Host galaxies of luminous quasars: population synthesis of optical off-axis spectra
aa r X i v : . [ a s t r o - ph . C O ] J un Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 1 November 2018 (MN L A TEX style file v2.2)
Host galaxies of luminous quasars: population synthesis of opticaloff-axis spectra
I. Wold, A. I. Sheinis, M. J. Wolf, and E. J. Hooper Dept of Astronomy, University of Wisconsin-Madison, 475 N. Charter St., Madison, WI 53706, USA
Accepted ... ; Received ... ; in original form ...
ABSTRACT
There is increasing evidence of a connection between AGN activity and galaxy evolution.To obtain further insight into this potentially important evolutionary phase, we analyse theproperties of quasar host galaxies. In this paper, we present a population synthesis model-ing technique for off-axis spectra, the results of which constrain host colour and the stellarages of luminous quasars ( M V ( nuc ) < − ). Our technique is similar to well establishedquiescent-galaxy models, modified to accommodate scattered nuclear light (a combinationof atmospheric, instrumental and host galaxy scattered light) observed off axis. In our model,subtraction of residual scattered quasar light is performed, while simultaneously modeling theconstituent stellar populations of the host galaxy. The reliability of this technique is tested viaa Monte-Carlo routine in which the correspondence between synthetic spectra with known pa-rameters and the model output is determined. Application of this model to a preliminary sam-ple of 10 objects is presented and compared to previous studies. Spectroscopic data was ob-tained via long-slit and integral-field unit observations on the Keck and WIYN telescopes. Weconfirm that elliptical quasar hosts are distinguishable (bluer) from inactive ellipticals in restframe B-V colour. Additionally, we note a trend for radio luminous ( L GHz & erg s − )quasars to be located in redder host galaxies in comparison to their less luminous radio coun-terparts. While the host colour and age of our radio luminous sample is in close proximityto the green valley, our radio faint sample is consistent with quiescent star-forming galaxies.However, further observations are needed to confirm these results. Finally, we discuss futureapplications for our technique on a larger sample of objects being obtained via SALT andWIYN telescope observing campaigns. Key words: galaxies:active – galaxies:evolution – galaxies:formation – quasars:general
It is well established that active galactic nucleus (AGN) activity isdue to accretion onto super massive black holes (SMBHs). Addi-tionally, SMBHs are found to be common, existing in most, if notall, massive galaxies. Observationally, the SMBH mass is found tobe tightly correlated to the bulge velocity dispersion raised to thefourth power (Gebhardt et al. 2000). However, the bulge extendswell outside the gravitational influence of the SMBH. This correla-tion has led to the hypothesis that the bulge and SMBH co-evolve(Kormendy 2000). Additionally, models of galaxy evolution mustinclude regulation mechanisms to adjust and quench star formationin massive galaxies after early times to prevent the over productionof bright galaxies (Benson et al. 2003).Evidence pointing toward the co-evolution of SMBHs / bulgesand the need for regulation mechanisms has led to models whichutilize AGN feedback to quench star formation. An example of amodel of relevance to our studies is the hydrodynamic simulationsof Hopkins et al. (2006). In this model, gas rich mergers inducegas inflow triggering both star formation and quasar activity. Some small percentage of AGN energy output is converted to thermalenergy, assisting in the stoppage of further star formation and ac-cretion. Simulations, which feature self-regulated SMBH growth,have been successful in reproducing various observables, such asthe bimodal colour-magnitude distribution (Cattaneo et al. 2006).However these models must operate on a very large range of scales( µpc - Mpc ) and must incorporate many physical processes. There-fore it is desirable to further constrain these models observationally.The goal of this work is to develop an off-axis method to providestellar age and host colour constraints during the quasar phase.Unfortunately, studies of the underlying stellar populationsare hampered by the overwhelming emission from the cen-tral quasar. Consequentially, progress has been slow and, insome cases controversial, in the struggle to place observationalconstraints on quasar host galaxies. One group of collabora-tors (McLure et al. 1999; Hughes et al. 2000; Nolan et al. 2001;Dunlop et al. 2003) believes that local quasar ( M V ( nuc ) < − . ) hosts are predominantly normal massive elliptical galax-ies. However, Hubble Space Telescope ( HST ) morphology(Hooper, Impey & Foltz 1997; Bennert et al. 2008), multicolour c (cid:13) Wold et al.
Object name redshift M V ( nuc ) M V ( host ) log ( L GHz ) erg s − HostMorphology3C 273 0.1583 -26.7 -23.2 44.1 E4C 31.63 0.2950 -25.1 -25.1 43.3 EPKS 1302-102 0.2784 -25.9 -22.9 43.0 EPKS 0736+017 0.191 -23.2 -22.6 43.0 EPKS 2135-147 0.2003 -24.7 -22.4 42.9 EPKS 2349-014 0.1740 -24.5 -23.2 42.5 EPG 1309+355 0.1840 -24.4 -22.8 41.3 SPHL 909 0.171 -24.1 -22.2 40.0 EPG 0052+251 0.1550 -24.1 -22.5 39.4 SPG 1444+407 0.2673 -25.3 -22.7 39.2 S
Table 1.
Test sample properties. M V ( nuc ) is the V-band absolute magnitude of the quasar. M V ( host ) is the V-band absolute magnitude of the host galaxies.Both of these quantities are derived from Bahcall et al. (1997) when possible, see Wolf & Sheinis (2008) for details. Radio luminosity is estimated utilizingthe NASA/IPAC Extragalactic Database and assuming spectral index of -0.5. Host morphology is obtained via Hamilton, Casertano & Turnshek (2002). ’E’denotes elliptical and ’S’ denotes spiral. imaging (Jahnke, Kuhlbrodt & Wisotzki 2004), and spectral stud-ies (Canalizo & Stockton 2001; Miller & Sheinis 2003) have foundevidence inconsistent with a population of normal relaxed ellipti-cals. In fact, an on-axis spectroscopic study probing the inner re-gion of nearby quasar hosts found young Sc-like stellar populationsin half of their sample (Letawe et al. 2007). The discrepancies inthese studies may be due to relatively small sample sizes, system-atics, and the different radii probed (Lacy 2006). Additional studyis required to develop a clear understanding of stellar properties ofnearby luminous quasar hosts.In Wolf & Sheinis (2008) we presented velocity dispersionmeasurements of a sample of nearby luminous quasar host galax-ies. Future work will examine the relation between SMBH massand velocity dispersion (Sheinis et al. 2010 in preparation). In thispaper, we develop a method to measure the stellar age and colourof nearby quasar host galaxies at off-axis radii of 9 to 15 kpc. Weobserve off-axis spectra to minimize the observed quasar emission,thereby maximizing signal-to-noise. However even in off-axis ob-servations, described in §
2, scattered light from the central sourcecan still contribute significantly to the observed light. Our tech-nique of removing this scattered light is similar to previous stud-ies (Boroson, Persson & Oke 1985; Miller & Sheinis 2003) and de-scribed in detail in § § § §
4. A comparison to recent spectroscopic and photo-metric studies is found in §
5. Summary and conclusions are pre-sented in §
6. Appendix A contains comments on individual objects.Throughout this paper a cosmology of H = 70 km s − Mpc − , Ω M = 0 . , and Ω Λ = 0 . is adopted. Our ongoing spectroscopic observing campaign has resulted ina current sample of 28 nearby luminous quasars. This sampleconsists of objects previously imaged by Bahcall et al. (1997),with the addition of a few objects from Dunlop et al. (2003) andGuyon, Sanders & Stockton (2006). We demonstrate the capabili-ties of our spectral synthesis model on a subset of 10 objects. Theseobjects have spectral data that display sufficient signal-to-noise andstellar continuum to constrain the stellar properties via the method discussed below. Additionally, this subset of ten objects has knownstellar velocity dispersions (Wolf & Sheinis 2008), eliminating apotential free parameter of the model. This relatively small sam-ple of quasars is not representative of the local quasar population,which consists of ∼
10% radio loud quasars. Our sample containssix radio loud and four radio quiet objects. We define an object asradio loud by the criteria established by Kellermann et al. (1994), L GHz > W Hz − , which is roughly . erg s − for ouradopted cosmology . All quasars are local z < . and luminous M V < − . The properties of our test sample are summarized inTable 1. Data acquisition and reduction is described in detail in Sheinis(2002), Miller & Sheinis (2003), and Wolf & Sheinis (2008). Insummary, seven of the ten objects were observed via the LowResolution Imaging Spectrograph (Oke et al. 1994) on the Keck10-m telescope. The remaining three objects were observed us-ing a 82 fiber integral field unit (IFU), SparsePak (Bershady et al.2005) which feeds the Bench Spectrograph on the 3.5-m WIYNtelescope. Short on-axis observations, typically 1-2 minutes forKeck and 30 minutes for WIYN, and longer off-axis (2 ′′ -4.5 ′′ fromthe nucleus) observations, typically 0.5 - 4 hours, were obtainedfor each object. The approximate off-axis slit and fiber positionswith respect to archival imaging is displayed in Fig. 1. The prop-erties of the off-axis pointings are summarized in Table 2. Keckobjects have a rest wavelength range of ∼ ∆ λ ∼ ˚A. WIYN objects have a rest wave-length range of ∼ ∆ λ ∼ ˚A. All spectra were corrected for Galactic extinction using thelaw of Cardelli, Clayton & Mathis (1989) and the A V values fromSchlegel, Finkbeiner & Davis (1998) as listed in the NASA/IPACExtragalactic Database (NED). Kellermann et al. (1994) observationally determined the luminositythreshold assuming a cosmology of H = 70 kms − Mpc − , q = 1 / .We solve for the observed flux at z = 0 . and then convert to our adoptedcosmology. The specific luminosity is multiplied by the observed frequencyto obtain the quoted 5 GHz luminosity. (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis Figure 1.
