Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A study of unified models of radio-loud active galactic nuclei through radio source evolution Xu, Lienong 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2016_november_xu_lienong.pdf [ 11.53MB ]
JSON: 24-1.0320834.json
JSON-LD: 24-1.0320834-ld.json
RDF/XML (Pretty): 24-1.0320834-rdf.xml
RDF/JSON: 24-1.0320834-rdf.json
Turtle: 24-1.0320834-turtle.txt
N-Triples: 24-1.0320834-rdf-ntriples.txt
Original Record: 24-1.0320834-source.json
Full Text

Full Text

A study of unified models ofradio-loud active galactic nucleithrough radio source evolutionbyLienong, XuB.Sc., Hong Kong University of Science and Technology, 2014A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Physics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)October 2016c© Lienong, Xu 2016AbstractThe study of Active Galactic Nuclei (AGN) unified models attempts to unitequasi-stellar objects (QSOs) and radio galaxies (RGs) through two orienta-tion effects: (1) from twin-jet relativistic ejection of synchrotron radiatingplasma along an axis; and from (2) a central dusty torus aligned at 90 ◦to the axis and hiding the black-hole (BH) accretion disk cone from side-onviewing. Random orientation of the axis to our line of sight provides suchorientation-dependent appearances that QSOs and RGs were originally clas-sified as entirely different. However, if unified models are correct, we gaindeeper theoretical understanding of the interplay among AGN, black-holesand host galaxy properties. This should provide an answer to what causesthe origin and diversity of AGN populations.The subset of AGN populations with powerful radio frequency emissionis called radio-loud (RL) AGN. The RL AGN unification scheme claimsthat powerful radio galaxies are the parent population of RL QSOs or BLLac objects. RL AGN unified models cannot ignore the impact of strongcosmic evolution of powerful radio sources. Moreover, radio-source countsof the populations differ at varying radio frequencies. Detailed modellingof the radio source evolution and radio luminosity function must be ac-counted for by cosmic evolution in order to properly interpret any unifiedmodel. Preliminary investigations of new techniques to incorporate updateddata are described here. These include both evolution-model tests and testson directly constructing the local radio luminosity function from new low-frequency data.Despite some successes, the subject of radio-loud AGN unified mod-elling remains active, since there are still unresolved issues, for example,contradictory results are revealed in existing samples that compare the pro-jected linear sizes of radio galaxies with radio-loud QSOs. Whether this ismerely the result of selection effects or there are complications in radio-loudunified models remains unclear. The surface-brightness sensitivity of new-generation low-frequency radio telescopes (e.g. the Murchison Wide FieldArray, MWA) may shed light on resolving these discrepancies.iiPrefaceThis thesis is original, unpublished, independent work by the author, Lienong, Xu.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 AGN types and classification . . . . . . . . . . . . . . . . . . 32.1 The AGN standard model . . . . . . . . . . . . . . . . . . . 32.2 Taxonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.3 More on AGN classification . . . . . . . . . . . . . . . . . . . 112.4 Unification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 A historical review of AGN unified models . . . . . . . . . 163.1 Establishment of superluminal motion . . . . . . . . . . . . . 163.2 Orientation effects, idea of the basic unified scheme . . . . . 203.3 Early attempt to unifying radio-loud and radio-quite QSOs . 223.3.1 SR unified model for radio-loud and quiet QSOs . . . 223.3.2 Discussion of Scheuer & Readhead (SR) model . . . . 223.4 Core-dominated QSOs vs. lobe-dominated QSOs . . . . . . . 243.4.1 Setting intrinsic core-to-extend flux ratio RT as pa-rameter . . . . . . . . . . . . . . . . . . . . . . . . . . 243.4.2 Radio source count as a constraint for the OB model 253.4.3 Validity of Orr & Browne (OB) model . . . . . . . . 263.5 Torus obscuration of the AGN hidden nuclear region . . . . . 273.6 More characteristics of radio-loud AGN anisotropy . . . . . . 30ivTable of Contents3.6.1 Radio jet sidedness . . . . . . . . . . . . . . . . . . . 303.6.2 Lobe depolarization asymmetry . . . . . . . . . . . . 313.7 Barthel’s model: unifying radio galaxies and radio QSOs . . 313.8 Shared properties between radio-loud QSOs and radio galax-ies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.8.1 Radio jets in both populations . . . . . . . . . . . . . 333.8.2 Extended radio luminosity . . . . . . . . . . . . . . . 333.8.3 Evolution similarity of radio luminosity function . . . 353.9 FR I radio galaxy – BL Lac object unification . . . . . . . . 373.9.1 Two paradigms of radio loud unification . . . . . . . 373.9.2 Observational resemblance of FR I galaxies and BLLac objects . . . . . . . . . . . . . . . . . . . . . . . . 383.10 The Urry & Padovani (UP) technique: the statistical frame-work of unified models . . . . . . . . . . . . . . . . . . . . . 383.10.1 Modelling of FRI and BL Lac luminosity functions . 393.10.2 Modelling of the FR II galaxy and RL QSO luminosityfunctions . . . . . . . . . . . . . . . . . . . . . . . . . 403.10.3 Unified Scheme of radio-loud population . . . . . . . 413.11 The Wall & Jackson model: modification of the radio loudAGN unified scheme . . . . . . . . . . . . . . . . . . . . . . . 413.11.1 Wall & Jackson (WJ) approach, evolution and radiosource count . . . . . . . . . . . . . . . . . . . . . . . 423.11.2 Re-classification and re-unification of FR galaxies . . 433.12 Development of the radio quiet AGN unified scheme . . . . . 453.13 Future development for the AGN unified models . . . . . . . 494 Main issues of unified models . . . . . . . . . . . . . . . . . . 504.1 Linear projected size and relative number fraction . . . . . . 504.1.1 Requirement of large deprojected size of parent pop-ulation . . . . . . . . . . . . . . . . . . . . . . . . . . 504.1.2 Barthel’s statistical test of RG-QSO unification . . . 514.1.3 Arguments against Barthel’s model . . . . . . . . . . 514.1.4 Possible explanations of the QSO fraction inconsis-tency . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.1.5 Receding torus model vs. exclusion of low-excitationradio galaxies . . . . . . . . . . . . . . . . . . . . . . 544.1.6 Radio samples with contradictory linear-size statistics 554.1.7 Testing the unified model by simulation of the linearsize distribution . . . . . . . . . . . . . . . . . . . . . 564.2 Dichotomy of radio-loud and radio-quiet QSO . . . . . . . . 58vTable of Contents4.3 More on FR I and FR II radio galaxies . . . . . . . . . . . . 594.4 Lack of parent population of type 2 radio-quiet QSOs . . . . 614.5 True type Seyfert 2 . . . . . . . . . . . . . . . . . . . . . . . 634.6 Torus complications . . . . . . . . . . . . . . . . . . . . . . . 634.7 AGN dichotomies by other physical issues . . . . . . . . . . . 654.8 Why low radio frequency work is important . . . . . . . . . . 685 Modelling of powerful radio source evolution . . . . . . . . 695.1 Brief description of the Wall, Pearson & Longair (WPL) tech-nique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695.1.1 Luminosity function and evolution function . . . . . . 695.1.2 WPL procedure: set-up of the local radio luminosityfunction . . . . . . . . . . . . . . . . . . . . . . . . . 705.1.3 WPL procedure: fitting the source count . . . . . . . 725.1.4 The method of MCMC and estimation of parameteruncertainties . . . . . . . . . . . . . . . . . . . . . . . 735.1.5 Data and radio samples . . . . . . . . . . . . . . . . . 765.2 Analysis of the evolution function model . . . . . . . . . . . 775.2.1 Investigating the models used in WPL . . . . . . . . 775.2.2 New but less successful models . . . . . . . . . . . . . 855.3 A new approach to evolution modelling . . . . . . . . . . . . 885.4 Gridding continued . . . . . . . . . . . . . . . . . . . . . . . 936 MWA GLEAM local radio luminosity function . . . . . . . 956.1 The MWA GLEAM 4-Jy sample . . . . . . . . . . . . . . . . 976.2 The subsample for the local radio luminosity function calcu-lation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 996.3 The GLEAM local radio luminosity function . . . . . . . . . 1036.4 Determining Vmax of each source . . . . . . . . . . . . . . . . 1056.5 Checks and comparisons . . . . . . . . . . . . . . . . . . . . 1066.6 Future work on the GLEAM 4-Jy bright-source sample . . . 1097 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111viList of Tables6.1 The MWA GLEAM 4-Jy source count data . . . . . . . . . . 996.2 The MWA GLEAM local bright radio source sub-sample. . . 1006.3 The MWA GLEAM local radio luminosity function data . . . 108viiList of Figures2.1 The AGN standard model with typical scale . . . . . . . . . . 42.2 The optical image of QSO 3C273 . . . . . . . . . . . . . . . . 52.3 QSO spectra energy distribution . . . . . . . . . . . . . . . . 62.4 Example of a Seyfert galaxy . . . . . . . . . . . . . . . . . . . 72.5 The spectral energy distribution of NGC 3783 . . . . . . . . . 72.6 Example of a FR I galaxy: 3C 272.1 . . . . . . . . . . . . . . 82.7 Example of a FR II galaxy: Cygnus A . . . . . . . . . . . . . 92.8 Example of a BL Lac object . . . . . . . . . . . . . . . . . . . 102.9 Brief AGN classification . . . . . . . . . . . . . . . . . . . . . 112.10 Various AGN spectra . . . . . . . . . . . . . . . . . . . . . . 122.11 Examples of radio source spectra . . . . . . . . . . . . . . . . 133.1 Schematic AGN unified models . . . . . . . . . . . . . . . . . 173.2 Superluminal apparent velocity as a function of viewing angle 183.3 Superluminal motion of 3C 279 . . . . . . . . . . . . . . . . . 193.4 Core jet structure of 3C 273 . . . . . . . . . . . . . . . . . . . 213.5 Scheuer, Readhead model plane . . . . . . . . . . . . . . . . . 233.6 Orr & Browne model source count . . . . . . . . . . . . . . . 253.7 Polarized broad Balmer lines of NGC 1068 . . . . . . . . . . . 283.8 Polarized broad Balmer lines in radio galaxies . . . . . . . . . 293.9 3C 175 with a one-sided jet . . . . . . . . . . . . . . . . . . . 303.10 Cumulative distribution of linear size of RG and QSO andexpected value of orientation division angle . . . . . . . . . . 323.11 Both large-scale and small-scale radio jet in Cygnus A . . . . 343.12 Comoving space density of QSOs . . . . . . . . . . . . . . . . 353.13 Comoving space density of steep-spectrum and flat-spectrumradio sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.14 Observed and predicted radio luminosity function of FRIs andBL Lac objects . . . . . . . . . . . . . . . . . . . . . . . . . . 393.15 Observed and predicted radio luminosity function of FRIIsand flat-spectrum radio QSOs . . . . . . . . . . . . . . . . . . 40viiiList of Figures3.16 Model and observed radio source count at 151 MHz . . . . . 423.17 Model and observed radio source count at 5 GHz . . . . . . . 433.18 Optical spectra of three 3C radio sources . . . . . . . . . . . . 443.19 Schematic unification of Seyfert galaxies . . . . . . . . . . . . 463.20 Ionization cone of NGC 5252 . . . . . . . . . . . . . . . . . . 473.21 HST image of NGC 4261 . . . . . . . . . . . . . . . . . . . . 484.1 Radio to optical flux ratio as a function of MB . . . . . . . . 584.2 The FR I / II break luminosity plot . . . . . . . . . . . . . . 604.3 A FR I galaxy (3C 465) with bended structure . . . . . . . . 614.4 The receding torus model . . . . . . . . . . . . . . . . . . . . 624.5 Double torus structure for Seyfert galaxy . . . . . . . . . . . 644.6 The clumpy torus model . . . . . . . . . . . . . . . . . . . . . 654.7 Radiative mode vs radio mode . . . . . . . . . . . . . . . . . 664.8 AGN classification plane by black-hole mass and accretion rate 675.1 Master luminosity distribution of Robertson (1973) all-skycatalogue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 755.2 WPL non-evolving model source count . . . . . . . . . . . . . 775.3 MCMC result for WPL non-evolving model . . . . . . . . . . 785.4 WPL model 4 source count and local radio luminosity func-tion plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 805.5 Space density enhancement of WPL model 4 . . . . . . . . . 815.6 Space density enhancement of WPL alternative model 4 . . . 825.7 Source count plot WPL model 5 . . . . . . . . . . . . . . . . 835.8 Local radio luminosity function and space density enhance-ment of WPL model 5 . . . . . . . . . . . . . . . . . . . . . . 845.9 Source count plot model A . . . . . . . . . . . . . . . . . . . . 855.10 MCMC output of WPL model A . . . . . . . . . . . . . . . . 865.11 Source count plot model B . . . . . . . . . . . . . . . . . . . . 875.12 Source count fit and local radio luminosity function of the2×2 grid model . . . . . . . . . . . . . . . . . . . . . . . . . . 895.13 Space density enhancement of 2×2 grid model . . . . . . . . . 905.14 Source count plot 3×3 grid model . . . . . . . . . . . . . . . . 915.15 Local radio luminosity function of the 3×3 grid model . . . . 925.16 Space density enhancement plot of the 3×3 grid model . . . . 936.1 MWA site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 956.2 MWA Gleam 4-Jy Sample . . . . . . . . . . . . . . . . . . . . 976.3 MWA GLEAM 4-Jy subsample source count . . . . . . . . . 98ixList of Figures6.4 MWA GLEAM 4-Jy subsample luminosity-redshift diagram . 1046.5 MWA GLEAM 4-Jy subsample V/Vmax plot . . . . . . . . . 1066.6 MWA GLEAM local radio luminosity function . . . . . . . . 