A PLETHORA OF ACTIVE GALACTIC NUCLEI AMONG Lyα GALAXIES AT LOW REDSHIFT^*

We observed 23 low-redshift LAEs in the EGS using Hectospec at the 6.5 m MMT, which has a 1° diameter field of view (Fabricant et al. 2005). We obtained 2 hr of spectroscopy on 2009 March 19, with spectral coverage from ∼3650 to 9200 Å, and spectral resolution of ∼5 Å. We used the External SPECROAD⁴ pipeline for data reduction (developed by J. Cabanela 2009, private communication). This pipeline performs standard spectroscopic reductions, including bias, dark and flat correction, and wavelength calibration (rms residuals ≈ 0.1 Å). We flux-calibrated the spectra using observations of Sloan Digital Sky Survey (SDSS) F-type stars. Further details on the observations and data reduction, as well as tabulated line fluxes, will be presented in a future paper (S. L. Finkelstein et al. 2009, in preparation).

The 23 LAEs varied in redshift from 0.2 ≲z≲ 0.45, thus we found 31 possible spectral lines which we could detect, from Mg ii λλ 2796, 2803 to [S ii] λ6733. We examined the expected position (using the redshift from Deharveng et al. (2008) as a first estimate) of each of the expected lines in each spectrum, and noted whether it was detected. We fit a Gaussian curve to each detected line, using the MPFIT IDL software package,⁵ obtaining the continuum flux, central wavelength, Gaussian σ, and line flux. Errors on each of these parameters were estimated via Monte Carlo simulations by varying each spectral data point within its photometric uncertainty. Table 1 lists the object redshifts as determined from this spectroscopy, which are consistent with (and refine) those published in Deharveng et al. (2008).

Lines from the broad-line region of AGNs have velocities ≳1000 km s⁻¹, while narrow-line region lines have widths ≲500 km s⁻¹. Depending on the orientation, one may only see one of these regions, leading to two separate classifications of AGNs: broad-lined AGNs (BLAGNs; also called Type 1 or Seyfert 1), and narrow-lined AGNs (NLAGNs; Type 2 or Seyfert 2). For each object, we measured the characteristic line velocity widths for all detectable, permitted transitions of H i, He i, and He ii. Only EGS24 has a velocity width >1000 km s⁻¹, with v = 1064 ± 17 km s⁻¹, and thus likely contains a BLAGN. The remaining objects have v< 300 km s⁻¹, implying that if an AGN is present in the rest of the sample, it is narrow-lined.

[Ne v] and He ii are two of the strongest indicators of AGN activity to which our data are sensitive. We detect [Ne v] λ3347 or [Ne v] λ3427 in EGS16, EGS17, EGS18, and EGS24 at >3σ significance, and in EGS14 and EGS20, at 2.9 and 2.0σ, respectively. The He ii λ4686 line is also an AGN indicator, and this line is observed at ⩾3σ significance in EGS2, and at 2.6σ in EGS24. We note that although He ii emission can also come from the winds of Wolf–Rayet stars, these lines are typically broad (e.g., Shapley et al. 2003), and thus that is likely not the case here. In addition, while the [Ne v] emission is most likely due to an AGN, there is the possibility that it arises due to shocks, although none of our objects solely identified by this line inhabit the BPT plane (see Section 3.3) near the shock models of Dopita & Sutherland (1995) (compare to van Dokkum et al. 2005).

[Ne iii] λ3869 is detected in numerous objects. However, this line may not conclusively point to an AGN, as its ionization potential is not much greater than [O iii]. As a control sample, we examined Sloan Digital Sky Survey spectroscopic data,⁶ to see where objects with detected [Ne iii] emission inhabit the BPT diagram. We found that at all fluxes objects with [Ne iii] emission populate both the star-forming and AGN sequences, implying that [Ne iii] emission is not conclusive evidence of an AGN.

In Figure 2, we plot ratios of Hα/[N ii] versus [O iii]/Hβ to diagnose the ionizing source within a galaxy (Baldwin et al. 1981). In this plane galaxies segregate nicely, with separate star-forming and AGN sequences. The contours represent the two sequences as defined using ∼10⁵ SDSS galaxies,⁷ with the lighter contours representing the star-forming sequence, and the darker contours representing the AGN sequence. We plot each of our objects⁸ on this plane using the various colored symbols. We also plot the theoretical maximum starburst line from Kewley et al. (2001) and the star formation/AGN demarcation line from Kauffmann et al. (2003). We estimated flux limits for undetected lines by placing mock emission lines in the spectrum, and decreasing the line strength until the detection significance dropped below 3σ, finding that 1σ line flux limits were typically ≈1.0 × 10⁻¹⁷ erg s⁻¹ cm⁻². In objects where any of these lines was undetected, we plot their 3σ upper limits.

