AKARI NEAR-INFRARED SPECTROSCOPY OF SDSS-SELECTED BLUE EARLY-TYPE GALAXIES

We selected 59 non-passive (i.e., currently star-forming or hosting an AGN) BEG targets based on the scheme of Lee et al. (2008) using the spectroscopic sample of galaxies in the SDSS Data Release 4 (DR4; Adelman-McCarthy et al. 2006). We adopted the physical parameters of the galaxies from several value-added galaxy catalogs (VAGCs) drawn from the SDSS: photometric parameters from the SDSS pipeline (Stoughton et al. 2002), structural parameters estimated by Park & Choi (2005) and Choi et al. (2007), and spectroscopic parameters from Max-Planck-Institute for Astrophysics (MPA)/Johns Hopkins University (JHU) VAGC (Kauffmann et al. 2003; Tremonti et al. 2004; Gallazzi et al. 2006). First, we selected the early-type galaxies from the SDSS galaxies using their distribution in the color, color gradient, and light-concentration parameter space (Park & Choi 2005). Next, we divided the selected early-type galaxies into REGs and BEGs, using the method of Lee et al. (2006), which is based on the color distribution of bright early-type galaxies as a function of redshift.⁶ The BEGs were classified using their flux ratios between several spectral lines (e.g., BPT diagram; Baldwin et al. 1981; Kewley et al. 2006) into passive, SF, Seyfert, and low ionization nuclear emission region (LINER) galaxies, among which passive BEGs were excluded. Finally, we dropped BEGs with conspicuous substructures or disk components in visual checks. Among the BEGs selected in this scheme, we chose the target objects as non-passive BEGs at 0.02 < z < 0.1 with K_s <14.4.⁷ The magnitude limit was applied to ensure the quality of the AKARI spectroscopy.

As a result, 59 BEGs were selected and proposed to be observed, but only 36 targets among them were successfully observed by the AKARI.⁸ Figure 1 shows the basic optical–NIR properties of the 36 BEGs in our sample: their magnitudes as a function of redshift, color–magnitude relation, and color–color relation, confirming that our sample BEGs are blue and relatively bright objects. Figure 2 displays the distribution of the BEGs in the line flux ratio diagrams, showing that the SF:Seyfert:LINER composition in our BEG sample is well balanced (11:15:10). In Figure 3, the atlas images of the BEGs retrieved from the SDSS are presented. The basic properties of the BEGs are summarized in Table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The AKARI⁹ is a Japanese infrared astronomy satellite from the Institute of Space and Astronautical Science (ISAS) of the Japanese Aerospace Exploration Agency (JAXA). It was launched on 2006 February 21 and ran out of its onboard coolant (helium) supply on 2007 August 26 after its successful operation. The observation of our BEG targets was conducted during the AKARI open-time operation, the post-helium phase of the AKARI observation, in which NIR (1.8–5.5 μm) imaging and spectroscopic observations are available.

The observation was conducted from 2008 November to 2009 August. The target BEGs were observed using the IRCZ4 Astronomical Observing Template (AOT), which is the spectroscopic AOT of the IRC. Between the grism and prism, we selected the grism spectroscopy that has better spectroscopic resolution. The 1σ noise equivalent flux density of the IRCZ4 grism is about 0.2 mJy, and its spectroscopic resolution is R ∼120. As the target positioning option, we used the point-source aperture (Np), because the typical effective radius of target BEGs is only several arcseconds. Among the successfully observed 36 targets, 29 objects were observed twice, while 7 targets were observed only once. The net exposure time for a single pointing observation is about 6 minutes (Onaka et al. 2007; Imanishi et al. 2008).

Unlike our expectation, two pointings per object were not enough to secure a sufficient signal-to-noise ratio (S/N) for most targets (half of the spectra have 3.5–3.9 μm continua with S/N <6), although the spectra of a few objects are relatively good. Thus, we applied the median stacking analysis technique for high S/N. The steps of the stacking analysis are as follows.

• 1.  
  We used the IRC Spectroscopy Toolkit for Phase 3 Data Version 20090211 provided by the AKARI team, for the basic processes such as dark subtraction, image flat fielding, spectral flat fielding, image combining, source detection and extraction, sky subtraction, wavelength calibration, and flux calibration.
• 2.  
  Using the SDSS spectroscopic redshift, we converted the observer-frame wavelength into the rest-frame wavelength for each object. Since the rest-frame wavelength ranges are different between objects, we trimmed each spectrum with the rest-frame wavelength range of 2.5–4.5 μm.
• 3.  
  All trimmed spectra were normalized for the median flux density at 3.5 μm < λ < 3.9 μm to be identical.
• 4.  
  We stacked the normalized spectra by taking a median. We also produced several median-stacked spectra in bins of several properties such as optical spectral type, optical color, and [O ii]-based specific star formation rate (SSFR).

