THE PROPERTIES OF POST-STARBURST QUASARS BASED ON OPTICAL SPECTROSCOPY

We investigate a sample of 38 objects spectroscopically selected from the Sloan Digital Sky Survey data release 3 (SDSS DR3; Abazajian et al. 2005) that meet the following criteria.

• 1.  
  Broad emission lines as defined by the SDSS DR3 online database (FWHM >1000 km s⁻¹).
• 2.  
  r model magnitudes ≲ 19.
• 3.  
  0.25 <z <0.45.
• 4.  
  S/N > 8 between rest wavelengths of 4150 and 4250 Å in the SDSS spectra.
• 5.  
  Summation of the Balmer absorption lines Hδ, Hζ, and Hη > 2 Å rest-frame equivalent width at a significance greater than 6σ.
• 6.  
  Hδ > 1 Å rest-frame equivalent width.
• 7.  
  Balmer break >0.9; based on the ratio of the fluxes at two 100 Å wide regions centered at rest wavelengths 4035 and 3790 Å.

The above criteria ensure that all objects in the sample are luminous AGNs with clear post-starburst stellar populations. We call these objects PSQs regardless of whether their AGN component alone exceeds a formal luminosity separating quasar from Seyfert galaxy. Table 1 characterizes some observational properties of the sample. C11 classified morphologies and measured quasar-to-host light contributions based on HST/ACS imaging of a subsample of 29 of these objects. The following sections provide details of our additional ground-based spectroscopy which we obtained in order to improve signal-to-noise ratio (S/N) and wavelength coverage compared to the SDSS spectra.

We obtained long-slit spectra of 12 PSQs and their companions using the Kitt Peak National Observatory (KPNO) 4 m Mayall telescope on the nights of 2006 May 20–23. We used the R-C spectrograph along with the 312 groove mm⁻¹ KPC-10A grating, in first order, producing a dispersion of 2.75 Å pixel⁻¹ and a resolution of 6.9 Å FWHM on the TK2B 2048 × 2048 CCD detector. The wavelength range covered is 3200 to 7200 Å. The detector read noise and gain are 4 e⁻ and 1.9 e⁻ ADU⁻¹, respectively. The long slit is covered at a scale of 069 pixel⁻¹; the slit width was 300 μm or 2''.

We applied a standard observation strategy with the exception of rotating the slit in order to observe the primary PSQ and its nearest bright companion within 20'' (to be presented in a future publication). We obtained three 1200 s exposures for each target PSQ. There were three instances where we observed the same target twice, changing the slit angle in order to observe a different companion (i.e., SDSS J115159.59+673604.8, SDSS J123043.41+614821.8, and SDSS J164444.92+423304.5). In these cases, we obtained six 1200 s exposures of the primary PSQ. For one source (SDSS J170819.80+603759.4), two of the three 1200 s exposures had low S/N due to poor conditions, and we use the data from a single exposure in our analysis.

Reductions of the spectra were carried out in a standard manner using the IRAF¹¹ software. After initial trimming, bias removal, and flat fielding, we extracted the one-dimensional cleaned spectrum using apall. We found the dispersion solution by identifying the lines of an He–Ne–Ar lamp. The spectra were flux calibrated using observations of several spectrophotometric standard stars (Massey et al. 1988). We median combined the individual spectrum with S/N weights using scombine. The combinations also include the SDSS spectrum in order to increase the S/N and the wavelength coverage of the resulting spectrum. Additional photometric corrections and combinations with other spectra were performed in a similar fashion as the Keck spectra and are described in Section 2.4.

Spectroscopic observations were carried out on 2005 November 1 and 2 with the Low-Resolution Imaging Spectrometer (LRIS; Oke et al. 1995) on the Keck I telescope. For the blue side (LRIS-B), we used the 600 groove mm⁻¹ grism blazed at 4000 Å, yielding a dispersion of 0.63 Å pixel⁻¹. For the red side (LRIS-R), we used the 400 groove mm⁻¹ grating blazed at 8500 Å, yielding a dispersion of 1.86 Å pixel⁻¹. The slit was 1'' wide, projecting to ∼7 pixels on the UV and blue-optimized CCD on LRIS-B and ∼5 pixels on the Tektronix 2048 × 2048 CCD on LRIS-R.