Sample of observed quasar hosts with approximate off-axis slit and fiber positions indicated in red. The slit lengths are much larger than indicated,extending approximately 7 ′ projected on the sky. For WIYN objects, only the off-axis fibers used for the host galaxy spectrum are shown. Objects are listedfrom left to right, top to bottom by most radio bright to least. All images are from HST (Bahcall et al. 1997) (23 ′′ × ′′ ) with the exception of 4C 31.63(10 ′′ × ′′ ) (Guyon et al. 2006) and PKS 0736+017 (30 ′′ × ′′ ) (Hutchings, Johnson & Pyke 1988). Where indicated, the 10 kpc scale is computed for Ω = 1 . and H = 100 km s − Mpc − (Bahcall et al. 1997). For our adopted cosmology this scale corresponds to ∼ . kpc.Pointing R obs (kpc) S/N ˚A − (5500-5700 ˚A) Quasar Scattering(3600-5900 ˚A) Telescope3C 273 4N 11.79 18.6 68.8% Keck4C 31.63 2N 8.74 11.5 50.5% Keck4C 31.63 2.5E 10.92 9.4 34.6% Keck4C 31.63 3S 13.11 10.1 38.7% KeckPKS 1302-102 2.3N 12.79 9.3 68.4% KeckPKS 0736+017 4.5NW 14.18 8.3 49.5% WIYNPKS 2135-147 3W 12.47 12.3 70.8% KeckPKS 2349-014 4N 14.12 14.8 21.8% KeckPKS 2349-014 3S 9.47 18.8 25.2% KeckPG 1309+355 4.5SW 13.76 13.4 41.3% WIYNPHL 909 4.5N 12.97 11.7 77.8% WIYNPG 0052+251 3S 9.65 23.0 64.6% KeckPG 1444+407 3S 15.07 8.8 69.6% Keck Table 2.
Off-axis spectral properties. Pointing designation indicates object and approximate off-axis position in arcseconds. R obs is the observed off-axisradius, as reported in Wolf & Sheinis (2008). Additionally, off-axis signal-to-noise, fraction of the observed off-axis spectrum which consists of scatteredlight, and telescope are shown. The scattered quasar light percentage is determined by the model output. Our spectral synthesis code is a modified version of the model firstdiscussed in Miller & Sheinis (2003), with the underlying stellarspectrum now modeled as a weighted summation of 15 instanta-neous starbursts of various post burst ages. We model the off-axisspectrum, which consists of a stellar component and a scatteredquasar component. Quasar light is scattered into the off-axis line ofsight due to atmospheric seeing effects, as well as a small contribu-tion from the intrinsic optical point-spread function and astrophys-ically via dust and gas in the intervening line of sight. The stel-lar and scattered quasar components are modeled simultaneouslyto achieve the best χ fit to the observed spectrum. Each modelspectrum is masked in the same manner as the data for which it isintended (see § M λ is of the form: M λ = X i =1 x i ssp iλ ! r λ + Q λ ξ λ (1)with the following definitions: • ssp iλ is the stellar component of the model which consistsof a reduced basis of 15 simple stellar populations (SSPs) fromthe Bruzual & Charlot (2003), BC03, synthesis model, degradedto the instrument resolution, assuming solar metallicity, Padova-1994 models, and a Chabrier initial mass function (Chabrier 2003).Extensive simulations by CF05, have found this reduced spectralbase ( t i =0.001, 0.00316, 0.00501, 0.01, 0.02512, 0.04, 0.10152,0.28612, 0.64054, 0.90479, 1.434, 2.5, 5, 11 and 13 Gyr) to reliablyproduce observed spectral features while limiting redundancies inquiescent galaxies. Due to the increased complexity introduced bythe scattered light subtraction, only solar metallicity is considered.Each SSP is normalized such that the area under the unmasked c (cid:13) , 000–000 Wold et al.
Figure 2.
Parameter space search for minimum χ value. The three densely populated regions consist of 20 simulated annealing runs each. As the number ofcalls, or calculations of χ , advances the ’tempature’ is reduced to more restrictive values. These regions are separated by 10 straight downhill searches. Thisentire procedure is performed 20 times independently to determine the model’s output. portions of the spectrum is unity. Normalization at a single wave-length ( λ =4020 ˚A, as prescribed by CF05) was also investigated.For our sample, the alteration of normalization convention does notaffect the results of our model more than ∼ σ . x i is the populationweighting factor. • r λ ≡ − . A λ is the extinction law used to model the insitu reddening. A λ ( A V , R V ) is defined by the Galactic law ofCardelli et al. (1989) assuming the mean value of R V (3 . in thediffuse ISM. A single extinction is assumed for the entire stellarcomponent. • Q λ is the on-axis observed spectrum of the quasar normalizedsuch that the area under the unmasked portions of the spectrum isunity. ξ λ ≡ a + a λ + a λ + a λ is the scattering efficiencycurve, which modifies the observed quasar spectrum with the goalof modeling the scattered quasar light observed off axis. We haveshown that this model component is necessary to reliably removescattered quasar light (Sheinis 2002). The model assumes no hostgalaxy contamination to the on-axis spectrum. The dominance ofthe high luminosity quasar ( M V < − ) on-axis makes this as-sumption acceptable.Finally, the individual pieces are combined as prescribed byequation (1) and the resulting model spectrum, M λ , is normalized,dividing by a single constant, such that the area under theunmasked portions of the spectrum is unity (the normalized modelis denoted M λn ). Our model contains 20 free parameters: (1) an extinction parame-ter A V , (15) SSP weighting factors x i , (4) the scattering efficiency curve coefficients, a [1:4] . The range of values to explore with themodel is constrained in the case of x i to be 0.0 to 1.0, where Σ x i = 1 . . The range of values for A V and a [1:4] were found in aniterative fashion. First the model was applied to the entire datasetwith parameter ranges informed from the work of Sheinis (2002)and Miller & Sheinis (2003). The range was then increased as nec-essary to ensure the best χ fit was not determined by the search-able parameter space. For example, the highest model A V outputobtained is 0.4; therefore, the model samples A V from 0.0 to 1.0.We have tested the sensitivity of the model output to these bound-ary conditions by doubling the range and re-running the model onthe observed data. No significant change in results is found. Having constructed a model, we then devised a reliable means tofind the minimum chi-squared fit. The χ fitting function is definedby: χ = X λ (cid:18) O λn − M λn σ λ (cid:19) (2) O λn is the observed off-axis spectrum, normalized in the samemanner as M λn , and σ λ is the observed spectral noise. The noisespectrum is generated by measuring the standard deviation of O λn at three to four locations devoid of prominent spectral features.These measurements are then extrapolated across the wavelengthrange of O λn by a low order fitting polynomial.The starting point in the parameter search is defined in thefollowing manner. The initial stellar component is set to be 100%5Gyr SSP. The reddening is randomly selected, constrained by its c (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis Figure 3.
