Table Of Content

Accepted for publicationin The Astrophysical Journal PreprinttypesetusingLATEXstyleemulateapjv.5/2/11 SPECTROSCOPYOF THE LARGEST EVER γ-RAY SELECTED BL LAC SAMPLE Michael S. Shaw1, Roger W. Romani1, Garret Cotter2, Stephen E. Healey1, Peter F. Michelson1, Anthony C. S. Readhead3, Joseph L. Richards3, Walter Max-Moerbeck3, Oliver G. King3, William J. Potter2 Accepted for publication inThe Astrophysical Journal ABSTRACT We report on spectroscopic observations covering most of the 475 BL Lacs in the 2nd Fermi LAT 3 catalog of AGN. Including archival measurements (correcting several erroneous literature values) we 1 now have spectroscopic redshifts for 44% of the BL Lacs. We establish firm lower redshift limits 0 via intervening absorption systems and statistical lower limits via searches for host galaxies for an 2 additional 51% of the sample leaving only 5% of the BL Lacs unconstrained. The new redshifts raise themedianspectroscopicz˜from0.23to0.33andincluderedshiftsaslargeasz =2.471. Spectroscopic n redshiftminimafrominterveningabsorbershavez˜=0.70,showingasubstantialfractionatlargezand a J arguing against strong negative evolution. We find that detected BL Lac hosts are bright ellipticals 2 with black hole masses M• ∼ 108.5−9, substantially larger than the mean of optical AGN and LAT FlatSpectrumRadioQuasarsamples. AslowincreaseinM• withz maybeduetoselectionbias. We ] find that the power-law dominance of the optical spectrum extends to extreme values, but this does E not stronglycorrelatewith the γ-rayproperties, suggestingthat strong beaming is the primary cause H of the range in continuum dominance. . Subject headings: BL Lacertae objects: general — galaxies: active — Gamma rays: galaxies — h quasars: general — surveys p - o 1. INTRODUCTION now decreased to 215 (19%), and the confirmed BLLs r st The Fermi Second Source Catalog (Nolan et al. 2012, have increased to 475 (42% of all 2LAC AGN). In§2,weoutlinethesampleproperties,datacollection, a 2FGL) lists the 1873 most significant sources detected [ by the Large Area Telescope (Atwood et al. 2009, LAT) and data reduction steps. We also summarize principal features of the spectra. In §3, we describe our spectro- duringFermi’sfirsttwoyearsofskysurveyobservations. 1 scopic constraints on the redshift, including a technique The majority of these sources are associated with jet- v to provide uniform redshift limits based on searches for dominated Active Galactic Nuclei, the so-called blazars, 3 hostgalaxyemission. In§4,wegiveestimatesoftheBLL many of which are bright, compact radio sources. There 2 3 are, in fact, 1121 such associations (1017 at |b| > 10◦), blackholemasses. Weturntocommentsontheprincipal BLL feature, the non-thermal dominance in the optical 0 collectedinthe SecondCatalogofAGN Detectedby the in §5, and conclude with general remarks in §6. . FermiLAT(Ackermann et al.2011,2LAC).TheseAGN 1 In this paper, we assume an approximate concordance are further classified as Flat-Spectrum Radio Quasars 0 cosmology – Ω = 0.3, Ω = 0.7, and H = 70 km s−1 (FSRQ) where the optical spectrum is dominated by m Λ 0 3 Mpc−1. thermal disk and broad-line region emission, BL Lacs 1 (BLL),where the opticalspectrumis dominatedby con- : v tinuumsynchrotronradiation,andacollectionofmiscel- 2. OBSERVATIONSANDDATAREDUCTION i laneous,mostlylowluminosityrelatedsources. In2LAC, 2.1. The BLL Sample X the sample included 410 BLL, 357 FSRQ, 28 AGN of BLLswereoriginallyidentifiedasopticalviolentlyvari- r other type (generally low z, lower luminosity Seyferts), a able AGN, and are often characterized by an optical and 326 AGN of (then) unknown type. continuum dominated by synchrotron emission. Their These ‘Blazars’ (BLL and FSRQ) are the brightest broad-band spectral energy distribution (SED) is de- extra-Galactic point sources in the microwaveand γ-ray scribed by a synchrotron component peaking in the far- bands; study of their population and evolution are cen- IR to X-ray bands and an Inverse Compton (IC) com- tral topics in high energy astrophysics. To support such ponent peaking in the MeV to TeV range. In the radio studieswehaveacquiredsensitivespectroscopicobserva- these sources display strong core dominance. Accord- tions of this sample. In a companion paper (Shaw et al. ing to the unified model (Urry & Padovani 1995) BLLs 2012, hereafter S12), we reported on measurements of are the beamed counterparts of the FR I radio galaxy a large fraction of the FSRQ. Here we concentrate on population, while the FSRQ are associated with FR IIs. the BL Lac objects. Our study has also found types However, the principal BLL characteristic, a dominant forsomeofthe unclassifiedblazars;the ‘unknowns’have and varying synchrotron/IC continuum, is a sign of a powerful jet whose emission is beamed closely toward 1DepartmentofPhysics/KIPAC,StanfordUniversity,Stanford, theEarthline ofsight. Thusthe distinctionbetweenthe CA94305 2DepartmentofAstrophysics,UniversityofOxford,OxfordOX1 traditionalBLL andthe FSRQ issensitiveto the precise 3RH,UK stateandorientationofthe jet(e.g.Giommi et al.2012) 3Department of Astronomy, CaliforniaInstitute ofTechnology, and,indeed,variationsinjetpowerordirectioncanbring Pasadena,CA91125 individual sources in or out of the BLL class (S12). 2 M. S. Shaw et al Our own assignment of the BLL label follows a prag- matic “optical spectroscopic” definition (Marcha et al. Table 1 SampleCompleteness 1996): these are blazars with no emission lines greater than 5 ˚A observed equivalent width, and a limit on any possible 4000 ˚A spectral break of < 40% (Marcha et al. Set Total Specz zmin Unk. z 2LAC 1121 1996; Healey et al. 2008; Shaw et al. 2009). For transi- BLL 475 209 241 25 tion objects with varying continuum emission, we retain LBL 72(21%) 35(20%) 36(23%) the BLL label if it has ever been confirmed to be a BLL IBL 91(26%) 41(24%) 49(31%) with a high quality spectrum, even if subsequently ob- HBL 187(53%) 98(56%) 73(46%) served in an FSRQ state. Within the BLL population Note. —BLLincludes65AGNsoclassifiedsince it is common to classify sources based on the peak fre- the 2LAC paper. 349 BLL have 2LAC SED sub- quency of their SED’s synchrotron component, as esti- classes; percentages give the breakdown. 174 of the mated from radio/optical/X-ray flux ratios, separating spectroscopic redshifts and 158 of the lower limits the sources into low-peak (ν < 1014Hz, LBL), in- havesubclasses; percentage breakdowns are given. peak termediate peak (1014Hz< νpeak < 1015Hz IBL), and vations of 278 BLL objects and 19 other Fermi blazars, high-peak (ν > 1015Hz, HBL) sources. We adopt notincluded in S12. We further analyzed75 SDSS spec- peak herethe LBL/IBL/HBLdesignationsfrom2LAC,which tra, treating them in the same manner as our new ob- provides such subclasses for 74% of all 2LAC BLL, and servations. Note that, unless we suspected a published 83% of those with spectroscopic redshifts (Table 1). redshift was erroneous, we generally did not obtain new The evolutionofBLL haslong beencontroversial,and spectra of many of the brighter, famous BLL with red- it has been claimed that they are predominantly a low shiftsintheliterature. Inseveralcaseswhenweobtained redshift population, showing strong ‘negative’ evolution new data it strongly contradictedthe literature redshift, (e.g.Beckmann et al.2003),especiallyfortheHBLclass eitherfromanewsecurespectroscopiczorfromaninter- (Rector et al.2000). Onechallengeto studiesofthe cos- vening absorber at larger z. In the end we retained 107 mologicalevolutionofthesesourcesisthedifficultyofob- redshiftsfromliteraturevalues;howeverforseveralearly tainingredshiftsfromtheirnearlyfeatureless,continuum BLLz’swehaveonlyinspectedplottedspectra(ofvary- dominated spectra. Indeed, many of the early studies ing qualities); we suspect that at least a few erroneous using X-ray or radio-selectedsamples had highly incom- values remain in this set. Our new data provide new plete redshift measurements, even though the samples spectroscopic redshifts for 102 objects and secure lower wereconfinedtorelativelybrightsources. Uncertaintyin limits for many more, as summarized in Table 1. This extrapolatingfrom the measuredset of redshifts compli- brings the 2LAC BLL sample to 44% redshift complete- cated population interpretations. In the Fermi era, this ness, and 95% completeness including redshift limits. issue becomes critical, as the large BLL contribution to theLATsourcepopulationandthehardBLLγ-rayspec- 2.2. Observations tra ensure that these sources are a major fraction of the The quest for high completeness has driven us to em- cosmic γ-ray background and may, indeed dominate the ploy medium and large telescopes in both hemispheres. LAT backgroundat high energies (Ajello et al, in prep). Observationswere obtainedwiththe MarcarioLowRes- Since the LAT provides a large, uniform, sensitivity olution Spectrograph (LRS) on the Hobby-Eberly Tele- limited (∼ flux limited) blazarsample, itprovidesa new scope (HET), with the ESO Faint Object Spectrograph opportunity to make progress on these questions. Im- and Camera (Buzzoni et al. 1984, EFOSC2) and ESO portant to interpreting the LAT blazars are the strong Multi-Mode Instrument (Dekker et al. 1986, EMMI) on correlations between the LAT-detected (IC) part of the the New Technology Telescope at La Silla Observa- SED and the synchrotron component covering the op- tory(NTT), with the GoodmanHighThroughputSpec- tical band. The synchrotron peak location determines trograph (GHTS) on the Southern Astrophysical Re- the sub-classification, but also correlates with the inten- search(SOAR)Telescope,withtheDoubleSpectrograph sity and LAT-band spectral index of the IC component, (DBSP) on the 200” Hale Telescope at Mt. Palomar, which in turn affects the depth of the LAT sample. Fur- withtheFOcalReducerandlowdispersionSpectrograph ther, as illustrated in this paper, the synchrotron peak (Appenzeller et al. 1998, FORS2) on the Very Large and intensity also affect the difficulty of optical spec- Telescope at Paranal Observatory (VLT), and with the troscopic measurements. Since we wish to recover the LowResolutionImagingSpectrograph(LRIS) atthe W. detailed evolution of the blazars, preferably also follow- M. Keck Observatory (WMKO). Observational configu- ing differences among the LBL/IBL/HBL subclasses (or rations and objects observed are listed in Table 2. even better, a physical parent property leading to these Sinceatthetimeofobservation,manyofthesesources subclasses),oneneedsadetailedtreatmentofboththeγ- were not classified, we often initially obtained only suf- ray(e.g.Ajello et al.2012a)andopticalselectioneffects. ficient S/N to identify the redshift of an FSRQ, or to We reserve such analysis for a future study, noting here firmly establish the source as a BLL. Also, with the va- only the most prominent trends in the measured optical riety of telescope configurations and varying observing sample. Of course, characterization and minimization conditions, the quality of the spectra are not uniform: of the optical biases are greatly aided by high redshift resolutions vary from 4 to 17˚A, exposure times from completeness, the goal of the present paper. 