We obtained between one or two exposures for each object, typically 1200 s each, dithering along the slit between exposures. The typical seeing for all observations was 07 in V. Two or three spectrophotometric standards from Massey et al. (1988) were observed at parallactic angle each night for flux calibration.

The position angles were selected so as to include nearby companion galaxies. The majority of objects were observed at low airmass to minimize the effects of differential atmospheric refraction. The only object that was not observed near transit was SDSS J040210.90−054630.3, which was observed at an airmass of 1.43.

The spectra were reduced with IRAF, using standard reduction procedures. After subtracting bias, dividing by a normalized halogen lamp flat-field frame and removing sky lines, we rectified the two-dimensional spectra and placed them on a wavelength scale using the least-mean-squares fit of cubic spline segments to identified lines in an Hg–Ne–Cd–Zn lamp. We calibrated the spectra using the spectrophotometric standards from Massey et al. (1988). The distortions in the spatial coordinate were removed with the IRAF apextract routines. For each slit position, we had two or three individual frames; we averaged the spatially corrected spectra using the IRAF task scombine.

Since we chose to orient the slit to also observe neighboring objects we did not observe at parallactic angle. We used the SDSS spectra to correct for atmospheric diffraction slit losses. First, we divide our spectra by the corresponding SDSS spectra and fit the result by a low-order Legendre polynomial. We then corrected our spectra by dividing by the fitted polynomial. The resulting corrected spectra were well matched to the flux calibration of the SDSS spectra.

We corrected for Galactic extinction using the Schlegel et al. (1998) maps and the IRAF task deredden which utilizes the Cardelli et al. (1989) extinction curves. We converted the spectra to rest frame using the IRAF dopcor task using SDSS redshifts. We note that we retained observed fluxes.

We have used a single uniform method to estimate the noise for each spectrum. For each pixel, we compute the root-mean-square value of the difference between the signal and the average using nine pixels centered on the input pixel. We use this estimated noise spectrum to weight individual points in the fitting procedure we describe below.

While our spectral fits are good in general, our model assumptions are too simple in some respects and their effects on our results should be considered more closely. In particular, metallicity can vary between galaxies and within galaxies. Dust, often associated with starbursts and sometimes AGNs, can also be present along the line of sight and affect model parameters. Finally, the original starburst event we model as an instantaneous burst certainly takes place in the environment of one or more galaxies with existing, older stellar populations. Ongoing star formation may be present as well. We briefly discuss these points below. However, each of these may represent more complex problems requiring more and higher S/N data to be more fully explored in future investigations.

Metallicity affects the spectra of stars. If we assumed a metallicity higher than solar we would fit a stellar population that is younger and less massive. Conversely, assuming a lower metallicity would result in our fitting a stellar population that is older and more massive. In the case of the PSQ prototype UN J1025−0040, the uncertainty in the age of the 400 Myr old stellar population was ±50 Myr assuming a reasonable range in metallicity (Brotherton et al. 1999). Their methodology is similar to ours and we expect similar uncertainties.

While our fits do not require reddening, dust may be present along the line of sight to the AGN and/or stellar component of our PSQs. Dust reddening makes a stellar population appear less luminous and older. Therefore, if significant dust is present the true stellar population age would be younger and the mass larger than our reported measurements. For a visual extinction of 0.1 mag a 422 Myr population appears both older and less luminous by ∼10%. The differences are less for older populations. The dust reddening laws toward AGNs and starbursts typically differ further complicating more detailed modeling (Calzetti et al. 1994; Richards et al. 2003).