Monte-Carlo simulation results for ideal noiseless off-axis spectra with 60% scattered quasar light contamination. 65 synthetic spectra, S λ , areanalysed by the model to determine the reliability of the results. For each S λ , the known young, intermediate and old population percentage and the known h log t ⋆ i L and rest frame B-V input is compared to the model output as represented by 65 plus signs in each box. Ideally, all points would fall on the diagonal,denoted by a dashed line. The mean offset from the ideal and the standard deviation of the model output are displayed numerically in the upper left corner andgraphically by the blue strip. range. The scattering coefficients are also randomly selected butthen scaled so that scattered quasar light comprises 60% of theoff-axis model spectrum. Parameters a [1:4] are further constrainedby rejecting any solutions which drive the resulting off-axis quasarflux, Q λ ξ λ , below zero.Simulated annealing optimization (Press 1992) is utilized tosearch the 20-D parameter space. Simulated annealing evaluates themerit function, χ in this case, and travels downhill in merit spaceto find a minimum within predefined tolerances. To help avoid lo-cal minima, a user defined parameter, traditionally called ’tempera-ture’, is set to determine the probability of travelling uphill in meritspace. The higher the temperature the more probability of travelinguphill in the merit space. A zero value for temperature correspondsto a straight downhill search.The utilization of this optimization routine, described below,is somewhat of an art. The temperature value must be decreasedslowly so that the merit space can be adequately sampled, yet fastenough to arrive at a timely solution. After extensive experimen-tation the following implementation has been found to produce fa-vorable results. First, twenty iterations of simulated annealing areperformed with linearly decreasing temperature. If an iteration failsto find a minimum χ value within a factor of two of the overallminimum then the next iteration is reset to the overall minimum;otherwise the next iteration starts where the previous iteration leftoff. At the completion of the simulated annealing runs, the ten bestdistinct solutions are used as input for straight downhill (tempera-ture = 0) runs. The current global minimum is then used as the inputto the next twenty iterations of simulated annealing. This procedureis carried out for a total of three loops; within each loop the tem-perature parameter is set to more restrictive values. An example ofthis parameter space search is shown in Fig. 2 for a typical χ mini-mization. Overall approximately 150,000 points are sampled in theparameter space. This entire procedure is performed 20 times in-dependently. The best χ value of these runs, which consists ofapproximately × calculations of χ , is defined as the modeloutput. Guided by the analysis of CF05, we limit the scope of our project toa coarse, but well recovered, description of the stellar populations.Like CF05, we only attempt recovery of ‘young’ (t <
100 Myr),‘intermediate’ (100 Myr t > x Y , x I , and x O , respectively). Additionally, wereport the mean light-weighted stellar age, h log t ⋆ i L = N ⋆ =15 X i =1 x i log t i (3)and the rest frame B-V colour as derived directly from the listedBC03 SSP values. These quantities have been studied extensivelyin quiescent galaxy samples and to a lesser extent in AGN hostgalaxy samples. Thus, we may assess the consistency of ourresults to previous AGN host studies. Furthermore, we maycompare these properties of quasar host galaxies to quiescentgalaxy samples. Analysis of other quantities of interest (see CF05)are deferred to future studies. The error in our method is estimated via a Monte-Carlo routine.Synthetic spectra, S λ , are generated with known parameters andthen degraded by noise. The same procedure used for the observeddata is applied to determine the agreement between the known in-put and model output. In other words, S λn replaces O λn in thefitting routine. S λ are constructed in the same manner as M λ withthe exception of metallicity, SSP base and added noise. Because theobserved galaxies will likely include non-solar metallicities, eventhough the model grid does not, the metallicity is allowed to devi-ate from solar, and is randomly selected from Z = 0 . , and . Z ⊙ . The SSP base is extended to include all 221 BC03 instanta-neous bursts assuming a Chabrier initial mass function. A syntheticspectrum is of the form: S λ = × X j =1 x j ssp jλ ! r λ + Q λ ξ λ (4)Poison noise is added until the synthetic spectra obtain the cor-responding observed noise as measured in the 5500-5700 ˚A spec-tral window. The 5500-5700 ˚A spectral range was selected becausethere are no prominent emission or absorption features; addition-ally, all observed spectra include this range. The value of A V and a [1:4] are randomly selected, constrained by the applicable range.Parameters a [1:4] are further constrained by rejecting any solu-tions which drive the resulting off-axis quasar flux, Q λ ξ λ , belowzero. The scattered quasar flux is then scaled to the contamina-tion percentage of interest. The value of x j is randomly selectedto uniformly sample each stellar age bin (young, intermediate, and To uniformly sample each stellar age bin the following procedure is em-c (cid:13) , 000–000
Wold et al.
Figure 4.
Model fit to the 3C 273 4N pointing. The observed off-axis rest frame spectrum is highlighted in cyan, with masked regions indicated in black and byvertical lines extending to zero flux; masked regions that are narrow, [OIII] at 5007 ˚A for example, are most easily seen by looking at the residuals at the bottomof the plot. The model output is over-plotted in red. The residual is shown with diamonds. Model components are also shown. The scattered light component isindicated by the thin black line and comprises 68.8% of the total observed flux. The stellar component is also shown, broken up into young (blue), intermediate(green), and old (magenta) bins. In this example, only the old stellar component, which comprises 86.2% of the stellar flux, is clearly visible. Results for allother pointings are found in Appendix A, Fig. A1. old). Thus, synthetic off-axis spectra are generated with generalizedmetallicity and SSP base. Sixty-five unique S λ are constructed totest the ability of the model to recover known input parameters. Wewill use this Monte-Carlo routine first to test the associated uncer-tainties in a noiseless test case ( § § § A concern regarding the use of our multi-component model is in-herent degeneracies. For instance, can a young stellar populationmimic scattered quasar light? The legitimacy of the model relies onthe subtle differences between the various components. The red-dening function, which only operates on the stellar component,should be distinct from the scattering function which operates onthe observed on-axis spectrum. The absorption features in the stel-lar component generally do not match quasar absorption lines. Thescattering function, applied to the quasar spectrum, uniquely has ployed: The first of 65 synthetic spectra, S λ , is randomly assigned a x O value between 0-100%. x I and x Y are then randomly assigned the remain-ing flux. The second (third) S λ is constructed similarly, except x I ( x Y )is first assigned a value between 0-100%. This cycle is repeated beginningwith the forth S λ . To further illustrate the procedure let us consider theconstruction of the first S λ . Suppose x O was assigned a value of 72%. Theprogram populates x O by randomly selecting a SSP and a metallicity andthen randomly assigning a percentage of the total x O value. This processcontinues until x O =72% to within ± x I and x Y are then populatedin the same manner. Thus, the stellar components of the synthetic spectraare constructed. broad-line components. However, can quantities of interest be reli-ably recovered given the many possible interactions between modelcomponents?To test the ideal performance of the model we con-ducted noiseless Monte-Carlo simulations. For this purpose, theSloan Digital Sky Survey (SDSS) composite quasar spectrum(Vanden Berk et al. 2001) was utilized as the on-axis quasar spec-trum, Q λ . The scattered quasar light was scaled to a value of 60%.The results of this test are shown in Fig. 3. Referring to the figure,input versus output values of x Y , x I , x O , h log t ⋆ i L and B-V areshown for 65 synthetic spectra. Ideally, all points would fall on thediagonal, denoted by a dashed line (where input equals output). Themean offset from the ideal, mean(output-input), plus the standarddeviation of the model output, ± stdev(output-input), are displayednumerically in the upper left corner and graphically by the bluestrip. The fitting routine reliably recovers h log t ⋆ i L and B-V in theideal case of noiseless spectra. The values of x Y , x I and x O areless well constrained, with systematic offsets of up to 6% and onesigma deviations of up to 13.4%. Additional Monte-Carlo tests were conducted to check for consis-tency in the model. Inputting simulated off-axis spectra consistingof 100% scattered quasar light and a signal-to-noise of 10, the al-gorithm reliably assigned no flux to the stellar component (65 iter-ations resulted in a mean result of 99.92 ± ± c (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis were not significantly affected by this alteration; for details see Ap-pendix A (PHL 909). Errors associated with the: 1) population synthesis model (BC03)(template mis-match), 2) inadequacy of the scattering efficiencycurve, 3) non-Poisson noise