We selected the median stacking instead of mean stacking because it removes unreal data points (hot or bad pixels) more effectively, whereas it also improves data quality similarly to that which is done by the mean stacking. Not only the total stacking (i.e., stacking all spectra into a single spectrum) but also the subsample stacking was carried out to compare the NIR spectral features between subsamples with different optical properties. We simply estimated the uncertainty of the stacked spectra by supposing that the spectral noise follows the Poisson error and that the S/N is accumulated during the stacking process. That is,

$$\begin{equation} ({\rm S/N})_s = \sqrt{\sum {({\rm S/N})_i}^2}, \end{equation} \tag{ 1 }$$

where (S/N)_s is the S/N of the stacked spectrum and (S/N)_i is the S/N of the individual spectra.

We used the power-law continuum +3.29 μm PAH emission (Gaussian) formula (Risaliti et al. 2006) to fit the stacked spectra:

$$\begin{equation} F_{\lambda } = A \lambda ^{\Gamma } + B e^{-\frac{\scriptstyle {(\lambda -3.29\,{\mu }{\rm m})^2}}{\scriptstyle {2{\sigma }^2}}}, \end{equation} \tag{ 2 }$$

where F_λ is the flux density, λ is the wavelength in units of μm, Γ is the continuum slope, σ is the Gaussian spread in units of μm, and A and B are amplitudes. The fitting was conducted in the wavelength range of 2.5–3.9 μm, because the Brα emission and CO absorption lines may affect the NIR spectra significantly at λ>4 μm (Imanishi et al. 2008; Risaliti et al. 2010).

BEGs are known to have not only very young stars but also a considerable amount of old stars (Menanteau et al. 2001a; Ferreras et al. 2005; Lee et al. 2010a). In our results, however, it is not easy to check this previous knowledge independently, because the NIR continuum slope (Γ) depends on both stellar and dust contents (dust either in host galaxies or in AGN tori). Figure 8 shows the variation of Γ as a function of age, metallicity, star formation history, and dust attenuation, using the models of Bruzual & Charlot (2003): simple stellar population (SSP), exponentially decreasing, and constant SFR models. In Figure 8, it is found that Γ becomes less negative as SF ended earlier, as age increases, as metallicity increases, and as dust increases. Due to this degeneracy, it is difficult to tell the dominant factor in determining Γ of BEGs. Even if we suppose that the NIR continuum of the BEGs is old-SSP-dominated from previous knowledge (the age of most stars in BEGs is about 10 Gyr; e.g., Ferreras et al. 2005; Lee et al. 2010a), the Γ of the BEGs can be reproduced using either the metal-rich SSP model without dust or the significantly dust-attenuated SSP model with solar metallicity.

Zoom In Zoom Out Reset image size

The effect of internal dust extinction is independently checked using the information extracted from the SDSS data; that is, using the Balmer decrement. Based on the formula of Calzetti et al. (1994), we derived the internal reddening E(B − V)_i (median value for each subsample) and compared them with the Γ and EW_3.29 of the stacked BEGs in Figure 9. It is found that the NIR spectral parameters have correlations with E(B − V)_i, in the sense that the Γ and EW_3.29 increase with increasing E(B − V)_i. The E(B − V)_i values of the BEGs are not small, but seem to be not large enough to explain the Γ of the BEGs in Figure 8 (the τ_V corresponding to E(B − V) = 0.4 is 1.14; De Ruyter et al. 2005). Thus, if we suppose moderate (not too strong) dust attenuation in the BEGs, Figure 8 indicates that the BEGs may have stellar populations with slightly high metallicity, although this estimation is strongly model-dependent.

Zoom In Zoom Out Reset image size

Meanwhile, BEGs are often thought to be the intermediate objects transforming from interacting galaxies or mergers to red elliptical galaxies or bulges of late-type galaxies (Lee et al. 2006, 2008; Kannappan et al. 2009). Known to have similar origins are Ultra-Luminous Infrared Galaxies (ULIRGs). ULIRGs are very luminous in the IR band due to vigorous starburst and/or AGN activity (Veilleux et al. 1995, 1999a), and are thought to be formed by strong interaction or merger of two disk galaxies (Sanders et al. 1988). Risaliti et al. (2006, 2010) estimated the Γ and EW_3.29 of some ULIRGs showing an interesting division between AGN-host ULIRGs and starburst ULIRGs.

Figure 10 shows the loci of our BEGs on the Γ–EW_3.29 plane, compared with the objects presented in Risaliti et al. (2010). In this figure, it is found that our BEGs have Γ and EW_3.29 values distinct from those of the ULIRGs, in the sense that the BEGs have largely negative Γ values (∼−2.5), while those of the ULIRGs are mostly larger than −2 (up to 6). The EW_3.29 of the BEGs is smaller than that of the ULIRGs on average, showing that the SF in the BEGs is not as active as that in the ULIRGs. The Γ value is expected to be largely positive when the NIR continuum is dominated by dust emission. In our BEG sample, however, even the stacked spectra using optical AGN or optical SF BEGs have largely negative Γ. This indicates that the dust amount in those BEGs may not be as large as that in ULIRGs, as shown already using the internal dust reddening derived from the Balmer decrement (the mean E(B − V) value of ULIRGs is larger than 1.0; Veilleux et al. 1999a, 1999b).