Our selection criteria ensure that an intermediate-age population is present and strong, but we also know that a more complex stellar population is likely present that includes both older and younger stars. We know that a few of the PSQs have morphologies that show knots of ongoing star formation (C11), suggesting that young stellar populations may be present. A similarly simple model, AGN+instantaneous burst, of the prototype PSQ, UN J1025-0040 (Brotherton et al. 1999) required revising to include a younger stellar population when high-resolution blue Hubble Space Telescope (HST) images were obtained (Brotherton et al. 2002). Because younger stellar populations have relatively weak spectral signatures and are difficult to identify with spectra alone in the presence of post-starburst populations, we want to suggest that this issue be kept in mind when interpreting spectral fitting results.

The imaging also shows galaxies that appear to be spirals with bulges, which typically have older populations, and ellipticals that appear to be post-merger remnants. The dominant intermediate age populations are most likely created as part of the merging process, but the older remain present. Older populations do have spectral signatures that are distinct (e.g., Mg i b absorption, larger Ca ii H & K ratios relative to Balmer line absorption), so we have a better chance of identifying objects that require a more complex stellar population in our fitting. Mg i b absorption is not readily apparent in individual spectra, but we note that it is visible in composite PSQ spectra we have examined.

We performed an additional set of fits in which we also included an older stellar population with an age of 5.6 Gyr. This additional component did not result in improved fits (as evaluated by χ² and examining by eye the absorption lines in detail). This could be the result of degeneracies inherent in fitting many components to a complex spectrum, something worth additional study, or because the older components do not contribute significantly to the observed flux. This does not mean that they are not present, only that the fits may be degenerate, or that the older components are weak. Others have investigated the issues involving fitting multiple-aged stellar components to a spectrum with a single-aged stellar synthesis model, e.g., Serra & Trager 2007; Trager & Somerville 2009, who have characterized some of the resulting biases and limitations.

We quantitatively investigated this effect ourselves with our own techniques by simulating spectra of a post-starburst plus old stellar component for a range of flux/mass ratios. We created our simulated spectra by combining ratios of 422 Myr old and 5.6 Gyr old Charlot & Bruzual instantaneous burst models and adding artificial noise consistent with our data. We then fit the resulting simulated spectra with a single-aged stellar component. The intermediate stellar age and mass were recovered to better than 10% as long as the older population was less than 70% of the stellar mass. At higher mass fractions, the intermediate age population fit was compromised and skewed the results to indicate older, more massive populations. The AGN component additionally greatly complicates the task and can mask the presence of different-aged stellar components. We conclude that our results are robust as long as the post-starburst population is massive enough to be dominant in flux, as it is at least in some cases (e.g., Hiner et al. 2012), but that higher S/N spectra (letting us measure the Mg i b feature more reliably) or another waveband of high-resolution imaging (e.g., Brotherton et al. 2002) will be required to definitively resolve this issue.

We additionally note that in our simulations, when a fit overestimates a post-starburst stellar population age, the equivalent widths of the Balmer lines are greatly overestimated (at least ∼30%), although as a complication the AGN component can dilute the equivalent widths. In our single intermediate-age population fits, the Balmer lines of the data are well characterized by our model and for only one object (SDSS J023700.30−010130.5) we note that there might be a significant overestimation. The good fits overall and the fact that we have used a single, consistent approach to fitting all objects suggests that even in the presence of some systematic issues, relative measurements are still likely meaningful and correlation analysis is of interest.

The issue of how to robustly fit complex spectra involving multiple stellar components of unknown metallicity, an AGN, and unknown and perhaps complex dust reddening, at a range of S/Ns, is complicated and deserving of additional deeper investigation and is beyond the scope of the present work. Our existing data are well and consistently fit by our simple model, and to investigate these more sophisticated models would require additional data. Even with more data, degeneracies may make it extremely difficult to find unique and perfect solutions. Again, our fits appear good and likely provide useful information about the PSQs, but keep in mind these caveats and how they may bias the results when interpreting our measurements.

