimating the intrinsic power of modestly obscured starburst activity.

Zoom In Zoom Out Reset image size

The L_3.3 PAH/L_IR ratios of LIRGs and ULIRGs are summarized in Column 4 of Tables 5 and 6, respectively. Figure 5(a) shows a plot of the L_3.3 PAH/L_IR ratios of the observed LIRGs and ULIRGs. As the figure indicates, the majority of starburst-classified LIRGs are distributed between the 50% and 100% lines, suggesting that the bulk of the infrared luminosities could be explained by PAH-probed modestly obscured starburst activity. On the other hand, only a very small fraction of the ULIRGs are above the 50% line. At face value, PAH-probed modestly obscured starbursts alone are insufficient to account for the large infrared luminosities of most ULIRGs. We note that the quality of the AKARI IRC 2.5–5 μm spectra of most of the LIRGs is high, because their relative proximity permits high continuum flux and high S/N to be achieved. For these objects, the 3.4 μm PAH sub-peaks are separated from the 3.3 μm PAH main peaks, and the contributions from the 3.4 μm PAH sub-peaks are removed from the final L_3.3 PAH values. ULIRGs are generally fainter than LIRGs at infrared 2.5–5 μm, simply because of the larger overall distances, and hence it is sometimes difficult to separate the 3.3 μm PAH peak and the 3.4 μm PAH sub-peak. Thus, the 3.4 μm PAH sub-peaks may contribute to the derived L_3.3 PAH values for some fraction of ULIRGs. This possible ambiguity may even reduce the true L_3.3 PAH/L_IR ratios for ULIRGs. The implied trend of the 3.3 μm PAH luminosity suppression in ULIRGs is therefore robust.

There are several plausible scenarios for the generally small observed L_3.3 PAH/L_IR ratios of ULIRGs. (1) The 3.3 μm PAH emission is more highly flux-attenuated in ULIRGs, because of the larger dust extinction, whereas the 8–1000 μm infrared fluxes are not significantly reduced. (2) Powerful AGN activity produces large infrared luminosities, but virtually no PAH emission in ULIRGs. (3) PAH emission from ULIRG starbursts is intrinsically weak. A normal starburst galaxy is observationally considered to be a composite of spatially well-mixed H ii regions, molecular gas, and photodissociation regions (Puxley 1991; McLeod et al. 1993; Forster Schreiber et al. 2001). If the starburst magnitudes per unit volume are larger in ULIRGs, more intense radiation fields could lead to greater destruction of 3.3 μm PAH carriers than in LIRG starbursts (Smith et al. 2007). Alternatively, under the effect of stronger radiation fields, a larger fraction of stellar UV photons could be absorbed by dust inside H ii regions (Abel et al. 2009), producing stronger infrared dust continuum emission, instead of creating photodissociation regions, which are the main source of the 3.3 μm PAH emission (Sellgren 1981). A higher fraction of UV photons could also be absorbed by dust, if the H ii regions are dustier in ULIRGs than in LIRGs (Luhman et al. 2003). In the third scenario, the L_3.3 PAH/L_IR ratios could decrease, even if the total amount of dust extinction is similar.

Regarding scenario (3), if the PAH emission from starbursts in ULIRGs is intrinsically more suppressed than in LIRGs, then even starburst-dominated ULIRGs should exhibit weak 3.3 μm PAH emission, and the upper envelope of the EW_3.3 PAH distribution should decrease in ULIRGs, compared with LIRGs. Figure 5(b) shows a plot of the EW_3.3 PAH distribution as a function of L_IR. There is no clear trend that the upper envelope systematically decreases for higher L_IR, and so at least for ULIRGs with minimum AGN contributions, the 3.3 μm PAH emission is as strong as in LIRG starbursts. Thus, scenario (3) is not strongly supported by our AKARI IRC observational data.

In scenario (1), the EW_3.3 PAH distribution should be relatively insensitive to dust extinction. This is because the 3.3 μm PAH and continuum emission are comparably flux-attenuated in a pure normal starburst galaxy with spatially well-mixed H ii regions/molecular gas/photodissociation regions. Specifically, the dust extinction of starbursts only decreases the L_3.3 PAH values, while the EW_3.3 PAH values remain relatively unchanged (Imanishi et al. 2006a, 2008). In scenario (2), the EW_3.3 PAH values diminish when the contributions from AGNs to the observed 2.5–5 μm fluxes become important. In Figure 5(b), many ULIRGs show small EW_3.3 PAH values, suggesting that scenario (2) occurs in at least a fraction of the observed ULIRGs.

