TESTING DIAGNOSTICS OF NUCLEAR ACTIVITY AND STAR FORMATION IN GALAXIES AT z > 1^*

Every massive nearby galaxy hosts a supermassive black hole (SMBH), and the mass of the SMBH correlates with the mass of the host galaxy bulge (Magorrian et al. 1998). Theoretical simulations suggest that this connection exists because past active galactic nucleus (AGN) phases of rapid SMBH growth were associated with periods of massive star formation (SF) in the host galaxy (e.g., Di Matteo et al. 2005; Hopkins et al. 2008). Observations of AGN frequency, including both weak Seyferts and powerful quasars, in different host galaxy types over the cosmic time can be used to directly test models of coupled SMBH–galaxy growth.

Selection by blue optical color, X-ray emission, or infrared (IR) color can be used to select powerful quasars to very high redshifts, but these methods are less effective for finding obscured or moderately accreting AGNs. Instead the most efficient way to find moderate-luminosity AGNs is by their unique emission line signature. Compared to typical SF processes, the higher-ionization radiation of an AGN tends to increase the ratios between rest-frame optical collisionally excited "forbidden" lines and hydrogen recombination lines (Baldwin et al. 1981; Kewley et al. 2006). In particular the line ratios f([O iii] λ5007)/f(Hβ), f([N ii] λ6584)/f(Hα), and f([S ii] λ6718 + 6731)/f(Hα) are typically used in the classic "BPT" and "VO87" diagnostics (Baldwin et al. 1981; Veilleux & Osterbrock 1987): the wavelength proximity of each line pair means the ratios are nearly insensitive to reddening. In the standard AGN unified model (Antonucci et al. 1993), these narrow emission lines can be detected even if the X-ray and ultraviolet (UV) ionizing radiation source is absorbed by anisotropic obscuration. The AGN line ratio signature also remains visible for SMBHs of moderately low accretion rates (L/L_Edd ∼ 10⁻³) which are otherwise dominated by their host galaxy starlight (e.g., Kauffmann & Heckman 2009).

There has been great success in using line ratio diagnostics to select AGNs and characterize their hosts at z ∼ 0 (e.g., Kauffmann et al. 2003a; Heckman et al. 2004; Yan et al. 2006; Schawinski et al. 2007). But extending a similar AGN census to z > 0.4 is difficult because the [N ii]/Hα and [S ii]/Hα ratios redshift to the IR, where ground-based spectroscopy has historically been expensive. A high [O iii]/Hβ ratio alone is degenerate between nuclear activity and metal-poor [H ii] regions, and the [N ii]/Hα or [S ii]/Hα line ratio is generally necessary to distinguish AGNs from inactive galaxies with low metallicity. Some authors have suggested combining [O iii]/Hβ with bluer emission line ratios, like [O ii] λ3726 + 3729/Hβ (Lamareille et al. 2004) or [Ne iii] λ3869/[O ii] (Trouille et al. 2011). However, these bluer lines are typically weaker, making them difficult to apply to distant galaxies. Instead it is possible to exploit the correlation of metallicity with color and stellar mass (Tremonti et al. 2004) to altogether eliminate a second line ratio. These modified AGN/SF diagnostics use [O iii]/Hβ to measure excitation, but replace the [N ii]/Hα ratio with rest-frame color (Yan et al. 2011) or stellar mass (Juneau et al. 2011). AGN selection using the "color-excitation (CEx)" and "mass-excitation (MEx)" methods agrees well with the classic line ratio diagnostics at z < 0.4 and X-ray selection at z < 1.

It is less clear if the BPT, VO87, CEx, and MEx AGN/SF diagnostics are directly applicable to higher redshift. SF processes at z > 1 are generally different than in nearby galaxies, with higher gas and dust fractions, younger stellar populations, lower metallicities, and higher star formation rates (SFRs; e.g., Papovich et al. 2005; Reddy et al. 2006; Tacconi et al. 2009; Shim et al. 2011). Some authors argue that the classical line ratio diagrams are unreliable at high redshift because local starburst galaxies with SF rates typical of z > 1 galaxies tend to have similarly high [N ii]/Hα and [O iii]/Hβ ratios (Liu et al. 2008; Brinchmann et al. 2008).

In this work, we directly test the effectiveness of the classical line ratio, color-excitation, and mass-excitation diagnostics for identifying AGNs at z ∼ 1.5. We use observations of 36 galaxies at 1.30 < z < 1.62 (mean $\bar{z}=1.52$) with the new MOSFIRE multi-object spectrograph (McLean et al. 2010, 2012) on the Keck telescope. The new MOSFIRE data are coupled to X-ray observations, previous spectroscopy, and multiwavelength photometry for stellar masses and rest-frame colors.