From the fitting results, we were able to estimate two fundamental physical AGN properties: the BH mass and Eddington ratio. We employed scaling relations that extend the reverberation mapping results of physical properties (i.e., M_BH and radius of the broad-line region) to the observable quantities (i.e., continuum luminosity and broad line FWHM) of single-epoch data to calculate BH mass. We used Mg ii, Hβ, Hα as well as estimated values of Hβ based on the measurements of Hα, being careful to discard data of low quality. All of our spectra have coverage at 5100 Å. We used the 5100 Å monochromatic luminosities based on the power-law components used in the fits as our AGN luminosities for the scaling relations below.

From the relation given by Vestergaard & Osmer (2009), we estimated M_BH based on the Mg ii broad line FWHM and the 5100 Å monochromatic luminosity:

$$\begin{equation} \frac{M_{{\rm BH}}\ (\hbox{Mg\,{\sc {ii}}})}{{M}_{\odot }} = 10^{6.96}\left[\frac{\hbox{FWHM Mg\,{\sc {ii}}}}{1000\ \hbox{km s}^{-1}}\right]^2\left[\frac{\lambda L_{\lambda }(5100\ \hbox{\AA})}{10^{44}\ \hbox{erg s}^{-1}}\right]^{0.5}. \end{equation} \tag{ 1 }$$

The scatter in the zero point of the relation is given to be 0.55 dex.

We used the relation

$$\begin{equation} \frac{M_{{\rm BH}}\ (\hbox{H}{\beta })}{M_{\odot }} = 10^{6.91\pm 0.02} \left[ \frac{\hbox{FWHM H}{\beta }}{1000\ \hbox{km s}^{-1}} \right]^2 \left[ \frac{\lambda L_{\lambda }(5100\ \hbox{\AA})}{10^{44}\ \hbox{erg s}^{-1}} \right]^{0.5} \end{equation} \tag{ 2 }$$

to calculate the BH mass based on the Hβ broad line FWHM and the 5100 Å AGN monochromatic luminosity (Vestergaard & Peterson 2006). The intrinsic scatter in the sample is described by 0.43 dex. We have given the relation in linear space as opposed to log space for ease of reading. We note that the relations given for Mg ii and Hβ are on the same mass scale.

We estimated the BH masses using Hα broad line FWHM and the 5100 Å AGN monochromatic luminosity according to the Greene et al. (2010) scaling relation:

$$\begin{eqnarray} \frac{M_{{\rm BH}}\ (\hbox{H}{\alpha })}{{M}_{\odot }} &=& (9.7 \pm 0.5) \times 10^6 \left[\frac{\hbox{FWHM H}{\alpha }}{1000\ \hbox{km s}^{-1}}\right]^{2.06 \pm 0.06} \nonumber\\ &&\times\left[\frac{\lambda L_{\lambda }(5100\ \hbox{\AA})}{10^{44}\ \hbox{erg s}^{-1}}\right]^{0.519 \pm 0.07}. \end{eqnarray} \tag{ 3 }$$

Though the scatter in this relation is not reported, typical values are ∼0.4–0.5 dex.

Since stellar absorption from the post-starburst population contaminates broad emission in Hβ, making the fits less reliable, and this stellar contamination is negligible for Hα we used Hα as a proxy for Hβ when Hα is present in the spectrum (Shen et al. 2011)

$$\begin{eqnarray} \log \left(\frac{\hbox{FWHM H}{\beta }}{\hbox{km s}^{-1}} \right) &=& (-0.11 \pm 0.03)\nonumber\\ && + (1.05 \pm 0.01) \times \log \left(\frac{\hbox{FWHM H}{\alpha }}{\hbox{km s}^{-1}} \right).\nonumber\\ \end{eqnarray} \tag{ 4 }$$

We then used the estimated Hβ values in conjunction with Equation (2) to calculate BH measurements for Hβ based on Hα. There are slight differences between the two formalisms of Greene et al. (2010) and Vestergaard & Peterson (2006); most notably between the prefactors and radius–luminosity (scaling of λL_λ) relations. However, the differences are small; for our sample the difference is less than 8%, much less than the intrinsic scatter.