We note that stellar photospheric continuum emission could also affect the EW_3.3 PAH values even for galaxies dominated by normal starbursts, if its contribution to the AKARI IRC 2.5–5 μm spectra is important and varies with different galaxies. The stellar photospheric emission shows a decreasing continuum flux with increasing wavelength at >1.8 μm (Sawicki 2002). The 3.5 μm to 2.5 μm flux ratio in F_ν is ∼ 0.75 (Γ = −0.85; F_ν∝ λ^Γ), which is insensitive to stellar ages, as long as they are >10 Myr (Bruzual & Charlot 1993; see also Figure 1 of Sawicki 2002). The stellar photospheric emission is bluer than the starburst-heated dust continuum emission (Γ ∼ 0; see Section 5.4), and so its contribution becomes relatively less important at longer wavelengths. Sources with the blue observed continuum emission at 2.5–5 μm are candidates largely contaminated by the stellar photospheric emission. Assuming that the stellar photospheric and starburst-heated hot dust continuum emission have Γ = −0.85 and Γ = 0, respectively, the stellar photospheric contribution to the 3.5 μm flux is estimated to be <30%, if the observed 3.5 μm to 2.5 μm flux ratio in F_ν is >0.9. Almost all sources meet this criterion (Figures 1 and 2), suggesting that the stellar photospheric contributions are not so important to largely decrease the EW_3.3 PAH values in the majority of the observed LIRGs and ULIRGs. Furthermore, given that the observed continuum emission of ULIRGs is generally redder than that of LIRGs (Section 5.4), this possible stellar photospheric contribution cannot explain the smaller EW_3.3 PAH trend in ULIRGs than LIRGs at all.

To find sources that may possess luminous AGNs on the basis of depressed EW_3.3 PAH values, we follow Imanishi et al. (2008). Sources with EW_3.3 PAH ≲ 40 nm are classified as galaxies that contain luminous AGNs, and the AGN contributions to the observed 2.5–5 μm fluxes are significant. Tables 5 and 6 (Column 6) indicate whether or not AGN signatures are found.

Based on this low EW_3.3 PAH method, 15 of 62 LIRG nuclei and 15 of 48 ULIRG nuclei display AGN signatures, after excluding additional interesting sources. Among the 15 LIRGs with detectable AGN signatures, 6 nuclei (MCG−03-34-064, NGC 5135, NGC 7469, IC 5298, NGC 7592W, and NGC 7674) are optically classified as Seyferts. For the ULIRGs, 9 of the 15 AGN-detected sources are optical Seyferts. After removing these optically identified AGNs surrounded by torus-shaped dusty media, nine LIRG nuclei and six ULIRG nuclei are possible containers for optically elusive luminous buried AGNs.

The low EW_3.3 PAH method provides clear AGN signatures if the observed 2.5–5 μm flux primarily originates in AGN-heated PAH-free hot dust emission. This can happen if (1) the surrounding starbursts are weak and/or (2) the AGN emission is not heavily obscured. However, luminous AGN signatures could be overlooked by this method if the AGN is dust-obscured (i.e., the AGN emission is flux-attenuated) to the point that the AGN contribution to the observed 2.5–5 μm flux is insignificant compared with weakly obscured starbursts in the foreground. Such a highly obscured AGN may be distinguished by measuring the optical depths of dust absorption features. While stellar energy sources and dust are spatially well mixed in a normal starburst (see Section 1), in an obscured AGN, the energy source (i.e., a mass-accreting compact SMBH) is more centrally concentrated than dust. In the former mixed dust/source geometry, there are upper limits for the optical depths of the dust absorption features found at 2.5–5 μm (Imanishi & Maloney 2003; Imanishi et al. 2006a), because the observed flux primarily originates from weakly obscured, less flux-attenuated foreground emission (which has weak dust absorption features), with a small contribution from highly obscured, highly flux-attenuated emission from the obscured regions (which should exhibit strong dust absorption features). On the other hand, in a centrally concentrated energy source geometry, the optical depths of the dust absorption features can be larger than the upper limits of the mixed dust/source geometry, because dust extinction and flux attenuation can be approximated by a foreground screen dust model. It has been determined that the optical depths of 3.1 μm ice-covered dust absorption features with τ_3.1 > 0.3 and 3.4 μm bare (ice-mantle-free) carbonaceous dust absorption features with τ_3.4 > 0.2 cannot be reproduced in the mixed dust/source geometry, if the Galactic elemental abundance patterns and dust compositions are assumed (Imanishi & Maloney 2003; Imanishi et al. 2006a).