Zoom In Zoom Out Reset image size

EGS1, EGS3, EGS14, EGS15, and EGS24 appear to unambiguously lie on the AGN sequence, and we henceforth classify them as AGNs. EGS24 has already been noted for its broad emission lines, as well as its [Ne v] and He ii emission, and EGS14 also has a 2.9σ [Ne v] detection. Although EGS1, EGS3, and EGS15 were not indicated by either of our previous classification methods, they are located very far from the star-forming sequence, and significantly higher than the maximum starburst curve. However, EGS1 appears to be in the LINER regime, and thus if an AGN is present, it may not be dominating the observed continuum and line fluxes. EGS16, EGS17, and EGS21 lie in regions which could correspond to either AGN or star-forming galaxies, as well as near the Kauffmann et al. (2003) demarcation line. EGS16 and EGS17 have detected [Ne v] emission. As EGS21 was not indicated by any of our previous methods, thus we also only consider it a possible AGN.

Baldwin et al. (1981) also derived classifications using lines of O i and [O ii], and their Figure 4, which plots O i λ6300/Hα versus [O ii]/[O iii], shows a clear separation in O i λ6300/Hα, with nearly all power-law ionization objects (i.e., AGN) having log(O i/Hα) ≳−1, and all star-forming objects having log(O i/Hα) ≲−1.5. We find that out of the five objects that have ⩾3σ detections of O i λ6300, EGS10 and EGS24 exhibit log(O i/Hα) indicative of AGN activity of −0.74 and −1.05, respectively. EGS10 was not indicated by any of our previous classification methods, thus we now regard it as a possible AGN.

Stern et al. (2005) defined a color–color region using colors from the Spitzer Infrared Array Camera (IRAC), which preferentially selects AGNs in the redshift range of our sample (20% contamination), due to the rising power-law spectrum (i.e., from warm dust) exhibited by these objects. We examined the AEGIS IRAC catalog and found that four of our objects were covered by the IRAC data: EGS1, EGS9, EGS15, and EGS27. All four were detected in all IRAC bands, while only EGS27 lies within the selection wedge, at ∼1σ from the edge. Thus, we regard EGS27 as a possible AGN. While other methods select the remaining three objects as AGNs, they all reside outside the wedge.

The Chandra AEGIS-X 200 ks catalog (v2.0; Laird et al. 2009) has limiting soft- and hard-band luminosities of L_{0.5−2 keV(SB)} = 1.5 × 10⁴⁰ and L_{2−10 keV(HB)} = 1.1 × 10⁴¹ erg s⁻¹ (z = 0.3), respectively; thus if our objects had typical AGN X-ray luminosities, we should be able to detect them. We examined the positions of each of the 11 objects that were covered by the Chandra observations, and found that only EGS9 has an X-ray counterpart within 5'' (a separation of r < 019), with full-band luminosity L_FB = 8.0 ± 1.5 × 10⁴¹ erg s⁻¹ (at z = 0.246). While star-forming galaxies typically show L_FB < 10⁴² erg s⁻¹ (Szokoly et al. 2004), EGS9's hardness ratio of −0.13, which when combined with the X-ray luminosity, satisfies the Szokoly et al. (2004) AGN criteria, thus we regard EGS9 as a probable AGN. The hardness ratio is consistent with the emission detected at 24 μm (see Section 3.5), which could imply that dust is obscuring some of the X-rays. An object similar to EGS9 at high redshift likely would not be selected as an AGN via its X-ray flux, as its X-ray/Lyα flux ratio of ∼1.2 is much less than the typical limit from the Large Area Lyman Alpha (LALA; Rhoads et al. 2000) survey of >10 (Wang et al. 2004).

Out of the 10 X-ray undetected LAEs, our previous methods have found possible AGNs in six of them. This can be understood given the unification picture of AGNs (e.g., Urry & Padovani 1995), where one has a direct line of sight to see the X-rays from the accretion disk. When observing from other angles, the accretion disk can be blocked by the torus of dust and gas surrounding the center. In this case, the dust emission from the torus may be re-emitted at long wavelengths, thus we examined the MIPS 24 μm catalog from the Far-Infrared Deep Extragalactic Legacy Survey (M. Dickinson et al. 2009, in preparation). Out of these six LAEs, we found that three had a MIPS counterpart (EGS2, EGS15, and EGS27; all nine MIPS detections out of the 13 LAEs covered are listed in Table 1), indicating that dust may be absorbing the X-rays.

To see whether any of the six X-ray undetected LAEs classified as AGNs harbor low-luminosity AGNs, we performed an X-ray stacking analysis. We found that the stack was marginally detected with an S/N ∼ 2.1. Assuming Γ = 1.4, the mean X-ray luminosity of the individually detected objects is L_SB (L_HB) = 1.39 ± 0.67 (1.28 ± 2.82) × 10⁴⁰ erg s⁻¹ at the median redshift of these six objects (z = 0.263). The S/N decreases to ∼1.8 when stacking all 10 X-ray undetected LAEs. This shows that the six AGNs are dominating the X-ray flux from all 10 undetected objects, and that they likely harbor low-luminosity AGNs, which are too faint to be individually detected in the AEGIS-X survey. Our NLAGN number density (see Section 4.1) of 7 × 10⁻⁵ Mpc⁻³ is consistent with this conclusion, as Hasinger et al. (2005) show that the number density of AGNs increases with decreasing luminosity, and n ∼ 1 × 10⁻⁵ Mpc⁻³ in their lowest luminosity bin (10⁴² ⩽ L_x ⩽ 10⁴³ erg s⁻¹).