We have given several measurements of the BH mass. Each line has its own particular issue. The blue side of the spectra suffer from S/N degradation which sometimes gives unreliable measurements for Mg ii. Occasionally, up to ∼50% of the Hβ broad emission can be contaminated by stellar absorption. Due to the redshift range of our sample, sometimes Hα is out of our observing window. In subsequent sections, we make comparisons using these values individually. However, we also give an adopted M_BH by applying the following prescription. When we have three reliable measurements we adopt the median of these values. For two good measurements, we give the mean of the values. If there is only one reliable value, we adopt this as our M_BH. There is one object (SDSS J020258.94−002807.5) for which Hα was not covered and both Mg ii and Hβ were unreliable.

We estimated the AGN bolometric luminosity using the measured rest-frame flux at 5100 Å from the fit to the power-law continuum and convert it into a total emitted luminosity using luminosity distances for our cosmology and an appropriate bolometric correction (f = 8.1; Runnoe et al. 2012):

$$\begin{equation} L_{{\rm bol}} = 4\pi D_L^2 f (1+z) \lambda F_{\lambda }(5100\ \hbox{\AA}). \end{equation} \tag{ 5 }$$

The Eddington ratio is given by L_bol/L_Edd where we use L_Edd = 1.51 × 10³⁸(M_BH/M_☉) erg s⁻¹ (Krolik 1999). Table 6 gives BH masses for each broad line measurement and their Eddington ratios.

Figure 2 shows where the different broad line width measurements (i.e., Mg ii, Hβ, Hα, and predicted Hβ) lie as a function of λL_λ(5100 Å). For reference we indicate lines of constant M_BH and L_bol/L_Edd based on Mg ii. Constant lines of M_BH and L_bol/L_Edd based on Hβ and Hα will vary slightly in slope and intercept.

Zoom In Zoom Out Reset image size

The post-starburst stellar populations can be described using two fundamental physical parameters, starburst age and mass. The post-starburst ages are known directly from the fitting results. The scale factor is a function of the mass and age of the starburst. The inputted PSQ spectra are in units of 10⁻¹⁷ erg s⁻¹ cm² Å⁻¹ while the template spectra are in units of L_☉/Å and scaled up by their age in Myr. Thus, we can derive the total starburst mass by using the scale factor and the outputted age:

$$\begin{eqnarray} \frac{M_{{\rm SB}}}{{M}_{\odot }}&=& \hbox{Raw scale} \times \hbox{Age (Myr)} \times 10^{-17} \hbox{erg s$^{-1}$cm$^{2}$\AA$^{-1}$}\nonumber\\ && \times \left[\frac{ 3.826 \times 10^{33} \hbox{erg s$^{-1}$\AA$^{-1}$} }{4\pi D_L^2 (1+z)}\right]^{-1}. \end{eqnarray} \tag{ 6 }$$

The expression in brackets is the solar flux at the luminosity distance, D_L. Thus, the ratio between the scaled model and the solar flux at D_L gives the starburst mass. Table 5 gives the starburst fitting results, derived starburst masses, and integrated luminosities.

We calculate Spearman-rank correlation coefficient matrices involving several hundred correlation tests between fitted and derived parameters for (1) the total sample, (2) the early-type and spiral morphological classifications, and (3) the disturbed and undisturbed classifications. We compute the probability of the correlations arising by chance and list those less than 1% in Table 7.

We find a number of strong correlations among quasar and starburst properties; however, many of these are due to selection effects which we will discuss below. Some merger induced evolutionary scenarios predict correlations among these properties of a more physical origin (Di Matteo et al. 2005). We find very few additional correlations of this type.