107xAcknowledgementsI thank Professor Jasper V. Wall for inviting me into the subject of radioastronomy and cosmology, and for suggesting this particular topic.Special thanks to Professor Carole Jackson and Dr. Thomas Franzen forsupplying the MWA GLEAM sample.xiChapter 1IntroductionActive galactic nuclei (AGN) emit electromagnetic radiation in possibly allof the electromagnetic spectrum. They are, so far, the most powerful butnon-explosive astronomical objects in the Universe. The AGN phenomenonrefers to the small energetic central region of a galaxy (less than a few lightdays) of objects. The extreme luminosity released by the AGN are believedto be due to the super massive (> 105M) black-hole (SMBH) accretiondisk system (Lynden-Bell, 1969, Begelman et al., 1984).So called radio-loud AGN are of special interest for multiple reasons.Typical radio loud (RL) sources include broad-line radio galaxies (BLRG),radio-loud quasi stellar objects (RL QSOs) and BL Lac objects. Theselatter (QSO + BL Lac) classes are frequently called ‘blazars’ collectively.Radio-loud AGN are the most energetic sub-types of AGN populations,with enormous luminosity at radio frequencies. The discovery of relativisticphenomena in RL QSOs seeded the idea of the AGN unified model, whichproposes that the variety of RL AGN morphology and phenomenology couldbe (at least partially) explained by orientation effects of anisotropic AGNemission. The current RL AGN unification picture incorporates: the ideaof flat-spectrum RL QSOs are beamed FR II radio galaxies, while the BLLac objects are aligned versions of FR I radio galaxies. FR I and FR II hererefer to the dichotomy in radio morphology originally noted by Fanaroff &Riley (1974).Because of the extreme luminosity of powerful radio sources, they areable to be observed in the distant Universe, out to very high redshifts. Con-sequently, extragalactic radio sources serve as probes of the high-redshiftearly Universe. The study of the space density of radio sources at differentepochs could reveal the growth of structure and black holes (BH) based onthe co-evolution of AGN, BH and host galaxies (Ferrarese & Merritt, 2000,Kauffmann et al., 2003). Determining (or modelling) the radio luminos-ity function therefore could provide profound information on cosmologicalevolution.1Chapter 1. IntroductionThis thesis is aimed at addressing the connection and interplay betweenthe radio luminosity function and the radio-loud AGN unified model andhow new generations of radio data could be incorporated into the study.The more basic issues of evolution and the local luminosity function must beaddressed as groundwork, before we understand the significance of apparentdiscord in current observations. The thesis presents an extensive review ofmodel development and the associated problems. It then makes a start onthe groundwork, with new studies of how to delineate radio-source evolutionand how to use low-frequency data in constructing the local radio luminosityfunction.This thesis will serve as a proposal for PhD research on elaborationof the unified model. The detailed physical theory of the SMBH, accretionand stellar spectral emission from AGN, and practical implementation ofradio surveys, are not covered in this thesis.The structure of the thesis is as follows: In Chapter 2, AGN are classi-fied using various characteristic. The basic idea and the brief history of theAGN unified model, especially for radio-loud AGN, are reviewed in Chap-ter 3. The current issues that argue against or complicate the classicalunified model are enumerated in Chapter 4. A much used modelling tech-nique is introduced and used to calculate various new models in Chapter 5,to explore radio-source evolution. A Murchison Wide-Field Array (MWA)bright-source radio sample is considered in Chapter 6. With the MWA sam-ple, a direct local radio luminosity function is established, as an exampleof how new low-frequency arrays may revolutionize unified-model studies.Finally, Chapter 7 summarizes the results of the study of unified model andradio-source evolution.2Chapter 2AGN types and classification2.1 The AGN standard modelGalaxies with AGN host perhaps the most bizarre and certainly the mostpowerful non-explosive astronomical events. The characteristic features ofAGN are briefly summarized here.High bolometric power: a literal explanation of why AGN or activegalaxies are called active. The luminosity of an AGN can be up to 1017L.Huge amounts of non-thermal and non-stellar emission: e.g.synchrotron emission, resulting in a featureless power law continuum fromradio, through infrared (IR) to soft X-ray frequencies.IR to gamma ray emission: IR continuum,and emission lines, spectral‘blue bump’ in the optical and point-like sources in X-ray are all present.Gamma rays are detected in highly energetic blazars, from the synchrotronself-Compton effect.Short time-scale variability and small emitting region: The vari-ability is generally more rapid at shorter wavelengths.Broad emission lines and blue ultraviolet (UV) excess: both areobservable in QSOs, sometimes with absorption-line emission as well.Detectable jets in optical, radio and X-rays: The jets are filled byionized gas at relativistic speeds, which transport energies from the centralnuclei to extended radio lobes (Scheuer, 1974, Blandford & Rees, 1974).Large redshifts: Showing the extragalactic nature: z > 0.1, Burbidge& Hewitt (1992), this is an important feature of QSOs.The general characteristics of AGN has been progressively establishedbased on the obscured AGN emission properties and the opening of all wave-bands. This thesis will not go through the comprehensive physics of AGN,however, details can be found in several review papers e.g. Osterbrock (1989,1991), Risaliti & Elvis (2004) and Netzer (2006).The major components of AGN include: a SMBH; central accretiondisk around this SMBH; broad emission-line region (BLR) on sub-parsecscales; narrow emission-line region (NLR) over several parsecs; relativistic32.2. Taxonomyradio jets; and a dusty obscuring torus of molecular gas extending toaround 100 pc. These are collectively called the standard model of AGN(Blandford et al., 1990, Robson, 1996, Peterson, 1997, Krolik, 1998, Ho,2008, Heckman & Best, 2014, Netzer, 2015). See Figure 2.1.Figure 2.1: AGN standard model with typical scales (figure from CalTechAy127 course lecture slides).Not every AGN type possesses all the standard components and charac-teristics mentioned above and the degree of the existence of a certain featurevaries as well. Otherwise there would not be such a diversity of AGN pop-ulations, even under unification models. This intrinsic scatter in propertiescomplicates the application of unification models.2.2 TaxonomyAGN or those classified as active galaxies constitute only a few percent ofall galaxies in the entire Universe (Jones et al., 2015). There are multipleaspects to classify AGN into their sub-types. For the purpose of this thesis,the AGN populations are primarily separated by their the radio loudness.42.2. TaxonomyQSO radio loudness QSOs are among the most distant and luminousobjects in the Universe. Historically, RL QSOs like 3C273 (Figure 2.2) werefirst detected (Schmidt, 1963) due to their extreme radio luminosity andwere named ‘radio stars’. Later optical identifications revealed that thehost objects of ‘radio stars’ were galaxies. Thus ‘radio stars’ were renamedas ‘quasi-stellar objects’ (QSOs) (Schmidt, 1969) 1.Figure 2.2: The bright radio-loud QSO 3C 273. The optical jet points northwest of the central object (figure from Hubble’s Wide Field and PlanetaryCamera 2 WFPC2).However, it turns out most QSOs detected by optical surveys have no1The term ‘quasi-stellar radio source’ (quasar) is still in use today. Quasars are definedas radio-loud QSOs. A radio-quiet quasar makes no sense however it is frequently seen inliteratures.52.2. Taxonomyradio counterpart. They are actually of ‘radio-quiet’ (RQ) nature (Sandageet al., 1965). Amongst all objects classified as QSO, only about 10 percent are radio loud (Kellermann et al., 1989, Padovani, 1999). Thoughtermed radio loud, the radio luminosity only account a few percent of totalluminosity of these AGN.It is worth noting that classifing QSOs as radio-loud and radio-quiet(Padovani, 1999, Padovani et al., 2014) is an over-simplification. The tradi-tionally radio-quiet AGN still have radio emission (Gregg et al., 1996), butat such a reduced level that there remains a real and physical dichotomycompared with RL QSOs (Kellermann et al., 1989). The dividing line is em-pirically set by where the radio to optical luminosity ratio Lradio/Lopt = 10(Stocke et al., 1992). Padovani (1999) proposed calling these two types“radio-strong” and “radio-weak”. In the context of this thesis, loud/strongand quiet/weak will be used interchangeably. The discussion on how the ra-dio loudness of QSOs affects the theory of unified models is further pursuedin Section 3.3 and Section 4.2.Figure 2.3: Schematic QSO spectral energy distribution (SED) from Elviset al. (1994).A typical QSO spectral energy distribution is shown in Figure 2.3. Inaddition to QSOs, other types of AGN are categorized as either in radio-loudor radio-quiet according to the following scheme.62.2. TaxonomyRadio-quiet sources: Radio-quiet QSOs (RQ QSOs), Seyfert galaxies,Low-Ionization Nuclear Emission Regions (‘LINERs’)Seyfert galaxies are luminous as galaxies but have relatively low opticalluminosity compared to RQ QSOs (Figure 2.4). They have QSO-like nucleiand broad continuum emission. Seyfert galaxies were detected and describedmuch earlier than RL QSOs (Seyfert, 1943). An example of a Seyfert galaxyspectral energy distribution is shown in Figure 2.5.Figure 2.4: Seyfert galaxies NGC 1068 and NGC 4151 (University of Al-abama, W. Keel; HST).Figure 2.5: An example of a Seyfert galaxy SED (Elvis et al., 1994).Spiral galaxy types occur as more than 95 per sent of Seyfert host galax-72.2. Taxonomyies. A review of Seyfert galaxy properties is provided by Weedman (1977).Seyfert population could be further divided into type 1 and type 2 by theiremission line properties (see Section 2.3).LINERs are another category of radio quiet AGN. They are galaxiescharacterized by narrow emission lines of low ionization state. Reviews ofLINER properties canbe found in Heckman (1980), Kewley et al. (2006) andHo (2008).Radio-loud species: radio-loud QSOs, radio galaxies, blazars includ-ing: Optically-Violently-Variable objects (OVVs) and BL Lac objects,Gigahertz-Peak-Spectrum sources (GPS) (O’Dea, 1998), compact steep-spectrum sources (CSS) (Saikia, 1995).Figure 2.6: M84, an example of an FR I galaxy from VLA observations byLaing & Bridle (1987).82.2. TaxonomyRadio galaxies (RGs) are galaxies (Bolton et al., 1949, Ryle & Hewish,1950) with strong synchrotron radio emission, commonly manifested in theirdouble radio lobes, which are able to extend to 100 of kiloparsecs fromtheir centres. The central nucleus can resemble a Seyfert galaxiy in opticalspectra. The visible galaxy is not out-shone by optical core emission. Thedominant host galaxy type is giant elliptical.Figure 2.7: Cygnus A (FR II galaxy) false colour image from the VLA(courtesy NRAO/AUI).There is a unique further classification of radio galaxies into FR I and FRII types (Fanaroff & Riley, 1974) by their extended radio lobe morphology.A source having lobes with hot spots (regions of highest surface brightness)appearing at the outer-most regions (edge-brightened) is of FR II class (Fig-ure 2.7), while FR I sources have edge-darkened lobe morphology (Figure2.6). The FR dichotomy closely coincides with luminosity classification,since FR II are the more luminous of the two.Other classification methods for radio galaxies are discussed in Section2.3.BL Lac objects are compact variable radio sources (Schmitt, 1968,Andrew et al., 1978) having QSO-like nuclei as their optical counterparts(Figure 2.8), but have a total or almost total lack of emission lines (Weiler& Johnston, 1980), such as in the prototype of the class BL Lacertae (Ver-92.2. Taxonomymeulen et al., 1995). The spectrum of BL Lac objects is dominated by afeatureless synchrotron continuum. BL Lacs are also OVV type due to theirshort time variability.BL Lac objects and flat-spectrum (see Section 2.3) RL QSOs are togethercalled blazars nowadays (Padovani, 1999). Blazars are characterized by theirhigh optical and radio polarization and rapid variability on time-scales asshort as one day (Miller et al., 1989, Bregman, 1990). They represent theextreme case of relativistic beaming in the unified model (Chapter 3).Figure 2.8: The BL Lac object H 0323+022 (z = 0.147) imaged at ESONTT (R filter).The term blazar was once reserved for special classes of QSOs withstrong variability or polarization such as highly-polarized QSO (HPQ) and102.3. More on AGN classificationOVV(Penston & Cannon, 1970, Burbidge & Hewitt, 1992). General reviewsof blazar properties can be found in Angel & Stockman (1980) and Bregman(1990).The categorization of AGN population by radio loudness can be brieflysummarized in Figure 2.9.Figure 2.9: Brief AGN classification (Padovani, 1999).2.3 More on AGN classificationAGN classification is traditionally based on observational properties, such asmorphology, size, broad-band optical/IR continuum emission, radio/opticalvariability, polarization, radio continuum spectrum, optical/IR emission-line strength and host galaxy type. Popular divisions that are related to theapplication of unified models are considered in this section.Broad (type 1) emission lines vs. narrow (type 2) emission lines. Ac-cording to the AGN standard model, the broad emission line region liesclose to the central BH accretion disk structure. The rapid circular motionsof the accretion disk surrounding the core BH produce Doppler-broadenedemission lines in the optical/IR band. The nominal definition of broad (per-mitted) emission lines is that the full width half maximum (FWHM) > 1000km s−1.The narrow (forbidden) line region lies outside the central BH-accretiondisk system, but within the structure of the host galaxy. This narrow-lineregion has typical line width of 300 km s−1 < FWHM < 1000 km s−1), andthe emission is generally isotropic.112.3. More on AGN classificationType 1 AGN are defined by the dual existence of broad-line (BL) andnarrow-line (NL) features. The main populations include Seyfert 1, RLQSOs and broad-line RGS of FR II type. Type 2s only exhibit narrow linespectra; most FR I radio galaxies and Seyfert 2s belong to this category (SeeFigure 2.10).Whether the difference between type 1 and type 2 is due to orientationis a key question for AGN unified models. More classification by variousemission line characteristic can be found in Baldwin et al. (1981), Lawrence(1987), Osterbrock (1989), Woltjer (1990), Boroson & Green (1992).There exists a spectral type “0’ of AGN, which involves radio-loud AGNwithout any spectral lines at all. Type 0 AGN mainly consists of blazars,i.e. BL Lacs and OVV.Figure 2.10: Various spectra of different AGN (M.Keel)Flat spectrum vs. steep spectrum. This division is generally de-pendent on radio power-law spectral index α (using the convention Sν ∝ να)within the radio-loud AGN population.Steep spectrum (SS) sources have spectral index α < −0.5 with mono-tonic spectra that generally steepenwith frequency, while flat spectrum (FS)ones are flatter than −0.5 (Fanti et al., 1984). In broader terms, any ra-dio spectrum is classified as “flat” spectrum if it does not obey the steep-122.3. More on AGN classificationspectrum classification (de Zotti et al., 2010). The shape of the spectrumhas different physical implications.Optically-thin synchrotron emission from a power-law relativistic electron-energy distribution produces a steep power law spectrum (Peterson, 1997).Steep spectra are commonly observed in low frequency (< 1 GHz) extended-lobe emission of radio sources. Empirically a mean spectral index of −0.8 isreported in steep-spectrum populations (Laing et al., 1983).Figure 2.11: Examples of different radio spectra. The upper right is atypical steep-spectrum source, while the lower right is a flat-spectrum sourcedue to the superposition at synchrotron self-absorption peaks at differentfrequencies (Verschuur & Kellermann, 1988).However, for higher radio frequencies (e.g. 5 GHz) surveys, the radioemission from objects with bright radio cores are almost flat or have mul-tiple humps (Marscher, 1995). The radio core is rather optically thick andso compact that only the very long baseline interferometry VLBI technique(Readhead & Wilkinson, 1978) is able to resolve its sub-parsec structure.The flatness of the spectra of the cores could be explained by the superposi-132.3. More on AGN classificationtion of synchrotron components that turnover (peak) at different frequencies(Cotton et al., 1980).On the other hand, the existence a small class called compact steep-spectrum objects (CSS) (Kapahi, 1981, Fanti et al., 1985) seems to be in-consistent with the spectral convention. In recent years, it has been sug-gested that CSS and GPS (having the synchrotron self-Compton effect in theGHz range) are likely to be very young radio galaxies (Bicknell et al., 1997,O’Dea, 1998, Snellen, 2008). Examples of such radio spectra are illustratedin Figure 2.11.Core-dominated vs lobe-dominated. This is another dichotomy ex-clusively for radio sources. In a resolved radio source that has both coreand lobe structure, the core-to-extended flux ratio can be calculated to dis-tinguish whether a source is core-dominated or lobe-dominated (Jackson &Wall, 1999).Radio galaxies with their gigantic steep-spectrum lobe emission are mostlylobe dominated (steep radio spectrum). QSOs were originally assumed tobe sources with only compact flat-spectrum cores until their extended radioemission was detected by Vary Large Array (VLA) (Browne et al., 1982).Together with the VLBI core observations, it then became sensible to dis-cuss lobe-dominated QSOs. In fact the dual existence of core-dominated andlobe-dominated QSOs was an original motivation for unified models (Orr &Browne, 1982), as we discuss later in Section 3.4. Note that blazars (BL Lac+ flat spectrum QSO) are exclusively core-dominated.High-excitation vs. low-excitation. This is a more modern classifi-cation of both radio-loud and radio-quiet AGN. Sources that contain weakor no narrow emission lines are considered as “low-excitation” (LE). High-excitation (HE) classes, on the other hand, show strong narrow lines (Lainget al., 1994) and may also have broad line emission. The lack of narrowionization lines indicates that there might be no QSO-like nuclear activityat the centres of these AGN. The physical explanation may be that there aredifferent ‘modes’ of accretion mechanism, having with different efficiencies(Heckman & Best, 2014).The existence of low-excitation radio galaxies (LERG) was confirmedby Hine & Longair (1979) and Laing et al. (1994) in the Third CambridgeCatalogue of Radio Sources (Laing et al., 1983). Most FR Is and local FRIIs are in the low excitation class. The high-excitation state correlates withradio luminosities for high redshift radio galaxies.LINERs, by definition are the radio quiet version of low-excitation AGN,142.4. Unificationand are mostly found in the local Universe (Ho, 2008).2.4 UnificationFactors like the appearance or non-appearance of emission lines, whetherthe emission lines are narrow and broad, the relative brightness of opticalcore to radio core and core dominance or lobe dominance in radio structure,can all be accounted by the AGN anisotropic emission with the orientationeffects – the ‘unified model’.Details of how the subtypes of AGN are connected or may even be intrinsi-cally the same are elaborated in the next two chapters.15Chapter 3A historical review of AGNunified modelsThe most basic assumption of AGN unified models is that the AGN mor-phology is anisotropic (Osterbrock, 1978), with twin relativistic jets ejectedfrom the central nucleus along the axis of rotation of the BH-accretion disksystem (Scheuer, 1974, Blandford & Rees, 1974, Bridle & Perley, 1984). Therandom distribution of the AGN’s jet axis angle relative to the line of sight isthe major factor determining the phenomenon that members from the samepopulation of AGN look completely different in each frequency window,and even at the same frequencies (Figure 3.1). Meanwhile, another fac-tor contributing to anisotropy is the axially-symmetric opaque dusty torus(Antonucci, 1983), which obscures the UV-optical-NIR emission from thecentral accretion disk when viewed side-on. The two anisotropy effects giverise to the various types of radio AGN observed. The purpose of unifiedmodels is therefore the re-assembly of the main subgroups of powerful AGNinto a single population.Astronomers pursuing unified models have successfully established theparadigm of unified model for the majority of the AGN classes after decadesof effort in observation and statistical analyses. The main understandingof the unification framework today includes radio-loud AGN by connect-ing FR I galaxies with BL Lac objects and FR II galaxies with radio-loudQSOs(Urry & Padovani, 1995, Jackson & Wall, 1999). The development ofunification in radio-quiet AGN mainly considers unifying Seyfert 1 galaxieswith Seyfert 2 galaxies (Antonucci, 1993).3.1 Establishment of superluminal motionAs soon as flux-varying radio-loud QSOs were discovered, Schmidt (1963),Rees (1966) and Rees & Simon (1968) proposed relativistic motion in thesesources to explain the rapid flux density variations. They pointed out thatif the relativistic object moves in a suitable direction relative to the core,163.1. Establishment of superluminal motionFigure 3.1: Graphical illustration of AGN structure and unified models. Atdifferent orientation angle, AGN will be observed as different populations.This figure specifically presents the radio-loud and radio-quiet unified model.Figure from Urry & Padovani (1995).173.1. Establishment of superluminal motionits transverse velocity will appear to observers to be faster than light. Fora moving component with velocity relative to the speed of light β = vcand corresponding gamma factor γ = (1− β2)− 12 , the maximum observedtransverse velocity could appear to be βγ with orientation angle sin θ = 1γ(Figure 3.2).Figure 3.2: Apparent velocity as a function of viewing angle, for differentrelativistic jet speeds (Kellermann & Pauliny-Toth, 1981).Thanks to VLBI measurements , the highly compact radio core struc-ture on the milliarcsec scale was successfully resolved (e.g. Readhead &Wilkinson, 1978). Superluminal expansion at the core of the brightest radioQSOs and strong double radio galaxies was continuously observed in 1970s(Cohen et al., 1971, 1977, Pearson et al., 1981, Cohen & Unwin, 1984, Unwinet al., 1985, Vermeulen & Cohen, 1994). For example, QSO 3C 345 has anapparent transverse velocity as high as 7c (Schraml et al., 1981) while 3C273 appears to travel at 2c (Cohen et al., 1971). See Figure 3.3 for a caseof superluminal motion observation.183.1. Establishment of superluminal motionFigure 3.3: Superluminal motion in radio-loud QSO 3C 279, observed byVLBA (Wehrle et al., 1998).193.2. Orientation effects, idea of the basic unified schemeThe relativistic radio-emitting jet theory (Blandford & Rees, 1978, Bland-ford & Ko¨nigl, 1979, Begelman et al., 1984) is the most convincing one toexplain the observed superluminal transverse velocity and therefore becameone of the fundamental ingredients of the standard AGN model. Not onlydoes it explained the rapid flux variations, but it also reconciles the inverseCompton problem and the observed high polarization. As the radiatingjet features travel close to speed of light, Doppler relativistic beaming with(Section 3.6.1) Doppler factor δ = γ−1(1− β cos θ)−1 can boost the bright-ness temperature significantly by Tb,obs = δ×Tb, even exceeding the limit ofsynchrotron self-Compton emission of about 1012 K (Kellermann & Pauliny-Toth, 1969).Apart from the existence of apparent superluminal motion itself, thestatistical study of superluminal effects is of great importance (Vermeulen& Cohen, 1994, Cohen et al., 2007). For instance, obtaining a reliable dis-tribution of γ among superluminal radio sources is the key to modellingthe relativistic beaming effect correctly from low to high radio frequencysamples (Urry et al., 1991, Wall & Jackson, 1997).3.2 Orientation effects, idea of the basic unifiedschemeThe discovery and study of superluminal motions of AGN core-jets (Fig-ure 3.4) provided the most straightforward evidence of orientation effects inradio-bright AGN, since one necessary requirement of observing superlumi-nal motion is that the relativistic jet should propagate close to the line ofsight. The jet axis of superluminal radio QSOs and blazars must point veryclose to the line of sight (Blandford & Rees, 1978), although it is possibleto have larger orientation angles and larger (thus rarer) gamma factors toproduce the same apparent velocity (Vermeulen & Cohen, 1994).Therefore in an ideal sample, with completely random orientations (prob-ability density p(θ) = sin θ, where 0 ≤ θ ≤ pi2 ), the fraction of superluminalsources ought to be a few percent, e.g. 2% if γ = 5 (Scheuer & Readhead,1979). However, superluminal expansion is commonly recorded in VLBImeasurements of radio QSOs (Cohen et al., 1977). Such regular occurrenceof superluminal motion in strong radio sources actually verifies the effect ofrelativistic beaming. The observed flux density of a point source is enhancedby δ3. In higher-frequency flux-limited samples, flat spectrum superluminalsources are preferentially selected by boosting rest frame flux densities above203.2. Orientation effects, idea of the basic unified schemethe survey limit, favouring selection of core-dominated objects.If common superluminal sources like RL QSOs and BL Lac objects areall pointing towards us, this immediately raises the question: what wouldRL QSOs look like when their relativistic jets have larger angle to the lineof sight?Figure 3.4: VLBI observation of RL QSO 3C 273 at 1.7 GHz, illustratingthe radio core and powerful jet structure observed by VLBA (Kovalev et al.,2016).Blandford & Rees (1978) were the first to recognize the possibility thatother regular AGN might demonstrate blazar properties if they are viewedalong the radio axes. Taking orientation as a key parameter, identifying off-axis counterparts and unifying them with beamed QSOs seeded the basicidea of AGN unified schemes, especially for the radio-loud types.Historically, the term “unified scheme” or “unified model” in AGN physicsrepresents the attempts to summarize and explain all the different AGN phe-nomenon by a small number of basic parameters. Usually unified models(or at least as many as possible) emphasize the notion of AGN orientationdependence (Peterson, 1997).213.3. Early attempt to unifying radio-loud and radio-quite QSOs3.3 Early attempt to unifying radio-loud andradio-quite QSOsMotivated by the strong anisotropy in superluminal sources, (Scheuer &Readhead, 1979, hereafter SR) attempted to connect radio-loud and radio-quiet QSOs as a single population and estimated the typical value of theLorentz factor in those sources. This was the first analytical attempt to testa unified model statistically and it was an ambitious one in the sense thatgrouping radio-quiet and radio-loud QSOs together was later referred to as“grand unification” (Urry & Padovani, 1995).3.3.1 SR unified model for radio-loud and quiet QSOsSR started with the assumption that the QSO’s optical flux is proportionalto the intrinsic flux of the central radio core and that the optical flux isindependent of orientation. The optically-selected QSOs form the randomlyoriented population, the adopted relativistic jet theory (Blandford & Rees,1978) accounts for the relativistic beaming of the central radio emission.The majority of the optically-selected QSOs present themselves as radioquiet, since their radio jet axis angles to the line-of-sight are systematicallylarge, showing only a marginal beaming effect. Only that small fraction ofQSOs that have orientation angle close to the line of sight will appear tobe flat-spectrum RL QSOs, exhibiting superluminal motion. To make theirmodel as simple as possible, SR further set all QSOs to share the same valueof γ. SR estimated that if γ is about 5, the observed fraction of RL QSOwill be 1% in optically-selected QSO samples (see Figure 3.5).Whether a source contains a relativistic radio core determines if thesource is beamed or not. This idea was applied to 3C radio sources withextended radio emission (classical radio doubles) as well in SR. The extendedemission was assumed to be orientation independent in their model. Thirtyper cent of the 3C radio sources with extended emission contain detectablecentral radio cores. They defined the core-to-extended flux ratio as R andestimated a range of 64 : 1 for R, which could explain the observed classicalradio double distribution if γ = Discussion of Scheuer & Readhead (SR) modelDespite the lack of validity of the simple Scheuer & Readhead (1979) ap-proach nowadays, it was an excellent start for the discussion of unified mod-els. Although it turned out to be far too simplistic, e.g. with a single γ223.3. Early attempt to unifying radio-loud and radio-quite QSOsfactor and perfect correlation of intrinsic radio and optical flux densities, itmade constructive progress in resolving the large scatter in radio luminosity(Schmidt, 1970) while incorporating the AGN relativistic beaming theory(Blandford & Rees, 1978).Figure 3.5: The ‘visibility’ plane from the Scheuer & Readhead (1979)model. The vertical axis represents the optical flux and the horizontal axisrepresents radio flux. A radio loud QSO is present only if its optical flux isabove the optical limit and the beamed radio flux surpasses the radio fluxlimit at the same time.For their unification scheme regarding radio loud/quiet QSOs, the twopopulations were later confirmed to be intrinsically bimodal (Peacock et al.,1986) from multiple analyses. The topic of grand unified models will be fur-ther elaborated in Section 4.2. For now, let us note that the most convincingevidence against the SR model is that compact flat-spectrum radio QSOsdo have extended radio structure detected by the VLA (Browne & Perley,1986, Murphy et al., 1993) and Multi-Element Radio Linked InterferometerNetwork (MERLIN Browne et al., 1982), using increased dynamic range.But such isotropic extended radio emission is much lower in luminosity inobserved radio quiet QSOs (Kellermann et al., 1989). Moreover, the pre-dicted QSO’s cumulative flux distribution from the SR model at 5 GHz is notconsistent with the Bright Quasar Survey (BQS) observations (Kellermann233.4. Core-dominated QSOs vs. lobe-dominated QSOset al., 1989).The sub-unification of 3C extended radio sources discussed by SR, onthe other hand, became more realistic with the inclusion of flat-spectrumradio-loud QSOs. Since the discovery that all strong radio sources haveextended isotropic emission, the radio-loud unification through treating core-dominance as an orientation indicator was substantially developed, e.g. byOrr & Browne (1982).Owing to SR’s pioneering work and the discovery of anisotropic emissionof radio AGN, the method of treating the jet orientation angle as a funda-mental parameter and consideration of superluminal motion set the basisof radio-loud AGN unified models. Extragalactic radio observations wereextensively conducted afterwards to obtain better and larger radio-sourcesamples e.g. the Sixth and Seventh Cambridge Surveys of radio sources (6C,7C), NRAO VLA Sky Survey (NVSS) and Faint Images of the Radio Skyat Twenty Centimeters (FIRST), which supported statistical and morpho-logical studies for the radio-loud AGN unification proposal.3.4 Core-dominated QSOs vs. lobe-dominatedQSOsFollowing SR’s idea, within the species of radio loud QSOs, the idea ofunifying core-dominated and lobe-dominated QSOs emerged in about 1982(Orr & Browne, 1982, hereafter OB) with the statistical examination of thetwin relativistic jet model (Blandford & Rees, 1974) as its theoretical foun-dation. The primary motivation was the detection of diffuse lobe emissionin compact flat spectrum radio QSOs (Browne et al., 1982), which led tothe hypothesis that the radio steep-spectrum lobe-dominated QSOs are theunaligned parent population of core-dominated QSOs. The SR model uni-fying radio loud and quiet QSOs was also proved not to work based on thisanalysis.3.4.1 Setting intrinsic core-to-extend flux ratio RT asparameterSimilar to the SR and Hine & Scheuer (1980) argument about sources withradio core, the OB model defined a parameter RT to quantify the ratioof intrinsic core to extended radio flux. The observed core to extendedratio R will be identical to the intrinsic one if the radio QSO lies exactlyon the sky plane. With a single value of RT and Lorentz factor γ, the243.4. Core-dominated QSOs vs. lobe-dominated QSOsobserved distribution of R could be derived assuming random orientationsfor the population. The OB model found a value of γ = 5 and RT = 0.024adequently fit the observed R distribution of 32 3C QSOs at 5 GHz (Jenkinset al., 1977).3.4.2 Radio source count as a constraint for the OB modelThe improvement of the OB model compared to the SR model was theconsideration of variation of RT with different observing frequencies andredshifts, which helped to fit the observed ratio of core-dominated QSOs tolobe-dominated ones. At different frequencies, the core flux is fixed, as