180s to 2400s, and telescope diameters from 3.58m to We have accordingly studied the 2LAC sample over a 10m. S/Nperresolutionelementvariesfrom10to>300. numberofyearswithawidevarietyoftelescopes,striving In a number of cases follow-on observations with higher to be as complete as possible. We here report on obser- S/Nand/orhigherspectralresolutionallowedustomore 3 carefullystudyconfirmedBLLlackingredshifts. Herewe ingIRAF’s de-reddeningfunctionandtheSchlegelmaps discuss the most constraining spectrum or spectrum av- (Schlegel et al. 1998). We made no attempt to remove erage for each source, referring to this as the ‘primary’ intrinsic reddening (i.e.: from the host galaxy). spectrum. Telluric templates were generated from the standard All spectra are taken at the parallactic angle, except star observations in each night, with separate templates for LRIS spectra using the atmospheric dispersion cor- for the oxygen and water line complexes. We cor- rector,whereweobservedinanorth-southconfiguration. rectedseparatelyforthetelluricabsorptionsofthesetwo Inafewcases,werotatedtheslitangletominimize con- species. Wefoundthatmosttelluricfeaturesdividedout tamination from a nearby star. At least two exposures well, with significant residuals only apparent in spectra aretakenofeverytargetforcosmicraycleaning. Typical with high S/N. On the HET spectra, residual second exposure times are 2x900s. order contamination prevented complete removal of the strong water band red-ward of 9000 ˚A. 2.3. Data Reduction Pipeline When we had multiple epochs of these final cleaned, Data reduction was performed with the IRAF pack- flux-calibrated spectra with the same instrumental con- age(Tody1986;Valdes1986)usingstandardtechniques. figuration, we checked for strong continuum variation. Data was overscan (where applicable) and bias sub- Spectrawithcomparablefluxeswerethencombinedinto tracted. Domeflatsweretakenatthe beginningofevery a single best spectrum, with individual epochs weighted night, the spectral response was removed, and all data by S/N. frames were flat-fielded. Wavelength calibration em- Due to variable slit losses andchanging conditions be- ployedarclampspectraandwasconfirmedwithchecksof tween object and standard star exposures, we estimated nightskylines. Weemployedanoptimalextractionalgo- that the accuracy of our absolute spectrophotometry is rithm(Valdes1992)tomaximizethefinalsignaltonoise. ∼ 30% (Healey et al. 2008), although the relative spec- ForHETspectra,carewastakentouseskywindowsvery trophotometry is considerably better. near the longslit target position so as to minimize spec- Fig. Set 1. Spectra troscopic residuals caused by fringing in the red, whose removal is precluded by the rapidly varying HET pupil. 2.4. General Trends Spectrawerevisuallycleanedofresidualcosmicraycon- tamination affecting only individual exposures. To illustrate the principal trends in the BLL spectra We performed spectrophotometric calibration using we refer the reader to Figure 1. By definition, the dom- standard stars from Oke (1990) and Bohlin (2007). In inant component is a power-law. J0516−6207, however, most cases standard exposures were available from the showsthatafterremovalofapower-lawweak,butbroad data night. On the queue-scheduled HET, and during CIV,CIII, and MgII features may occasionallybe seen our queue-scheduled VLT observations, standards from in high S/N spectra. Here the equivalent widths (< 1 subsequentnightsweresometimesused. Atallothertele- ˚A) are sufficiently small to secure the identification as scopes, multiple standard stars were observed per night a BLL. However, should the continuum fade by ∼ 10×, under varying atmospheric conditions and different air- thiswouldbeclassified(atthatepoch)asanFSRQ.Such masses. The sensitivity function was interpolated be- transition objects support the idea of a Blazar contin- tween standard star observations when the solution was uum,ratherthantwodistinctpopulations(Fossati et al. found to vary significantly with time. 1998; Ghisellini & Tavecchio 2008). In S12, we reported For blue objects, broad-coverage spectrographs can significant broad line measurements for 5 of our BLL, suffer significant second order contamination. In par- including J0516-6207. ticular, the standard HET configuration using a Schott While the power-laws of most BLL are very blue, like GG385 long-pass filter permitted second-order effects J0516−6207,afew like J1849+2748,appearintrinsically redward of 7700 ˚A. The effect on object spectra were flat or red, even after correcting for Galactic extinction. small, but for blue WD spectrophotometric standards, This may plausibly be a sign of a synchrotron compo- secondorder corrections were needed for accurate deter- nent peaking near the optical, but might also indicate minationofthesensitivityfunction. Thiscorrectionterm incomplete extinction correction, with residual redden- wasconstructedfollowingSzokoly et al.(2004). Inaddi- ingcausedbydustnotintheSchlegel et al.(1998)maps. tion, since BLL spectra are generally simple power laws, It could also be intrinsic host extinction. we used our objects to monitor second order contami- The Galactic reddening can be very severe. For nation and residual errors in the sensitivity function. In J2001+4352 (upper right) we show both the highly extremecases,wefitanaveragedeviationfrompowerlaw extincted, pre-correction spectrum and the blue post- acrossallobjectsinanight,andtreateditasacorrection correction power law. This source is in a direction of to our spectrophotometric calibrations. This resulted in known high AV = 1.75. Such extincted power-law spec- excellent, stable response functions for the major data traprovideanexcellentopportunityforISMstudies: The sets. featuresseenafterde-extinctionanddivisionbythebest- Spectra were corrected for atmospheric extinction us- fit power law (lower panel) are all interstellar in origin ingstandardvalues. WefollowedKrisciunas et al.(1987) – Galactic H and K, Na I 5892, and a series of diffuse forWMKOLRISspectra,andusedthemeanKPNOex- interstellar bands, as described in Yuan & Liu (2012). tinction table from IRAF for P200 DBSP spectra. Our For J0124−0625 (upper left) the residual absorption NTT,VLT,SOAR,andHET spectradonotextendinto featuresareintergalacticinorigin. Redwardof3900˚Awe the UV and so suffer only minor atmospheric extinc- detectanumberofmetal-linesystems,bluewardonesees tion. Thesespectrawerealsocorrectedusingthe KPNO the onset of strong Lyman-α forest absorptions. These extinction tables. We removed Galactic extinction us- features determine a redshift z =2.117 (one of the high- 4 M. S. Shaw et al Table 2 ObservingConfigurations Telescope Instrument Resolution SlitWidth Objects Filter λmin λmax ˚A Arcseconds ˚A ˚A HET LRS 15 2 41 GG385 4150 10500 HET LRS 8 1 8 GG385 4150 10500 NTT EFOSC2 16 1 31 - 3400 7400 NTT EMMI 12 1 1 - 4000 9300 Palomar200” DBSP 5/15 1 4 - 3100 8100 Palomar200” DBSP 5/15 1.5 5 - 3100 8100 Palomar200” DBSP 5/9 1.5 42 - 3100 8100 SOAR GHTS 6 0.84 2 - 3200 7200 VLT FORS2 11 1 14 - 3400 9600 VLT FORS2 17 1.6 16 - 3400 9600 WMKO LRIS 4/7 1 90 - 3100 10500 WMKO LRIS 4/9 1 40 - 3100 10500 Note. — ForDBSPandLRIStheblueandred channelsaresplit bya dichroicat 5600 ˚A; thelisted resolutions are for blueand red side, respectively. est in our BLL sample). The lack of similar Lyα forest light,whichthesignificantemissionfeaturesallidentified absorption in many of our other high S/N, high resolu- as z =0 stellar or ISM features. tionspectraallowsusto placestatisticalupper limits on Inafewcases,objectspreviouslycatalogedasBLLsdo the redshift as described in §3.3. nothavesufficientS/NinourspectraforadefinitiveBLL Finally at lower right we see two BLL with significant classification. For J0801+4401, we find that undetected flux from the host. In J2042+2426, the galaxy provides broadlinescouldhaveanequivalentwidthaslargeas9.5 about a third of the total flux and is easily visible in the ˚A; for J0209-5229the limit is 5.6 ˚A, for J1311+0035the raw spectrum. This is still safely a BLL, and we can limit is 8.0 ˚A and for J1530+5736we could have missed measure the continuum contributionby a ‘Non-Thermal linesasstrongasEW=5.5˚A.AshigherS/Nspectroscopy Dominance’ (see §5, here NTD=1.38). For J2055−0021, would likely confirm the archival BLL designations, we the host is swamped by the core synchrotron emission consider them BLLs for the purposes of this paper. (NTD=47.5) and the galaxy features are visible only af- Five of the BLL described here had high significance ter subtracting the best-fit power law, as in the lower broadlinedetectionsandhavealreadybeendescribedin rightpanel. The flux increaseinthe blue appearsdue to S12;were-measurethesespectrahereforauniformBLL residualfew-%fluxingerrors(here,likelyincompletecor- treatment. InJ0847−2337andJ0430−2507,thefluxand rection for atmospheric extinction), rather than intrin- spectral index measurements differ from the S12 values. sic emission. Despite the careful calibration, such resid- This is because in the present analysis we first subtract ual fluxing issues persist in several spectra. However, the host galaxy flux, and calculate the flux and spectral the high-pass filtering described in §3.4 ensures that our index of the remaining non-thermal component. In S12, measurements of, and bounds on, host galaxy flux are no such correction was attempted. almost completely immune to such residual calibration artifacts. We find that a number of BLL show visible 3. MEASURINGBLLREDSHIFTS hostgalaxycomponents,allconsistentwithgiantellipti- TheopportunitytoadvanceourunderstandingofBLL cals (Urry et al. 2000). We discuss the flux distribution evolutionwiththelarge,fluxlimitedFermisampleisim- of these host galaxies in §3.7. portant (Ajello et al. 2012b). Yet, despite the substan- tial telescope resources and careful analysis summarized 2.5. Individual Objects above,many BLL did not yield directspectroscopic red- A number of BLL reductions required special treat- shifts,duetotheextremeweaknessoftheiremissionlines ment. Forexampleafewobjectsclearlyrequiredchanges (Sbarufatti et al. 2005b) and lack of clear host features. to the Schlegel map A , so that the de-extinction re- Therefore we collect here both the direct redshift mea- V sulted in clean power laws. For J0007+4712, A was surements and quantitative constraints on the allowed V increased from 0.3 to 0.8 and for J1941−6211 from 0.3 redshift range for our observed BLL. to 1.0. Conversely we decreased the A of J1603−4904 V from 7.8 to 5.0 and J2025+3343, from 6.15 to 5.0. We 3.1. Emission Line Redshifts checked the recent recalibration of the Galactic extinc- We visually inspected all spectra for AGN emission tionmaps(Schlafly & Finkbeiner2011),butdidnotfind linefeatures,andhostgalaxyabsorptions. Spectroscopic large changes, so these extinction features affecting our redshiftsaremeasuredbycross-correlationanalysisusing BLL are probably on scales below the map resolution. thervsaopackage(Kurtz & Mink 1998). Werequireone J1330+7001wasobservedoffoftheparallacticangle— emissionlinetobepresentatthe>5σlevel,andasecond the ensuing drop in blue flux is not intrinsic to the sys- linepresentatthe>3σlevel—significancesaremeasured tem, and our power law is fit only redward of 5000 ˚A. byfittingaGaussiantemplateinthesplottool;weallow We thus remove the broadband residual in our power thewidthandamplitudeofthe Gaussiantovary,butfix law divided spectrum in Figure 1. J1829+2729 and a the center at the redshift derived by rvsao’s xcsao rou- nearby star were spatially unresolved in our data. The tine. For this study, we limited our search to typically presentedspectrumisacompositeofstarlightandquasar strong emission lines known to be present in some BL 5 Figure 1. Spectra of Fermi BLLs. Each object is presented twice – directly in the upper panel, and then with the best-fit power law ‘removed’ (generally by division, but for composite spectra by subtraction). This residual is plotted in the lower panel. SDSS spectra, whilediscussedinTable3,arenotreplottedinthisFigure. Thesixsamplespectrahereillustratemajortrends. Similarfiguresforthefull BLLsampleareavailableinthe electroniceditionofthis journal. Therewemarkonlythetwo linesusedto securezspec;hereother lines of interest aremarkedand the two qualifyinglines aremarkedinred. z limitsgivenare the mostconstraining limitspresented—2 digits aregivenforlimitsfromnon-detectionofhostgalaxies;3digits,forinterveningabsorptionsystems. Themarkedabsorptionsystemisthe bestspectroscopiclimit;insomecases,astrongerhostgalaxylimitispresented. SeeTable3fortheprecisespectroscopiclimit. Lacs: Broad emission from C IV (1549, 1551 ˚A), C III termined by a significant (> 3σ) detection of a host (1909 ˚A), MgII (2796, 2799, 2804 ˚A), Hγ (4340 ˚A), Hβ galaxy, as will be described in §3.4. A few redshifts (4861 ˚A), and Hα (6563 ˚A) and narrow emission from require further note: In J0124-0625 and J1451+5201, [O II] (3727, 3729 ˚A), [O III] (4959, 5007 ˚A), and [N II] we identify the redshift by a Lyα and C IV absorption system at the onset of the Lyα forest. In J0434-2015, (6549, 6583 ˚A). While other species exist in our spec- we identify a single strong feature with [O II], consis- tra, these here listed are sufficient for definite redshift tent with weak Mg II and Ca H/K absorptions. For IDs. Velocities are not corrected to helio-centric or LSR J1728+1215, we find strong Mg II, confirmed by [O II] frames. at the same z in an archival spectrum. In J2152+1734, In many cases, spectroscopic redshifts are further de- 6 M. S. Shaw et al we identify a strong feature with Mg II confirmed by a require the stronger (bluer) line to have > 5σ signifi- significant [OII] detection