Zoom In Zoom Out Reset image size

In short, the Γ and EW_3.29 features of the BEGs are not easy to interpret, because they are affected by the combined effects of age, metallicity, star formation history, and dust attenuation. However, based on some previous knowledge, Balmer line information, and a few assumptions, the BEGs are thought to be old-SSP-dominated metal-rich galaxies with moderate dust attenuation. The dust attenuation in the BEGs may originate from recent SF or AGN activity. This interpretation is consistent with the previous understanding of stellar populations in BEGs (that is, mostly old stars + partially young stars; Ferreras et al. 2005; Lee et al. 2010a). There is a possibility that ULIRGs are the progenitors of BEGs in the merging/interacting phase (Veilleux et al. 2002), but any clear evidence of the close relationship between BEGs and ULIRGs is not yet found in our results based on the NIR spectroscopy.

As introduced in Section 1, NIR spectroscopy provides interesting clues to the internal evolutionary process of the BEGs, related to their SF or AGN activity. Those clues become more useful when combined with the optical properties of the BEGs. Figure 11 compares the stacked spectra in bins of several properties on the Γ–EW_3.29 plane, giving a summary of the key results in Figures 4–7. The EW_3.29 reflects the current SF of the BEGs well (Figure 11(c)), but it is noted that the Seyfert BEGs have EW_3.29 larger than that of the LINER BEGs. As mentioned at the end of Section 4, this result is not well explained simply by the current consensus that AGN activity destroys PAH particles.

Zoom In Zoom Out Reset image size

In addition to EW_3.29, another key parameter is Γ. As shown in Figure 8, Γ is commonly affected by age, metallicity, and dust. However, the bluest BEGs with the largest (least negative) Γ in Figure 11(b) are not easily understood by the age or metallicity effects, because blue galaxies tend to be young or metal-poor, which will make Γ more negative, unlike the bluest BEGs. Thus, the Γ difference between the bluest BEGs and the other BEGs seems to be mainly due to the difference in their dust contents. It is also noted that the optical SF BEGs have small (largely negative) Γ in Figure 11(d), which shows good consistency with the SF history dependence of Γ, indicating that those SF BEGs are not very dusty. That is, several AGNs with dusty tori in the bluest BEG subsample seem to mainly contribute to the largest (least negative) Γ.

In Figure 11(c), the BEGs with large SSFR_[O ii] have Γ intermediate between those of the bluest BEGs and optical SF BEGs. These are because the optical color, SSFR, and optical spectral type of the BEGs do not tightly correlate: some non-SF BEGs have large SSFR_[O ii] or blue ^0.1(u −  r) color. These results lead us again to the conclusion that the SF and the AGN activity in BEGs are not necessarily contradictory to each other. The coexistence of SF and AGN is not an entirely new discovery (e.g., Tzanavaris & Georgantopoulos 2007; Treister et al. 2009), but currently it is widely believed that the AGN activity tends to destroy PAH particles and truncate SF (Voit 1992; Weinmann et al. 2006; Schawinski et al. 2007; Tortora et al. 2009). In our result, however, Seyfert BEGs have larger EW_3.29 than LINER BEGs, in spite of the fact that Seyferts show stronger AGN activity than LINERs.

This result may be explained by various ways. Smith et al. (2007) and Imanishi et al. (2008) showed that PAHs can survive even if they are close to AGNs when the AGNs are not dust-free, which gives an answer to how the AGN activity and PAH emission coexist. However, it does not sufficiently explain why stronger AGNs have stronger PAH emission features. One possibility is that Seyfert BEGs and LINER BEGs may not be the separate branches in BEG evolution. Satisfying our results, one possible scenario of BEG evolution is as follows (see also Lee et al. 2010a).

•  Ph.1  
  SF BEG phase. Triggered by some mechanisms (possibly interaction or merger; Lee et al. 2006), active (but not very dusty) SF occurs in the progenitor of a BEG, originally consisting of old stars in the main. The NIR continuum is very similar to that of old passive galaxies, but the PAH emission is clearly detected.
•  Ph.2  
  Seyfert BEG phase. After the central black hole in a BEG increases its mass by gas accretion, the AGN activity starts. The SF in the BEG starts to be suppressed by the AGN feedback, but at this time, both AGN and SF activities coexist. The NIR continuum is contaminated by the AGN (Γ increases), and the PAH emission becomes weaker than that of SF BEGs.
•  Ph.3  
  LINER BEG phase. The AGN feedback has truncated most SF by removing ambient gas in the BEG. Since this makes the gas accretion into the central black hole stop also, the AGN activity becomes weak as a natural result. The NIR continuum is intermediate between SF BEGs and Seyfert BEGs, and the PAH emission feature is the weakest among the three phases.

After these three phases, the BEGs may evolve into REGs or bulges of late-type galaxies via the phase of passive BEGs (Lee et al. 2006, 2008; Kannappan et al. 2009). The proposed scenario is consistent with the findings about the relationship between Seyferts and LINERs in several recent studies (Schawinski et al. 2007, 2009; Hickox et al. 2009; Lee et al. 2010a).