it ispostulated to be a flat-spectrum source. The extended emission is assumedto have a spectral index−1.0 (Laing & Peacock, 1980). Whether a radio loudQSO is core-dominated or lobe-dominated depends on the comparison of thetotal radio flux at two very different frequencies. The redshift distributionof QSOs was approximated by a low-frequency-sample histogram.Figure 3.6: Model prediction of the source count made by OB’s versionof relativistic beaming at 2.7 GHz for different values of Lorentz factor.The source count at 5 GHz is qualitatively the same as at 2.7 GHz (Orr &Browne, 1982)253.4. Core-dominated QSOs vs. lobe-dominated QSOsWith a given value of R and redshift z at a particular frequency ν,the number of QSOs N(R, ν, z) (proportional to number of possible sourcesabove the survey limit) and the differential probability of having a sourceat R are calculated. After integrating over redshift, R and selecting a valuefor the only unknown parameter γ, the fraction of core-dominated QSOsF can be obtained and compared with the observations. OB applied thistechnique to seven different radio surveys at different frequencies and got aweighted mean result of 4.7 for γ. If redshift is not integrated over, the core-dominated QSO fraction as a function of redshift F (z) can be calculatedinstead. The estimated F (z) increases at higher redshift, consistent withobservation (Ulvestad et al., 1981, Kapahi & Kulkarni, 1986). This is aconsequence of relativistic beaming, which boosts QSOs into core-dominatedversions in a flux-limited high-frequency sample.Lastly, if the unified scheme is right for QSOs, then the QSO contributionto the radio source count at both low and high frequency should be closelyrelated. Starting from the 408-MHz total source count (Wall et al., 1980),OB estimated the source count from QSOs alone. Again by integratingover redshift and R with fixed value of trial γ, the source count at higherfrequencies 2.7 GHz and 5 GHz could be constructed. The best fit of thehigh frequency source count (e.g. Condon & Ledden, 1982) was achievedby setting γ = 5 (See Figure 3.6).3.4.3 Validity of Orr & Browne (OB) modelOrr & Browne (1982) summarized the arguments for their support of theunified scheme which includes:(1) an explanation of superluminal motion;(2) the existence of diffuse emission in the flat-spectrum superluminalsources;(3) more rapid variability in flat-spectrum sources;(4) consistent QSO statistics, source counts and QSO fractions.Kapahi & Saikia (1982) followed the assumptions of the OB model andchecked the correlation among orientation-dependent quantities. It wasshown that the QSO core-fraction emission anti-correlates with the observedQSO linear size, while the core-fraction emission positively correlates withthe lobe hot-spot-to-core distance ratio and observed degree of misalign-ment. All of these correlations were consistent with the orientation unifiedscheme of OB.Browne & Perley (1986) compiled 135 VLA flat-spectrum QSOs whosestatistics of relativistic beaming were in fair agreement with the OB model263.5. Torus obscuration of the AGN hidden nuclear regionas well. In addition, the core-dominated QSOs in this sample are moreasymmetric than the lobe-dominated ones. X-ray emission in radio-loudQSOs, which is assumed to be another beamed component, also tightlycorrelates with the radio emission (Browne & Murphy, 1987).Small-number statistics were still a limitation of the OB analysis andlack of large values of R caused the Lorentz factor γ to be less constrained.Browne & Perley (1986) showed that single values of γ and RT is not appro-priate to fit the observations. A uniform distribution of γ from 3 to 11 wasrequired to fit the observed distribution of Murphy et al. (1993) for VLAcore-dominated QSOs.Nevertheless the main arguments of the OB model are still qualitativelyvalid today (Urry & Padovani, 1995, Singal & Laxmi Singh, 2013) and theirstatistical methods inspired the following studies on unified models, espe-cially at the stage of unifying radio galaxies with QSOs in 1990s (Vermeulen& Cohen, 1994, Wall & Jackson, 1997).The OB model was the most compelling version of a unified schemefor radio-loud QSOs before the incorporation of radio galaxies (Scheuer,1987b, Barthel, 1989). It turned out that both lobe-dominated QSOs andcore-dominated QSOs represent increasingly aligned versions of radio galax-ies (Urry & Padovani, 1995), providing a better and more general unifiedscheme. It took ten more years for the picture to mature to the general radiogalaxy (RG) – QSO unified model, during which more features of anisotropicemission were discovered in AGN and more similarities were found betweenthe two populations.3.5 Torus obscuration of the AGN hiddennuclear regionThe discovery that AGN and QSO central-core spectral emission (opticaland UV) is blocked by an obscuring dusty torus is the second telling evi-dence of the anisotropic characteristic of AGN. It is just as important as therelativistic beaming effect and sets up the framework for a radio-quiet AGNunified model as well.Seyfert galaxies are classified as low optical-luminosity versions of radio-quiet QSOs (see Section 2.2). The presence of broad-line emission deter-mines whether a Seyfert galaxy is type 1 or type 2 (section 2.3). Osterbrock(1978) recognized there might be orientation effects caused by obscurationof the central broad emission-line region in Seyfert and radio galaxies.It was the detailed optical spectra-polarimetry of NGC1068 that con-273.5. Torus obscuration of the AGN hidden nuclear regionfirmed the obscuration effect (Antonucci & Miller, 1985, Miller et al., 1991).In this Seyfert 2 galaxy, a linearly polarized (16%) weak broad emission linespectrum was detected (see Figure 3.7). The featureless continuum polariza-tion is wavelength independent, which indicates that the polarized spectrumis scattered by free electrons. The polarized emission phenomenon could beexplained by dust obscuration of the AGN central broad-line region plus thescattering medium above the disk plane, which mirrors the central broadline spectrum into the line of sight (Antonucci, 1993).Therefore Seyfert 2 galaxies are merely high-inclination versions of Seyferts,which is nothing but the basic hypothesis for the radio-quiet AGN unifiedscheme. Similar polarization emission was found in other Seyfert 2 galaxies(e.g. Miller & Goodrich, 1990, Tran et al., 1992).Figure 3.7: Broad emission lines seen in polarized light in a Seyfert 2galaxy NGC 1068. Image from Miller et al. (1991) for spectropolarimetryand Capetti et al. (1995) for HST image.Apart from radio-quiet Seyfert galaxies, optical polarized emission wasalso observed in radio galaxies such as 3C 234 and 3C 282 (Antonucci,1984, Tran et al., 1995). The polarization is perpendicular to the radio jetaxis shows QSO-like nuclear luminosity and spectrum. Other examples ofpolarized radio galaxies were reported by Bailey et al. (1986), di SeregoAlighieri et al. (1988), Tadhunter et al. (1992) and Tran et al. (1998), as283.5. Torus obscuration of the AGN hidden nuclear regionsummarized in Antonucci (1993), Wills (1999), Tadhunter (2008); see Figure3.8. According to Morganti et al. (1992), the closest AGN, Centaurus A, anFR I radio galaxy, turns out to contain a core with BL Lac features.The effect of torus obscuration could be inferred from many other as-pects rather than just polarimetric evidence. For instance, Lawrence & Elvis(1982) showed that hard X-ray luminosities of Seyfert 2 galaxies are signifi-cantly lower than for Seyfert 1s. Such attenuation of the X-ray emission inSeyfert 2s could be explained by large opacity of a dusty torus.Figure 3.8: Illustrations of polarized emission in radio galaxy 3C 343.1,compared with a Seyfert 2 galaxy M51 (Tran et al., 1998).The undoubted existence of obscuring tori not only contributed to thestandard model of AGN but also helped to develop the orientation unifiedscheme for radio quiet AGN (Antonucci, 1993, Ogle et al., 2006, Netzer,2015) and stimulated the possibilities to connect radio galaxies to radioloud QSOs.293.6. More characteristics of radio-loud AGN anisotropy3.6 More characteristics of radio-loud AGNanisotropyBefore formally proceeding to the RG-QSO unification, an explanation isrequired of more aspects of orientation effects in RGs and QSOs.3.6.1 Radio jet sidednessLarge-scale (kiloparsec) two-sided jets are detected more often in weak radiogalaxies, while parsec-scale one-sided jets (Figure 3.9) are almost exclusivelyin radio-loud QSOs (Bridle & Perley, 1984, Bridle et al., 1994). Amongdifferent theories to explain this, relativistic beaming of the radio jet is themost natural explanation, (rather than alternatives like asymmetry of energydissipation/transport), because it explains superluminal motion and rapidvariability together (Blandford & Rees, 1978).Figure 3.9: The QSO 3C175 with one-sided radio jet at 5 GHz using theVLA (Bridle, Hough,Lonsdale, Burns and Laing).If the radio jet is intrinsically two-sided and identical in the rest frame,due to relativistic bulk motion the relativistic Doppler effect would favourthe forward jet and boost the flux by a factor of ( 1γ(1−β cos θ))3+α, while the303.7. Barthel’s model: unifying radio galaxies and radio QSOsreceding jet would be de-boosted by ( 1γ(1+β cos θ))3+α (Rybicki & Lightman,1979). With limited dynamic range, the counter-jet is very likely to beunobservable, causing the observed jet one-sidedness (Bridle & Perley, 1984).The ubiquitous single-sided jets in radio-loud QSOs indicate that their radioaxes are aligned systematically closer to our line of sight.3.6.2 Lobe depolarization asymmetryIt was reported by Laing (1988) and Garrington et al. (1991) that the degreeof depolarization is asymmetric in classical FR II type double-lobe radiosources having large-scale one-sided jets at long wavelengths (from 6cm to20cm). In 49 of 69 sources including both FR II galaxies and QSOs, the radiolobe with greater depolarization lies on the opposite side to the one-sidedjet. This is known as the Laing–Garrington effect.The depolarization asymmetry could be explained by differential Fara-day rotation (Burn, 1966). The closer radio lobes, due to less interveningmedium along the line of sight, show lower degree of depolarization, fullyconsistent with a one-sided jet appearing on the same (near) side of thesource.3.7 Barthel’s model: unifying radio galaxies andradio QSOsMotivated by a series of observations, including relativistic kinematics (Bland-ford & Rees, 1978), the existence and statistics of superluminal motion (Co-hen & Unwin, 1984), small- and large-scale one-sided radio jets (Bridle &Perley, 1984), hidden broad-line regions in type 2 AGN (Antonucci & Miller,1985), lobe depolarization asymmetry (Laing, 1988), several peoplem no-tably Peacock (1987), Scheuer (1987a) and Barthel (1989) came to proposethe unified scheme for radio galaxies and radio loud QSOs. Barthel claimedthat RGs and RL QSOs are viewed differently only because of orientationof the source axis to our line of sight. He showed that his model was statis-tically in agreement with QSOs having a broad-ine region observed with asingle value of torus opening angle.Apart from explaining the existing RL QSO phenomenology, anotherpurpose of Barthel’s model was to account for the large de-projected sizeand ubiquitous jet one-sidedness of QSOs. If the Orr & Browne (1982)model is entirely correct, lobe-dominated QSOs would then form the entireparent population of radio loud QSOs. Due to random orientation, there313.7. Barthel’s model: unifying radio galaxies and radio QSOsshould be many more QSOs lying closer to the sky plane as lobe-dominated,having larger linear extent and more symmetric two-jet morphology. How-ever the largest observed size of QSO 4C 34.47 (Barthel et al., 1989), is stilla superluminal source, which indicates it has to orient its jet axis close tothe line of sight. The almost universal one-sided jet property among QSOsconfirmed that all QSO populations are favourably oriented, contrary to theOB model.The hypothesis of treating powerful radio galaxies as the parent popu-lation of radio-loud QSOs resolves these issues. Jet one-sidedness becomesexpected as QSOs are favourably orientated in Barthel’s model. Findinglarge-size QSOs is also no longer required since radio galaxies participate toform the parent population with the large linear sizes.Figure 3.10: Results from Barthel’s model. On the left is the cumulativedistribution of observed linear sizes of radio sources. The dashed line rep-resents QSO and solid line denotes RG, which have larger linear dimensionstatistically. The cone division angle is obtained from number statistics,assuming a single value for cone dividing angle (Barthel, 1989).Barthel’s statements are mainly based on the prediction of linear sizedistribution for RGs and RL QSOs (Figure 3.10). The best unbiased sam-ple at that time was 3CR (Laing et al., 1983) because of its low selectionfrequency (151 MHz). The QSOs considered are only of the steep-spectrumtype. Nevertheless this did support such unified scheme with a cone openingangle of 45◦, even though other studies showed results contrary to Barthel’sunification scheme, e.g. Singal (1993b). Detailed arguments of Barthel’smodel and the QSO linear size debate are presented in Section 4.1. The323.8. Shared properties between radio-loud QSOs and radio galaxiesissue of linear size in RG-QSO unification remains outstanding.3.8 Shared properties between radio-loud QSOsand radio galaxiesBarthel’s model is developed based on the presence of anisotropic RL-QSOradiation, morphology and kinematics. Meanwhile it has gained supportfrom observed similarities between RGs and RL QSOs. If the quantityobserved is believed to be isotropic, it should be comparable between RGsand QSOs. Otherwise if the observable is orientation-dependent, differencesare expected between the two classes of source.3.8.1 Radio jets in both populationsThe appearance of radio jets in radio galaxies is a straightforward clue forBarthel’s unified scheme. For example, twin radio jets, which connect itscentral core to the radio lobes, were first detected in Cygnus A by the VLA(Perley et al., 1984). The parsec-scale radio jet (VLBI) is well aligned withthe kpc-scale jet (VLA) in radio galaxies, demonstrating relativistic kine-matics (Scheuer, 1987b, Venturi et al., 1994, Bicknell, 1994). These radio-jetobservations (e.g. Cygnus A in Figure 3.11) resemble the jets in bright QSOs(Bridle & Perley, 1984), consistent with the RG-QSO unification idea.Moreover, large-scale radio jet velocities can also be measured (Barthel,1989). Although statistics are small, lobe-dominated QSOs show less super-luminal velocities than core dominated QSOs (Hough & Readhead, 1987),which agrees with the expected anti-correlation of inclination angles andsuperluminal velocities.3.8.2 Extended radio luminosityUlvestad et al. (1981) showed that compact radio sources have strong ex-tended radio emission. In OB’s model analysis, the assumption that theextended radio emission in radio-loud QSOs is isotropic was shown by theVLA observation of extended emission in flat-pectrum QSOs (Browne et al.,1982). Therefore if radio galaxies and RL QSOs are intrinsically the same,diffuse low-frequency lobe emission in the extended structure should be com-parable in radio luminosity or span the same range (Urry & Padovani, 1995).Antonucci & Ulvestad (1985) found that extended radio luminositiesof 49 radio galaxies and RL QSOs actually overlap, consistent with theirassumption that core-dominated QSOs (blazars in their notation) are normal333.8. Shared properties between radio-loud QSOs and radio galaxiesFigure 3.11: The radio galaxy Cygnus A has both large scale and smallscale radio