Our selection criteria provides an efficient means for selecting the brightest systems with both AGNs and post-starburst populations. However, this then creates some artificial correlations between parameters. For example, the starburst and AGN must be comparable in luminosity in order for one not to swamp out light from the other, thus giving rise to a significant correlation of the luminosities of both. We note that the mean AGN-to-total light, f_nuc, is 0.55 with only a standard deviation of 0.13. This leaves us with a limited parameter space to explore.

4.1.1. Derived AGN Properties

Figure 3 shows our strongest correlation among AGN properties. The M_BH increases as L/L_Edd decreases, which probably arises as a result of two effects. The first effect is the dearth in massive BHs with high accretion rates at this redshift, which is the result of cosmic downsizing (Heckman et al. 2004). The second cause is the lack of lower mass black holes with low accretion rates which fail to make the luminosity cut. The combination of these two effects, demographics plus selection effects, leads to a strong inverse correlation. However, this does not rule out the possibility that there is an underlying physical correlation, but confirming this would require a more sophisticated sample selection.

Zoom In Zoom Out Reset image size

4.1.2. Narrow-line Ratios

Narrow emission lines can be powered by several sources of photoionizing radiation, including young O and B stars and the central AGN. Comparing the relative strengths between lines with high- and low-ionization potentials tells us about the shape of the photoionizing continuum and thus diagnoses the source. AGNs have harder continua (relatively more high-energy photons) while star-forming continua are softer (relatively more low-energy photons).

We find a number of strong correlations among narrow-line ratios that can be interpreted in terms of the relative contributions of AGNs and star formation. In particular, the highly ionized [O iii] λ5007 emission line is more prominent in AGN spectra but other lines may also be diagnostic. For example, Figure 4 shows the traditional Baldwin et al. (1981, hereafter BPT) diagram, [O iii]/Hβ versus [N ii]/Hα flux ratios. Diagrams such as this one help us understand the nature of relative contributions of ionizing radiation from different sources. PSQs fall along and above nearly the full extent of the mixing line, which indicates the presence of an AGN but a wide range in the relative contribution of star formation.

Zoom In Zoom Out Reset image size

4.1.3. Starburst Properties

The strongest correlation among starburst properties is between the age and mass of the post-starburst population (Figure 5). Just as in the case of AGN properties, Section 4.1.1, this correlation likely arises from a combination of demographics and selection effects. There is a lack of objects at large mass and young age. Such objects must exist and are likely dust enshrouded, seen in the infrared as luminous infrared galaxies (LIRGs; Sanders & Mirabel 1996) or may be found as post-starburst galaxies if there is a significant time delay before the onset of AGN activity (see, e.g., Wild et al. 2010; Schawinski et al. 2009). The envelope in the upper left is the result of a selection effect: our luminosity limit.

Zoom In Zoom Out Reset image size

4.1.4. AGN versus Starburst Properties

We investigate the relationships among the morphology and fundamental (fitted and derived) properties of the AGN, narrow-line emission, and starburst. Visual classifications were performed by three of the authors and the results are presented in Cales et al. (2011). Generally, arms and bars distinguish spiral galaxies while smooth, somewhat featureless (i.e., lacking arms/bars), light distributions identify early-type galaxies. We note that the early-type hosts are allowed to show tidal features although sometimes disturbances make classification difficult, resulting in an indeterminate classification. In Table 8, we list the mean, standard deviation, and number of objects for our basic properties for the total sample, and early-type and spiral (including "probable" spirals) subpopulations. When reliable statistical differences in sample means exist at P the Values below the 1% level (better than 2.58σ) we include the nonparametric Mann–Whitney statistic and its associated P-Value and give their distributions in Figure 6. For reference, the mean, standard deviation, and number of objects for the total sample are also given.

Zoom In Zoom Out Reset image size

Due to the small number of objects in the subpopulations and the lack of significant differences below the 1% level between the disturbed and undisturbed populations, we do not present population statistics for classifications of this type.