and 4) the application of a single sim-plistic extinction law are not represented in our Monte-Carlo simu-lations . Despite these deficiencies, similarly constructed quiescentmodels have demonstrated consistency when compared to compet-ing techniques (e.g. see CF05). In § In this section we apply our model to the test sample to constrainthe stellar age and host colour. We then conduct 10 (one for eachpointing) tailored Monte-Carlo simulations to estimate the reliabil-ity of our results.The following caveats should be considered in the applicationof our model to the test sample. For each object, concurrent on andoff-axis observations have been obtained, with the exception of the3S pointing of 4C31.63 ( △ t ∼ . yr ). The possible variability ofthis quasar over this time period is not accounted for by the model.Line-of-sight stellar motions are not modeled for Keck objects. Ourinstrument resolution, σ v ∼ km s − , being greater or of theorder of the measured velocity dispersions (Wolf & Sheinis 2008)should result in minimal errors. For WIYN objects, with σ v ∼ km s − , the measured velocity dispersions are used to smooth themodel stellar component, ssp iλ . Narrow emission line regions andsky residuals are uniquely masked for each object, giving these re-gions no weight in the χ fitting. The broad components, and insome cases entire broad lines, are left unmasked, though most orall of the H-alpha region is masked out in many objects due tosignificant sky residuals in the red end of the spectrum. Typically, ∼ spectral data points are fit for Keck objects; ∼ spec-tral data points are fit for WIYN objects. Fig. 4 depicts the result ofthe method applied to 3C 273 4N. The results for all other off-axispointings are presented in the appendix, Fig. A1. Table 3 summa-rizes these findings.The reliability of these results are then determined with tai-lored simulations. Synthetic spectra are generated with featuresrepresentative of the observed spectra by (1) degrading S λ to theobserved noise level, (2) applying the same mask constructed forthe observed spectrum, (3) using the appropriate observed on-axisspectrum as Q λ and (4) scaling the scattered quasar light to theobserved contamination level as determined by the model output .An example of a simulation with the same on-axis spectrum, noise The contamination percentage is well recovered as tested by the Monte-Carlo routines; for example, the contamination percentage for PKS 0736-017 is recovered within ± . %. This observation has the most limitingsignal-to-noise and provides an estimate of the error upper limit. characteristics, applied mask and percent scattered quasar light asthe observed off-axis spectrum of PKS 0736+017 is displayed inFig. 5 top row. The bottom row of Fig. 5 presents the simulationresults for PG 1309+355. The PKS 0736+017 off-axis pointing hasthe worst signal-to-noise, while the PG 1309+355 off-axis point-ing has ∼ average signal-to-noise and quasar contamination. Theresults of all simulations are shown in Table 3. Throughout the restof the paper, mean offsets are applied to model results and Monte-Carlo one sigma errors are quoted. We compare our spectroscopically derived rest frame
B-V hostcolours to the available results from imaging studies. These imag-ing studies spatially subtract the point spread function of the AGNto reveal the host galaxy. Our method, which spectrally subtractsthe scattered quasar light, provides a complimentary technique toconstrain the host
B-V colours.Our off-axis method probes a population of luminous quasarsunattainable by most other spectroscopic techniques. The stellarproperties of this population have yet to be determined for a largewell defined sample. This deficiency is due to the overwhelmingemission from the central quasar. Previous off-axis studies (e.g.Hughes et al. 2000) have dealt with this problem by carefully se-lecting distant ( ∼ ′′ ) off-axis slit positions. This can significantlyreduce the typical scattered light observed off axis ( ∼ ∼
80% and2) more freedom in slit positioning. Furthermore, our method drawsfrom established quiescent population synthesis models to more ro-bustly model the stellar content and to estimate the reliability of ourresults.Recent advances in on-axis observational techniques haveallowed for the study of luminous quasars (Jahnke et al. 2007;Letawe et al. 2007). For objects with M V ( nuc ) > M V ( host ) ,our method can compliment these studies by observing off-axis( & kpc ) stellar content with favorable signal-to-noise. We notethat on-axis studies require nucleus-to-host ratios of . to detectstellar absorption features (L07). Objects such as NAB 0205+02,HE 0530-3755 and 3C 273 cannot be observed on-axis due to thiscriterion. Off-axis studies do not have this limitation. Thus, ourmethod is needed to obtain optical stellar spectra for these objects.In the following subsections we compare our study to previ-ous imaging and spectroscopic work in more detail. Furthermore,we quantify how our sample compares to applicable SDSS AGNstudies. Host
B-V colours have been reported for nearby ( z < . )quasars by Jahnke et al. (2004), hereafter J04, and for nearby( z < . ) BL Lacs by Hyv¨onen et al. (2007), hereafter H07.J04, studying multicolour data from a sample of AGN distributedaround the classical dividing line in luminosity between Seyfert 1galaxies and quasars, found bulge dominated hosts to have bluercolours than expected compared to their quiescent counterparts;disc hosts were also on average bluer, but not by a significantamount. H07, also studying multicolour data, found their exclu-sively elliptical BL Lac hosts to be on average bluer than expectedcompared to their quiescent equivalent. Table 4 shows the mean B-V colours for the host galaxies in J04 and H07, as well as in c (cid:13) , 000–000 Wold et al.
Pointing Young stellpop - % flux Inter stellpop - % flux Old stellpop - % flux h log t ⋆ i L B - V
3C 273 4N 13.8 (-3.0 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ∆ ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Table 3.
Modeled off-axis stellar populations, h log t ⋆ i L and rest frame colour are shown. Host stellar population offset and ± σ error are shown in parenthesesas estimated with Monte-Carlo simulations. ∆ denotes the observation with non-concurrent on and off-axis observations. Figure 5. Top row : Monte-Carlo simulation for PKS 0736+017 (off-axis pointing of most limiting signal-to-noise).
Bottom row : Monte-Carlo simulationfor PG 1309+355 (off-axis pointing of ∼ average signal-to-noise and quasar contamination). 65 synthetic spectra are analysed by the model to determine thereliability of the results. For each S λ , the known young, intermediate and old population percentage and the known h log t ⋆ i L and rest frame B-V input iscompared to the model output as represented by 65 plus signs in each box. Ideally, all points would fall on the diagonal, denoted by a dashed line. The meanoffset from the ideal and the standard deviation of the model output are displayed numerically in the upper left corner and graphically by the blue strip. our sample and for quiescent galaxies of various morphologies(from Fukugita, Shimasaku & Ichikawa 1995). In agreement withJ04 and H07, we find the rest frame
B-V colour for the ellipticalhost galaxies of our sample to be bluer than one would expect giventheir quiescent morphological counterparts. This provides confir-mation that elliptical quasar hosts are distinguishable in
B-V colourfrom inactive elliptical galaxies using a spectroscopic technique.Spirals hosts for our sample are found to be consistent with Scgalaxies which have a
B-V colour of ∼ + ’), and quiescent COMBO-17early-type galaxies (black ’ ∗ ’) are shown in relation to J04 hosts(blue lower-case letters; ’s’ for spiral or disc, ’e’ for elliptical mor-phology), H07 hosts (red ’ ǫ ’), and this work (green upper-case let-ters; ’S’ for spiral or disc, ’E’ for elliptical morphology). Quiescentearly and late-type galaxy designations are determined by the U-V vs M V cut as proposed by Bell et al. (2004) for z=0.25. A typicalerror bar is shown for our sample in the lower right of Fig. 6.Although we confirm bluer elliptical host galaxy colours, wefind that our sample is on average much redder than J04. This off-set may be due to the different effective radii probed since our method is only sensitive to the stellar content observed off axis.For example, if nuclear starbursts were a common occurrence inquasar hosts, then one might expect the outer stellar populationsto exhibit redder colours. However, results from Kauffmann et al.(2003) suggest that the different radii probed may not be responsi-ble for this observed offset. Studying a large sample of obscuredAGN hosts, they find that star formation in the hosts of power-ful AGN (approximately one magnitude fainter than a typical ob-ject in this study) is not concentrated primarily in the nuclear re-gions; it is spread out over scales of at least several kilo-parsecs.Furthermore, the host galaxies in the H07 study exhibit a widerange of host colour gradients, with the majority displaying bluercolours off axis. This negative colour gradient is also observedfor radio galaxies (Govoni et al. 2000) and quiescent ellipticals(Peletier, Valentijn & Jameson 1990). Assuming these conclusionshold true for quasar hosts, one would expect our sample to be bi-ased toward bluer colours rather than the observed red offset.An alternative explanation for our redder host colours com-pared to the J04 study is that our sample is on average more radio c (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis Figure 6.