In Table 9 (Column 6), sources with τ_3.1 > 0.3 and/or τ_3.4 > 0.2 are indicated as possible hosts for obscured AGNs with centrally concentrated energy source geometry. Three LIRGs and 16 ULIRGs meet the criteria, of which one LIRG (IC 5298) and two ULIRGs (IRAS 12072−0444 and 16156+0146) are optically classified as Seyfert 2s. Thus, the remaining two LIRGs and 14 ULIRGs are buried AGN candidates. The number and fraction of sources with large τ_3.1 and/or τ_3.4 is substantially larger for ULIRGs than for LIRGs, suggesting that the 2.5–5 μm ULIRG emission generally comes from more obscured regions.

The continuum slope Γ (F_ν ∝ λ^Γ) can also be used to find candidates for obscured AGNs with coexisting strong starbursts in the foreground (Risaliti et al. 2006, 2010; Sani et al. 2008). The logic is basically the same. The 2.5–5 μm continuum emission from a normal starburst with mixed dust/source geometry cannot be so red, while that from an obscured AGN can become very red (Risaliti et al. 2006). Figure 6(a) shows the distribution of Γ for the LIRGs and ULIRGs observed in this paper. In Figure 6(b), the same distribution is shown with the ULIRGs studied by Imanishi et al. (2008) included. A larger value of Γ indicates a redder, rising continuum with increasing wavelength. In both plots, a larger fraction of the ULIRGs are distributed in the red, large Γ tails compared to the LIRGs, again suggesting that the infrared 2.5–5 μm emission from ULIRGs is generally more obscured than that of LIRGs. After excluding the very red outliers, most of the LIRGs and ULIRGs have Γ = −1 to 1, with a median value of Γ ∼ 0 (Tables 5, 6, and 10). This Γ value may be regarded as a typical value for a starburst.

Zoom In Zoom Out Reset image size

If we interpret values of Γ > 1 as obscured AGN signatures, then only six LIRGs correspond to this case, of which two sources (MCG−03-34-064 and NGC 7674) are optical Seyferts. 11/34 (32%) of optically non-Seyfert ULIRGs and 9/14 (64%) of optically Seyfert ULIRGs in the IRAS 1 Jy sample have large Γ values (>1), indicative of obscured AGNs. Among the additional interesting sources, red continuum emission is seen in four sources (IRAS 00183−7111, 06035−7102, 20100−4156, and 20551−4250). Tables 5 and 6 (Column 8) summarize the detection or non-detection of AGN signatures, based on Γ values.

In previous research (Risaliti et al. 2006, 2010; Sani et al. 2008), Γ is defined by F_λ ∝ λ^Γ and Γ_prev = −0.5 is assumed to be a typical value for a weakly obscured starburst/AGN. Because Γ_ours, as defined in our paper (F_ν ∝λ^Γ), is related to Γ_prev by Γ_ours = Γ_prev + 2, the typical Γ_ours values of 0 (−1 to 1) correspond to Γ_prev = −2 (−3 to −1), which is smaller (bluer) than the previously assumed values for weakly obscured starbursts/AGNs.

Figure 6(c) compares the Γ and EW_3.3 PAH values. Red continuum sources with large Γ values are preferentially distributed in the low EW_3.3 PAH regions, usually occupied by AGN-important galaxies, demonstrating that large Γ values can be a good criterion for finding obscured AGNs. However, there are many sources that have low EW_3.3 PAH values, and yet exhibit blue continuum emission (small Γ). Obviously, weakly obscured AGNs cannot be detected by large Γ values.