in archival spectroscopy. cance, and the second line to have >3σ significance. In For a few objects only a single emission line was mea- a few cases, one component of an otherwise strong dou- sured with high S/N. In general we use the lack of oth- blet was affected by skylines, telluric features or cosmic erwise expected features to identify the species and the rays. In these cases, another expected feature from the redshift with high confidence. Nevertheless, a few red- absorption complex (e.g.: a Fe II line) detected at > 3σ shifts have some systematic uncertainty and are marked qualified the system. We further require the doublet ra- by a ‘:’ in Table 3. We briefly discuss these cases here. tio to be consistent (within errors) to a value between For J0007+4712,we derive a redshift from the clear on- 2:1 and 1:1. setoftheLyman-αforestandreportonlytwosignificant Our principal goal is not an absorption line study. figures. InJ0212+2244,wedetermine a tentativez from Thus we concentrated on the longest wavelength (high- weak Ca H, K and g-band absorptions. For J0439-4522, est z) candidate system and measured sufficient lines to we identify the one strong emission feature as C IV; in- confirmthez (i.e: twosignificantlines). Aftervalidation terveningabsorptionexcludes a MgII identification, but we continued to search for higher z until no candidates aless likely CIIIidentificationatz ∼1.45is notconclu- passed the significance test. We therefore believe that sively ruled out. J0629−1959 presents broad but weak we have found the highest z absorption system in each emission at the redshift of the highest z metal line ab- of our spectra strong enough to reach the 5σ/3σ criteria sorption system (1.724). We thus identify this, tenta- above. tively, as the object’s true z. For J0709−0255,we iden- Since we see the onset of the Lyman-α forest in our tify the strong feature at 9200 ˚A with [OIII] by the line highestredshiftobjects,theredendoftheforestprovides shape; an [O II] identification at z ∼ 0.84 is not ex- a strict lower limit on redshift.This can be higher than cluded. ForJ0825+0309,wefindsignificant[OIII]emis- that inferred from the reddest metal line system. sionat5007˚A(andpossible,butnotsignificantemission Welistthesespectroscopicminimumz’saszmin,when at4959˚A), ata z consistentwith an MgII feature iden- available, in Table 3. tified in Stickel et al. (1993). We find weak features in 3.3. Redshift Upper Limits J1231+2847 at the SDSS z, but they have low signifi- cance. For J1312−2156, we find a plausible Mg II fea- We can use the absence of Lyman-α absorptions to ture; this single line identification is in a small allowed providestatistically-basedupperlimits onz forallBLLs redshift range (z ∼ 1.6), other identifications for this without redshift. The exclusion z depends on the max line are spectroscopically excluded. In J1754-6423, we spectral range, resolution and S/N of the particular ob- tentatively identify emission at ∼ 6300 ˚A with Mg II— servation, but is generally 1.65<z <3.0. higherz redshiftsareexcludedby the lackofLyαforest. To quantify the upper bound, we need the expected In J2116+3339’s spectrum, a significant broad emission densityofLyαforestabsorbers. Penton et al.(2004)find feature is identified with C IV, consistent with a weak that for rest EW ≥ 0.24 ˚A, dN/dz ∼ 40 at z = 1.6, bump in the far blue at Lyα. Nevertheless a lower z varying with redshift as logdN/dz ∝ 1.85log(1 + z). redshift is possible if the purported Lyα line is not real. We follow Weymann et al. (1998) for the EW scaling: J2208+6519presentsonestrong,broademissionfeature, dN/dz ∝ e−(EWrest−0.24)/0.267. As the S/N and resolu- tentatively identified as Mg II—a C IV identification at tion of our data vary, we generally measured a conser- z ∼1.8 is not excluded. vativeuniform3σ sensitivityfornarrowLyαabsorptions 3.2. Intervening Absorbers 100˚Afromtheblueendofourspectra. TypicalEWlim- its in this range were ∼ 0.2−1.0 ˚A. Given this density Forsome BLLs,the corelightpassesnear aninterven- wesolveforthe ∆z rangegivingaPoissonprobabilityof ing galaxy on its way to Earth. At small radii one can 0.32 (ie: 1σ) for detecting no absorbers, obtaining: encounterlowexcitationcloudsinthegalaxy’shalo,giv- ing absorption doublets from Mg II at (2795.5, 2802.7) λ −1.85 ˚A.LargerimpactparameterscansampleCIVat(1548.2, ∆z =0.167·e(EWrest−0.24)/0.267)· min (1) (cid:18)1215(cid:19) 1550.77)˚A. Insome low excitation(i.e.: MgII) systems, we also see absorption from Fe II at (2344.2, 2374.4, whereλ istheeffectiveblueendofthespectrum(gen- min 2382.7, 2586.6, 2600.1) ˚A. Finally, for our highest red- erally3150–4200˚A),andEWisthemeasuredequivalent shift BLL we Lyman-α absorption at 1215 ˚A for the width limit. Thus, we infer a maximum source redshift metal line systems, as well as onset of the Lyman-α for- z = (λ /1215)−1+∆z. In some cases, the S/N max min est. is too low at the blue end of the spectrum. We then For unsaturated absorptions, the doublet ratio for measure EW limits closer to the sensitivity peak of the Mg II and C IV is 2:1, with the blue line dominant. spectrum. Of course, with a larger λ for the effective min In saturated absorptions, the ratio is 1:1 (Nestor et al. spectrum end Equation 1 gives less constraining upper 2005; Michalitsianos et al. 1988). limits. We visually search all spectra for candidate doublets, If the actual blazar redshift z is very close to z as max and follow Nestor et al. (2005) in employing a quantita- estimatedabove,thenitsUVradiationmayphoto-ionize tive test of the significance of each candidate. We used Lyα clouds along the line of sight, postponing the onset a two Gaussian template with wavelength spacing scal- of the forest and artificially lowering z . In practice, max ing with z, but free amplitudes, and fit the equivalent the effect is usually small (∆z ∼ 0.01−0.02) except max width and error of each component, using the splot tool for large z when our bound is generally not of in- max in iraf. For the candidate to qualify as a detection we terest. We follow Bajtlik et al. (1988) to estimate this 7 ‘proximity effect’ correction, by computing based on a Levenberg-Marquardtfitter. To model host slit losses, we assume an r = f (1+z )5 (1+z)1/2−1 2 10kpc de Vaucouleurs profile with Sersic index 1/4 ν Lyα ω(z )= Lyα 4πJ (1+z) (cid:20)(1+z)1/2−(1+z )1/2(cid:21) (O’Dowd & Urry 2005) and account for the individual ν Lyα observations’ slit width and seeing profile (measured (2) from the core full width at half maximum, FWHM). where f is the blazar flux at the absorption Lyman limit (atνz ) and J = 10−21.5ergcm−2s−1Hz−1sr−1 Since we have employed optimal extractions of the BLL Lyα ν spectra, our effective aperture along the slit varies, but is the cosmic ionizing flux, conservatively estimated for we estimate a typical width of ≈2× the spatial FHWM our redshift range. We compute f from the power law ν achievedduringourspectralintegration. Accordinglywe fit in Table 3, and increase the blazar redshift z until estimate host slit losses through a rectangular aperture ω(z )<1. We thus adopt these revised z =z and Lyα max of the slit width × twice the spectrum FWHM. Inferred quote these corrected upper limits in Table 3. host fluxes are corrected for these slit losses. ResultsofsamplefitsareshownasthebluedotsinFig- 3.4. Host Galaxy Fitting ure2wherethefitamplitudeofthehostgalaxytemplate Ithasbeen claimedthat BLLac objectsarehostedby is plotted against trial redshift. giant elliptical galaxies with bright absolute magnitude 3.5. Power Law Fit – M =−22.9±0.5 in our cosmology (Urry et al. 2000; R Sbarufatti et al. 2005b). We reportthe power law fluxes and spectralindices of If we adopt the common assumption that these are thebestfittothede-extinctedspectruminTable3. The standardcandles,wecanestimatetheredshiftoftheBLL fluxisgiveninunitsofLog10−28ergcm−2s−1Hz−1asob- bydetectingsuchgalaxies. Inimagingstudies,onelooks served at 1014.7Hz (∼5980 ˚A), the center of our typical in the wings of the BLL for the host galaxy flux, and spectralrange. TheindexαismeasuredF ∝να. These ν comparesthattothestandardcandlefluxatvariousred- values may be combined with multi-wavelength data to shifts(Sbarufatti et al.2005a;Meisner & Romani2010). study the continuum SED of the blazars in our sample. In spectroscopic studies, one typically looks for individ- Since the statistical errors on the fit are, in general, un- ual absorption features (i.e.: H, K, g-band). One can physically small, we follow S12 in estimating errors on also use the lack of such lines as evidence that the BLL the spectral index by independently fitting the red and is at higher redshift (Sbarufatti et al.2005b; Shaw et al. blue halves of the spectrum. Note that large errors bars 2009). generally indicate a relatively poor fit to a simple power With high S/N spectra, however,one canobtain more law rather than large statistical errors. The statistical stringent limits by using the entire elliptical template errors on the F amplitude are also small; we convolve ν ratherthanjustone(orafew)lines. Plotkin et al.(2010) thesewithourestimated30%overallfluxinguncertainty developedatechniqueoffittingforhostgalaxiesinSDSS (Healey et al.2008),whichdominatesinnearlyallcases. BLL spectra. We expand here on that method for our For objects with high significance (> 3σ) detections more heterogeneous spectra. of galaxies, as described in §3.7, we report the best fit Our spectra come from a variety of spectrographs in power law from the simultaneous power-law/host fit in disparateobservingconditionsandwefindlowfrequency §3.4. systematic fluctuations in many of the fits. These are Whenwehaveobservedobjectsatmultipleepochs,we likely caused by imperfect spectrophotometric fluxing also fit a power law to the other, non-primary spectra. and second order contamination as discussed in §2.3. Thesefluxesvarysubstantially,someby morethan10×. These effects can dominate over real galaxy features. In Figure 3, we show the distribution of f /f flux max min Using SciPy’s Signal Processing routines4, we construct ratios. This is well-described by an α=1.29 power law. a bandpass Kaiser window from 220˚A to 1.5× the Epochs from our fiducial spectra are listed in Table 3. Nyquist frequency. We apply that window as an effec- 3.6. Testing the Standard Candle Assumption tive high pass filter both to our spectra and to the tem- plates,tomitigatethislow-frequencynoisebeforefitting BLL with a redshift and a secure (> 3σ) host de- (Kaiser & Schafer 1980). tection can be used to test the uniformity of the host We test possible redshifts z on a grid scaled to the luminosities. There are 59 such BLL in our sample. i spectrograph resolution with constant spacing in logz. We derive synthetic R-band magnitudes by applying a This grid is thus z = (2∆λ + 1)i − 1, where ∆λ is Kron-CousinsRfiltertoourspectra(Meisner & Romani i λ0 2010). The results are shown as a histogram in Figure the pixel scale of the spectrograph. For each trial z i we fit the power law F ∝ να index and flux and 4. We find < MR >= −22.5, down ∼ 0.4 magnitudes ν fromM =−22.9foundinSbarufatti et al.(2005b). We the amplitude of a redshifted elliptical template. This R findasimilarspreadinluminosity(∼±0.5magnitudes). host template is generated from the PEGASE model Whenweseparatethehostmeasurementsforlower-peak (Fioc & Rocca-Volmerange 1997) tables and evolved to sources (LBL+IBL) we find that they have a median lowz fromz =2,asin O’Dowd & Urry(2005). Foruni- luminosity ∼ 0.3 magnitudes fainter than that of our formity,we here use the same template for all z , anddo i HBL.Unfortunately,wedonothavesufficientLBL+IBL not perform any evolution corrections. Our fit minimizes χ2 with three free variables at each hosts to test for such differences at high significance. Past studies differ: Urry et al. (2000) found no signif- trial z . We employ the scipy.optimize.leastsq routine i icant offset in the host magnitudes of HBL and LBL, 4 Documentation and more information available at butSbarufatti et al.(2005b)notedthathigherpeakHBL http://docs.scipy.org/doc/scipy/reference/signal.html tend to have more luminous hosts. 