The significant differences for fundamental parameters within the morphology subclasses are as follows.

• 1.  
  AGNs. The PSQs hosted in early-type galaxies have statistically significant higher total luminosities than spiral-hosted PSQs, and this appears to be somewhat more driven by differences in AGN luminosity although starburst luminosities also differ. This is supported by various methods of measuring the PSQ luminosity and its components (i.e., L_Tot, L_AGN, and L_SB; see Figures 6(a)–(c)). We note that AGN luminosity increases with BH mass and/or accretion rate. While there are no statistically significant differences between the subpopulations for Eddington fraction, BH mass, or starburst mass, it is possible that the AGN with early-type hosts tend to have higher luminosities because they have higher Eddington ratios (Figures 6(d)–(f)).
• 2.  
  NLR. Figure 6(e) shows that the PSQs hosted in early-type galaxies have larger [O iii]/Hβ than spiral-hosted PSQs, and we interpret this as differences in the hardness of the ionizing continuum reflecting the relative contributions of the AGN and star formation. The spiral PSQs have more ongoing star formation.
• 3.  
  Starburst. The PSQs hosted in early-type galaxies tend to have younger starburst ages from our fitting results than the PSQs hosted by spirals (Figure 6(f)). This may be interpreted in different ways. Some possible explanations of this are: (1) the starburst luminosities are smaller in the spiral-hosted PSQs and our single age fits may be skewed by the presence of an older stellar population, (2) chance snapshot of unassociated starburst and AGN activity in spiral hosts with large starburst ages, or (3) longer timescales between triggering of the AGN and starburst in spiral PSQs. There may be no clear interpretation of this effect without additional data.

PSQs appear to display both merger-driven and secular forms of galaxy evolution. They are a heterogeneous population hosted by both early-type and spiral galaxies, some of which appear to be results of major mergers while others exist in isolated systems. C11 postulates that the early-type PSQs may be the low-z analogs of luminous merger-induced z ∼ 2 quasars. Furthermore, at least some of the spiral PSQs are likely undergoing secular evolution. This conclusion is consistent with current theoretical frameworks which argue for two fundamentally different fueling mechanisms (i.e., merger induced versus secular) responsible for mutual SMBH–bulge growth at the characteristic dividing line of quasar-Seyfert luminosities (Hopkins & Hernquist 2009).

We explored issues of the relative contributions of AGNs and star formation to powering narrow-emission lines previously in the context of the BPT diagram. We have also found differences between a number of properties and morphological subtype. We can combine this information to create observational selection criteria to select PSQs as a function of galaxy host type without requiring high spatial resolution imaging. With the exception of a rare outlier, we can use a combination of Hβ/[O iii], [O iii]/[O ii], and L_Tot to distinguish host galaxy type. Figure 7 shows several plots involving these quantities and where the objects of different classification fall. Using these we establish the following parameters to select PSQs by morphological subtype:

• 1.  
  Early-type: L_Tot > 10^43.85 erg s⁻¹ + Hβ/[O iii] < 0.4;
• 2.  
  Spiral: L_Tot < 10^43.95 erg s⁻¹ + [O iii]/[O ii] < 2.0.

While it has been suggested that major-mergers are not important drivers of AGN activity in the local downsized universe (Cisternas et al. 2011), such objects do exist primarily among quasars and may be valuable to study as analogs of luminous high-redshift quasars (Canalizo et al. 2007; Bennert et al. 2008, 2010). Studies of local AGNs (z ∼ 0.05) have also found two modes of growth distinguished by morphology. Schawinski et al. (2010b) find that the least massive SMBH population of early-type hosts is growing while the most massive SMBH population of late-type hosts is growing. Furthermore, Schawinski et al. (2010a) trace a correlation between the merger fraction of early-type hosts and an evolutionary sequence involving starbursts and AGNs, such that, even in the local universe, early-type galaxies are still merger induced. This has important ramifications for future studies involving galaxy evolution and should guide the path of such studies.