To relate Γ value to obscuration, we estimate the dust extinction toward the observed 2.5–5 μm continuum emission regions, based on the observed optical depths of the 3.1 μm ice-covered dust (τ_3.1) and 3.4 μm bare carbonaceous dust absorption features (τ_3.4). The τ_3.1 value reflects the column density of dust grains covered with an ice mantle deep inside molecular clouds and is roughly proportional to the dust extinction (τ_3.1/A_V = 0.06; Tanaka et al. 1990; Smith et al. 1993; Murakawa et al. 2000). The τ_3.4 value can be used to estimate the column density of bare, ice-mantle-free dust grains in the diffuse interstellar medium (outside molecular clouds) and is correlated with the dust extinction (τ_3.4/A_V = 0.004–0.007; Pendleton et al. 1994; Imanishi et al. 1996; Rawlings et al. 2003). Because the 3.4 μm dust absorption feature is absent from ice-covered dust grains (Mennella et al. 2001), we can estimate the total column densities of dust, including both ice-covered and ice-mantle-free, from the combination of τ_3.1 and τ_3.4 values. Assuming that A_V = τ_3.1/0.06 + τ_3.4/0.006, the continuum slope Γ and dust extinction toward the 2.5–5 μm continuum emission regions (A_V) are compared in Figure 6(d). For sources with undetectable τ_3.1 and τ_3.4, we tentatively adopt A_V = 0 mag. It can be seen that all the sources with A_V > 40 mag show red continua (Γ > 1), supporting the hypothesis that large Γ values are caused by dust obscuration.

Here, we summarize the fraction of sources for which AGN signatures are observed using at least one of the methods previously discussed (low EW_3.3 PAH, large τ_3.1 and/or τ_3.4, and large Γ), after excluding the additional interesting sources. The low EW_3.3 PAH method is relatively sensitive to weakly obscured AGNs, while both the large τ and large Γ techniques can selectively detect highly obscured AGNs. Thus, these methods complement each other for the purpose of finding AGNs.

AGN signatures are found in 17/62 (27%) LIRG nuclei. The AGN detection rate is 6/8 (75%) for optically Seyfert LIRG nuclei and 11/54 (20%) for optically non-Seyfert LIRG nuclei.

For ULIRGs, the fraction of AGN-detected sources is 29/48 (60%) for all nuclei, 10/14 (71%) for nuclei classified as optically Seyfert, and 19/34 (56%) for nuclei classified as optically non-Seyfert. The detection rate for buried AGNs in galaxies classified as optically non-Seyfert is higher in ULIRGs (19/34; 56%) than in LIRGs (11/54; 20%). For optically identified AGNs (optical Seyferts) of LIRGs and ULIRGs combined, our AKARI IRC 2.5–5 μm spectroscopic methods recover AGN signatures in 16/22 (73%). A small fraction of optically identified AGNs does not exhibit clear AGN signatures according to our methods (see also Nardini et al. 2010), possibly because of a relatively small covering of dust around a central AGN. For such an AGN, the NLRs develop so well that optical AGN detection becomes easy, while AGN-heated hot dust emission (on which our infrared AGN diagnostic is based) is not so strong.

For the additional interesting sources, AGN signatures are detected in 7/8 (88%) by at least one method. Because these sources are selected for their strong dust absorption features and/or AGN signatures at other wavelengths, the high AGN detection rate is exactly what we would expect.

In many LIRGs and a small fraction of bright ULIRGs with high S/N, strong Brα (λ_rest = 4.05 μm) emission lines are clearly seen, particularly for PAH-strong, starburst-important galaxies. The Brβ (λ_rest = 2.63 μm) emission lines are often recognizable in sources with very strong Brα emission with larger equivalent widths. Although an AGN can also produce Brα and Brβ emission, their equivalent widths are expected to be low, because of dilution by strong, AGN-heated hot dust continuum emission. Thus, Brα and Brβ emission from strong Br emission sources is assumed to originate mostly from starburst activity. Both lines are relatively strong, and their dust extinction effects are much smaller than the widely used hydrogen recombination lines, such as Lyα (λ_rest = 0.12 μm) in the UV range, or Hα (λ_rest = 0.66 μm) and Hβ (λ_rest = 0.49 μm) in the optical range. Thus, the Brα and Brβ emission lines could provide a good tracer for probing the properties of starbursts in the dusty LIRG and ULIRG populations, by substantially reducing the uncertainties of dust extinction corrections.