8 M. S. Shaw et al Figure 2. Best fit host galaxy fluxes as a function of trial zi (blue dots). Our estimate of the 2σ local systematic flux error is shown by the blue line (see text). The red bands give the 1σ range about the expected host flux for MR =−22.5±0.5. Vertical green dashed lines give our zmin(−22.5) limit on the redshift. For comparison, solid green lines give the spectroscopic redshifts, when measured. For J2055−0021 and J0148+0129 these are consistent; the former has a high significance host detection (Figure 1), while for the latter the higherzredshiftisfromemissionlines. J1534+3715isoneof9objectswithlimitsinconsistentwithspectroscopicredshifts. Here,thehost galaxyissub-luminous(MR=−21.87±0.16)andthus (just)missedbythistechnique(Prob=0.15). Two of the high significance host galaxies have imag- assumption M =−22.9±0.5, for more direct compari- R ing magnitudes reported in Sbarufatti et al. (2005b). sontopreviouswork,butwerecommendadoptionofthe For J1442+1200, we measure M = −22.99 ± 0.14; less stringent M =−22.5±0.5 redshift constraints. R R Sbarufatti et al. (2005b) reported M = −22.77. For R J1428+4240, we measure MR = −22.78 ± 0.13; 3.7. Lower Limits from Non-detections of Host Galaxy Sbarufatti et al. (2005b) find M = −22.75. These are R Weusetheresultsofthefittingin§3.4andthecalibra- consistent within measurement errors, a good check of tionin§3.6toconstraintheredshiftofthehost. Ateach our slit-loss corrections and magnitude estimates. trial redshift, our fitter reports a flux (f ±σf) for the Overall,theLATBLLsamplethusrepresentsafainter host galaxy. This is to be comparedwith the model flux host population of than those studied in previous work. from the redshifted standard candle elliptical template Conceivablyahigher(LBL+IBL)/HBLratioinoursam- (f ±∆f ). ple causes part of the difference (although we remain M M While we have greatly decreased the effect of the low HBL dominated). However, we suspect that our rather frequency noise in our spectral fits using the high-pass exhaustive 8-m class campaign, skipping most objects filter, we still find that the statistical errors on the fit withredshiftsalreadyintheliterature,selectsforfainter host galaxy fluxes are unrealistically small; these flux host galaxies than in the past. Thus we may be probing estimates remain dominated by systematic effects. We fainteronthetruehostluminositydistribution;the BLL therefore construct an error estimate ∆f for each z by forwhichwewerenotabletoprovidehostdetectionsmay i measuringthedispersionoffluxestimatesfornearbyred- then be similarly under-luminous compared to previous shift bins. This is computed from a sample of the 30 studies. Atrueevaluationoftheintrinsichostluminosity nearest f(z), skipping 5 bins on each side of our test z distribution,aswellasanydependenceonsubclasstype, i to minimize high pass correlation. After σ-clipping, the willrequireacarefulassessmentoftheparentpopulation fitfluxdistributionsarewellbehavedandweusetheseto (γ-ray) and host detection selection biases. compute a 2-σ upper limit (scaledfromthe rms)ateach In the restofthis section, we conservativelyadoptour z . Vectorsoftheseupperlimitsareshownbythejagged M = −22.5±0.5 estimate. We do also report (Table i R blue lines in Figure 2. These have captured the local ef- 3) more aggressive redshift limits based on the common fective noise quite well and so we adopt these vectors as 9 z. Thuswecompute aprobabilitythatthefitflux f and error∆f areconsistentwiththeexpectedmodelfluxf M and its uncertainty at the given z as ∞ Prob(z;f,∆f)=C G(f′ ,∆f′ )·G(f,∆f)df′ (3) Z M M M fM where G(x,σ ) is a Gaussian of width σ centered at x. x x The normalization C is chosen such that Prob = 1 for f = 0 (i.e.: any f is acceptable for a model of zero M flux). Thisisaconservativechoiceasitdoesnotexclude over-luminous hosts. For example, when we assume a model M = −22.5±0.5 this probability also finds any R f consistent with M = −22.9±0.5 to be acceptable. R When the fitter returns an unphysical negative f, we evaluate the probability for f =0 and the local ∆f. Thisprobabilitybecomessubstantialforz nearagood candidate redshift. It also grows as the sensitivity of our host search drops at large z. We thus calculate a minimum redshift(z ) as the lowestredshiftfor min(−22.5) which Prob≥0.17 (i.e.: 1σ). We list these values in Ta- ble3. Forcomparison,wealsogivez calculated min(−22.9) Figure 3. Histogram of maximum flux ratios. For each object inthesamefashion,assumingamodelMR =−22.9±0.5. wherewehavecollectedmultiplespectra,wecomputethefluxratio Comparison between the spectroscopic detections and between the power law components of the brightest and faintest z suggest that the M = −22.5 value is most con- observed epoch. These are plotted as a log-log histogram. The min R sistent with observed detections, and lower bounds (as dashedlineisthebest-fitpowerlawwithindexα=1.29. Theinset histogram shows the distribution of time between observations – expectedfrom§3.6). Werecommenduseoftheseconser- typically∼1year vative lower limits. Note also that the vector Prob(z), oncenormalizedwithanappropriatepriorandtruncated at z , can be used as a PDF for the BLL redshift. max 3.8. Redshift Distribution AsseeninFigure5,archivalredshiftmeasurementsfor BLL are dominated by low values (z˜= 0.23). Our new spectroscopic redshifts have extended the population to higherz,withsomeobjects’redshiftsatz ≫1. Still,the objectswithredshiftsremaindominatedbylowz. Inthe new spectroscopicredshifts we find z˜=0.33. We believe there is a significantbias to low redshift in both of these samples,astheweaklowEWemissionorabsorptionfea- tures of our typical BLLs with known redshift are easier to detect at low z. InFigure5wealsoshowtwosetsofredshiftlowerlim- its. Foreveryobjectinoursample,wecanderiveahost- detectionlimit(z˜=0.41). Theselowerlimitsonredshift arestillbiasedlow: asevidentfromFigure2wearemost sensitiveto galaxiesatlow z. Nevertheless,they suggest that these objects are not consistent with the spectro- scopic redshifts (the K-S test gives probability < 10−11 of consistency with archival redshifts). The absorption line limits we have for some objects (described in §3.2) Figure 4. MeasuredBLLhostabsolutemagnitudes(Requivalent give further evidence for a population of BLLs at high atz=0). Blackhistogram: allBLLhostswithspectroscopicred- redshift (z > 1). Together, these results strongly im- shift. Aweightedχ2fitgivesMR=−22.5±0.5,ourbestestimate ply that previous BLL studies suffered important biases fortheluminosityofBLLhostgalaxies. Red: hostswithhighsig- nificance (>3σ) detections. GreenandBluesub-histogramsshow due to shallow samples with large redshift incomplete- highsignificanceLBL+IBLandHBLhosts,respectively. ness preventing detection of bright, but high z BLL. The inset shows the spectroscopic redshifts and the effective 2σ confidence limits. redshiftslimits forhigh-peaked(HBL) andlower-peaked Fit fluxes well abovethese 2σ limits denote likely host (LBL+IBL) sources classified in 2LAC. We see that the detections. Indeed, we found that this automatic pro- lower-peaked detections extend to higher z, as might be cessing was quite effective at flagging candidate z for expected if these sources are more luminous and have i hostdetection. Here,however,wefocusonhowwellafit a less dominant synchrotron continuum. However, the fluxf withlocaleffectiveerror∆f canbeusedtoexclude openhistogramsoflimitsremindusthatbothsub-classes ahostofthe expectedmagnitude f atthe testredshift still suffer substantial redshift incompleteness, and the M 10 M. S. Shaw et al LBL+IBL) we fill these in with blue or green, respec- tively. For comparison, we show the 1σ spread of virial- estimateBHmassesfromopticallyselectedSDSSquasars from Shen et al. (2011) (gray band) and masses of the Fermi FSRQs in S12 (red points). Of course, if BLL hosts really are standard candle ellipticals, then the Mbulge−M• relationimpliesconstantblackholemasses. The masses corresponding to the standard M = −22.9 R and our revised −22.5 are shown by dashed lines. Interestingly, our BLL M• estimates increase with z much as the optical QSO or FSRQ. Of course, we only plothighsignificancehostdetectionshere,andlowlumi- nosity hosts at high z are increasingly difficult to detect (unless the core luminosity decreases). Accordingly, as for the QSO, we suspect that the bulk of this trend is due to selection effects. In the case of the FSRQ, S12 argued that the offset to smaller black hole mass was at least partly due to a preferentially polar view of an equatoriallyflattened broadline region,withthe projec- tion decreasing the observed kinematic line width and the average virial mass estimate. Like γ-ray selected FSRQ, BLL are Doppler-boosted along our line of sight (Urry & Padovani 1995). However since the host flux is nearly isotropic,we expect little alignmentbias in our M• estimates. Thus,itisunclearwhethertheBLLoffset Figure 5. BLL redshift and redshift lower limit distributions. tolargerblackholemassesisrealorselectiondominated. Within each δz = 0.2 bin, we show (left to right) the Archival z, ournewspectroscopicz,lowerlimitz’sfromhostfittingandlower In a study of BLL hosts detected in the SDSS limitz’sfrominterveningabsorptionlinesystems. Thelimitsshow Le´on-Tavareset al.(2011)foundnosignificantdifference that the spectroscopic redshift samples, particularly the archival between the masses of the central black holes of HBL sample,areselectionbiasedtolowz. Thesubframeshowsthered- andLBL.Infigure6theHBLmassesarehoweverbiased shifts(filledhistograms)andlowerlimits(openhistograms)forthe LBL+IBL (green) and HBL(blue). Again, both limithistograms upwards with respect to the lower-peak BLL black hole extendtohigherz. TheLBL+IBLsampleextends somewhatfur- masses. This is of course just a restatement of the off- therthantheHBL. set in host luminosity seen in Figure 4. Unfortunately, missing redshifts for both sub-classes extend substan- we cannot claim that this is a physical difference as the tially higher than those in hand. A re-appraisal of the trend is precisely what would expect from selection bias BLL population, properly including the new redshifts, z if HBL have brighter non-thermal cores. constraints and remaining selection biases is needed to Ideally we could use these black hole masses to ex- test whether either subclass is still consistentwith nega- plorethe relationshipbetweenthe BLLs andthe general tive cosmologicalevolution. QSOpopulation. Largeblackholemasses,ifnotinduced solely by selection bias would imply a late stage ofAGN 4. BLACKHOLESANDHOSTGALAXIES evolution. The black hole mass is often comparedto the The masses of the central black holes provide impor- source luminosity to characterize the state of the accre- tant insight into the cosmic evolution of various AGN tion in Eddington units. However,with the exception of classes. These are most easily estimated by the virial the few BLL for which we see broadlines (which seldom technique (cf., Shen et al. 2011). In S12, we adopted have a significantly host detection), the observed flux is this method to give mass estimates for the Fermi FS- so dominated by beamed jet emission that quoting the RQs. For BLL, the lack of high S/N broad line mea- accretion luminosity in Eddington units is not feasible. surements precludes such estimates. However, we have 5. NON-THERMALDOMINANCE measured a number of host magnitudes in §3.4; since these are ellipticals, this is all ‘bulge,’ and we can apply In S12, we introduced the non-thermal dominance an M −L relation to estimate the hole mass. We follow (NTD) as a quantitative measure of how much the op- Gu¨ltekin et al. (2009): tical is contaminated by non-thermal synchrotron emis- sion. We here extend that analysis to BLLs. M LV For most BLL, the dominant ‘thermal’ contribution is log =(8.95±0.11)+(1.11±0.18)log (cid:18)M⊙(cid:19) (cid:18)1011L⊙,V (cid:19) thehostgalaxy,notthebigbluebump. Wethereforeset (4) NTD ≡ Fcore/Fhost where both fluxes are measured at whereM istheblackholemass,andL istheluminosity 5100˚A–thesamewavelengthastheHβ continuummea- V in a V filter [log(LV/L⊙,V) = 0.4(4.83−MV0,bulge)]. To surements for FSRQs. The wavelength choice is impor- convert our fit template spectrum amplitudes to consis- tant for BLL NTD measurements, since the host galaxy tent V magnitudes, we integrate our template spectrum is much redder than the continuum-dominated core; an over the Hubble F555W filter (Lauer et al. 2005) as in NTD measurement just above (redward of) the 4000 ˚A §3.6. breakwould typically give values &4× larger. Measure- The masses from Equation 4 are plotted as circles ments below the break would, of course, diverge even in Figure 6. When the sub-class is known (HBL or more.