jets, resolved by VLA and VLBI at multiple frequencies. Thejets are well aligned and point to the extended radio lobes. Image fromKrichbaum et al. (1998)343.8. Shared properties between radio-loud QSOs and radio galaxiesRGs or (lobe-dominated) QSOs viewed along their radio axes. In the 3CR(Laing et al., 1983), 2-Jy sample (Wall & Peacock, 1985) and MRC samples(McCarthy et al., 1998) the unbeamed radio luminosities of RG and QSOsdo share the same range, as expected by the Barthel model.3.8.3 Evolution similarity of radio luminosity functionThe discussion of radio source evolution can be traced back to the 1960s(Ryle & Clarke, 1961, Longair, 1966, Schmidt, 1968). This subject devel-oped together with radio astronomy itself. The evolutionary behaviour ofextragalactic radio sources was intensively explored in early years even with-out consideration of AGN unified models.Figure 3.12: The space density of Parkes flat-spectrum QSOs (big dotsfit by thick line) as a function of redshift (Shaver et al., 1999). The othersymbols are optical-selected QSO samples listed in Shaver et al. (1999).The main methodology was to start with flux-limited radio samples thathave complete optical identifications to obtain the full information on radioluminosity and redshift for the sample. Then different evolutionary modelscould be proposed (Wall et al., 1980) to explain the radio sample’s luminosity353.8. Shared properties between radio-loud QSOs and radio galaxiesand redshift distribution, by fitting the radio source count and local radioluminosity function (Condon, 1984b,a, Peacock, 1985).In modelling the radio source evolution, estimates of the approximateform of the radio luminosity function at different redshift epoch could there-fore be constructed (e.g. Wall et al., 1980, Jackson & Wall, 1999). A con-sensus in all the radio-source evolution models was that the radio luminos-ity function or simply comoving number density of luminous radio sourcesundergoes strong evolution with increasing cosmological redshift (Longair,1966, Wall et al., 1980), the degree of evolution depending strongly on radiopower.Figure 3.13: Dunlop & Peacock (1990) have shown that both flat-spectrumradio sources and steep-spectrum radio sources also undergo a space densitydecline after peak epoch.All evolution models predict a levelling off or decline of space density atz ≈ 2. The behaviour of this redshift cutoff (Sandage, 1972, Shaver et al.,1996) was noted in the luminosity function of QSOs (either optical or radioselected), which reached a maximum comoving spatial number density atredshift around z = 2, “the quasar epoch”. Later studies showed the decline363.9. FR I radio galaxy – BL Lac object unificationafterward (Windhorst, 1984, Warren et al., 1988, Dunlop & Peacock, 1990,Wall et al., 2005, Richards et al., 2006). See Figure 3.12 for a summary.Bright core-dominated QSOs are exclusively flat-spectrum radio sources.By their dramatically beamed core radio luminosity, they are detectable overa wider range of redshift and are more complete in radio samples. Peacock(1985) model led a redshift cutoff for flat-spectrum radio sources.However for the steep-spectrum radio sources, which are mainly radiogalaxies and steep-spectrum lobe-dominated QSOs, the attempt to deter-mine a redshift cutoff was inconclusive due to the sample detection limita-tions, the radio K-correction in particular. Dunlop & Peacock (1990) for thefirst time showed the possibility of steep-spectrum radio sources also hav-ing a redshift cutoff. Moreover, before reaching the cutoff redshift, Dunlop& Peacock (1990) showed that the space density of both flat- and steep-spectrum powerful radio sources increased similarly by orders or magnitudefrom the local universe to redshift z = 2 (Figure 3.13).This follows from the prediction for the unified model (Peacock, 1987),since if QSOs (flat-spectrum) and radio galaxies (steep-spectrum) are intrin-sically the same, their evolutionary trends should be similar. After Dunlop& Peacock (1990), more observational evidence has confirmed the redshiftcutoff for both types (e.g. Shaver et al., 1999, Jackson et al., 2003).3.9 FR I radio galaxy – BL Lac object unificationThe difference between BL Lac objects and RL QSOs in the optical is thatBL Lac objects have featureless continua (Figure 2.10) with little or noemission-line radiation (Antonucci, 2012). As a subset of blazars, BL Lacobjects do have relativistic kinematics (Blandford & Rees, 1978), and hencethe orientation of their radio axis is also favourably aligned with the line ofsight. Early on (Browne, 1983), the possibility of BL Lac object and radiogalaxy unification was pointed out through a discussion of the relative spacedensity and relativistic beaming of the populations.3.9.1 Two paradigms of radio loud unificationRadio galaxies are not a homogeneous population in a morphological sense.The division of FR type radio galaxies (Fanaroff & Riley, 1974) has yetbeen emphasized. However the distinction between FR I and FR II radiogalaxies is essential in modern theories of unified models. Brief observationaldifferences of the two FR types have been mentioned in Section 2.2. Whilea detailed discussion on issues like physical cause is reserved for Section 4.3373.10. The Urry & Padovani (UP) technique: the statistical framework of unified modelsIn Barthel (1989), only FR II galaxies are proposed (and studied) tobe the parent population of radio-loud QSOs. The beamed versions of low-power FRI galaxies are believed to be BL Lac objects (Browne, 1989, Morriset al., 1991, Urry & Padovani, 1995).3.9.2 Observational resemblance of FR I galaxies and BLLac objectsSubsequent observational evidence has kept the BL Lac unification idealooking promising. The isotropic extended radio luminosities of BL Lacobjects and low luminosity FR I galaxies are well matched between 1023to 1026 WHz−1 (Wardle et al., 1984, Antonucci & Ulvestad, 1985). LocalBL Lac objects tend to live in rich cluster environments (Falomo et al.,1993), similar to FRI galaxies. The type of host galaxy for BL Lacs tendsto be elliptical, as seen in a sample of 34 BL Lac objects above 1 Jy at5 GHz (Stickel et al., 1991). Following Antonucci & Miller (1985) using aspectrapolarimetry technique, BL Lac features are shown in Centaurus Aby its highly anisotropic ionizing continuum (Morganti et al., 1992).On the other hand, almost no polarized broad emission-line detectionsamong FR I galaxies (Antonucci, 2012) safely excludes the possibility ofunification of FR Is with QSOs (all of which show broad lines). Murphy et al.(1993) pointed out that the core-to-extended ratios are indistinguishablebetween BL Lac objects and radio loud QSOs, while the redshift distributionfor both are different, which argues against an alternative theory of treatingBL Lac objects as increasingly beamed versions of RL QSOs.3.10 The Urry & Padovani (UP) technique: thestatistical framework of unified modelsWith the observational features described in previous sections, the hypoth-esis of the dual unified scheme, i.e. FR I radio galaxies connected withBL Lac objects and FR II radio galaxies with RL QSOs, demanded a betteranalytical check, through applying relativistic beaming to examine the num-ber density between beamed and unbeamed populations as Orr & Browne(1982). Such work was first undertaken in statistical studies by (Urry &Padovani, 1995, hereafter UP).They extensively explored the feasibility of radio-loud unification throughluminosity function modelling combined with relativistic beaming.383.10. The Urry & Padovani (UP) technique: the statistical framework of unified models3.10.1 Modelling of FRI and BL Lac luminosity functionsIn UP’s framework, the comparison of FR I and BL Lac objects was appliedin the X-ray domain, and then in the radio (Urry & Padovani, 1994). TheX-ray parent FR I luminosity function was set up using two methods. Thefirst was from direct X-ray observation of a complete sample of radio-selectedFRIs (Fabbiano et al., 1984). The second independent estimation was toconvert the radio luminosity function into X-ray using the radio to X-raycorrelation (Padovani & Urry, 1990). These two estimations of FR I X-rayluminosity functions match well.Figure 3.14: Relativistic beaming result of Urry & Padovani (1994), whichshows that the predicted BL Lac radio luminosity function from the FR I lu-minosity function matches the observed BL Lac luminosity function beamedfrom the FR I luminosity function.The beaming of the luminosity function was carried out using the tech-nique of Urry & Shafer (1984), assuming a power-law form for the unbeamedluminosity function. The beamed luminosity function, which contains theextended isotropic emission and Doppler-boosted core, would form a doublepower law. Based on the derived X-ray luminosity function, the X-ray countof BL Lac objects could be calculated by adopting an appropriate cosmolog-ical model and fit with observations from EMSS, EXOSAT and HEAO1 A2,(detailed references in Padovani & Urry, 1990). A ratio of BL Lac to FR I393.10. The Urry & Padovani (UP) technique: the statistical framework of unified modelsnumber density 1:7 with the optimal Lorentz factor ≈ 3 and intrinsic core-toextended ratio ≈ 0.1 were obtained by Padovani & Urry (1990). Later thederived BL Lac X-ray luminosity function was directly compared with thepublished observations (Wolter et al., 1994) with good agreement.Similarly the radio luminosity function of FR I sources was again de-rived by Urry et al. (1991). Using the same beaming formalism, the radioluminosity function of BL Lac objects was calculated and fit to the obser-vations (Stickel et al., 1991). The radio luminosity function comparison ofFR I galaxies and BL Lacs is shown in Figure 3.14. Instead of using a singleLorentz factor, a range of values from 5 to 32 with a power law γ distributionproduced an acceptable fit.3.10.2 Modelling of the FR II galaxy and RL QSOluminosity functionsFigure 3.15: Same technique of relativistic beaming applied to the FRIIand QSO radio luminosity functions (Padovani & Urry, 1992). The dashedline is the observed FRII luminosity function. Dots stand for the observedflat spectrum QSO luminosity function. Both observed luminosity functionsare de-evolved to the local Universe values. The solid line is the model fittingof the flat spectrum QSO luminosity function.The overall methodology for the UP version of a unified model for pow-403.11. The Wall & Jackson model: modification of the radio loud AGN unified schemeerful radio sources is almost the same as for FR I–BL Lac objects (Padovani& Urry, 1992). There are two complicatoins for the QSO unification. Thefirst one is the differential evolution of powerful radio sources (Longair, 1966,Schmidt, 1968, Wall et al., 1980). Therefore, for the luminosity functions de-rived from Padovani & Urry (1992) using the 2-Jy sample at 2.7 GHz (Wall& Peacock, 1985), they needed to de-evolve back to a local radio luminosityfunction by assuming their modelling of evolution.The other issue is the number of populations considered in the unifiedscheme. Here the radio sample was classified into three categories, namelyFR II galaxies, steep-spectrum QSOs and flat-spectrum QSOs, an attemptto merge the idea of Orr & Browne (1982) and Barthel (1989) together.By beaming the FR II luminosity function to fit the flat-spectrum QSOluminosity function, a range of 5 to 40 in Lorentz factor with a power lawdistribution was calculated (see Figure 3.15). From the relative numberdensity estimation, the derived cone division angle by UP for radio galaxyand steep-spectrum QSO is 38◦, and 14◦ for steep-spectrum QSO and flat-spectrum QSOs.3.10.3 Unified Scheme of radio-loud populationThe success of the UP approach fitting the data for both FR Is and FR IIsgreatly enhanced the believability of the proposed unified model for radio-loud species (Browne, 1983, Scheuer, 1987b, Peacock, 1987, Barthel, 1989) toa highly credible level. The review paper of UP systematically summarizedthe history, observation, physical and statistical models of radio loud unifiedschemes. Liu & Zhang (2007) and Cara & Lister (2008) provided moreprecise parameter estimation in the UP unified scheme.3.11 The Wall & Jackson model: modification ofthe radio loud AGN unified schemeWall & Jackson (1997), Wall & Jackson (1999) and Jackson & Wall (1999)also carried out a unified scheme, concentrating on source count fitting inthe analytic framework of Wall et al. (1980), rather than luminosity-functionfitting as in UP.413.11. The Wall & Jackson model: modification of the radio loud AGN unified scheme3.11.1 Wall & Jackson (WJ) approach, evolution and radiosource countWJ started with carefully modelling the parent population of radio galaxiesthrough a luminosity-dependent density evolutionary form. The input dataof the luminosity distribution is chosen from a complete flux-limited samplewith redshift data, the 178 MHz 3CR (Laing et al., 1983) steep-spectrumradio galaxies, transposed to 151 MHz. The luminosity distribution andevolution properties of FR galaxies are treated separately as powerful FRII galaxies require significant evolution while FRI evolve mildly (Wall &Jackson, 1997). With the Wall et al. (1980) technique, the radio luminos-ity function for both FR galaxy types at any arbitrary redshift could beconstructed by modelling radio source evolution and 151-MHz radio sourcecount data from 3CRR (Laing et al., 1983) and 6C (Hales et al., 1988) asshown in Figure 3.16.Figure 3.16: Source count fitting at 151 MHz from Jackson & Wall (1999).The low frequency source count is contributed by both FR radio galaxytypes; starburst galaxies do not contribute above a level of about 10 mJy.After determining the space density of radio sources at lower frequency,Monte Carlo simulation could be run to randomly orient the radio sources,with radio flux density values at higher frequencies calculated by adopting423.11. The Wall & Jackson model: modification of the radio loud AGN unified schemeparameters for Lorentz factor and intrinsic core-to-extended ratio (Orr &Browne, 1982). The radio luminosities at 5 GHz are assumed to be domi-nated by the flat-spectrum QSO population. The modelled high frequencysource counts by relativistic beaming are further compared with the ob-served 5-GHz source count (Jackson & Wall, 1999). Lorentz factors of 8.5were optimal for FR IIs and 15.0 for FR Is.The major difference between the WJ and UP approaches is in the fittingcontent. Multi-frequency radio source-count fitting is highlighted in WJ,while UP directly fit he beamed population’s luminosity functions. The re-liability of both approaches depends on the correct modelling of radio sourceevolution. See Chapter 5 for detailed discussion of radio source evolution.Figure 3.17: Source count fitting at 5 GHz from Jackson & Wall (1999),obtained using relativistic beaming of FR radio galaxies.3.11.2 Re-classification and re-unification of FR galaxiesThe one-to-one correspondence of FR radio galaxies to blazars was com-plicated by the observational emergence of a non-trivial number of low-excitation FR II galaxies, as pointed out both by Urry & Padovani (1995)and Wall & Jackson (1997). This type of FR II galaxy is defined by its lack433.11. The Wall & Jackson model: modification of the radio loud AGN unified schemeof emission lines (Figure 3.18 bottom spectrum) and is commonly observedin the low-redshift universe (Laing et al., 1994, Barthel, 1994).This indicates that at least some of the FR II galaxies do not containQSO broad-line regions in their nuclei. Alternatively FR II galaxies thatlack polarized broad emission lines and FR I galaxies are collectively a classof low-excitation radio galaxies (LERG), mentioned in Section 2.3.Figure 3.18: Different type emission lines in the 3CR sample from Lainget al. (1994): (a) broad line, 3C 67; (b) narrow line, 3C 171; (c) weak lineor low excitation, 3C 132.Wall & Jackson (1997) and Jackson & Wall (1999) therefore suggested amodification of the radio loud unified scheme by treating all LERGs as theparent population of BL Lac objects. Such adjustment could further resolvethe lack of predicted FR I galaxy numbers as the parent population required443.12. Development of the radio quiet AGN unified schemeby BL Lac objects (Owen et al., 1996). Consistent FR II lobe morphologyaround some BL Lac objects with similar radio luminosity (Murphy et al.,1993, Kharb et al., 2010) additionally supports the modified unified model.Jackson & Wall (1999) also showed that low excitation FRIIs undergo littleor no cosmic evolution, just as for FR Is.The rest of the FR II galaxies, which do contain the broad line regions(BLR) and are hence defined as high-excitation radio galaxies (HERGs),remain to be unified with RL QSOs. The effect of this adjustment wasagain examined in detail by Jackson & Wall (1999), considering luminosityfunction and relativistic beaming together.Antonucci (2012) pointed out this new dichotomy to remind those whowere still challenging the fundamental UP radio-loud unification. For in-stance, Kharb et al. (2010) showed that the parent population of BL Lacobjects would probably be the entire FR Is plus part of FR IIs, by the ob-servation of extended FR II structure in BL Lac objects, was a threat tothe Urry & Padovani (1995) picture. However this issue had already beenrevised by Jackson & Wall (1999).Insights about the FR dichotomy are further discussed in Chapter 4.3.Physical difference and connections of FR types are critical for understand-ing the central engines of AGN.3.12 Development of the radio quiet AGNunified schemeThough the unified model of radio-quiet (radio-faint) AGN is not the focusof this thesis, it is still of great importance to introduce the major resultsand development of this subject for at least three reasons.