We study the emission line ratios, rest-frame colors, stellar masses, and X-ray properties of 36 galaxies in the Great Observatories Origins Deep Survey-South field (GOODS-S; Giavalisco et al. 2004). The targets were selected from Hubble Space Telescope (HST) Wide Field Camera 3 (WFC3) G141 grism observations as part of the 3D-HST survey (Brammer et al. 2012). The initial selection required F140W < 24 and a detected [O iii] emission line in the redshift range 1.3 < z < 1.7. The flux-limited sample includes a mix of star-forming, AGN, and composite galaxies. New Keck/MOSFIRE H-band spectra were obtained for the [N ii]/Hα and [S ii]/Hα line ratios. The GOODS-S field also includes tremendously deep optical and IR photometry (Dahlen et al. 2010) and 4 Ms of X-ray coverage (Xue et al. 2011). Details on the WFC3 grism, MOSFIRE, and photometric data are provided below. Table 1 includes the full suite of line ratios, colors, and masses for the 36 galaxies. ACS viz color-composite images of eight representative galaxies are shown in Figure 1. In general the sample includes disk dominated and clumpy morphologies without strong point sources, similar to the clumpy galaxies and AGN hosts of Bournaud et al. (2012).

Zoom In Zoom Out Reset image size

2.1. Keck/MOSFIRE

MOSFIRE observations were performed on 2012 September 14–15 and October 10, with three H-band masks in the GOODS-S field. All targets were observed in two dither positions within 07 slits, with total on-target exposure times of 2 hr each. The resulting wavelength coverage is 1.46 < λ < 1.81 μm with a spectral resolution of R ∼ 3200 (∼5 Å per resolution element). Spectra were reduced, sky subtracted, wavelength calibrated, and one-dimensional extracted using the public MOSFIRE data reduction pipeline.⁸ Redshifts were found using the Specpro⁹ software (Masters & Capak 2011). The wavelength calibration was very accurate, with shifts of Δz ≲ 0.001 from previous spectroscopic redshifts. The spectra were not flux calibrated: flux calibration is unnecessary for ratios of closely separated emission lines.

Figure 2 shows the sensitivity of MOSFIRE, for both emission lines and continuum measurements, in our 2 hr exposures. The left panel shows the emission-line signal-to-noise ratio (S/N) with line flux and SFR, using the Kennicutt (1998) relation to convert line flux at z = 1.5 to SFR. The right panel shows the continuum S/N (measured over a resolution element of 5 pixels) with F140W magnitude from 3D-HST observations. Continuum S/N is translated to the minimum rest-frame equivalent width (EW₀) for a 3σ line detection at z = 1.5, and F140W magnitude is converted to stellar mass using the relation log (M_*) = 21.3 − 0.5m_F140W (derived using the stellar masses from spectral energy distribution, SED, fitting in Section 2.3). In general, our sensitivity measurements agree well with the estimates provided on the MOSFIRE Web site.¹⁰

Zoom In Zoom Out Reset image size

2.2. HST/WFC3 G141 Slitless Grism

2.3. Ancillary Photometry

We compare the classical line ratio with the color-excitation and mass-excitation AGN/SF diagnostics, using emission line ratios measured from the spectra and photometry-derived rest-frame colors and stellar masses. The MOSFIRE H-band spectra provide [N ii], [S ii], and  Hα fluxes, while  Hβ and [O iii] are measured from the WFC3 grism spectra. Examples of the emission-line fitting are shown in Figure 3.

Zoom In Zoom Out Reset image size

For the high-resolution MOSFIRE spectra, a continuum is fit across the emission line regions by splining the 50 pixel smoothed continuum. The emission line intensities are then computed simply as the sum of the continuum-subtracted flux in each of the wavelength regions 6556 < λ < 6570 ( Hα), 6578 < λ < 6592 ([N ii]), and 6711 < λ < 6740 (both [S ii] lines).

Line flux measurements in the low-resolution WFC3 grism spectra are more difficult because the  Hβ and [O iii] lines are blended. For the grism spectra, we subtract a linear continuum over the wavelength region 4750 < λ < 5120 and then fit three Gaussians, each of which are restricted to be within 20 Å (∼1200 km s⁻¹) of the line centers (4861 Å, 4959 Å, and 5007 Å). The [O iii] λ5007 flux is computed as 3/4 of the sum of the blended [O iii] Gaussians (Storey & Zeippen 2000).