Brα (λ_rest = 4.05 μm) is observable from the ground in the longest wavelength component of the L-band (2.8–4.2 μm) Earth's atmospheric window for only the nearest sources with small redshifts. However, for such nearby sources, Brβ (λ_rest = 2.63 μm) emission is not covered by the ground-based L band. Even if the Brα line is covered, it is difficult to measure with ground-based spectroscopy because of the considerable amount of Earth's atmospheric background noise in the 4.05–4.2 μm range. If a galaxy is redshifted, the Brβ emission line enters the L-band atmospheric window. However, for such redshifted sources, the Brα emission is shifted beyond the L-band's longest wavelength range. Consequently, in ground-based spectroscopy, it is impossible to observe both the Brα and Brβ emission lines. The simultaneous coverage of 2.5–5 μm provided by AKARI IRC is quite unique, in that it enables the comparison of the Brα and Brβ emission lines.

For sources in which both the Brα and Brβ emission lines are detected, the observed Brβ/Brα luminosity ratios are compared with the value predicted by case B theory (∼0.64 for 10⁴ K; Wynn-Williams 1984) and are found to agree to within a factor of 2 (Tables 7 and 8). The lowest and highest observed values deviate from the theoretical value by a factor of ∼2 (VV 114E and IC 694) and ∼1.5 (NGC 232, NGC 6285, and Mrk 273), respectively. Given the possible presence of flux measurement uncertainties (due to the faintness of Brβ), we see no clear evidence that the Brβ/Brα luminosity ratio deviates from the case B prediction. Even if the factor of ∼ 2 for the lower value is due to dust extinction, assuming the dust extinction curve A_λ∝ λ^−1.75 (Cardelli et al. 1989), the estimated dust extinction is A_V < 16 mag. Thus, the strong Br emission lines are most likely to originate from modestly obscured (A_V < 30 mag) starbursts, as probed by the 3.3 μm PAH emission features (Section 5.1).

For this reason, the Brα emission luminosity could also furnish an independent probe for modestly obscured starburst magnitudes. In starburst galaxies, the luminosity ratio of the optical Hα and far-infrared (40–500 μm) emission is estimated to be 0.57 × 10⁻² (Kennicutt 1998). Given that the far-infrared luminosity is only slightly smaller than the infrared (8–1000 μm) luminosity in high-luminosity starbursts (Sanders & Mirabel 1996), and assuming a Brα/Hα luminosity ratio of ∼0.02, as predicted by case B (10⁴ K; Wynn-Williams 1984), we obtain a Brα to infrared luminosity ratio of L_Brα/L_IR ∼ 1 × 10⁻⁴ for a starburst.

Figure 7 compares the L_Brα/L_IR and L_3.3 PAH/L_IR ratios for Brα-detected LIRGs and ULIRGs. The source distribution lies along the dotted line, with some scatter, suggesting that both the Brα and 3.3 μm PAH emission provide similar starburst contributions to the infrared luminosities of these LIRGs and ULIRGs. Hence, both the Brα and 3.3 μm PAH emission can be used as independent measures of the magnitudes of modestly obscured starbursts. There may be a greater number of sources above the dashed line than that below the line, indicating that for some fraction of sources, the Brα emission tends to provide a higher starburst contribution than the 3.3 μm PAH emission. The fraction of such sources is higher in ULIRGs (Figure 7). Although AGN contribution to the Brα emission line is a possibility, this trend could be due to selection bias, because only ULIRGs with clearly detectable, strong Brα emission lines are plotted. In fact, for the modestly obscured (A_V < 30 mag) starbursts probed with the 3.3 μm PAH and Brα emission, dust extinction reduces the L_3.3 PAH/L_Brα flux ratios by only ∼30%, if the dust extinction curve A_λ ∝λ^−1.75 (Cardelli et al. 1989) is adopted.

Zoom In Zoom Out Reset image size

For LIRGs and ULIRGs with low EW_3.3 PAH values, we can reasonably assume that the bulk of the observed 2.5–5 μm continuum fluxes originate in AGN-heated hot dust continuum emission. By applying the correction for flux attenuation by dust extinction, we can estimate the extinction-corrected intrinsic luminosity of AGN-heated hot dust emission. Conversion from this type of luminosity to the intrinsic primary energetic radiation luminosity of an AGN is straightforward if the AGN is surrounded by a large column of dust in virtually all directions, because energy is conserved (without escape) from the inner hot dust regions to the outer cool dust regions. In a pure AGN with this type of geometry, and without strong starbursts, the luminosity of the AGN-heated hot dust emission must be comparable to the primary energetic radiation of the AGN (Figure 2 of Imanishi et al. 2007a). Buried AGNs in optically non-Seyfert LIRGs and ULIRGs could be modeled by this geometry, as a first approximation.