ebook img

Spectroscopy of The Largest Ever Gamma-ray Selected BL Lac Sample PDF

1.6 MB·English
by  Michael S. Shaw
#additional_collections #journals #arxiv
Save to my drive
Quick download
Download
Upgrade Premium
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Spectroscopy of The Largest Ever Gamma-ray Selected BL Lac Sample

Accepted for publicationin The Astrophysical Journal PreprinttypesetusingLATEXstyleemulateapjv.5/2/11 SPECTROSCOPYOF THE LARGEST EVER γ-RAY SELECTED BL LAC SAMPLE Michael S. Shaw1, Roger W. Romani1, Garret Cotter2, Stephen E. Healey1, Peter F. Michelson1, Anthony C. S. Readhead3, Joseph L. Richards3, Walter Max-Moerbeck3, Oliver G. King3, William J. Potter2 Accepted for publication inThe Astrophysical Journal ABSTRACT We report on spectroscopic observations covering most of the 475 BL Lacs in the 2nd Fermi LAT 3 catalog of AGN. Including archival measurements (correcting several erroneous literature values) we 1 now have spectroscopic redshifts for 44% of the BL Lacs. We establish firm lower redshift limits 0 via intervening absorption systems and statistical lower limits via searches for host galaxies for an 2 additional 51% of the sample leaving only 5% of the BL Lacs unconstrained. The new redshifts raise themedianspectroscopicz˜from0.23to0.33andincluderedshiftsaslargeasz =2.471. Spectroscopic n redshiftminimafrominterveningabsorbershavez˜=0.70,showingasubstantialfractionatlargezand a J arguing against strong negative evolution. We find that detected BL Lac hosts are bright ellipticals 2 with black hole masses M• ∼ 108.5−9, substantially larger than the mean of optical AGN and LAT FlatSpectrumRadioQuasarsamples. AslowincreaseinM• withz maybeduetoselectionbias. We ] find that the power-law dominance of the optical spectrum extends to extreme values, but this does E not stronglycorrelatewith the γ-rayproperties, suggestingthat strong beaming is the primary cause H of the range in continuum dominance. . Subject headings: BL Lacertae objects: general — galaxies: active — Gamma rays: galaxies — h quasars: general — surveys p - o 1. INTRODUCTION now decreased to 215 (19%), and the confirmed BLLs r st The Fermi Second Source Catalog (Nolan et al. 2012, have increased to 475 (42% of all 2LAC AGN). In§2,weoutlinethesampleproperties,datacollection, a 2FGL) lists the 1873 most significant sources detected [ by the Large Area Telescope (Atwood et al. 2009, LAT) and data reduction steps. We also summarize principal features of the spectra. In §3, we describe our spectro- duringFermi’sfirsttwoyearsofskysurveyobservations. 1 scopic constraints on the redshift, including a technique The majority of these sources are associated with jet- v to provide uniform redshift limits based on searches for dominated Active Galactic Nuclei, the so-called blazars, 3 hostgalaxyemission. In§4,wegiveestimatesoftheBLL many of which are bright, compact radio sources. There 2 3 are, in fact, 1121 such associations (1017 at |b| > 10◦), blackholemasses. Weturntocommentsontheprincipal BLL feature, the non-thermal dominance in the optical 0 collectedinthe SecondCatalogofAGN Detectedby the in §5, and conclude with general remarks in §6. . FermiLAT(Ackermann et al.2011,2LAC).TheseAGN 1 In this paper, we assume an approximate concordance are further classified as Flat-Spectrum Radio Quasars 0 cosmology – Ω = 0.3, Ω = 0.7, and H = 70 km s−1 (FSRQ) where the optical spectrum is dominated by m Λ 0 3 Mpc−1. thermal disk and broad-line region emission, BL Lacs 1 (BLL),where the opticalspectrumis dominatedby con- : v tinuumsynchrotronradiation,andacollectionofmiscel- 2. OBSERVATIONSANDDATAREDUCTION i laneous,mostlylowluminosityrelatedsources. In2LAC, 2.1. The BLL Sample X the sample included 410 BLL, 357 FSRQ, 28 AGN of BLLswereoriginallyidentifiedasopticalviolentlyvari- r other type (generally low z, lower luminosity Seyferts), a able AGN, and are often characterized by an optical and 326 AGN of (then) unknown type. continuum dominated by synchrotron emission. Their These ‘Blazars’ (BLL and FSRQ) are the brightest broad-band spectral energy distribution (SED) is de- extra-Galactic point sources in the microwaveand γ-ray scribed by a synchrotron component peaking in the far- bands; study of their population and evolution are cen- IR to X-ray bands and an Inverse Compton (IC) com- tral topics in high energy astrophysics. To support such ponent peaking in the MeV to TeV range. In the radio studieswehaveacquiredsensitivespectroscopicobserva- these sources display strong core dominance. Accord- tions of this sample. In a companion paper (Shaw et al. ing to the unified model (Urry & Padovani 1995) BLLs 2012, hereafter S12), we reported on measurements of are the beamed counterparts of the FR I radio galaxy a large fraction of the FSRQ. Here we concentrate on population, while the FSRQ are associated with FR IIs. the BL Lac objects. Our study has also found types However, the principal BLL characteristic, a dominant forsomeofthe unclassifiedblazars;the ‘unknowns’have and varying synchrotron/IC continuum, is a sign of a powerful jet whose emission is beamed closely toward 1DepartmentofPhysics/KIPAC,StanfordUniversity,Stanford, theEarthline ofsight. Thusthe distinctionbetweenthe CA94305 2DepartmentofAstrophysics,UniversityofOxford,OxfordOX1 traditionalBLL andthe FSRQ issensitiveto the precise 3RH,UK stateandorientationofthe jet(e.g.Giommi et al.2012) 3Department of Astronomy, CaliforniaInstitute ofTechnology, and,indeed,variationsinjetpowerordirectioncanbring Pasadena,CA91125 individual sources in or out of the BLL class (S12). 2 M. S. Shaw et al Our own assignment of the BLL label follows a prag- matic “optical spectroscopic” definition (Marcha et al. Table 1 SampleCompleteness 1996): these are blazars with no emission lines greater than 5 ˚A observed equivalent width, and a limit on any possible 4000 ˚A spectral break of < 40% (Marcha et al. Set Total Specz zmin Unk. z 2LAC 1121 1996; Healey et al. 2008; Shaw et al. 2009). For transi- BLL 475 209 241 25 tion objects with varying continuum emission, we retain LBL 72(21%) 35(20%) 36(23%) the BLL label if it has ever been confirmed to be a BLL IBL 91(26%) 41(24%) 49(31%) with a high quality spectrum, even if subsequently ob- HBL 187(53%) 98(56%) 73(46%) served in an FSRQ state. Within the BLL population Note. —BLLincludes65AGNsoclassifiedsince it is common to classify sources based on the peak fre- the 2LAC paper. 349 BLL have 2LAC SED sub- quency of their SED’s synchrotron component, as esti- classes; percentages give the breakdown. 174 of the mated from radio/optical/X-ray flux ratios, separating spectroscopic redshifts and 158 of the lower limits the sources into low-peak (ν < 1014Hz, LBL), in- havesubclasses; percentage breakdowns are given. peak termediate peak (1014Hz< νpeak < 1015Hz IBL), and vations of 278 BLL objects and 19 other Fermi blazars, high-peak (ν > 1015Hz, HBL) sources. We adopt notincluded in S12. We further analyzed75 SDSS spec- peak herethe LBL/IBL/HBLdesignationsfrom2LAC,which tra, treating them in the same manner as our new ob- provides such subclasses for 74% of all 2LAC BLL, and servations. Note that, unless we suspected a published 83% of those with spectroscopic redshifts (Table 1). redshift was erroneous, we generally did not obtain new The evolutionofBLL haslong beencontroversial,and spectra of many of the brighter, famous BLL with red- it has been claimed that they are predominantly a low shiftsintheliterature. Inseveralcaseswhenweobtained redshift population, showing strong ‘negative’ evolution new data it strongly contradictedthe literature redshift, (e.g.Beckmann et al.2003),especiallyfortheHBLclass eitherfromanewsecurespectroscopiczorfromaninter- (Rector et al.2000). Onechallengeto studiesofthe cos- vening absorber at larger z. In the end we retained 107 mologicalevolutionofthesesourcesisthedifficultyofob- redshiftsfromliteraturevalues;howeverforseveralearly tainingredshiftsfromtheirnearlyfeatureless,continuum BLLz’swehaveonlyinspectedplottedspectra(ofvary- dominated spectra. Indeed, many of the early studies ing qualities); we suspect that at least a few erroneous using X-ray or radio-selectedsamples had highly incom- values remain in this set. Our new data provide new plete redshift measurements, even though the samples spectroscopic redshifts for 102 objects and secure lower wereconfinedtorelativelybrightsources. Uncertaintyin limits for many more, as summarized in Table 1. This extrapolatingfrom the measuredset of redshifts compli- brings the 2LAC BLL sample to 44% redshift complete- cated population interpretations. In the Fermi era, this ness, and 95% completeness including redshift limits. issue becomes critical, as the large BLL contribution to theLATsourcepopulationandthehardBLLγ-rayspec- 2.2. Observations tra ensure that these sources are a major fraction of the The quest for high completeness has driven us to em- cosmic γ-ray background and may, indeed dominate the ploy medium and large telescopes in both hemispheres. LAT backgroundat high energies (Ajello et al, in prep). Observationswere obtainedwiththe MarcarioLowRes- Since the LAT provides a large, uniform, sensitivity olution Spectrograph (LRS) on the Hobby-Eberly Tele- limited (∼ flux limited) blazarsample, itprovidesa new scope (HET), with the ESO Faint Object Spectrograph opportunity to make progress on these questions. Im- and Camera (Buzzoni et al. 1984, EFOSC2) and ESO portant to interpreting the LAT blazars are the strong Multi-Mode Instrument (Dekker et al. 1986, EMMI) on correlations between the LAT-detected (IC) part of the the New Technology Telescope at La Silla Observa- SED and the synchrotron component covering the op- tory(NTT), with the GoodmanHighThroughputSpec- tical band. The synchrotron peak location determines trograph (GHTS) on the Southern Astrophysical Re- the sub-classification, but also correlates with the inten- search(SOAR)Telescope,withtheDoubleSpectrograph sity and LAT-band spectral index of the IC component, (DBSP) on the 200” Hale Telescope at Mt. Palomar, which in turn affects the depth of the LAT sample. Fur- withtheFOcalReducerandlowdispersionSpectrograph ther, as illustrated in this paper, the synchrotron peak (Appenzeller et al. 1998, FORS2) on the Very Large and intensity also affect the difficulty of optical spec- Telescope at Paranal Observatory (VLT), and with the troscopic measurements. Since we wish to recover the LowResolutionImagingSpectrograph(LRIS) atthe W. detailed evolution of the blazars, preferably also follow- M. Keck Observatory (WMKO). Observational configu- ing differences among the LBL/IBL/HBL subclasses (or rations and objects observed are listed in Table 2. even better, a physical parent property leading to these Sinceatthetimeofobservation,manyofthesesources subclasses),oneneedsadetailedtreatmentofboththeγ- were not classified, we often initially obtained only suf- ray(e.g.Ajello et al.2012a)andopticalselectioneffects. ficient S/N to identify the redshift of an FSRQ, or to We reserve such analysis for a future study, noting here firmly establish the source as a BLL. Also, with the va- only the most prominent trends in the measured optical riety of telescope configurations and varying observing sample. Of course, characterization and minimization conditions, the quality of the spectra are not uniform: of the optical biases are greatly aided by high redshift resolutions vary from 4 to 17˚A, exposure times from completeness, the goal of the present paper. 