(1) Radio-quiet AGN are the dominant population (90%) of the entireAGN zoo (Kellermann et al., 1989).(2) The methodology of studying RL or RQ unification could be mu-tually referential, e.g. the scattered polarized broad-line emission studies(Antonucci & Miller, 1985).(3) The simple point that both RL and RQ species are vital for the studyof AGN physics.The radio-quiet AGN mainly include Seyfert galaxies, radio quiet QSOsand LINERs (Section 2.2). The unified model of radio quiet AGN (Figure453.12. Development of the radio quiet AGN unified scheme3.19), mainly considering Seyfert 2 galaxies as the parent population ofSeyfert 1s, was suggested by Osterbrock (1978).The crucial evidence for this hypothesis is the polarized broad-line emis-sion detection in the Seyfert 2 type galaxy NGC 1068 (Miller & Antonucci,1983, Antonucci, 1983, Antonucci & Miller, 1985, Code et al., 1993), indi-cating an obscuring torus blocking the central broad-line region (See Figure3.7).Figure 3.19: Graphical representation of the Seyfert galaxy unified scheme.Image from Pogge and Peterson (1997).Observation in other spectral regions supported the Seyfert galaxy uni-fication, requiring the obscuring torus structure. For example, Lawrence(1987) pointed out that Seyfert 1s have stronger variability and strongercontinuum from FIR to X-ray than Seyfert 2s. Kinney et al. (1991) showedthat the UV spectra of Seyfert 2s are fainter in magnitude but similar inshape to Seyfert 1s. Moreover, the torus attenuation effect in X-rays is463.12. Development of the radio quiet AGN unified schemealso observed in Seyfert 2s (Lawrence & Elvis, 1982, Smith & Done, 1996,Maiolino et al., 1998, Risaliti et al., 1999).The torus could still have large opacity in the presumably isotropic far-infrared emission (Krolik & Begelman, 1988, Pier & Krolik, 1992) and theorientation dependency of FIR emission is indeed found in radio-loud pop-ulations (Heckman et al., 1992).Figure 3.20: Narrow-line ionization cone structure is shown in the colourimage from Tadhunter & Tsvetanov (1989), with the optical continuum asyellow contours. Blue contours are VLA 1.4-GHz data (Wilson & Tsvetanov,1994).Other straightforward evidence of anisotropic emission in Seyfert galaxiesprovides more insight into the unified scheme. The direct detection of broadIR lines in Seyfert 2s is reported by Goodrich et al. (1994). Pogge (1988) andWilson & Tsvetanov (1994) reported an ionization cone structure extendingfrom the nuclei of NGC1068 and NGC525 (Figure 3.20). The gas within473.12. Development of the radio quiet AGN unified schemethe cone that radiates UV emission lines is ionized by the central continuumaccretion disk. The conical structure (Tadhunter & Tsvetanov, 1989, Wilsonet al., 1993) could imply the effect of a central torus blocking effect forSeyfert galaxies. HST direct imaging of NGC 4261 (Figure 3.21) did directlyreveal an extended disk obscuration structure (Jaffe et al., 1993, Ferrareseet al., 1996).Compared with radio-loud unified models, the orientation effect is moredominant in the torus-blocking effect than relativistic beaming, due to thelack of powerful radio jets. Further details of radio-quiet AGN unified mod-els can be found in several review papers (e.g. Antonucci, 1983, Miller &Goodrich, 1990, Antonucci, 1993, Axon, 2001, Netzer, 2015). Potential is-sues and problems of radio-quiet unified models will be further discussed inChapter 4.Figure 3.21: HST Planetary Camera image of NGC4261, showing diskobscuration around the central black hole. (Jaffe et al., 1993, Ferrareseet al., 1996).483.13. Future development for the AGN unified models3.13 Future development for the AGN unifiedmodelsIt took about three decades to establish and develop the classical model forthe AGN unified scheme of both radio-quiet and radio-loud types. The fun-damental framework was constructed iteratively by continuous observationand physical / statistical modelling.In the twenty-first century, it is expected that more diverse and powerfulobservational data at multi-band and multi-frequency will be used to checkand debate the validity and completeness of the classical unified model.Whether there will be a new theory of AGN unified models, revolution-izing the subject remains unknown. Involving more theoretical quantitieslike black hole mass, accretion mode to classify types of AGN activity andmorphology, seems to be one promising approach (Heckman & Best, 2014).There are plenty of issues pending, such as complications in the torus struc-ture (Elitzur, 2012). Related discussions are further elaborated in the nextchapter.Works has continued on refining the models and more review papersregarding AGN unification are available (Dopita, 1997, Cohen et al., 1999,Wills, 1999, Veron-Cetty & Veron, 2000, Axon, 2001, Urry, 2007, Tadhunter,2008, Beckmann & Shrader, 2012, Bianchi et al., 2012a).49Chapter 4Main issues of unified modelsThis section considers the often debated issues confronting or challenging thecurrent unified model picture. Some subjects are of great current interest,like the linear size inconsistency (Section 4.1), which plans to explore in myPhD research.4.1 Linear projected size and relative numberfraction4.1.1 Requirement of large deprojected size of parentpopulationAn unambiguous and simple prediction of the unified scheme is that thelinear projected size of the QSO populations should be statistically smallerthan parent populations, since their jet axes are closer to the line of sight.This also provides a criterion for finding the parent population of QSOs.The successful candidates for the parent population must therefore possesslarger linear sizes. Historically, the detection of large size radio-loud QSOs(e.g. 4C 34.47 of size 560h−1kpc, Ω0 = 1 Barthel et al., 1989) made therequired size of parent populations significantly larger.For the radio-loud species, radio galaxies (RGs) are the most suitablecandidate for being the parent population of radio-loud QSOs from manyperspectives as discussed above. In terms of linear size, Miley (1980) summa-rized the linear sizes of extragalactic radio sources as having a wide span,from 3C 346 (comparable to the size of host galaxy) up to 3C 236 (at 4Mpc). Gathering early radio surveys like B2 and 3C, he concluded thatedge brightened radio doubles (FR IIs) have a median size of 170 kpc, whileone-sided sources (radio QSOs) appear to be smaller, with typical sizes of afew tens of kpc. FR I morphology sources have a larger uncertainty in size,with median value about 150 kpc. These data are consistent with the radioloud unification model, although RGs had not been linked to RL QSOs atthat time.504.1. Linear projected size and relative number fraction4.1.2 Barthel’s statistical test of RG-QSO unificationAs the 3CR sample (Laing et al., 1983) became more completely identified,it started to be feasible to perform complete statistical tests for examiningthe linear size distribution of RGs and RL QSOs. The low frequency (178MHz) nature of the flux-limited 3CR sample ensured that the radio ob-jects are selected by the orientation-independent diffuse radio lobe emission.This made it the best available unbiased sample for testing the unificationhypothesis.Barthel (1989) carefully studied RGs and RL QSOs in the 0.5 < z < 1range from the 3CR sample. The reason for excluding z > 1 was becausethere was a small number of sources not yet identified. At z < 0.5 locally,there are too few RL QSOs. And so z < 0.5 is removed to avoid smallnumber statistics. Barthel (1989) picked powerful RGs (> 1026 WHz−1)with FRII morphology. There are finally 12 RL QSOs and 30 powerfulRGs in the 0.5 < z < 1 sub-sample. The radio luminosities are compa-rable. Barthel constructed the cumulative apparent linear-size distributionfor both species, which illustrated that projected sizes of RGs are statisti-cally larger than RL QSOs. The median size ratio is 150 kpc/67.5 kpc = 2.2.Based on the assumption that this 3CR subsample sources being completelyrandomly oriented, Barthel further derived a cone division angle if 44◦ andthe theoretical (expected) median size ratio of 1.8.The apparent linear size foreshortening effect and the relative numberfraction were clearly identified in Barthel’s analysis and were utilized as onepiece of evidence to strongly support the unification scheme for powerfulRGs and RL QSOs. Since this phenomenon is straightforward to observe andanalyze, much discussion and debate has occurred among different authorswith different surveys.4.1.3 Arguments against Barthel’s modelLawrence (1991) analyzed the same 3CR sample using all the 172 sources.It was shown that the fraction (relative number ratio) of narrow-line RG tototal RG plus RL QSOs (with an average value of 0.71) decreases with radioluminosity at 178 MHz. This is due to the observation that narrow lineradio galaxies (NLRGs) appeared to be rarer at low luminosity and higherredshift. The non-constant relative number was inconsistent with the sim-plest unified scheme with a single value of the cone division angle (Barthel,1989). Lawrence proposed that the obscuring torus is geometrically thick atlow radio powers and becomes to geometrical thin at high powers. This so514.1. Linear projected size and relative number fractioncalled ‘receding torus model’ is able to explain the decline in the RG fractionwith increasing radio power (Hill et al., 1996, Arshakian, 2005).Singal (1993b) followed this issue by excluding 24 FR I and 8 compactsteep-spectrum sources, since he claimed that they are intrinsically differ-ent from FR IIs. Nine flat spectrum sources were also ruled out, as theyare selected by their core emission in 3CR. Singal curtailed the subsampleto 131 sources. However there is no close similarity shown in the redshiftdistribution of 99 RGs and 32 RL QSOs, contrary to the hypothesis thatthey are the same population viewed from different orientations. Further-more, Singal decomposed the subsample into three redshift bins, z < 0.5 ,0.5 < z < 1 and z > 1. He argued that if the unified scheme works, thesame statistics should be observed, independent of redshift and luminos-ity. Applying Barthel’s method, Singal calculated the expected value of themedian-size ratio and cone-angle of each bin and compared the calculatedsize ratio with the observed ones. It turned out that in all bins, linear sizesof QSOs are less than RGs. But only in the moderate redshift bin is thecalculated median size ratio 2.2 comparable to the observed value 1.8 (closeto that of Barthel, 1989).Both Lawrence (1991) and Singal (1993b) found the cone angle derivedfrom the relative number fraction appears to be increasing with redshift.Hence they proposed cosmic evolution of cone angle to account for such aneffect. But this cannot totally solve the discrepancy. RGs outnumber RLQSOs significantly at the lowest redshifts. If this is due to small cone angles,then the median size of RGs should be statistically larger than RL QSOs.However from observations, the extra size is least significant at z < 0.5.On the other hand, the fraction varies with different samples. In the muchfainter 1-Jy sample (Allington-Smith, 1982), the total RL QSO frequency isonly 15 per cent, less than the 30 percent value in 3CR.In Singal (1993a), more evidence was provided against Barthel’s sim-ple unified model. A mixed sample was compiled of 789 sources from the1-Jy sample (Allington-Smith, 1982), Third Bologna sample (McCarthy,1991), Molonglo sample (McCarthy et al., 1991) and 3CR sample. By directstatistics and modelling of the size evolution, Singal showed, at all redshiftsthat the size of a RG correlates with its luminosity, while RL QSO sizesanti-correlate with luminosity. At any luminosity, while the size of RGs de-creases rapidly with redshift, RL QSOs only reduce in size mildly. Theseobservations appear inexplicable if RGs and RL QSOs come fundamentallyfrom a single population.524.1. Linear projected size and relative number fraction4.1.4 Possible explanations of the QSO fractioninconsistencyVarious explanations have been proposed to reconcile the problem of linearsize and relative number fraction. Barthel (1994) answered some points con-cluded by Singal. He emphasized that completeness is the most importantproperty for a sample. Large-size high-redshift RGs are the hardest objectsto identify. Missing these RGs will cause an apparent discrepancy of thesize-redshift correlation in two classes. As for the inconsistency of RL QSOfraction values, Barthel argued that it was due to small number statisticsof the sample size. Even within the 3CR sample itself, at z > 1, settingthe flux limit 1 Jy deeper alters the value of the QSO fraction from 53 percent to 44 per cent. Therefore it is reasonable to obtain different statisticsacross different samples, e.g. the 1-Jy sample (Singal, 1993b). Constructingsamples of larger size with minimal selection effects is always the ultimateway to reduce the uncertainty. Lastly, in the lowest redshift bins, thereexist a substantial number of FR II RGs which do not harbour a hiddenbroad-line region (low-excitation RG). Such an RG type should be treatedas contamination in the unified scheme and taken out of the analysis, whichtakes care of the problem of too many parent RGs at low redshifts.Urry & Padovani (1995) admitted that Singal’s work was a challenge tothe unification scheme. However, UP pointed out that about half of thesource redshifts in the mixed sample of Singal (1993a) are estimated fromthe luminosity-redshift correlation. Determination of source morphologyand size were not precise, due to poor resolution. These two factors reducedthe credibility of the power/redshift–size correlation Singal obtained.Moreover, the 3CR sample was ideal for obtaining samples of RGs