Errors in line fluxes and ratios are calculated by bootstrapping 10,000 realizations of the resampled data. These errors accurately quantify the difficulty in line fitting when there is significant contamination by sky lines in MOSFIRE or nearby objects in the WFC3 grism. We use 1σ upper limits for undetected emission lines.

Figure 4 presents the color-excitation and mass-excitation diagrams along with the traditional BPT ([O iii]/Hβ versus [N ii]/Hα and [O iii]/Hβ versus [S ii]/Hα) AGN/SF diagnostics for the z ∼ 1.5 galaxies. The solid lines give the AGN/SF separation line defined for each diagram (CEx: Yan et al. 2011; MEx: Juneau et al. 2011; BPT and VO87: Kewley et al. 2006), with AGNs having log([O iii]/Hβ) above this line and star-forming galaxies lying below. Each panel also includes a comparison sample of z < 0.3 galaxies from the Sloan Digital Sky Survey (SDSS; York et al. 2000) MPA-JHU value-added DR7 catalog.¹² These z < 0.3 galaxies have emission line and stellar mass measurements described by Tremonti et al. (2004) and Kauffmann et al. (2003b), with rest-frame magnitudes calculated using the kcorrect IDL software (Blanton & Roweis 2007). The z ∼ 1.5 galaxies are color-coded according to their Kewley et al. (2006) AGN/SF classification as follows.

• 1.  
  X-ray sources: filled red circles for AGNs and filled blue diamonds for galaxies (as classified by Xue et al. 2011).
• 2.  
  SF galaxy, blue crosses: log([O iii]/Hβ) < 1.3 + 0.61/[log([N ii]/Hα) − 0.05].
• 3.  
  Composite SF+AGN, open green diamonds: log([O iii]/Hβ) > 1.3 + 0.61/[log([N ii]/Hα) − 0.05] and log([O iii]/Hβ) < 1.19 + 0.61/[log([N ii]/Hα) − 0.47], from Kauffmann et al. (2003a).
• 4.  
  AGN, open red circles: log([O iii]/Hβ) > 1.19 + 0.61/[log([N ii]/Hα) − 0.47] and log([O iii]/Hβ) > 0.76 + 1.89 log([S ii]/Hα).
• 5.  
  LINER/Shock: log([O iii]/Hβ) > 1.19 + 0.61/[log([N ii]/Hα) − 0.47] and log([O iii]/Hβ) < 0.76 + 1.89 log([S ii]/Hα).

Zoom In Zoom Out Reset image size

None of the z ∼ 1.5 galaxies meet the LINER/Shock criteria, although the elevated [S ii]/Hα ratios of several z ∼ 1.5 galaxies classified as star forming (by their [N ii]/Hα and [O iii]/Hβ ratios) support the conclusion of Yan & Blanton (2012) that such systems are excited by shocks rather than AGN ionization.

Galaxies at z ∼ 1.5 have typically higher [O iii]/Hβ ratios than z < 0.3 galaxies, in agreement with previous z > 1 studies (e.g., Liu et al. 2008; Wright et al. 2010; Trump et al. 2011). However, this is not necessarily because z ∼ 1.5 galaxies are more likely to be AGNs: low-metallicity star-forming galaxies (in the upper left of each panel in Figure 4) are also more likely at z > 1. An elevated [O iii]/Hβ ratio is insufficient for AGN identification at z > 1.

Determining the effectiveness of each AGN/SF diagnostic requires knowledge of the "true" AGN population. The X-ray data provide an independent AGN/SF classification method. For the four X-ray-detected galaxies, the X-ray classifications of Xue et al. (2011) generally agree with the Kewley et al. (2006) classifications: the three X-ray AGNs have emission line ratios indicative of either AGNs or composite AGN+SF systems, while the X-ray galaxy has star-forming emission-line ratios. The classic AGN/SF line ratio classifications remain accurate for the four X-ray-detected galaxies at z ∼ 1.5.