However, for Seyfert-type AGNs, which are thought to be surrounded by torus-shaped dusty media, the discussion is more complex. If the solid angle of the torus, viewed from the central compact SMBH, does not vary with the distance from the center, the AGN-originated 2.5–5 μm continuum luminosity (hot dust of the inner part) should be similar to the luminosity at longer infrared wavelengths (cool dust on the outside). However, if the torus is strongly flared (Wada et al. 2009) or warped (Sanders et al. 1989), this assumption is no longer valid, and the luminosity at the longer wavelengths could be greater than the AGN-originated 2.5–5 μm continuum luminosity. Furthermore, a significant fraction of the primary AGN radiation may escape without being absorbed by dust, and such an emission component cannot be probed by infrared observations. To estimate the primary energetic radiation luminosity of an AGN from the AGN-originated 2.5–5 μm continuum luminosity, we adopt a correction factor of 5 (L_AGN/L_{2.5−5 μm} = 5) for sources optically classified as Seyferts (Risaliti et al. 2010).

We choose λ_rest = 3.5 μm as representative of the 2.5–5 μm continuum emission and estimate νF_ν(3.5 μm) or equivalently λF_λ(3.5 μm), for sources with EW_3.3 PAH < 40 nm. The 3.5 μm continuum flux attenuation is estimated from the τ_3.4 and τ_3.1 values, rather than the continuum slope Γ, because the intrinsic Γ value for unobscured AGNs could have high uncertainty (Section 5.4). The dust extinction is derived from the formula A_V = τ_3.1/0.06 + τ_3.4/0.006 (see Section 5.4). The flux attenuation at 3.5 μm, relative to the optical V band (0.55 μm), is observationally found to be A_3.5 μm/A_V = 0.03–0.06 (Rieke & Lebofsky 1985; Nishiyama et al. 2008, 2009). By adopting A_3.5 μm/A_V = 0.05, we obtain A_3.5 μm = 0.83 × τ_3.1 + 8.3 ×τ_3.4.

Figure 8 compares the extinction-corrected intrinsic AGN luminosities and observed infrared luminosities of LIRGs and ULIRGs with EW_3.3 PAH < 40 nm. Only a small fraction of the sources exhibit intrinsically luminous AGNs, which could quantitatively account for the observed infrared luminosities. However, we note that all but one of them are 3.4 μm absorption-detected sources. As the formula of A_3.5 μm = 0.83 ×τ_3.1 + 8.3 × τ_3.4 indicates, correction of the flux attenuation at 3.5 μm is very sensitive to the τ_3.4 values, and yet the detection rate of this 3.4 μm bare carbonaceous dust absorption feature is limited, partly because the feature is intrinsically weak (small oscillator strengths) and/or can be veiled by the 3.4 μm PAH emission sub-peak for PAH-detected sources. Thus, it is fair to assume that the estimated intrinsic AGN luminosities for sources with non-detectable 3.4 μm absorption features are only lower limits, and the actual AGN luminosities could be much higher. To clarify this possible ambiguity, sources with detectable and non-detectable 3.4 μm absorption features are distinguished by different symbols in Figure 8. For LIRGs and ULIRGs optically classified as Seyferts, the estimated AGN luminosity is a significant fraction (>10%) of the observed infrared luminosity in most cases. For buried AGN candidates in optically non-Seyfert LIRGs and ULIRGs without detectable 3.4 μm absorption features, the buried AGNs could explain at least a few to 10% of the observed infrared luminosities.

Zoom In Zoom Out Reset image size

In this regard, the non-detection of the 3.4 μm bare carbonaceous dust absorption feature in the ULIRG IRAS 00397−1312 is puzzling. This galaxy displays a very red continuum, but no detectable 3.3 μm PAH emission (Figure 2), both of these factors indicating an energetically important buried AGN. In this ULIRG, the abundance of carbonaceous dust (= the carrier of the 3.4 μm absorption feature) may be suppressed for some reason. At the same time, the intrinsic AGN luminosities for 3.4μm absorption-detected sources are often greater than the observed infrared luminosities (Figure 8) where the carbonaceous dust abundance may be enhanced and the τ_3.4 value per unit dust extinction could be higher than in the Galactic diffuse interstellar medium. In this case, the intrinsic AGN luminosity could be overestimated, if we are based on the τ_3.4 values and the Galactic extinction curve. Possible variation of the abundance of carbonaceous dust as the carrier of the 3.4 μm absorption feature could introduce additional ambiguity into the estimates of the intrinsic AGN luminosities obtained from the AKARI IRC 2.5–5 μm spectra.