Rest frame colour-magnitude diagram for local quiescent galaxies and active host galaxies. Black plus signs indicate (z < < ǫ ’. Quasar hosts from this work are indicated by green uppercase letters (’S’ for spiral or disc, ’E’ for elliptical morphology). A typical error bar is shown for our sample in the lower right. ’1’ and ’2’ denote objects withmultiple pointings, 4C 31.63 and PKS 2349-014 respectively. ∆ denotes the observation with non-concurrent on and off-axis observations. loud . The only J04 object which overlaps in B-V colour with ourmain sample is their most radio loud. The two objects from oursample which overlap in colour with the main J04 sample are alsoour most radio quiet. This relation of colour to radio luminosity isshown in Fig. 7. Above L GHz ∼ erg s − , the host galax-ies are systematically redder. Given the correlation between radioluminosity and bulge mass found for our sample (Wolf & Sheinis2008, L GHz ∼ M . bulge ), L GHz = 10 erg s − correspondsto M bulge ∼ . M ⊙ . The rms scatter of the correlation is 1.09dex. This trend for radio loud objects to be located in redder hostsis not surprising; Best et al. (2005) found that radio-loud AGN arepreferentially located in older more massive galaxies. However thedistinct transition we see in our sample should be investigated on alarger sample.How can this be understood in terms of stellar ages? Themean light-weighted age is correlated to various colours (e.g., u-r , Mateus et al. 2006). We verify this correlation for our colour of Radio luminosity is estimated utilizing NED and NVSS (Condon et al.1998). A spectral index of -0.5 is assumed. Upper limits are based on thesensitivity of NVSS ∼ B - V
E Quasars (this work) 0.78 ± ± ± ± L GHz > erg s − (this work plus J04) 0.77 ± L GHz < erg s − (this work plus J04) 0.50 ± ± Table 4.
Available rest frame
B-V host mean colours and quiescent meancolours. Reported errors are standard deviations of the mean. ’E’ denoteselliptical and ’S’ denotes spiral. interest,
B-V , by constructing 65 test spectra, Fig.8. Using this re-lation we convert the
B-V colours of J04 to mean light-weightedages, Fig. 9. While the observed colour trend is more pronounced,it is found that all h log t ⋆ i L values above L GHz =10 erg s − are older than values below this threshold. Previous studies (CF05;Mateus et al. 2006) have studied the mean light-weighted age andfound this quantity to be frequently associated with the mostrecent epoch of starbursts. Assuming major merger progenitorsfor our quasar hosts with a significant past starburst (e.g., seeHopkins et al. 2006), h log t ⋆ i L may indicate the approximate tim-ing of the merger event. It should be noted that our definition of h log t ⋆ i L differs slightly from the definition adopted by CF05 andMateus et al. (2006) due to different SSP normalization conven- c (cid:13) , 000–000 Wold et al.
Figure 7.
Radio luminosity as a function of rest frame
B-V colour. Greenupper case letters indicate the results of this study, blue lower case lettersindicate the results of J04. ’1’ and ’2’ shown above denote objects withmultiple pointings, 4C 31.63 and PKS 2349-014 respectively. ∆ denotesthe observation with non-concurrent on and off-axis observations. H07’sBL Lacs are not shown due to the expected significant beaming of radioemission. Figure 8.
Relation between mean light-weighted stellar age and rest frame
B-V colour derived from 65 randomly constructed stellar components, P i =1 x i ssp iλ . tions (see § h log t ⋆ i L results by ∼ σ or less.To estimate this time-scale we divide the combined sampleinto two groups; a radio bright sample with L GHz > erg s − displaying an average B-V colour of 0.77 and a radio faint samplewith L GHz < erg s − displaying an average B-V colour of0.50 (average colour for this grouping is also reported in Table 4). L GHz is found to correspond to bulge mass and may provide amore significant division than morphology whose classification ishampered by quasar point source subtraction and un-modeled tidalfeatures. We find that, on average, h t ⋆ i L = 2 . +0 . − . Gyrs for the h t ⋆ i L ≡ h log t ⋆ i Figure 9.
Radio luminosity as a function of mean stellar age weighted byflux. Green upper case letters indicate the results of this study, blue lowercase letters indicate the results of J04. ’1’ and ’2’ shown above denote ob-jects with multiple pointings, 4C 31.63 and PKS 2349-014 respectively. ∆ denotes the observation with non-concurrent on and off-axis observations. radio bright sample and h t ⋆ i L = 570 +140 − Myrs for radio faintsample.
This result for radio loud objects compares remarkably well to re-cent findings of Canalizo et al. (2006) and Bennert et al. (2008).Canalizo et al. (2006) report preliminary deep Keck LRIS spectralanalysis of luminous z ∼ . quasar hosts which finds evidence formajor starburst episodes with ages ranging from 0.6 Gyr to 2.2 Gyr.Incidentally at least two of the host galaxies studied by Canalizo &Stockton (in preparation) overlap with our sample, PKS 0736+017and PHL 909. They find that both of these objects located in ourradio bright group have massive 2.2 Gyr starbursts, consistent withour colour and mean light-weighted age estimate. However, see Ap-pendix A for an inconsistency in the derived stellar populations forPHL 909. Bennert et al. (2008), again studying PKS 0736+017 andPHL 909 along with three other hosts in deep HST ACS images,find shells and tidal tails indicative of merger events. ConsultingN-body simulations they conclude that in general the observed finestructure can be explained by either a recent minor merger or anolder major merger ∼ ∼ Gyr poststarburst quasar are discussedextensively in Canalizo et al. (2007) and Bennert et al. (2008). Insummary the currently estimated (optically luminous) quasar life-time ( − yrs (Yu & Tremaine 2002)) appears to be too shortto be triggered at the same time as the starburst. Bennert et al.(2008) conclude that this may imply a scenario in which quasaractivity is significantly delayed or a scenario in which intermit-tent quasar activity encompasses a longer duration. Alternatively, amore recent minor merger could explain the observed quasar. Ourresults do not rule out this scenario. However, the average mass-to-light ratio for our sample being ∼
10 (Wolf & Sheinis 2008), con-strains any recent (t ≪ c (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis The off-axis ( ′′ ) spectroscopic study of Hughes et al. (2000, here-after H00) consists of three matched subsamples of nine radio quietquasars (RQQs), ten radio loud quasars (RLQs), and seven ra-dio galaxies (RGs). Similar to our study, all quasars are luminous( M V . − ) and local ( . z . ).H00 determine the 4000 ˚A break strength, an indicator ofthe mean stellar age, for each object. The average values for RQQhosts, RLQ hosts, and RGs all indicate younger mean stellar agesthan local inactive elliptical galaxies, in agreement with our study.However, H00 notes wide scatter in their measured values. Addi-tionally, measurements indicative of younger ages were typicallyassociated with either poor signal-to-noise data or with spectrathat show significant scattered quasar light. Thus H00 defer fur-ther analysis of the stellar populations to the followup spectro-photometric modeling conducted by Nolan et al. (2001, hereafterN01).N01 develop two population synthesis models. The first modelconsists of a 0.1 Gyr SSP and a second SSP component whose ageand contribution are allowed to vary to obtain the best χ fit. Likeour study, solar metallicity is assumed. However, two SSPs are usedto construct the stellar population in contrast to our model whichincorporates a base of 15 BC03 SSPs. N01 also develop a threecomponent model which uses the observed nuclear spectrum of theRQQ, 0054+144, to account for scattered quasar light detected offaxis. While our approach also uses observed nuclear light to modeloff-axis scattered light, our model does not assume a universal nu-clear spectrum. The observed nuclear spectrum is matched to theoff-axis observation (e.g. the model input for the quasar contamina-tion in the 3C 273 off-axis spectrum is the observed 3C 273 nuclearspectrum). Additionally, it should be noted that whenever possible(9 out of 10 pointings) we have observed on and off-axis obser-vations simultaneously to avoid variation in the quasar