180s to 2400s, and telescope diameters from 3.58m to We have accordingly studied the 2LAC sample over a 10m. S/Nperresolutionelementvariesfrom10to>300. numberofyearswithawidevarietyoftelescopes,striving In a number of cases follow-on observations with higher to be as complete as possible. We here report on obser- S/Nand/orhigherspectralresolutionallowedustomore 3 carefullystudyconfirmedBLLlackingredshifts. Herewe ingIRAF’s de-reddeningfunctionandtheSchlegelmaps discuss the most constraining spectrum or spectrum av- (Schlegel et al. 1998). We made no attempt to remove erage for each source, referring to this as the ‘primary’ intrinsic reddening (i.e.: from the host galaxy). spectrum. Telluric templates were generated from the standard All spectra are taken at the parallactic angle, except star observations in each night, with separate templates for LRIS spectra using the atmospheric dispersion cor- for the oxygen and water line complexes. We cor- rector,whereweobservedinanorth-southconfiguration. rectedseparatelyforthetelluricabsorptionsofthesetwo Inafewcases,werotatedtheslitangletominimize con- species. Wefoundthatmosttelluricfeaturesdividedout tamination from a nearby star. At least two exposures well, with significant residuals only apparent in spectra aretakenofeverytargetforcosmicraycleaning. Typical with high S/N. On the HET spectra, residual second exposure times are 2x900s. order contamination prevented complete removal of the strong water band red-ward of 9000 ˚A. 2.3. Data Reduction Pipeline When we had multiple epochs of these final cleaned, Data reduction was performed with the IRAF pack- flux-calibrated spectra with the same instrumental con- age(Tody1986;Valdes1986)usingstandardtechniques. figuration, we checked for strong continuum variation. Data was overscan (where applicable) and bias sub- Spectrawithcomparablefluxeswerethencombinedinto tracted. Domeflatsweretakenatthe beginningofevery a single best spectrum, with individual epochs weighted night, the spectral response was removed, and all data by S/N. frames were flat-fielded. Wavelength calibration em- Due to variable slit losses andchanging conditions be- ployedarclampspectraandwasconfirmedwithchecksof tween object and standard star exposures, we estimated nightskylines. Weemployedanoptimalextractionalgo- that the accuracy of our absolute spectrophotometry is rithm(Valdes1992)tomaximizethefinalsignaltonoise. ∼ 30% (Healey et al. 2008), although the relative spec- ForHETspectra,carewastakentouseskywindowsvery trophotometry is considerably better. near the longslit target position so as to minimize spec- Fig. Set 1. Spectra troscopic residuals caused by fringing in the red, whose removal is precluded by the rapidly varying HET pupil. 2.4. General Trends Spectrawerevisuallycleanedofresidualcosmicraycon- tamination affecting only individual exposures. To illustrate the principal trends in the BLL spectra We performed spectrophotometric calibration using we refer the reader to Figure 1. By definition, the dom- standard stars from Oke (1990) and Bohlin (2007). In inant component is a power-law. J0516−6207, however, most cases standard exposures were available from the showsthatafterremovalofapower-lawweak,butbroad data night. On the queue-scheduled HET, and during CIV,CIII, and MgII features may occasionallybe seen our queue-scheduled VLT observations, standards from in high S/N spectra. Here the equivalent widths (< 1 subsequentnightsweresometimesused. Atallothertele- ˚A) are sufficiently small to secure the identification as scopes, multiple standard stars were observed per night a BLL. However, should the continuum fade by ∼ 10×, under varying atmospheric conditions and different air- thiswouldbeclassified(atthatepoch)asanFSRQ.Such masses. The sensitivity function was interpolated be- transition objects support the idea of a Blazar contin- tween standard star observations when the solution was uum,ratherthantwodistinctpopulations(Fossati et al. found to vary significantly with time. 1998; Ghisellini & Tavecchio 2008). In S12, we reported For blue objects, broad-coverage spectrographs can significant broad line measurements for 5 of our BLL, suffer significant second order contamination. In par- including J0516-6207. ticular, the standard HET configuration using a Schott While the power-laws of most BLL are very blue, like GG385 long-pass filter permitted second-order effects J0516−6207,afew like J1849+2748,appearintrinsically redward of 7700 ˚A. The effect on object spectra were flat or red, even after correcting for Galactic extinction. small, but for blue WD spectrophotometric standards, This may plausibly be a sign of a synchrotron compo- secondorder corrections were needed for accurate deter- nent peaking near the optical, but might also indicate minationofthesensitivityfunction. Thiscorrectionterm incomplete extinction correction, with residual redden- wasconstructedfollowingSzokoly et al.(2004). Inaddi- ingcausedbydustnotintheSchlegel et al.(1998)maps. tion, since BLL spectra are generally simple power laws, It could also be intrinsic host extinction. we used our objects to monitor second order contami- The Galactic reddening can be very severe. For nation and residual errors in the sensitivity function. In J2001+4352 (upper right) we show both the highly extremecases,wefitanaveragedeviationfrompowerlaw extincted, pre-correction spectrum and the blue post- acrossallobjectsinanight,andtreateditasacorrection correction power law. This source is in a direction of to our spectrophotometric calibrations. This resulted in known high AV = 1.75. Such extincted power-law spec- excellent, stable response functions for the major data traprovideanexcellentopportunityforISMstudies: The sets. featuresseenafterde-extinctionanddivisionbythebest- Spectra were corrected for atmospheric extinction us- fit power law (lower panel) are all interstellar in origin ingstandardvalues. WefollowedKrisciunas et al.(1987) – Galactic H and K, Na I 5892, and a series of diffuse forWMKOLRISspectra,andusedthemeanKPNOex- interstellar bands, as described in Yuan & Liu (2012). tinction table from IRAF for P200 DBSP spectra. Our For J0124−0625 (upper left) the residual absorption NTT,VLT,SOAR,andHET spectradonotextendinto featuresareintergalacticinorigin. Redwardof3900˚Awe the UV and so suffer only minor atmospheric extinc- detectanumberofmetal-linesystems,bluewardonesees tion. Thesespectrawerealsocorrectedusingthe KPNO the onset of strong Lyman-α forest absorptions. These extinction tables. We removed Galactic extinction us- features determine a redshift z =2.117 (one of the high- 4 M. S. Shaw et al Table 2 ObservingConfigurations Telescope Instrument Resolution SlitWidth Objects Filter λmin λmax ˚A Arcseconds ˚A ˚A HET LRS 15 2 41 GG385 4150 10500 HET LRS 8 1 8 GG385 4150 10500 NTT EFOSC2 16 1 31 - 3400 7400 NTT EMMI 12 1 1 - 4000 9300 Palomar200” DBSP 5/15 1 4 - 3100 8100 Palomar200” DBSP 5/15 1.5 5 - 3100 8100 Palomar200” DBSP 5/9 1.5 42 - 3100 8100 SOAR GHTS 6 0.84 2 - 3200 7200 VLT FORS2 11 1 14 - 3400 9600 VLT FORS2 17 1.6 16 - 3400 9600 WMKO LRIS 4/7 1 90 - 3100 10500 WMKO LRIS 4/9 1 40 - 3100 10500 Note. — ForDBSPandLRIStheblueandred channelsaresplit bya dichroicat 5600 ˚A; thelisted resolutions are for blueand red side, respectively. est in our BLL sample). The lack of similar Lyα forest light,whichthesignificantemissionfeaturesallidentified absorption in many of our other high S/N, high resolu- as z =0 stellar or ISM features. tionspectraallowsusto placestatisticalupper limits on Inafewcases,objectspreviouslycatalogedasBLLsdo the redshift as described in §3.3. nothavesufficientS/NinourspectraforadefinitiveBLL Finally at lower right we see two BLL with significant classification. For J0801+4401, we find that undetected flux from the host. In J2042+2426, the galaxy provides broadlinescouldhaveanequivalentwidthaslargeas9.5 about a third of the total flux and is easily visible in the ˚A; for J0209-5229the limit is 5.6 ˚A, for J1311+0035the raw spectrum. This is still safely a BLL, and we can limit is 8.0 ˚A and for J1530+5736we could have missed measure the continuum contributionby a ‘Non-Thermal linesasstrongasEW=5.5˚A.AshigherS/Nspectroscopy Dominance’ (see §5, here NTD=1.38). For J2055−0021, would likely confirm the archival BLL designations, we the host is swamped by the core synchrotron emission consider them BLLs for the purposes of this paper. (NTD=47.5) and the galaxy features are visible only af- Five of the BLL described here had high significance ter subtracting the best-fit power law, as in the lower broadlinedetectionsandhavealreadybeendescribedin rightpanel. The flux increaseinthe blue appearsdue to S12;were-measurethesespectrahereforauniformBLL residualfew-%fluxingerrors(here,likelyincompletecor- treatment. InJ0847−2337andJ0430−2507,thefluxand rection for atmospheric extinction), rather than intrin- spectral index measurements differ from the S12 values. sic emission. Despite the careful calibration, such resid- This is because in the present analysis we first subtract ual fluxing issues persist in several spectra. However, the host galaxy flux, and calculate the flux and spectral the high-pass filtering described in §3.4 ensures that our index of the remaining non-thermal component. In S12, measurements of, and bounds on, host galaxy flux are no such correction was attempted. almost completely immune to such residual calibration artifacts. We find that a number of BLL show visible 3. MEASURINGBLLREDSHIFTS hostgalaxycomponents,allconsistentwithgiantellipti- TheopportunitytoadvanceourunderstandingofBLL cals (Urry et al. 2000). We discuss the flux distribution evolutionwiththelarge,fluxlimitedFermisampleisim- of these host galaxies in §3.7. portant (Ajello et al. 2012b). Yet, despite the substan- tial telescope resources and careful analysis summarized 2.5. Individual Objects above,many BLL did not yield directspectroscopic red- A number of BLL reductions required special treat- shifts,duetotheextremeweaknessoftheiremissionlines ment. Forexampleafewobjectsclearlyrequiredchanges (Sbarufatti et al. 2005b) and lack of clear host features. to the Schlegel map A , so that the de-extinction re- Therefore we collect here both the direct redshift mea- V sulted in clean power laws. For J0007+4712, A was surements and quantitative constraints on the allowed V increased from 0.3 to 0.8 and for J1941−6211 from 0.3 redshift range for our observed BLL. to 1.0. Conversely we decreased the A of J1603−4904 V from 7.8 to 5.0 and J2025+3343, from 6.15 to 5.0. We 3.1. Emission Line Redshifts checked the recent recalibration of the Galactic extinc- We visually inspected all spectra for AGN emission tionmaps(Schlafly & Finkbeiner2011),butdidnotfind linefeatures,andhostgalaxyabsorptions. Spectroscopic large changes, so these extinction features affecting our redshiftsaremeasuredbycross-correlationanalysisusing BLL are probably on scales below the map resolution. thervsaopackage(Kurtz & Mink 1998). Werequireone J1330+7001wasobservedoffoftheparallacticangle— emissionlinetobepresentatthe>5σlevel,andasecond the ensuing drop in blue flux is not intrinsic to the sys- linepresentatthe>3σlevel—significancesaremeasured tem, and our power law is fit only redward of 5000 ˚A. byfittingaGaussiantemplateinthesplottool;weallow We thus remove the broadband residual in our power thewidthandamplitudeofthe Gaussiantovary,butfix law divided spectrum in Figure 1. J1829+2729 and a the center at the redshift derived by rvsao’s xcsao rou- nearby star were spatially unresolved in our data. The tine. For this study, we limited our search to typically presentedspectrumisacompositeofstarlightandquasar strong emission lines known to be present in some BL 5 Figure 1. Spectra of Fermi BLLs. Each object is presented twice – directly in the upper panel, and then with the best-fit power law ‘removed’ (generally by division, but for composite spectra by subtraction). This residual is plotted in the lower panel. SDSS spectra, whilediscussedinTable3,arenotreplottedinthisFigure. Thesixsamplespectrahereillustratemajortrends. Similarfiguresforthefull BLLsampleareavailableinthe electroniceditionofthis journal. Therewemarkonlythetwo linesusedto securezspec;hereother lines of interest aremarkedand the two qualifyinglines aremarkedinred. z limitsgivenare the mostconstraining limitspresented—2 digits aregivenforlimitsfromnon-detectionofhostgalaxies;3digits,forinterveningabsorptionsystems. Themarkedabsorptionsystemisthe bestspectroscopiclimit;insomecases,astrongerhostgalaxylimitispresented. SeeTable3fortheprecisespectroscopiclimit. Lacs: Broad emission from C IV (1549, 1551 ˚A), C III termined by a significant (> 3σ) detection of a host (1909 ˚A), MgII (2796, 2799, 2804 ˚A), Hγ (4340 ˚A), Hβ galaxy, as will be described in §3.4. A few redshifts (4861 ˚A), and Hα (6563 ˚A) and narrow emission from require further note: In J0124-0625 and J1451+5201, [O II] (3727, 3729 ˚A), [O III] (4959, 5007 ˚A), and [N II] we identify the redshift by a Lyα and C IV absorption system at the onset of the Lyα forest. In J0434-2015, (6549, 6583 ˚A). While other species exist in our spec- we identify a single strong feature with [O II], consis- tra, these here listed are sufficient for definite redshift tent with weak Mg II and Ca H/K absorptions. For IDs. Velocities are not corrected to helio-centric or LSR J1728+1215, we find strong Mg II, confirmed by [O II] frames. at the same z in an archival spectrum. In J2152+1734, In many cases, spectroscopic redshifts are further de- 6 M. S. Shaw et al we identify a strong feature with Mg II confirmed by a require the stronger (bluer) line to have > 5σ signifi- significant [OII] detection in archival spectroscopy. cance, and