We selected buried AGN candidates based on low EW_3.3 PAH values, large τ_3.1 and τ_3.4 values, and large Γ values in the AKARI IRC 2.5–5 μm spectra. Although the presence of a buried AGN is a natural explanation for these observed properties, alternative scenarios are also possible and cannot be ruled out completely. For example, extreme starbursts which are exceptionally more centrally concentrated than the surrounding molecular gas and dust (Figure 1(e) of Imanishi et al. 2007a) and consist of H ii regions only, with virtually no molecular gas or photodissociation regions, could also produce weak PAH emission, strong dust absorption features, and red continuum slopes. It has been argued that extreme starbursts do not readily occur in ULIRGs (Imanishi et al. 2007a, 2010a; Imanishi 2009), because of the extremely high emission surface brightness required, which is much higher than the value normally sustained by starburst phenomena (Thompson et al. 2005). Additionally, a normal starburst nucleus with a mixed dust/source geometry and a large amount of foreground screen dust in an edge-on host galaxy (Figure 1(d) of Imanishi et al. 2007a) could produce large τ_3.1, τ_3.4, and Γ values, although a low EW_3.3 PAH value is not explained in this way. To investigate whether our buried AGN interpretation is generally persuasive, we compare our results with published data for several selected well-studied sources.

Since LIRGs are generally closer than ULIRGs, high-quality data are available at other wavelengths and can be used to detect the presence of buried AGNs. In Tables 5 and 9, 17 LIRG nuclei display AGN signatures. After excluding six optically identified AGNs (i.e., optical Seyferts; MCG−03-34-064, NGC 5135, NGC 7469, IC 5298, NGC 7592W, and NGC 7674), 11 buried AGN candidates remain. These are NGC 232, VV 114E, NGC 2623, IC 2810, IC 694, NGC 3690, IRAS 12224−0624, IC 860, NGC 5104, IRAS 15250+3609, and NGC 7771, of which buried AGN signatures are found in VV 114E, NGC 2623, and NGC 3690 by more than one of the methods based on the AKARI IRC 2.5–5 μm spectra. The nuclei of these three LIRGs are thus particularly strong buried AGN candidates. In fact, for NGC 2623 and NGC 3690, X-ray observations strongly suggest the presence of Compton-thick (N_H > 10²⁴ cm⁻²) AGNs (Della Ceca et al. 2002; Maiolino et al. 2003; Zezas et al. 2003; Ballo et al. 2004). The presence of an obscured AGN in NGC 2623 is also supported by the detection of a high-excitation, mid-infrared [Ne v] 14.3 μm forbidden emission line (Evans et al. 2008). For VV 114E, Le Floc'h et al. (2002) argued the presence of a luminous buried AGN on the basis of the strong, featureless mid-infrared 15 μm continuum emission. Thus, for all three of the particularly strong buried AGN candidates (VV 114E, NGC 2623, and NGC 3690), there are independent buried AGN arguments in other works.

Among other buried AGN candidates, IC 694, the pair galaxy of NGC 3690 in the Arp 299 merging system, may also contain an obscured X-ray-emitting AGN (Zezas et al. 2003; Ballo et al. 2004). IRAS 12224−0624 exhibits strong core radio emission suggestive of an AGN (Hill et al. 2001). The infrared >5 μm spectrum of IRAS 15250+3609 is typical of buried AGNs (Spoon et al. 2002; Armus et al. 2007). These supporting results demonstrate the reliability of our AKARI IRC 2.5–5 μm spectroscopic technique as a tool for finding optically elusive buried AGNs.

Unlike LIRGs, the literature contains relatively few references to buried AGN signatures for members of our ULIRG sample. Because most ULIRGs are more distant than LIRGs, the detection of AGN signatures generally becomes difficult by X-ray observation (the most powerful AGN search tool), simply due to the lack of sensitivity of the currently available X-ray satellites. Even with this difficulty, Teng et al. (2008) detected AGN signatures in our buried AGN candidate IRAS 04103−2838 by deep X-ray observations. In another buried AGN candidate, IRAS 12127−1412, Spoon et al. (2009) found AGN signatures via high-velocity neon forbidden emission lines in the mid-infrared range.