spectrumwhich could hamper an accurate scattered-light subtraction. As afinal point, we allow for the continuum shape of the nuclear lightobserved off axis to be altered by the scattering efficiency curve,which we have shown is necessary for reliable scattered light re-moval (Sheinis 2002).Interpretation of the N01 model output indicates that the stel-lar mass of all three subsamples studied (RQQ, RLQ, and RG) isdominated by old stars of age 8-14 Gyr. This result is used to sup-port the claim that quasar hosts are “to the first order, indistinguish-able from ’normal’ quiescent giant elliptical galaxies.” If we takeN01’s derived age to be representative of the mean mass-weightedstellar age, then it is not clear that this quantity can be used to differ-entiate between late and early type galaxies. Mateus et al. (2006),studying 50,000 luminous galaxies from the SDSS, show that thedistribution of mean mass-weighted stellar ages is not bimodal, asboth star-forming and passive galaxies have formed a large frac-tion of their stellar mass at early times ( ∼ . Gyr in the past).We may comment on the result of N01 by referring to the mass-to-light ratio derived from the measured stellar velocity dispersion.Wolf & Sheinis (2008) found that the average mass-to-light ratiofor our quasar host sample is ∼ , which supports the domi-nance by mass of old stars. However, as stated in § h log t ⋆ i L determination indicates that quasar hosts are bluer /younger than quiescent ellipticals. Radio loud in H00 is defined as L GHz > W Hz − sr − , whichis roughly consistent with our adopted definition. Sample h log t ⋆ i L Quasars L GHz > erg s − (this work plus J04) 9.4 ± L GHz < erg s − (this work plus J04) 8.8 ± Table 5.
Comparison of derived h log t ⋆ i L values to those found in theSDSS study conducted by M06. Reported errors are standard deviations ofthe mean. J07 present a method that spatially subtracts the quasar point spreadfunction from the on-axis spectrum to extract the host galaxy light.Unlike our method, which is constrained to collect light & kpc offaxis, this technique allows for the inner regions of the host galaxyto be probed. Like our study, the J07 test sample consists of nearby( z < . ) quasars. However, this sample consists of quasars that aretypically less luminous in optical and radio bands than our study.J07 model 13 host spectra with a two component BC03 SSP model.Eight of the spectra were deemed trustworthy, yielding a typicallight-weighted stellar age of 1-2 Gyr. For a rough comparison, oursample average h t ⋆ i L is approximately 2 Gyr. If we limit our sam-ple to radio quiet objects, to more closely match the sample of J07,we find an average h t ⋆ i L of 1 Gyr. Confidence in this comparisonis hampered by the small number of objects probed and the samplemismatch: our targets are typically more luminous.Despite this mismatch, there is one object that is modeled byboth groups, the radio loud quasar PKS 1302-102. While J07 finda stellar age < Myr , our results indicate a predominantly old, > Gyr stellar population. This disagreement may indicate a radialdependence on the stellar populations or model systematics.L07, utilizing an on-axis method complementary to the J07technique, examine the host galaxies of 20 luminous quasars whoseredshifts and nuclear optical luminosities are well matched to oursample. However, their optically selected sample contains far fewerRLQs. L07 measure diagnostic absorption and emission lines andthen compare these to known quiescent galaxy values. Half oftheir sample was found to have young Sc-like stellar populations.The sample examined by our study indicates a redder and henceolder stellar population than a typical Sc galaxy. Our studies areagain brought into closer agreement if we consider only our RQQhosts. However, no secure agreement can be claimed due to oursmall RQQ sample. Further complicating the comparison, L07 alsomodel PKS 1302-102 and find evidence for a young Sc-like stellarpopulation. Additional interpretation of this disagreement is dis-cussed in Appendix A.
The Sloan Digital Sky Survey (SDSS) has been utilized tostudy AGN hosts by a number of groups. These studies ben-efit from a large well defined sample. Narrow-line AGN stud-ies (Kauffmann et al. 2003, hereafter K03; Zakamska et al. 2003;Mateus et al. 2006, hereafter M06) and the broad-line AGN studyby Vanden Berk et al. (2006, hereafter V06) are compared to thiswork below.K03 study the host galaxies of 22,623 local type 2 AGN. Mod-ulo deficiencies in AGN unification models, quasar hosts can bestudied via their narrow-line counter-parts. From 4000 ˚A break c (cid:13) , 000–000 Wold et al. measurements, K03 find that high luminosity AGN have youngermean stellar ages than normal early-type galaxies. Additionally,they find a large fraction of powerful type 2 AGN have experi-enced significant starbursts in the past 1-2 Gyr. This is similarto our findings: h t ⋆ i L = 2 . +0 . − . Gyrs for the radio bright sam-ple and h t ⋆ i L = 570 +140 − Myrs for radio faint sample. However,there is little overlap in luminosity between our samples. The K03powerful (L[OIII] ∼ erg s − ) AGN are about one magnitudefainter than our sample’s nuclear luminosities. A smaller more lu-minous sample of SDSS narrow-line AGN has been examined byZakamska et al. (2003). In qualitative agreement with K03, they toofind evidence for relatively blue / young host galaxies.M06 divide a sample of 50,000 SDSS galaxies into star-forming, passive, and AGN hosts spectral classes. The AGN probedby this study are comparable in luminosity to the K03 sample. M06find a clear bimodal distribution for the mean light-weighted stel-lar age among star-forming and passive galaxies. As summarized inTable 5, median values of h log t ⋆ i L = 8 . for star-forming galax-ies and h log t ⋆ i L = 9 . for passive galaxies are found. The greenvalley is located at h log t ⋆ i L ∼ . and the median h log t ⋆ i L for AGN hosts is 9.66. The M06 median h log t ⋆ i L for AGN hostsis heavily weighted toward the more numerous low luminosity ob-jects. In agreement with K03, they find that the typical h log t ⋆ i L decreases for more luminous AGN. Our radio bright sample with h log t ⋆ i L ∼ . ± . is in close proximity to the green valley(slightly younger and bluer). We find the radio faint sample with h log t ⋆ i L ∼ . ± . to fall very close to the peak of star-formingdistribution. Qualitatively we find this to be in overall agreementwith our colour-magnitude diagram, Fig. 6.V06 study the host galaxies of 4,666 local broad-line AGN.Like K03 and M06, they find that higher luminosity AGN hosts arebluer /younger than normal early-type galaxies. This sample con-tains ∼ quasars with luminosities comparable to our study.However, the model employed by V06 fails when the on-axis hostflux fraction falls below 10% or when the signal-to-noise is lessthan 10. These criteria prevent the study of ∼ of their lumi-nous objects and would exclude the majority of the objects (7 of10) in this study. We have presented an off-axis technique to spectroscopically con-strain the colour and the stellar ages of quasar ( M V ( nuc ) < − )host galaxies. Our method draws heavily from the quiescent galaxymodel of CF05, utilizing a basis of BC03 SSPs and Cardelli et al.