the second line to have >3σ significance. In For a few objects only a single emission line was mea- a few cases, one component of an otherwise strong dou- sured with high S/N. In general we use the lack of oth- blet was affected by skylines, telluric features or cosmic erwise expected features to identify the species and the rays. In these cases, another expected feature from the redshift with high confidence. Nevertheless, a few red- absorption complex (e.g.: a Fe II line) detected at > 3σ shifts have some systematic uncertainty and are marked qualified the system. We further require the doublet ra- by a ‘:’ in Table 3. We briefly discuss these cases here. tio to be consistent (within errors) to a value between For J0007+4712,we derive a redshift from the clear on- 2:1 and 1:1. setoftheLyman-αforestandreportonlytwosignificant Our principal goal is not an absorption line study. figures. InJ0212+2244,wedetermine a tentativez from Thus we concentrated on the longest wavelength (high- weak Ca H, K and g-band absorptions. For J0439-4522, est z) candidate system and measured sufficient lines to we identify the one strong emission feature as C IV; in- confirmthez (i.e: twosignificantlines). Aftervalidation terveningabsorptionexcludes a MgII identification, but we continued to search for higher z until no candidates aless likely CIIIidentificationatz ∼1.45is notconclu- passed the significance test. We therefore believe that sively ruled out. J0629−1959 presents broad but weak we have found the highest z absorption system in each emission at the redshift of the highest z metal line ab- of our spectra strong enough to reach the 5σ/3σ criteria sorption system (1.724). We thus identify this, tenta- above. tively, as the object’s true z. For J0709−0255,we iden- Since we see the onset of the Lyman-α forest in our tify the strong feature at 9200 ˚A with [OIII] by the line highestredshiftobjects,theredendoftheforestprovides shape; an [O II] identification at z ∼ 0.84 is not ex- a strict lower limit on redshift.This can be higher than cluded. ForJ0825+0309,wefindsignificant[OIII]emis- that inferred from the reddest metal line system. sionat5007˚A(andpossible,butnotsignificantemission Welistthesespectroscopicminimumz’saszmin,when at4959˚A), ata z consistentwith an MgII feature iden- available, in Table 3. tified in Stickel et al. (1993). We find weak features in 3.3. Redshift Upper Limits J1231+2847 at the SDSS z, but they have low signifi- cance. For J1312−2156, we find a plausible Mg II fea- We can use the absence of Lyman-α absorptions to ture; this single line identification is in a small allowed providestatistically-basedupperlimits onz forallBLLs redshift range (z ∼ 1.6), other identifications for this without redshift. The exclusion z depends on the max line are spectroscopically excluded. In J1754-6423, we spectral range, resolution and S/N of the particular ob- tentatively identify emission at ∼ 6300 ˚A with Mg II— servation, but is generally 1.65<z <3.0. higherz redshiftsareexcludedby the lackofLyαforest. To quantify the upper bound, we need the expected In J2116+3339’s spectrum, a significant broad emission densityofLyαforestabsorbers. Penton et al.(2004)find feature is identified with C IV, consistent with a weak that for rest EW ≥ 0.24 ˚A, dN/dz ∼ 40 at z = 1.6, bump in the far blue at Lyα. Nevertheless a lower z varying with redshift as logdN/dz ∝ 1.85log(1 + z). redshift is possible if the purported Lyα line is not real. We follow Weymann et al. (1998) for the EW scaling: J2208+6519presentsonestrong,broademissionfeature, dN/dz ∝ e−(EWrest−0.24)/0.267. As the S/N and resolu- tentatively identified as Mg II—a C IV identification at tion of our data vary, we generally measured a conser- z ∼1.8 is not excluded. vativeuniform3σ sensitivityfornarrowLyαabsorptions 3.2. Intervening Absorbers 100˚Afromtheblueendofourspectra. TypicalEWlim- its in this range were ∼ 0.2−1.0 ˚A. Given this density Forsome BLLs,the corelightpassesnear aninterven- wesolveforthe ∆z rangegivingaPoissonprobabilityof ing galaxy on its way to Earth. At small radii one can 0.32 (ie: 1σ) for detecting no absorbers, obtaining: encounterlowexcitationcloudsinthegalaxy’shalo,giv- ing absorption doublets from Mg II at (2795.5, 2802.7) λ −1.85 ˚A.LargerimpactparameterscansampleCIVat(1548.2, ∆z =0.167·e(EWrest−0.24)/0.267)· min (1) (cid:18)1215(cid:19) 1550.77)˚A. Insome low excitation(i.e.: MgII) systems, we also see absorption from Fe II at (2344.2, 2374.4, whereλ istheeffectiveblueendofthespectrum(gen- min 2382.7, 2586.6, 2600.1) ˚A. Finally, for our highest red- erally3150–4200˚A),andEWisthemeasuredequivalent shift BLL we Lyman-α absorption at 1215 ˚A for the width limit. Thus, we infer a maximum source redshift metal line systems, as well as onset of the Lyman-α for- z = (λ /1215)−1+∆z. In some cases, the S/N max min est. is too low at the blue end of the spectrum. We then For unsaturated absorptions, the doublet ratio for measure EW limits closer to the sensitivity peak of the Mg II and C IV is 2:1, with the blue line dominant. spectrum. Of course, with a larger λ for the effective min In saturated absorptions, the ratio is 1:1 (Nestor et al. spectrum end Equation 1 gives less constraining upper 2005; Michalitsianos et al. 1988). limits. We visually search all spectra for candidate doublets, If the actual blazar redshift z is very close to z as max and follow Nestor et al. (2005) in employing a quantita- estimatedabove,thenitsUVradiationmayphoto-ionize tive test of the significance of each candidate. We used Lyα clouds along the line of sight, postponing the onset a two Gaussian template with wavelength spacing scal- of the forest and artificially lowering z . In practice, max ing with z, but free amplitudes, and fit the equivalent the effect is usually small (∆z ∼ 0.01−0.02) except max width and error of each component, using the splot tool for large z when our bound is generally not of in- max in iraf. For the candidate to qualify as a detection we terest. We follow Bajtlik et al. (1988) to estimate this 7 ‘proximity effect’ correction, by computing based on a Levenberg-Marquardtfitter. To model host slit losses, we assume an r = f (1+z )5 (1+z)1/2−1 2 10kpc de Vaucouleurs profile with Sersic index 1/4 ν Lyα ω(z )= Lyα 4πJ (1+z) (cid:20)(1+z)1/2−(1+z )1/2(cid:21) (O’Dowd & Urry 2005) and account for the individual ν Lyα observations’ slit width and seeing profile (measured (2) from the core full width at half maximum, FWHM). where f is the blazar flux at the absorption Lyman limit (atνz ) and J = 10−21.5ergcm−2s−1Hz−1sr−1 Since we have employed optimal extractions of the BLL Lyα ν spectra, our effective aperture along the slit varies, but is the cosmic ionizing flux, conservatively estimated for we estimate a typical width of ≈2× the spatial FHWM our redshift range. We compute f from the power law ν achievedduringourspectralintegration. Accordinglywe fit in Table 3, and increase the blazar redshift z until estimate host slit losses through a rectangular aperture ω(z )<1. We thus adopt these revised z =z and Lyα max of the slit width × twice the spectrum FWHM. Inferred quote these corrected upper limits in Table 3. host fluxes are corrected for these slit losses. ResultsofsamplefitsareshownasthebluedotsinFig- 3.4. Host Galaxy Fitting ure2wherethefitamplitudeofthehostgalaxytemplate Ithasbeen claimedthat BLLac objectsarehostedby is plotted against trial redshift. giant elliptical galaxies with bright absolute magnitude 3.5. Power Law Fit – M =−22.9±0.5 in our cosmology (Urry et al. 2000; R Sbarufatti et al. 2005b). We reportthe power law fluxes and spectralindices of If we adopt the common assumption that these are thebestfittothede-extinctedspectruminTable3. The standardcandles,wecanestimatetheredshiftoftheBLL fluxisgiveninunitsofLog10−28ergcm−2s−1Hz−1asob- bydetectingsuchgalaxies. Inimagingstudies,onelooks served at 1014.7Hz (∼5980 ˚A), the center of our typical in the wings of the BLL for the host galaxy flux, and spectralrange. TheindexαismeasuredF ∝να. These ν comparesthattothestandardcandlefluxatvariousred- values may be combined with multi-wavelength data to shifts(Sbarufatti et al.2005a;Meisner & Romani2010). study the continuum SED of the blazars in our sample. In spectroscopic studies, one typically looks for individ- Since the statistical errors on the fit are, in general, un- ual absorption features (i.e.: H, K, g-band). One can physically small, we follow S12 in estimating errors on also use the lack of such lines as evidence that the BLL the spectral index by independently fitting the red and is at higher redshift (Sbarufatti et al.2005b; Shaw et al. blue halves of the spectrum. Note that large errors bars 2009). generally indicate a relatively poor fit to a simple power With high S/N spectra, however,one canobtain more law rather than large statistical errors. The statistical stringent limits by using the entire elliptical template errors on the F amplitude are also small; we convolve ν ratherthanjustone(orafew)lines. Plotkin et al.(2010) thesewithourestimated30%overallfluxinguncertainty developedatechniqueoffittingforhostgalaxiesinSDSS (Healey et al.2008),whichdominatesinnearlyallcases. BLL spectra. We expand here on that method for our For objects with high significance (> 3σ) detections more heterogeneous spectra. of galaxies, as described in §3.7, we report the best fit Our spectra come from a variety of spectrographs in power law from the simultaneous power-law/host fit in disparateobservingconditionsandwefindlowfrequency §3.4. systematic fluctuations in many of the fits. These are Whenwehaveobservedobjectsatmultipleepochs,we likely caused by imperfect spectrophotometric fluxing also fit a power law to the other, non-primary spectra. and second order contamination as discussed in §2.3. Thesefluxesvarysubstantially,someby morethan10×. These effects can dominate over real galaxy features. In Figure 3, we show the distribution of f /f flux max min Using SciPy’s Signal Processing routines4, we construct ratios. This is well-described by an α=1.29 power law. a bandpass Kaiser window from 220˚A to 1.5× the Epochs from our fiducial spectra are listed in Table 3. Nyquist frequency. We apply that window as an effec- 3.6. Testing the Standard Candle Assumption tive high pass filter both to our spectra and to the tem- plates,tomitigatethislow-frequencynoisebeforefitting BLL with a redshift and a secure (> 3σ) host de- (Kaiser & Schafer 1980). tection can be used to test the uniformity of the host We test possible redshifts z on a grid scaled to the luminosities. There are 59 such BLL in our sample. i spectrograph resolution with constant spacing in logz. We derive synthetic R-band magnitudes by applying a This grid is thus z = (2∆λ + 1)i − 1, where ∆λ is Kron-CousinsRfiltertoourspectra(Meisner & Romani i λ0 2010). The results are shown as a histogram in Figure the pixel scale of the spectrograph. For each trial z i we fit the power law F ∝ να index and flux and 4. We find < MR >= −22.5, down ∼ 0.4 magnitudes ν fromM =−22.9foundinSbarufatti et al.(2005b). We the amplitude of a redshifted elliptical template. This R findasimilarspreadinluminosity(∼±0.5magnitudes). host template is generated from the PEGASE model Whenweseparatethehostmeasurementsforlower-peak (Fioc & Rocca-Volmerange 1997) tables and evolved to sources (LBL+IBL) we find that they have a median lowz fromz =2,asin O’Dowd & Urry(2005). Foruni- luminosity ∼ 0.3 magnitudes fainter than that of our formity,we here use the same template for all z , anddo i HBL.Unfortunately,wedonothavesufficientLBL+IBL not perform any evolution corrections. Our fit minimizes χ2 with three free variables at each hosts to test for such differences at high significance. Past studies differ: Urry et al. (2000) found no signif- trial z . We employ the scipy.optimize.leastsq routine i icant offset in the host magnitudes of HBL and LBL, 4 Documentation and more information available at butSbarufatti et al.(2005b)notedthathigherpeakHBL http://docs.scipy.org/doc/scipy/reference/signal.html tend to have more luminous hosts. 