The energetic importance of AGNs as a function of galaxy infrared luminosity has previously been investigated by a number of researchers (Tran et al. 2001; Imanishi 2009; Veilleux et al. 2009; Nardini et al. 2009, 2010; Valiante et al. 2009; Imanishi et al. 2010a). However, these studies focused on ULIRGs with L_IR ⩾ 10¹² L_☉, simply because energy diagnostics of LIRGs with L_IR = 10¹¹–10¹² L_☉ are still relatively scarce (Brandl et al. 2006; Imanishi et al. 2009; Pereira-Santaella et al. 2010). The infrared emission of LIRGs is apparently more extended than that of ULIRGs, so that slit spectroscopy using ground-based spectrographs with large telescopes and the Spitzer IRS are not adequate to fully cover the emission from LIRGs. A second possible reason for the scarcity of LIRG studies is that while there are numerous indications that luminous buried AGNs are present in at least some fraction of ULIRGs (Sanders et al. 1988a; Soifer et al. 2000), and thus there is some motivation for seeking AGNs through detailed infrared spectroscopy, such AGN implications are generally rarer in optically non-Seyfert LIRGs than in ULIRGs (Soifer et al. 2001).

With the acquisition of the AKARI IRC 2.5–5 μm slit-less spectra for a large sample of LIRGs, systematic investigations of LIRG energy sources are now possible, enabling us to understand the role of both optically identified and optically elusive AGNs in galaxies with a wide infrared luminosity range, from LIRGs to ULIRGs. It is known that the fraction of optically identified AGNs (optical Seyferts) increases with increasing galaxy infrared luminosity (Veilleux et al. 1999; Goto 2005). Figure 9 shows a plot of the fraction of sources with detectable buried AGN signatures⁸ as a function of galaxy infrared luminosity, including both LIRGs and ULIRGs. Only optically non-Seyfert LIRGs and ULIRGs are plotted, and optical Seyferts are excluded. In the AKARI IRC results, the number of LIRGs with L_IR < 10¹² L_☉ is sufficient (50 sources), but the sample is not statistically complete for ULIRGs with 10¹² L_☉ ⩽ L_IR < 10^12.3 L_☉ (43 sources) and L_IR ⩾ 10^12.3 L_☉ (30 sources). In the Spitzer IRS results, although the sample is statistically complete for ULIRGs with 10¹² L_☉ ⩽ L_IR < 10^12.3 L_☉ (54 sources) and L_IR ⩾ 10^12.3 L_☉ (31 sources), the number of LIRGs (L_IR < 10¹² L_☉) is limited (18 sources). Thus, these results complement each other. Although the samples are not exactly the same, both the AKARI and Spitzer results show that the energetic importance of buried AGNs gradually increases with increasing galaxy infrared luminosity, ranging from LIRGs to ULIRGs. Accordingly, the existing implications that the AGN–starburst connections are luminosity-dependent in ULIRGs (Imanishi et al. 2010a; Nardini et al. 2010) can be extended to LIRGs as well.

Zoom In Zoom Out Reset image size

The AGN luminosity increases when the mass accretion rate onto an SMBH increases. The luminosity dependence of AGN–starburst connections suggests that more (less) infrared luminous galaxies currently are actively (sluggishly) accreting mass onto SMBHs. More luminous buried AGNs with higher SMBH mass accretion rates could provide stronger feedback to the surrounding host galaxy, because the buried AGNs are still surrounded by a large amount of gas and dust. As Imanishi et al. (2010a) have argued, galaxies with greater infrared luminosities not only demonstrate the relatively higher energetic importance of buried AGNs but also indicate higher absolute star formation rates, and thus are likely to be the progenitors of more massive galaxies with larger stellar masses. The recently discovered galaxy downsizing phenomenon, in which massive red galaxies have completed their major star formation more quickly in an earlier cosmic age (Cowie et al. 1996; Heavens et al. 2004; Bundy et al. 2005), is widely argued to have resulted from stronger AGN feedback in the past (Granato et al. 2004; Sijacki et al. 2007; Booth & Schaye 2009). Our new AKARI results further indicate a possible connection between buried AGN feedback and the origin of the galaxy downsizing phenomenon.