andsteep-spectrum RL QSOs only. The population of flat-spectrum QSOs,which orient with smallest angles according to the unified scheme, are funda-mentally absent in the statistics. This causes the expected median size ratioof RGs to QSOs in the 3CR survey to be less significant. If using an existingsample with enough flat spectrum RL QSOs, like the 2-Jy sample at 2.7 GHz(Wall & Peacock, 1985) instead, the QSO number fraction will be too high,since the 2-Jy survey sample is biased towards core-dominated (beamed)radio sources. Only a large sample size, low-frequency selected sample canavoid small-number statistics (in flat-spectrum sources) and meanwhile se-lect sources of independent orientation.Urry & Padovani (1995) emphasized that the expected de-projectionsize ratio is only a factor of 2, while the intrinsic size scatter of powerfulradio sources is many times larger. Therefore observing RGs smaller than534.1. Linear projected size and relative number fractionRL QSOs is certainly acceptable. The existing largest projected size of aQSO (around 1 Mpc) was not a problem since superluminal motion doesnot necessarily require an extremely small orientation angle. The actualde-projected size will not be very much larger and still less than the largestRG. Similar to Barthel (1994), UP stated that after excluding the LERGs,the QSO fraction across different redshift bins does not vary significantly.Apart from examining the sample selection, there are also more phys-ical models aimed at reconciling the apparent linear size problem. Gopal-Krishna et al. (1994) considered a misalignment between the radio axis andvisibility cone axis by 20◦ to 30◦ to explain the radio size inconsistency. Fal-cke et al. (1995) and Gopal-Krishna et al. (1996) elaborated on the idea ofthe receding-torus model. Gopal-Krishna et al. parametrized their model sothat the torus opening angle of a radio source increases with the initial radiopower. This naturally explained why at low powers the RGs outnumber RLQSOs, due to a smaller cone opening angle. This model also alleviates thedifference of the size-luminosity correlation of the populations reported bySingal (1993a).Singal carried out further studies claiming that the QSO fraction is de-creasing with flux density; but the ratio of RG to RL QSO is also decreasingtowards lower flux (Singal, 1996), a combination that cannot be explainedby orientation unified model. Padovani (1997) argued that all the abnor-mal linear size observations were well reconciled by the receding torus modelwhich provides the non-trivial probability of observing RL QSOs larger thanRGs. Kellermann & Wall (1987) and Jackson & Wall (1999), on the otherhand, argued that the decrease of QSO fraction to lower flux densities isactually predicted by the unified model.4.1.5 Receding torus model vs. exclusion of low-excitationradio galaxiesWith more available radio samples (3CR, 6C, 7C) and optical observationsfor support, Willott et al. (2000) calculated the QSO fraction as a functionof radio power. It was shown that above log P > 26.5, the QSO fractionis independent of radio power with a corresponding cone angle of 53 ◦. Atlower radio power, the QSO fraction is much lower. Both the receding torusmodel and the exclusion of low excitation RGs (believed to be fundamen-tally different from powerful narrow line RGs), are compelling explanationsfor the observed behaviour of the QSO fraction. Grimes et al. (2004) per-formed principle component analysis on the same three radio samples andconstructed the generalized luminosity function. It was shown that with the544.1. Linear projected size and relative number fractionacceptance of two distinct populations of RGs, the idea of a receding torusis no longer required to explain the QSO fraction at low radio powers.Ogle et al. (2006) presented Spitzer mid-infrared (MIR) data on the3CR sample. There are 55 per cent of RGs in the sample without significantemission at 15µm, confirming the existence of RGs with no hidden QSOs atlow radio powers. For the other 45 per cent of the RGs, direct observation ofpolarized broad emission lines confirmed that they do contain hidden broad-line regions and these RGs are regarded as the genuine parent populationof RL QSOs.Antonucci (2012) mentioned Singal’s linear size issue at the lowest red-shift bin z < 0.5 with his judgement on the two most constructive solutions.For the receding torus model, this is actually comparing older QSOs, whichevolved their size over time with younger RGs within the same redshift bin.This explains the observed phenomenon, but relies on more assumptionscompared with the LERG exclusion model. The exclusion of LERGs (typ-ically with small linear size) reduces both the QSO fraction and linear sizeinconsistencies at the same time and is supported by latest observations ontorus and hidden QSOs in 3CR radio sources (Ogle et al., 2006, van derWolk et al., 2010).Singal (2014) eventually conceded that the exclusion of the LERGs wouldget the relative number ratio of QSO to RG to be consistent among redshiftbins and make the redshift distribution of both population indistinguishable.However, at z < 0.5 the foreshortening size problems are still not completelyresolved, even though LERGs do possess smaller linear sizes. The mediansize ratio at z < 0.5 is 0.8, less than the expected minimum linear size ratiovalue 1.15 (when only sources lying exactly on the sky plane are consideredas RGs), which is inadequate to demonstrate the foreshortening effect.4.1.6 Radio samples with contradictory linear-size statisticsAfter results from the 3CR sample, Boroson (2011) found observationalresults directly opposite to the predictions of linear-size foreshortening, usingthe Westerbork Northern Sky Survey (WENSS) at 325 MHz, matched withthe Sloan Digital Sky Survey (SDSS). In his sample, 86 radio sources abovelog P > 26.5 objects were selected in the redshift range 0.1 < z < 0.5.Samples were further matched with NVSS (Condon et al., 1998) or FIRST(Becker et al., 1995) at 1.4 GHz. 34 broad-line sources in Boroson’s samplehave a median size 350 kpc, while the median size of the 52 narrow-linesources is 200 kpc. The median size is obviously opposite to that of thesimple unified scheme prediction of Barthel (1989).554.1. Linear projected size and relative number fractionSingal & Laxmi Singh (2013) explored a larger (494 sources) and deeper(Slim = 1 Jy) independent MRC sample at 408 MHz (McCarthy et al., 1998).Only steep-spectrum powerful sources (log P ≥ 25) were considered. Dueto incomplete redshift information, angular size is directly utilized in thesize analysis. Singal & Laxmi Singh (2013) argued that using angular size isequivalent to using linear size, provided that the redshift distributions of RGand QSO are similar. In this MRC sample, there is almost no distinctionbetween the cumulative size distribution of RGs and QSOs. The similarityis universal, even when splitting the sources in different flux density bins,again contrary to Barthel (1994). For size comparison, the sample was againseparated into z < 0.5, 0.5 < z < 1, z > 1 redshift bins. The overall QSOangular sizes are bigger than RGs by a factor of 1.3, although at z > 1 RGsare double the angular size of QSOs.Another sample (Singal & Singh, 2013) composed of 88 RGs and 43QSOs, was selected from the Best, Rottgering and Lehner (BRL) sample(Best et al., 1999) at 408 MHz. The selection procedure was set to be thesame as that of 3CR. Once again in this sample, the size distribution doesnot show significant linear-size overszize of RGs except at z > 1. Thoughthe anomaly at z < 0.5 could be cancelled if half of the RGs in this range areindeed smaller LERGs, it cannot resolve the lack of size foreshortening inthe 0.5 < z < 1 bin. At this moderate redshift, the QSO fraction is already50 percent. In order to offset the size of RGs, some RGs in this bin have tobe assumed to be LERGs, but this will make the QSO fraction even largerand make the cone opening angle too large at moderate redshifts.4.1.7 Testing the unified model by simulation of the linearsize distributionInspired by these puzzling radio size observations, DiPompeo et al. (2013)addressed the lack of size foreshortening by running statistical simulations.They carefully tested the significance of the intrinsic size scatter of radiosources for the size foreshortening. This was a possible explanation of thelack of foreshortening originally proposed by Urry & Padovani (1995), butnever systematically studied. The previous calculations of expected fore-shortening factors were based on the simplification perhaps too ideal thatradio size is the same among the whole population. However this assump-tion still makes sense statistically. If the sample size is sufficiently large,the sample’s size distribution of RGs and QSOs will be identical if they aretruly selected from a single population. Since in reality the radio samplesare still small (approximately 200), it could simply be chance that the size564.1. Linear projected size and relative number fractiondistribution of RGs and QSOs in real samples do not match the predictions.The samples studied in DiPompeo et al. (2013) included those of Barthel(1989), Singal & Laxmi Singh (2013) and Boroson (2011), in which onlyBarthel (1989) produced the expected size foreshortening for QSOs. Toconstruct the intrinsic size distribution, each object in each sample wasrandomly assigned a cone opening angle and the corresponding deprojectedlinear size calculated from this angle. This process was repeated 1000 timesfor each sample to set up the underlying size distribution. The resultingdistribution is further parametrized into functional forms that could be usedin later simulations.To simulate the hypothetical samples, real redshifts and radio types (RGor QSO) were used for each object, with the linear size randomly drawnfrom the parametrized size distribution. The projected size was obtainedby assigning a random orientation angle. Here different cone opening angle(30◦, 45◦, 60◦ or no difference) were tested. Each radio sample was simulated105 times and the median size ratio was calculated for each iteration.From the above method, every repeated sample will have the same red-shift distribution and the same number of RGs and QSOs. There is a secondtype of resampling in DiPompeo et al. (2013), by requiring equal numbersof RG and QSO in each of the three radio samples. There will no longer bethe same redshift distribution between iterations. The purpose of this stepis merely to check whether relative numbers will affect the median size ratiomarkedly. There were five main results:(1) A distribution of intrinsic size does affect the statistics of the observedratio, but it is insensitive to the exact form of the distribution.(2) If the orientation unification is correct, it is not surprising to observean expected size ratio as in Barthel (1989).(3) Even if there is no restriction on viewing angle for RG and QSOs, itis still possible to observe RGs to be larger than QSOs.(4): Allowing a wide range of size distribution, the probability of findingQSOs having larger size (as Singal & Laxmi Singh, 2013, Boroson, 2011,did) is highly unlikely under the basis of the unification model.(5): A reconciliation of the three results under unification was morefortuitous. Although DiPompeo et al. (2013) questioned the result of Singal& Singh (2013) by using apparent angular sizes, a convincing explanationfor these observations under the unification model is still not available.I will consider the classification of radio galaxies and QSOs in more detail(parent and beamed population) in order to address the issue of linear-sizedistribution in my PhD project.574.2. Dichotomy of radio-loud and radio-quiet QSO4.2 Dichotomy of radio-loud and radio-quiet QSOThere has been long debate over the wide spread of radio luminosities ofQSOs, from radio-quiet to radio-loud. The hypothesis that there is just onehomogeneous population of QSOs having a wide scatter of radio luminositieswas proposed by Schmidt (1970), based on an apparent correlation betweenoptical and radio luminosity. This was actually the basic assumption of theSR model (section 3.3).Figure 4.1: The majority of sources whose radio to optical ratio is between0.1 to 1 are regarded as “radio quiet”. Sources with ratio above 100 arethe “radio loud” types. Image and data from the VLA observations of thePalomar Bright Quasar Survey (Kellermann et al., 1989)The other possibility is that there are two distinct populations of QSOs:the radio-loud QSOs and radio-quiet QSOs. This latter view has proved tobe more correct. Observationally, there are extended radio halos detected inRL QSOs by the VLA (Browne & Perley, 1986). The soft X-ray spectral in-dex is flatter among radio-loud type (Wilkes & Elvis, 1987, della Ceca et al.,1994). Further support comes from their statistical properties. Strittmatteret al. (1980) and Kellermann et al. (1989) have shown that the distribu-tion of radio to optical flux in QSOs is bimodal (Figure 4.1). The bimodaldistribution was also confirmed by the FIRST survey (Becker et al., 1995)584.3. More on FR I and FR II radio galaxiesin sources with QSO identifications. Miller et al. (1990) pointed out thatthe two populations are clearly divided by radio power, either radio loud(> 1025 WHz−1) or radio quiet (< 1024 WHz−1).From the perspective of cosmic evolution, RL and RQ QSOs differ signif-icantly as well. Padovani et al. (2014) have shown that the µJy QSOs in theChandra Deep Field South (CDFS) have different evolutionary behaviour.By adopting a pure luminosity evolution model, the radio loud populationevolves negatively (the higher the redshift, the lower the luminosity) with aflatter luminosity function shape, while the radio-quiet sources evolve posi-tively, with a steeper luminosity function.Physically, what causes the dichotomy of QSOs in radio emission and howthe radio jets are produced initially remain active topics. Blandford et al.