8 M. S. Shaw et al Figure 2. Best fit host galaxy fluxes as a function of trial zi (blue dots). Our estimate of the 2σ local systematic flux error is shown by the blue line (see text). The red bands give the 1σ range about the expected host flux for MR =−22.5±0.5. Vertical green dashed lines give our zmin(−22.5) limit on the redshift. For comparison, solid green lines give the spectroscopic redshifts, when measured. For J2055−0021 and J0148+0129 these are consistent; the former has a high significance host detection (Figure 1), while for the latter the higherzredshiftisfromemissionlines. J1534+3715isoneof9objectswithlimitsinconsistentwithspectroscopicredshifts. Here,thehost galaxyissub-luminous(MR=−21.87±0.16)andthus (just)missedbythistechnique(Prob=0.15). Two of the high significance host galaxies have imag- assumption M =−22.9±0.5, for more direct compari- R ing magnitudes reported in Sbarufatti et al. (2005b). sontopreviouswork,butwerecommendadoptionofthe For J1442+1200, we measure M = −22.99 ± 0.14; less stringent M =−22.5±0.5 redshift constraints. R R Sbarufatti et al. (2005b) reported M = −22.77. For R J1428+4240, we measure MR = −22.78 ± 0.13; 3.7. Lower Limits from Non-detections of Host Galaxy Sbarufatti et al. (2005b) find M = −22.75. These are R Weusetheresultsofthefittingin§3.4andthecalibra- consistent within measurement errors, a good check of tionin§3.6toconstraintheredshiftofthehost. Ateach our slit-loss corrections and magnitude estimates. trial redshift, our fitter reports a flux (f ±σf) for the Overall,theLATBLLsamplethusrepresentsafainter host galaxy. This is to be comparedwith the model flux host population of than those studied in previous work. from the redshifted standard candle elliptical template Conceivablyahigher(LBL+IBL)/HBLratioinoursam- (f ±∆f ). ple causes part of the difference (although we remain M M While we have greatly decreased the effect of the low HBL dominated). However, we suspect that our rather frequency noise in our spectral fits using the high-pass exhaustive 8-m class campaign, skipping most objects filter, we still find that the statistical errors on the fit withredshiftsalreadyintheliterature,selectsforfainter host galaxy fluxes are unrealistically small; these flux host galaxies than in the past. Thus we may be probing estimates remain dominated by systematic effects. We fainteronthetruehostluminositydistribution;the BLL therefore construct an error estimate ∆f for each z by forwhichwewerenotabletoprovidehostdetectionsmay i measuringthedispersionoffluxestimatesfornearbyred- then be similarly under-luminous compared to previous shift bins. This is computed from a sample of the 30 studies. Atrueevaluationoftheintrinsichostluminosity nearest f(z), skipping 5 bins on each side of our test z distribution,aswellasanydependenceonsubclasstype, i to minimize high pass correlation. After σ-clipping, the willrequireacarefulassessmentoftheparentpopulation fitfluxdistributionsarewellbehavedandweusetheseto (γ-ray) and host detection selection biases. compute a 2-σ upper limit (scaledfromthe rms)ateach In the restofthis section, we conservativelyadoptour z . Vectorsoftheseupperlimitsareshownbythejagged M = −22.5±0.5 estimate. We do also report (Table i R blue lines in Figure 2. These have captured the local ef- 3) more aggressive redshift limits based on the common fective noise quite well and so we adopt these vectors as 9 z. Thuswecompute aprobabilitythatthefitflux f and error∆f areconsistentwiththeexpectedmodelfluxf M and its uncertainty at the given z as ∞ Prob(z;f,∆f)=C G(f′ ,∆f′ )·G(f,∆f)df′ (3) Z M M M fM where G(x,σ ) is a Gaussian of width σ centered at x. x x The normalization C is chosen such that Prob = 1 for f = 0 (i.e.: any f is acceptable for a model of zero M flux). Thisisaconservativechoiceasitdoesnotexclude over-luminous hosts. For example, when we assume a model M = −22.5±0.5 this probability also finds any R f consistent with M = −22.9±0.5 to be acceptable. R When the fitter returns an unphysical negative f, we evaluate the probability for f =0 and the local ∆f. Thisprobabilitybecomessubstantialforz nearagood candidate redshift. It also grows as the sensitivity of our host search drops at large z. We thus calculate a minimum redshift(z ) as the lowestredshiftfor min(−22.5) which Prob≥0.17 (i.e.: 1σ). We list these values in Ta- ble3. Forcomparison,wealsogivez calculated min(−22.9) Figure 3. Histogram of maximum flux ratios. For each object inthesamefashion,assumingamodelMR =−22.9±0.5. wherewehavecollectedmultiplespectra,wecomputethefluxratio Comparison between the spectroscopic detections and between the power law components of the brightest and faintest z suggest that the M = −22.5 value is most con- observed epoch. These are plotted as a log-log histogram. The min R sistent with observed detections, and lower bounds (as dashedlineisthebest-fitpowerlawwithindexα=1.29. Theinset histogram shows the distribution of time between observations – expectedfrom§3.6). Werecommenduseoftheseconser- typically∼1year vative lower limits. Note also that the vector Prob(z), oncenormalizedwithanappropriatepriorandtruncated at z , can be used as a PDF for the BLL redshift. max 3.8. Redshift Distribution AsseeninFigure5,archivalredshiftmeasurementsfor BLL are dominated by low values (z˜= 0.23). Our new spectroscopic redshifts have extended the population to higherz,withsomeobjects’redshiftsatz ≫1. Still,the objectswithredshiftsremaindominatedbylowz. Inthe new spectroscopicredshifts we find z˜=0.33. We believe there is a significantbias to low redshift in both of these samples,astheweaklowEWemissionorabsorptionfea- tures of our typical BLLs with known redshift are easier to detect at low z. InFigure5wealsoshowtwosetsofredshiftlowerlim- its. Foreveryobjectinoursample,wecanderiveahost- detectionlimit(z˜=0.41). Theselowerlimitsonredshift arestillbiasedlow: asevidentfromFigure2wearemost sensitiveto galaxiesatlow z. Nevertheless,they suggest that these objects are not consistent with the spectro- scopic redshifts (the K-S test gives probability < 10−11 of consistency with archival redshifts). The absorption line limits we have for some objects (described in §3.2) Figure 4. MeasuredBLLhostabsolutemagnitudes(Requivalent give further evidence for a population of BLLs at high atz=0). Blackhistogram: allBLLhostswithspectroscopicred- redshift (z > 1). Together, these results strongly im- shift. Aweightedχ2fitgivesMR=−22.5±0.5,ourbestestimate ply that previous BLL studies suffered important biases fortheluminosityofBLLhostgalaxies. Red: hostswithhighsig- nificance (>3σ) detections. GreenandBluesub-histogramsshow due to shallow samples with large redshift incomplete- highsignificanceLBL+IBLandHBLhosts,respectively. ness preventing detection of bright, but high z BLL. The inset shows the spectroscopic redshifts and the effective 2σ confidence limits. redshiftslimits forhigh-peaked(HBL) andlower-peaked Fit fluxes well abovethese 2σ limits denote likely host (LBL+IBL) sources classified in 2LAC. We see that the detections. Indeed, we found that this automatic pro- lower-peaked detections extend to higher z, as might be cessing was quite effective at flagging candidate z for expected if these sources are more luminous and have i hostdetection. Here,however,wefocusonhowwellafit a less dominant synchrotron continuum. However, the fluxf withlocaleffectiveerror∆f canbeusedtoexclude openhistogramsoflimitsremindusthatbothsub-classes ahostofthe expectedmagnitude f atthe testredshift still suffer substantial redshift incompleteness, and the M 10 M. S. Shaw et al LBL+IBL) we fill these in with blue or green, respec- tively. For comparison, we show the 1σ spread of virial- estimateBHmassesfromopticallyselectedSDSSquasars from Shen et al. (2011) (gray band) and masses of the Fermi FSRQs in S12 (red points). Of course, if BLL hosts really are standard candle ellipticals, then the Mbulge−M• relationimpliesconstantblackholemasses. The masses corresponding to the standard M = −22.9 R and our revised −22.5 are shown by dashed lines. Interestingly, our BLL M• estimates increase with z much as the optical QSO or FSRQ. Of course, we only plothighsignificancehostdetectionshere,andlowlumi- nosity hosts at high z are increasingly difficult to detect (unless the core luminosity decreases). Accordingly, as for the QSO, we suspect that the bulk of this trend is due to selection effects. In the case of the FSRQ, S12 argued that the offset to smaller black hole mass was at least partly due to a preferentially polar view of an equatoriallyflattened broadline region,withthe projec- tion decreasing the observed kinematic line width and the average virial mass estimate. Like γ-ray selected FSRQ, BLL are Doppler-boosted along our line of sight (Urry & Padovani 1995). However since the host flux is nearly isotropic,we expect little alignmentbias in our M• estimates. Thus,itisunclearwhethertheBLLoffset Figure 5. BLL redshift and redshift lower limit distributions. tolargerblackholemassesisrealorselectiondominated. Within each δz = 0.2 bin, we show (left to right) the Archival z, ournewspectroscopicz,lowerlimitz’sfromhostfittingandlower In a study of BLL hosts detected in the SDSS limitz’sfrominterveningabsorptionlinesystems. Thelimitsshow Le´on-Tavareset al.(2011)foundnosignificantdifference that the spectroscopic redshift samples, particularly the archival between the masses of the central black holes of HBL sample,areselectionbiasedtolowz. Thesubframeshowsthered- andLBL.Infigure6theHBLmassesarehoweverbiased shifts(filledhistograms)andlowerlimits(openhistograms)forthe LBL+IBL (green) and HBL(blue). Again, both limithistograms upwards with respect to the lower-peak BLL black hole extendtohigherz. TheLBL+IBLsampleextends somewhatfur- masses. This is of course just a restatement of the off- therthantheHBL. set in host luminosity seen in Figure 4. Unfortunately, missing redshifts for both sub-classes extend substan- we cannot claim that this is a physical difference as the tially higher than those in hand. A re-appraisal of the trend is precisely what would expect from selection bias BLL population, properly including the new redshifts, z if HBL have brighter non-thermal cores. constraints and remaining selection biases is needed to Ideally we could use these black hole masses to ex- test whether either subclass is still consistentwith nega- plorethe relationshipbetweenthe BLLs andthe general tive cosmologicalevolution. QSOpopulation. Largeblackholemasses,ifnotinduced solely by selection bias would imply a late stage ofAGN 4. BLACKHOLESANDHOSTGALAXIES evolution. The black hole mass is often comparedto the The masses of the central black holes provide impor- source luminosity to characterize the state of the accre- tant insight into the cosmic evolution of various AGN tion in Eddington units. However,with the exception of classes. These are most easily estimated by the virial the few BLL for which we see broadlines (which seldom technique (cf., Shen et al. 2011). In S12, we adopted have a significantly host detection), the observed flux is this method to give mass estimates for the Fermi FS- so dominated by beamed jet emission that quoting the RQs. For BLL, the lack of high S/N broad line mea- accretion luminosity in Eddington units is not feasible. surements precludes such estimates. However, we have 5. NON-THERMALDOMINANCE measured a number of host magnitudes in §3.4; since these are ellipticals, this is all ‘bulge,’ and we can apply In S12, we introduced the non-thermal dominance an M −L relation to estimate the hole mass. We follow (NTD) as a quantitative measure of how much the op- Gu¨ltekin et al. (2009): tical is contaminated by non-thermal synchrotron emis- sion. We here extend that analysis to BLLs. M LV For most BLL, the dominant ‘thermal’ contribution is log =(8.95±0.11)+(1.11±0.18)log (cid:18)M⊙(cid:19) (cid:18)1011L⊙,V (cid:19) thehostgalaxy,notthebigbluebump. Wethereforeset (4) NTD ≡ Fcore/Fhost where both fluxes are measured at whereM istheblackholemass,andL istheluminosity 5100˚A–thesamewavelengthastheHβ continuummea- V in a V filter [log(LV/L⊙,V) = 0.4(4.83−MV0,bulge)]. To surements for FSRQs. The wavelength choice is impor- convert our fit template spectrum amplitudes to consis- tant for BLL NTD measurements, since the host galaxy tent V magnitudes, we integrate our template spectrum is much redder than the continuum-dominated core; an over the Hubble F555W filter (Lauer et al. 2005) as in NTD measurement just above (redward of) the 4000 ˚A §3.6. breakwould typically give values &4× larger. Measure- The masses from Equation 4 are plotted as circles ments below the break would, of course, diverge even in Figure 6. When the sub-class is known (HBL or more.

See more

The list of books you might like

Upgrade Premium
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.
DMCA & Copyright: Dear all, most of the website is community built, users are uploading hundred of books everyday, which makes really hard for us to identify copyrighted material, please contact us if you want any material removed.
Copyright @ PDFdrive.to, 2022 | We love our users