(1989) dust extinction in a similar fashion. Complicating our χ fit-ting routine is the residual scattered quasar light (a combination ofatmospheric, instrumental and host galaxy scattered light) whichmust be accounted for in the off-axis spectrum. Scattered light ismodeled by altering the the observed nuclear spectrum by a low or-der polynomial, while simultaneously fitting the constituent stellarpopulations of the host galaxy. Furthermore, Monte-Carlo simula-tions are tailored to each observed pointing. Thus, the ability of themodel to recover known parameters from synthetic spectra is de-termined. It is found that quasar host B-V colour and h log t ⋆ i L arewell recovered. These parameters are then compared to previousstudies. Overall consensus is found giving further credence to ourmodel (however, see Appendix A for discussion of individual casesthat potentially disagree with previous results).Our method probes a population of luminous quasars unattain-able by most other techniques. The stellar properties of this popu- lation have yet to be determined for a large well defined sample.For the smaller sample ( N obj ∼ ) studies that can access thisluminous regime, our method can be used in conjunction to gainfurther insight. For example, our technique can observe the off-axis( R obs & kpc ) stellar content of objects (nucleus-to-host ratio & B-V colour.Additionally, we note a trend for radio luminous quasars to be lo-cated in redder host galaxies in comparison to their less radio lu-minous counterparts. However, general conclusions await furtherobservations. Currently we have 10 additional WIYN off-axis ob-servations in the process of being reduced. Furthermore, time hasbeen allotted on the WIYN telescope to increase our total analyzedsample size to ∼ objects. This sample will consist of roughlyan equal number of RLQs and RQQs and allow for 1) more rigor-ous comparison to other studies due to the increase in sample size,2) further verification of our technique with the targeting of L07overlap objects (specifically HE 0914-0031 and HE 0956-0720),3) investigation of the trends found in our preliminary sample. Inthe near future, the Robert Stobie Spectrograph (RSS) on the 11-mSouthern African Large Telescope (SALT) will be employed in ourongoing observational campaign. This phase of the project will in-crease the observable sample by its location in the Southern hemi-sphere and by allowing for higher redshift objects to be targeted( z ∼ . ).Future work will build on the current model to analyze thenarrow-line emission. This will provide insight into the star forma-tion rate, the emission mechanism, the metallicity, and the stellarcontent (e.g. see L07). The accomplishment of this program willprovide important observational constraints on luminous quasars. ACKNOWLEDGEMENTS
The authors would like to thank Joseph Miller, Roberto Cid Fer-nandes, and Laura Trouille for insightful conversations on variousaspects of this work. We are also in debt to Amy Barger for use ofcomputing resources. Some of the data presented herein were ob-tained at the W. M. Keck Observatory, which is operated as a scien-tific partnership among the California Institute of Technology, theUniversity of California, and the National Aeronautics and SpaceAdministration. The Observatory was made possible by the gen-erous financial support of the W. M. Keck foundation. Data werealso obtained at the WIYN Observatory, which is a joint facilityof the University of Wisconsin-Madison, Indiana University, YaleUniversity, and the National Optical Astronomy Observatories.
APPENDIX A: COMMENTS ON INDIVIDUAL HOSTS
In this section, we describe the stellar populations found for in-dividual hosts and compare these to previous studies. Objects arelisted in order of greatest 5 GHz luminosity to least, as in Table1. Unless otherwise mentioned: young ≡ (t <
100 Myr), interme-diate ≡ (100 Myr t ≡ (t > c (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis By convention, application of offsets may not force the percent fluxcontribution to be negative or greater than 100%. For output param-eters without offsets applied see Table 3. Fig. 4 and Fig. A1 are alsoshown without offsets applied. As mentioned in §
3C 273 (PG 1226+023) – The host galaxy is classi-fied as an elliptical galaxy via HST imaging (Bahcall et al.1997; Hamilton et al. 2002). The off-axis spectral analysis ofBoroson et al. (1985) qualitatively found evidence for a significantcontribution from an old stellar population. Our results, 88.8 ± σ error, a young stellar component is found by the routine.Residual OII and OIII narrow line emission is noted.
4C 31.63 (Q2201+315) – 4C 31.63 is a radio loud quasarwhose host galaxy has been classified as an elliptical by HST imag-ing (Bahcall et al. 1997; Hamilton et al. 2002) and by near-infraredadaptive optics imaging (Guyon et al. 2006) . Our results for theNorthern and Eastern pointings are consistent with a predominantlyold stellar population. Toward the south we find evidence for an in-termediate stellar population not found in the other pointings. Notethat the 3S pointing has non-concurrent on and off-axis observa-tions, see § PKS 1302-102 (HE 1302-1017) – The host galaxy of thisradio loud quasar has been classified as a disturbed ellipticalby various studies (Hutchings & Neff 1992; Bahcall et al. 1997;Hamilton et al. 2002; Guyon et al. 2006). Jahnke et al. (2007) andLetawe et al. (2007) have found evidence from on-axis spec-troscopy for a young spiral-like stellar population. Jahnke et al.(2007) note that there are almost no stellar absorption lines withthe exception of a weak Ca II K and Mg I line. This is not the casein our observed off-axis spectrum; Ca II K λ and H λ are prominent with clear Mg I b λ and Na D λ absorp-tion features. The stellar component is found to be best fit by a88.9 ± ∼ ∼ to reliably detect absorption featureswith the on-axis method (L07). Regardless, the observed prominentCa H and K absorption lines support the claim that a significant oldpopulation is present at the radius probed ∼ Kpc. Residual OIIand OIII narrow line emission is noted.
PKS 0736+017 – The host galaxy has been classified as el-liptical (Wright, McHardy & Abraham 1998 McLure et al. 1999;Falomo & Ulrich 2000; Hamilton et al. 2002). Others describethe host as a disturbed elliptical based on near-infrared imaging(Dunlop et al. 1993) and deep HST imaging (Bennert et al. 2008).Off-axis spectroscopic studies have found a stellar component in-dicative of a 12 Gyr population (Hughes et al. 2000; Nolan et al.2001). Canalizo & Stockton (in preparation) find a 2.2 Gyr star-burst in this host (Bennert et al. 2008) which falls in our ’old’ age bin. Our results indicate 100 ± PKS 2135-147 – The host galaxy of this radio loud quasarhas been classified as elliptical by Bahcall et al. (1997) andHamilton et al. (2002). Our results are consistent with the expectedstellar population of an elliptical galaxy. We find 100 ± PKS 2349-014 – The host galaxy of this radio loud quasar hasbeen classified as a highly disturbed elliptical by many imagingstudies (Bahcall et al. 1997; Guyon et al. 2006). An off-axis spec-troscopic study by Nolan et al. (2001) found a dominant old stellarpopulation of 12 Gyr with a low level young population. Off-axisstudies (Miller & Sheinis 2003; Wolf & Sheinis 2008), utilizing thesame data analysed in this paper, have also found evidence for apredominately old stellar population with a lesser young popula-tion. Our results are consistent with a predominantly old populationby flux. The northern pointing (14.12 kpc), probing large tidal arms,is consistent with an entirely old stellar population; whereas thecloser-in Southern pointing (9.47 kpc) we find evidence for a smallintermediate population (16.1 ± . ). Multiple residual narrowemission lines are noted for this object. PG 1309+355 – HST imaging studies classify this object asa spiral (Bahcall et al. 1997; Hamilton et al. 2002), while the near-infrared study of Guyon et al. (2006) classifies the host as an elon-gated elliptical. If this object is truly a spiral, then this is a very rareobject displaying a large radio luminosity, just below our radio-loudthreshold. Our results indicate a predominantly old population. Nosignificant narrow emission lines are noted.
PHL 909 (0054+144) – The host is an elliptical galaxyas classified by the imaging studies of Bahcall et al. (1997) andHamilton et al. (2002) . The near-infrared imaging of Dunlop et al.(1993) detects extended emission toward a companion. Recentdeep HST imaging (Bennert et al. 2008) detects tidal tails andshells. Barthel (2006), analysing the far-infrared / radio correla-tion, finds evidence of recent or on going star formation. Off-axis spectroscopic studies have either classified the stellar popula-tion as old (Nolan et al. 2001) or have reported ambiguous results(Hughes et al. 2000). Canalizo & Stockton (in prep.), analysing ∼ ± ± . Ifour observations are co-spatial with those of Canalizo & Stockton, As listed in Table 1 and Table 2, the original error simulations found thefollowing offsets and ± σ errors: − . ± . for young, +4 . ± . for intermediate, +3 . ± . for old, +0 . ± . for B-V colour, and +0 . ± . for h log t ⋆ i L . The steep blue scattering simulations found: − . ± . for young, − . ± . for intermediate, +4 . ± . for old, +0 . ± . for B-V colour, and +0 . ± . for h log t ⋆ i L .c (cid:13) , 000–000 Wold et al.