We assign confidence levels to the classified AGNs, with objects indicated by two or more methods given a classification grade of "A"; objects with unambiguous classification via only one method are given a grade of "B"; and objects with a possible classification with one method are given a grade of "C." As shown in Table 1, we have four grade A, six grade B, and four grade C AGNs. Taking A and B classifications as being reliably indicative of AGN activity, we find an AGN fraction of 43% (10/23) in low-redshift LAEs. Including class C AGNs, this could be as high as 61% (14/23), while neglecting class B and C AGNs provides a minimum value of 17% (4/23). A classification of A or B calls into doubt the stellar population modeling results of Finkelstein et al. (2009a) for that particular object.

Applying our AGN fraction (0.435) and SF fraction (0.565) to the 38/39 EGS LAEs at 0.2 ⩽z⩽ 0.35 from Deharveng et al. (2008), we find total numbers of 38 total LAEs, 16.5 LAEs with NLAGN, and 21.5 SF-dominated LAEs, corresponding to number densities of 1.6 × 10⁻⁴ Mpc⁻³, 7.0 × 10⁻⁵ Mpc⁻³, and 9.1 × 10⁻⁵ Mpc⁻³, respectively. From their photometric sample, Ouchi et al. (2008) found LAE number densities of 5.1 × 10⁻⁴ Mpc, 1.6 × 10⁻⁴ Mpc, and 4.3 × 10⁻⁴ Mpc⁻³ at z = 3.1, 3.7, and 5.7, respectively. Compared to our number density for all star-forming LAEs, we find that, on average, LAEs are ∼4 times rarer than at high redshift. This is a rough comparison, as Deharveng et al. (2008) is a factor of ∼7 more sensitive in luminosity to typical high-redshift studies. On the flip side, many of the AGNs here have low L_x, and thus would not be flagged as AGNs at high redshift.

Using X-ray data, typical AGN fractions in LAE samples at z > 3 range from ∼0% to 1% (Malhotra et al. 2003; Wang et al. 2004; Gawiser et al. 2007; Ouchi et al. 2008), increasing to 5% at z∼ 2.25 (Nilsson et al. 2009), implying that the AGN fraction in LAEs may be evolving below z = 3. However, these studies only probed intermediate X-ray luminosities, thus low-luminosity AGNs like we find in this Letter may have been missed. Dawson et al. (2004) and Wang et al. (2009) examined z∼ 4.5 LAEs with rest-frame ultraviolet spectroscopy, and also found little evidence for AGN activity, with Wang et al. reporting an upper limit on the AGN fraction of <17% from the upper limit on the C iv λ1549 emission line. Line widths can also be used at high redshift; however, BLAGNs appear to be rare in high-redshift LAEs (e.g., Dawson et al. 2004).

Our AGN fraction of 43⁺¹⁸₋₂₆% for z∼ 0.3 LAEs is significantly higher than high-redshift measurements. While at first glance this is a fair comparison, as high-redshift LAEs lack BLAGNs, and Deharveng et al. (2008) removed all BLAGNs from their sample (8 out of 47 original EGS LAEs were removed as BLAGNs, corresponding to n_BLAGN = 2.1 × 10⁻⁵ Mpc⁻³ over 0.2 ⩽z⩽ 0.35; J.-M. Deharveng 2009, private communication), only three of our methods (line widths, X-rays, and high-ionization emission) have been used in LAEs at high redshift. With only these methods we obtain an AGN fraction of 26%, better suited to comparisons with high redshift. When restricted to measurements available at high redshift, we thus miss from ∼15% to 40% of the AGNs in our sample, implying that high-redshift LAE AGN fraction measurements may be lower limits.

The AGN X-ray luminosities in our sample are ∼100 times lower than current sensitivities at high redshift. Thus, it is unclear if the LAE AGN fraction is evolving: if the low-luminosity AGNs we detect at z∼ 0.3 also exist at high redshift, we cannot yet detect them (although the increase in LAE AGN fraction at z∼ 2 compared to z > 3 may be due to the decreased luminosity distance). Given the observed downsizing trend of AGN luminosity with decreasing redshift (e.g., Cowie et al. 2003), the higher LAE AGN fraction at z∼ 0.3 is likely a combination of evolution, coupled with the presence of undetected low-luminosity AGNs at high redshift. Even if a substantial number of low-luminosity AGNs similar to their z∼ 0.3 cousins exist at high redshift, their contribution to the total Lyα flux is minor, scaling from the X-ray luminosity and the Lyα to X-ray ratios of Seyferts (Kriss 1984). This implies that the Lyα flux is star formation dominated. Nonetheless, infrared spectroscopy and deeper X-ray imaging are necessary to detect the true high-redshift LAE AGN fraction, and to quantify the AGN contamination. Some of this capability will be available in the not too distant future with the James Webb Space Telescope.