The X-ray detection of galaxies in or near the "composite" region, with line ratios intermediate between AGN and SF, implies that these objects do in fact host accreting AGNs. These data are in contrast with Liu et al. (2008) and Brinchmann et al. (2008), who argued that AGN/SF composite galaxies at z > 1 are starbursts with unusual gas properties and no significant AGNs. Galaxies in the composite region may host starbursts, but the X-ray data imply that they also host genuine AGNs. Indeed, many studies suggest a connection between SF activity and AGN accretion (e.g., Kauffmann et al. 2003a; Trump et al. 2012). The analyses of Wright et al. (2010) and Trump et al. (2011), with spatially resolved emission line ratio gradients, provide additional evidence that z > 1 AGN/SF composite galaxies host nuclear activity.

The lack of an X-ray detection in the other 32 galaxies does not mean that they lack AGNs. X-ray surveys are less sensitive to moderate-luminosity AGNs in galaxies of lower stellar masses (Aird et al. 2012), and all three X-ray-detected AGNs have the highest stellar masses in our sample (log (M_*/M_☉) > 10.7). X-ray surveys are also insensitive to heavily obscured AGNs. Because the Kewley et al. (2006) classifications match the X-ray data for the high-mass X-ray-detected galaxies, we conclude that these classifications are robust for z > 1 galaxies. Of the 32 X-ray undetected galaxies, those identified as AGNs by their line ratios are presumably undetected because they have low stellar masses or are X-ray obscured.

The total number of AGNs from the combined X-ray and line ratio classifications is 11/36, with an additional 10/36 AGN/SF composite galaxies. Thus, about 1/3 of our z > 1 emission-line galaxies are AGN dominated, and nearly ∼2/3 have some AGN component. This is significantly higher than the ∼15% AGNs and composite fraction at z ∼ 0 (Kauffmann et al. 2003a), although this may be partly driven by our emission-line selection.

The mass-excitation method remains generally consistent with the BPT/VO87 and X-ray classifications at z ∼ 1.5. Juneau et al. (2011) emphasize that a probabilistic approach is more meaningful with the MEx method than simply classifying individual galaxies by the AGN/SF dividing line. Following Juneau et al. (2011), the MEx probability analysis indicates that ∼18/36 galaxies are AGNs, matching the high AGN fraction of the X-ray and BPT/VO87 classifications.

Meanwhile, the color-excitation method, as calibrated at z ∼ 0, identifies only 7 of the 12 AGNs classified by X-rays or line ratios. This is because a z ∼ 1.5 galaxy has significantly higher SFR and lower metallicity and is consequently bluer than a z ∼ 0 galaxy of the same mass (Erb et al. 2006; Mannucci et al. 2010). While the change in metallicity is apparent in the increasing [O iii]/Hβ ratio, the change in SFR is not. Shifting the original AGN/SF division of Yan et al. (2011) by 0.2 mag provides a much more effective separation between AGNs from SF galaxies at z ∼ 1.5. Our data suggest the following recalibrated color-excitation AGN/SF demarcation at z > 1:

$$\hbox{$\log (\hbox{[{\rm O}\,{\sc iii}]}/\hbox{{\rm H}$\beta $}) = {\rm max}[1.2-1.2(U-B)_0,-0.1].$}$$

We use some of the first Keck/MOSFIRE data to test various diagnostics of AGN/SF activity in 36 emission-line galaxies at z ∼ 1.5. Although only four sources are X-ray detected, their X-ray data suggest that the classic emission-line ratio diagnostic remains effective at z > 1, and that "composite" galaxies (of intermediate AGN/SF classification) do in fact host accreting AGNs. We find that nearly ∼2/3 of our z ∼ 1.5 emission-line galaxies have some AGN contribution detected by X-rays or line ratios. Among alternative AGN/SF diagnostics, the mass-excitation method remains consistent with X-ray and line ratio classification at z ∼ 1.5, but the color-excitation method does not. We suggest a recalibration for AGN/SF classification with the color-excitation diagnostic at z > 1. A larger set of galaxies and emission line measurements will allow a more detailed calibration and estimation of the AGN fraction at z > 1.

J.R.T. and the authors from UCSC acknowledge support from NASA HST grants GO-12060.10-A and AR-12822.03, Chandra grant G08-9129A, and NSF grant AST-0808133. N.P.K. acknowledges support from NSF grant AST-1106171. This work uses data from the 3-D HST Treasury Survey (Program ID 12177). We owe tremendous gratitude to the MOSFIRE commissioning team for development and support of a spectacular instrument. We also thank Greg Wirth, Marc Kassis, Jim Lyke, and the staff of Keck observatory for excellent support while observing. We wish to recognize and acknowledge the very significant cultural role that the summit of Mauna Kea has within the indigenous Hawaiian community: we are fortunate to have the opportunity to conduct observations from this mountain.