Figure A1.
Model fit to the observed off-axis spectra. The meanings of the symbols and colours are the same as in Fig. 4.c (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis Figure A1. continuedc (cid:13) , 000–000 Wold et al.
Figure A1. continued c (cid:13) , 000–000 ost galaxies of luminous quasars: population synthesis Figure A1. continuedc (cid:13) , 000–000 Wold et al. then the difference in the results may simply be a σ outlier arisingfrom separate observations and different analysis methods. Regard-less, construction of a solar BC03 stellar spectrum consistent withthe findings of Canalizo et al. (2006) results in values for the meanlight-weighted age and B-V colour that are consistent (within 1 σ ) ofour findings. Therefore, we believe that the main discussion regard-ing the more reliably recovered parameters, colour and h log t ⋆ i L ,are robust. PG 0052+251 – The host galaxy associated with this radioquiet quasar is a spiral as classified by various imaging studies(Bahcall et al. 1997; Hamilton et al. 2002). Our results are consis-tent with a significant young population of 35.3 ± ± . %. Multiple residual narrow emis-sion lines are noted for this object. PG 1444+407 – The host of this radio quiet quasar is classi-fied as a spiral by Hamilton et al. (2002). Bahcall et al. (1997) notethat the host has the appearance of an elliptical; however the lightprofile is best fit by an exponential disc model. Additionally, Ho(2005) estimate a significant star formation rate (19.4 M ⊙ yr − )based on the [O II] luminosity. Our results indicate a 60.8 ± ± REFERENCES
Bahcall, J. N., Kirhakos, S., Saxe, D. H., & Schneider, D. P. 1997,ApJ, 479, 642Barthel, P. D. 2006, A&A, 458, 107Bell, E. F., et al. 2004, ApJ, 608, 752Bennert, N., Canalizo, G., Jungwiert, B., Stockton, A., Schweizer,F., Peng, C. Y., & Lacy, M. 2008, ApJ, 677, 846Benson, A. J., Bower, R. G., Frenk, C. S., Lacey, C. G., Baugh,C. M., & Cole, S. 2003, ApJ, 599, 38Bershady, M. A., Andersen, D. R., Verheijen, M. A. W., Westfall,K. B., Crawford, S. M., & Swaters, R. A. 2005, ApJS, 156, 311Best, P. N., Kauffmann, G., Heckman, T. M., Brinchmann, J.,Charlot, S., Ivezi´c, ˇZ., & White, S. D. M. 2005, MNRAS, 362,25Boroson, T. A., Persson, S. E., & Oke, J. B. 1985, ApJ, 293, 120Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Canalizo, G., & Stockton, A. 2001, ApJ, 555, 719Canalizo, G., Stockton, A., Brotherton, M. S., & Lacy, M. 2006,New Astronomy Review, 50, 650Canalizo, G., Bennert, N., Jungwiert, B., Stockton, A., Schweizer,F., Lacy, M., & Peng, C. 2007, ApJ, 669, 801Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345,245Cattaneo, A., Dekel, A., Devriendt, J., Guiderdoni, B., & Blaizot,J. 2006, MNRAS, 370, 1651Chabrier, G. 2003, PASP, 115, 763Cid Fernandes, R., Mateus, A., Sodr´e, L., Stasi´nska, G., & Gomes,J. M. 2005, MNRAS, 358, 363Condon, J. J., Cotton, W. D., Greisen, E. W., Yin, Q. F., Perley,R. A., Taylor, G. B., & Broderick, J. J. 1998, AJ, 115, 1693Dunlop, J. S., Taylor, G. L., Hughes, D. H., & Robson, E. I. 1993,MNRAS, 264, 455Dunlop, J. S., McLure, R. J., Kukula, M. J., Baum, S. A., O’Dea,C. P., & Hughes, D. H. 2003, MNRAS, 340, 1095Falomo, R., & Ulrich, M.-H. 2000, A&A, 357, 91 Fukugita, M., Shimasaku, K., & Ichikawa, T. 1995, PASP, 107,945Gebhardt, K., et al. 2000, ApJ, 539, L13Govoni, F., Falomo, R., Fasano, G., & Scarpa, R. 2000, A&A,353, 507Guyon, O., Sanders, D. B., & Stockton, A. N. 2006, New Astron-omy Review, 50, 748Hamilton, T. S., Casertano, S., & Turnshek, D. A. 2002, ApJ, 576,61Ho, L. C. 2005, ApJ, 629, 680Hooper, E. J., Impey, C. D., & Foltz, C. B. 1997, ApJ, 480, L95Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T., Robertson,B., & Springel, V. 2006, ApJS, 163, 1Hopkins, P. F., Cox, T. J., Kereˇs, D., & Hernquist, L. 2008, ApJS,175, 390Hughes, D. H., Kukula, M. J., Dunlop, J. S., & Boroson, T. 2000,MNRAS, 316, 204Hutchings, J. B., Johnson, I., & Pyke, R. 1988, ApJS, 66, 361Hutchings, J. B., & Neff, S. G. 1992, AJ, 104, 1Hyv¨onen, T., Kotilainen, J. K., Falomo, R., ¨Orndahl, E., & Pur-simo, T. 2007, A&A, 476, 723Jahnke, K., Kuhlbrodt, B., & Wisotzki, L. 2004, MNRAS, 352,399Jahnke, K., Wisotzki, L., Courbin, F., & Letawe, G. 2007, MN-RAS, 378, 23Kauffmann, G., et al. 2003, MNRAS, 346, 1055Kellermann, K. I., Sramek, R. A., Schmidt, M., Green, R. F., &Shaffer, D. B. 1994, AJ, 108, 1163Kormendy, J. 2000, arXiv:astro-ph/0007400Lacy, M. 2006, Astrophysics Update 2, 195Letawe, G., Magain, P., Courbin, F., Jablonka, P., Jahnke, K.,Meylan, G., & Wisotzki, L. 2007, MNRAS, 378, 83Mateus, A., Sodr´e, L., Cid Fernandes, R., Stasi´nska, G.,Schoenell, W., & Gomes, J. M. 2006, MNRAS, 370, 721McLure, R. J., Kukula, M. J., Dunlop, J. S., Baum, S. A., O’Dea,C. P., & Hughes, D. H. 1999, MNRAS, 308, 377Miller, J. S., & Sheinis, A. I. 2003, ApJ, 588, L9Nolan, L. A., Dunlop, J. S., Kukula, M. J., Hughes, D. H., Boro-son, T., & Jimenez, R. 2001, MNRAS, 323, 308Oke, J. B., et al. 1994, SPIE, 2198, 178Peletier, R. F., Valentijn, E. A., & Jameson, R. F. 1990, A&A, 233,62Press, W., Teukolsky, S., Vetterling, W., & Flannery, B. 1992, Nu-merical Recipes in C: The Art of Scientific Computing (Cam-bridge, MA: Cambridge Univ. Press)Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500,525Sheinis, A. I. 2002, Ph.D. Thesis, Univ. CaliforniaTremonti, C. A. 2003, Ph.D. Thesis, Johns Hopkins Univ.Vanden Berk, D. E., et al. 2001, AJ, 122, 549Vanden Berk, D. E., et al. 2006, AJ, 131, 84Wolf, C., et al. 2004, A&A, 421, 913Wolf, M. J., & Sheinis, A. I. 2008, AJ, 136, 1587Wright, S. C., McHardy, I. M., & Abraham, R. G. 1998, MNRAS,295, 799Yu, Q., & Tremaine, S. 2002, MNRAS, 335, 965Zakamska, N. L., et al. 2003, AJ, 126, 2125 c (cid:13)000