(1990) noted that ratio loudness as a phenomenological parameter actuallydepends on more fundamental physical quantities like Eddington ratio (AGNaccretion luminosity scaled by Eddington luminosity) and central black holemass (as discussed further in Section 4.7).4.3 More on FR I and FR II radio galaxiesThough primarily separated by the morphological criteria described in Sec-tion 2.3, FR type galaxies differ significantly in other aspects. In radioluminosity, as emphasized in the original discovery paper (Fanaroff & Riley,1974), there is a dividing line at radio luminosity of ≈ 5 × 1025 WHz−1 .Almost all FR II galaxies have a higher luminosity than this value and mostFR I galaxies luminosities lie below this level. Owen & Ledlow (1994) gen-eralized the division into a bivariate radio and optical luminosity diagram(Figure 4.2).As for QSO radio loudness, FR galaxies differ in cosmic evolution, mainlydue to the fact that the evolution of radio sources is function of radio power(Longair, 1966, Wall et al., 1980). Low-power FR Is undergo mild evolutionwhile FR IIs evolve strongly (Wall & Jackson, 1997). In a more recentstudy (Sadler et al., 2007), within a moderate redshift range (z < 0.7),the evolution of low-and high-power radio sources were found to still bedistinguishable. But the low-power RGs (FR Is) do have significant positiveevolution rather than none or negative evolution. The same result wasobtained by Donoso et al. (2009) for redshifts between 0.4 < z < 0.8.594.3. More on FR I and FR II radio galaxiesFigure 4.2: FR break as a function of both optical R-band magnitude andradio luminosity at 1.4 GHz (Owen & Ledlow, 1994).The physical dichotomy of FR radio galaxies is still an active field ofresearch. The debate is framed around whether the cause is environmentalor intrinsic. For instance, FR Is tend to have more distorted or plume-likeradio structures (Figure 4.3), since their radio jets are subsonic and henceeasier to interact with the ambient medium, while the jet speeds of FR IIsare often supersonic (Kembhavi & Narlikar, 1999). This distinction couldbe caused by either differences in nature of the central engine or differencesin the intergalactic medium.Baum et al. (1995) summarized the findings of FR dichotomy, favouringthe explanation of central engine differences, from a comprehensive studyof correlations of emission-line luminosity, host galaxy magnitude and radioluminosity.With the recognition of lack of excitation in some FR IIs (Laing et al.,1994), radio galaxies are now reclassified by their optical excitation level.The physical distinction caused by excitation state due to central enginedifferences appears to decide the FR dichotomy. It has now been collectively604.4. Lack of parent population of type 2 radio-quiet QSOsconfirmed that the excitation state is connected to the accretion efficiencyof AGN (Heckman & Best, 2014).Figure 4.3: The bending FR I galaxy 3C 465 from the VLA (Eilek et al.,1984).A recent view of our understanding of the FR dichotomy is in Saripalli(2012). Some peculiar radio galaxies are referred as “FR 0” type. FR 0galaxies are similar to local FRI galaxies but with a higher core-to-extendedflux ratio (Baldi et al., 2015).4.4 Lack of parent population of type 2radio-quiet QSOsA major achievement of radio-quiet unified models is the framework ofSeyfert unification, with Seyfert 1 galaxies representing the pole-on versionsof Seyfert 2s from the direct detection of broad-line regions. Meanwhile,the radio-quiet QSOs are apparently the brighter counterpart of Seyfert 1galaxies.614.4. Lack of parent population of type 2 radio-quiet QSOsHowever, with the significant number of low-luminosity Seyfert 2 galaxiesdetected (Risaliti et al., 1999), the high-power counterpart, i.e, type 2 radio-quiet QSOs are for some reason missing from observations (Peterson, 1997,Veron-Cetty & Veron, 2000).The lack of detection could be simply due to the limitation that type 2QSOs are systematically very faint due to heavy obscuration. With the helpof O[III] luminosity from unobscured narrow-line regions, SDSS does find887 type 2 QSOs with redshift z < 0.83 (Reyes et al., 2008).Figure 4.4: Graphical illustration of the receding torus model to explainthe lack of type 2 QSOs at high luminosity (Hill et al., 1996).The receding-torus model for radio-quiet AGN (Hill et al., 1996) seemsto be a way to escape the requirement of large numbers of type 2 QSOs. Thereceding torus picture states that the inner radius of the torus sublimatesmore with increasing radio luminosity (Figure 4.4). Hence at high luminosi-ties, the fraction of type 1 QSO (or equivalently the probability detecting thebroad-line region of an AGN) would be statistically higher (Bennert et al.,2006).An alternative approach is to search for candidate type 2 QSOs. Type 2radio-quiet QSOs, if they indeed exist, should show themselves through largeamounts of infrared emission from the torus dust. Therefore the ultra lumi-624.5. True type Seyfert 2nous infrared galaxies (ULIRG, LIR > 1012L) and sub-millimetre galaxies(Blain et al., 2002) which show AGN-related activity may be promoted tobe the type 2 QSOs (Veilleux et al., 2009).4.5 True type Seyfert 2There is another issue in radio-quiet unification: not every Seyfert 2 galaxyshows polarized broad emission lines (Miller et al., 1990). The fraction ofSeyfert 2 with BLRs is hard to determine (50 percent in the 12 µm sampleof Tran, 2001) as the Seyfert samples are always preferentially selected viahigher luminosity in emission lines.Detailed observations in the optical and X-rayx confirm the existenceof these true Seyfert 2s, without BLR and obscuration (Ghosh et al., 2007,Bianchi et al., 2012b). The true type 2 Seyferts are exceptions in the radioquiet unified scheme, an issue which has not yet been properly addressed(Netzer, 2015).4.6 Torus complicationsThe torus is a crucial component of AGN, as discussed in Section 2.1. Thetorus itself emits (re-radiates) mainly anisotropic infrared radiation from itsmolecular and dusty material by absorbing nuclear luminosity. Furthermore,viewed from side-on, it blocks the emission from the central broad-line re-gion. The classical unification scheme (Antonucci, 1993, Urry & Padovani,1995) has shown that the incorporation of torus structure into the unifiedmodel is necessary.With more detailed infrared observations, modern theories of more com-plex torus structures have been developed based on the classical view, e.g.Pier & Krolik (1992). The complications of torus structure are more rele-vant in the radio-quiet unified scheme, since radio-quiet AGN do not haveanisotropy caused relativistic radio jets (Netzer, 2015). There are even stud-ies about sequential evolution of torii (Haas et al., 2003).Receding torus model The receding torus model was introduced inradio-loud and radio-quiet unification models separately. It is employed toexplain the non-uniform QSO fraction for radio-loud classes (Section 4.1.4).Additionally in Seyfert unification (Lawrence, 1987), it can reconcile thelack of detection of type 2 QSOs.634.6. Torus complicationsWith the notion of the clumpy torus and covering factors, the recedingtorus could be generalized to the phenomenon that the covering factor de-creases with the AGN luminosity, as reported in recent study by Mateoset al. (2016).Multiple torii on different scales Maiolino et al. (1995) reported thelack of 10 µm emission in Seyfert 2s and Risaliti et al. (1999) showed thata large fraction of local Seyfert 2s are Compton-thick, with column den-sity NH > 10−23cm−2. Both imply the presence of a 10s-of-parsec-scaleCompton-thick gas regions in the central BLR, blocking IR emission lines(Figure 4.5).Figure 4.5: A second 10-pc torus inferred from the high column density ofNH in Seyfert 2s (Risaliti et al., 1999).Clumpy torus model The classical torus model (Pier & Krolik, 1992,Efstathiou & Rowan-Robinson, 1995) considers a uniform, solid and non-scattering structure with clear inner and outer radius and height. It broadlyexplains the infrared emission features in Seyfert 1 and 2 galaxies.However, this cannot resolve the Si 10 µm puzzle. The 10 µm emissionis indeed observed in Seyfert 2s, with absorption features in optically-thickedge-on tori. But emission features are not shown in Seyfert 1s (Roche644.7. AGN dichotomies by other physical issueset al., 1991). The classical torus model could only suppress silicon emissionin a fine-tuned range of parameters. Moreover, the observed FIR emissionis much broader than the model prediction.Rowan-Robinson (1995) proposed that clumpiness could reduce both dif-ficulties within a spherical geometry. It was sometime later that Nenkovaet al. (2002) proposed the clumpy torus structure, which successfully mod-elled the infrared spectral energy distribution with the same geometry viewedat different angles (Figure 4.6).Figure 4.6: The clumpy torus model for modelling IR spectral energydistribution. Whether an AGN is type 1 or type 2 now depends on thecovering factor (Nenkova et al., 2002, Elitzur, 2012).The clumpy torus structure may allow finite but non-zero probability toobserve a type 1 AGN at high inclination and observe a type 2 AGN whenthe object is pole-on (Elitzur, 2012). Orientation then is no longer the onlyfactor deciding the AGN type. The covering factor is defined as the fractionof the sky an AGN could see. An AGN will probably be preferentially oftype 1 if it has a small covering factor. AGN with large covering factortend to be type 2. Seyfert galaxies are reported to be consistent with thisexpectation by Ramos Almeida et al. (2011). However, there are still manycovering-factor related issues not yet fully resolved (Netzer, 2015).4.7 AGN dichotomies by other physical issuesPhysical understanding of AGN classification is of great importance to AGNtheory. These physical quantities include radio jet power, black hole mass,Eddington ratio and black hole spin, quantities that can be calculated or654.7. AGN dichotomies by other physical issuesestimated by correlation from the observables e.g. radio luminosity and hostgalaxy magnitude (Ghisellini & Celotti, 2001).Classifying AGN by underlying physical quantities could shed light onthe most relevant questions (Blandford et al., 1990) such as: How are su-permassive black holes formed? What triggers the AGN phase or activity?How are the radio jets formed? How does AGN feedback operate?Accretion mode A new physical dimension of AGN classification via ac-cretion rate has recently impacted the subject. This concerns the differencein AGN accretion, resulting in different fuelling and feedback mechanisms(Falcke et al., 2004, Ho, 2008, Trump et al., 2011, Antonucci, 2012).Figure 4.7: Schematic (left) of radiative mode (high accretion efficiency)AGN and (right) radio mode (low accretion efficiency) AGN, plus the ori-entation effect (Heckman & Best, 2014).For instance, in the FR classification, the accretion difference is illus-trated by excitation states from observational data (Laing et al., 1994).The dichotomy of the excitation states is generalized by Heckman & Best(2014) for the entire AGN population. In particular, the high-excitationobjects refer to AGN that dominate in radiative energy (hence called the664.7. AGN dichotomies by other physical issues“radiative accretion mode”) by efficient black-hole accretion (accretion rategreater than 1 percent of Eddington limit). The low-excitation sources havelower accretion efficiency with an advection-dominated accretion flow mech-anism (“radio mode”). The dominant energy output is the kinetic radio jet(Figure 4.7).It has been shown (Best & Heckman, 2012) that, in radio-loud popu-lations, HERGs evolve strongly at all radio luminosities, with lower stellarand black hole masses. The LERGs, on the other hand, show little or nocosmic evolution, which is consistent with the results favoured by the dual-population unified scheme (Jackson & Wall, 1999).Figure 4.8: AGN classification by accretion rate, on top of the classicalorientation unified scheme. XRBs refers X-ray binaries (Falcke et al., 2004).Black hole spin & radio loudness The accretion-mode classificationcannot address the presence of powerful radio jets. Radiative-mode AGNcould be either radio-loud QSO or radio-quiet sources. Theoretically, ra-dio loudness is attributed to the black-hole spin or angular momentum,as indicated in Figure 4.8 (Blandford et al., 1990, Wilson & Colbert, 1995,Blandford, 1999). The radio-loud AGN tend to appear in massive black-holesystems with significant spin; details can be found in Sikora et al. (2007)and Chiaberge & Marconi (2011).674.8. Why low radio frequency work is importantGrand unified model The ultimate purpose of any sort of unified modelis always an attempting to use the minimal number of parameters to ex-plain all the morphological diversity of AGN. The so-called grand unifica-tion specifically refers to the AGN unification by radio loudness. The veryfirst attempt by Scheuer & Readhead (1979), who intended to unify radioloudness by orientation and relativistic beaming effects, turned out to beinadequate. But the grand unified model returned to prominence with thestudy of accretion and black hole spin in the theoretical AGN dichotomy(Meier, 2002).4.8 Why low radio frequency work is importantNew low-frequency radio observations, together with new methods of clas-sifying radio-loud and radio-quiet populations, re-ignite the study of unifiedmodels. There are multiple aspects to attack. In the remainder of thisthesis, we make a start on two of these:(1) a determination of the low-frequency radio luminosity function (theorientation independent parent population);(2) modern statistical methods to explore the cosmic evolution of thelocal radio luminosity function.Coming back to the orientation unified model, if two observationallydistinct populations are believed to be identical, then any orientation inde-pendent quantity they have has to be indistinguishable statistically.Within radio-loud AGN, the low-frequency radio extended continuumemission is still believed to be the best orientation-independent indicator,since narrow-line regions are located outside the central torus and suffer theleast extinction. Comprehensive study of the radio-loud unified scheme willrequire larger samples of low frequency radio surveys such as MWA GLEAM(see Chapter 6).68Chapter 5Modelling of powerful radiosource evolutionTo rigorously study the AGN unified model, the issue of the strong cosmicevolution of powerful radio sources has to be considered. Distinct radiosource populations show different cosmic evolution. But if two populationare intrinsically the same, their space densities should undergo similar evolu-tionary behaviour, as Dunlop & Peacock (1990), Rigby et al. (2011) showedfor radio galaxies and flat-spectrum QSOs.In this chapter, a classical evolution modelling technique (Wall et al.,1980, hereafter WPL) is employed to check and test various forms of ra-dio source evolution. With modern statistical techniques such as MonteCarlo Markov chain (MCMC), estimation of optimized parameter valuesand uncertainty in the parameters can be obtained to validate any adoptedevolution model.5.1 Brief description of the Wall, Pearson &Longair (WPL) techniqueThe primary purpose of the WPL methodology is to construct the radioluminosity function and to model the form of the radio source evolution. Theprocedure described in this section is consistent with the original paper (Wallet al., 1980). Following the prescribed procedure, the method of MCMC isimplemented as an extension of this numerical scheme.5.1.1 Luminosity function and evolution functionThe radio luminosity function ρ(Pν , z) represents the comoving space densityof radio sources at a given redshift z and monochromatic power (luminosity)Pν . The radio sources could be of a specific population or several typesof population together. In this thesis, the source type is specified to besteep-spectrum radio sources at a fixed particular low frequency, with the695.1. Brief description of the Wall, Pearson & Longair (WPL) techniqueradio luminosity function denoted as ρν(Pν , z). The spectral index of radioemission is also assumed to be a single value overall. For simplicity, thenotation ρ(P, z) is adopted for the rest of the chapter.The radio luminosity function ρ(P, z) at any a particular redshift canbe factorized by the evolution function F (P, z) multiplied by the local radioluminosity function (LRLF) ρ0(P ) = ρ(P, z = 0):ρ(P, z) = F (P, z)ρ0(P, z). (5.1)The evolution function F (P, z) traces how the space density changes as afunction of redshift and radio power from local information at zero redshift.There are two main types of evolution, density evolution and luminosityevolution; and the above factorization can encompass both. Physically ofcourse there is a difference, but mathematically the two forms of evolutionare actually indistinguishable as they simply correspond to translationalmotion of the curve of the luminosity function.5.1.2 WPL procedure: set-up of the local radio luminosityfunctionTo begin the WPL procedure, a flux-limited (S0) complete steep-spectrumradio source sample is prepared from observational data. For each ra-dio source in the sample, the absolute monochromatic radio luminosity Pν(WHz−1sr−1) is calculated by its observed specific radio flux density Sν ,spectral index α and redshift z, obtained from spectroscopic measurements:Pν = dc(z)2Sν(1 + z)1+α = dL(z)2Sν(1 + z)−1+α. (5.2)In the above equation, the convention of definition of spectral index adoptedis Sν ∝ ν−α. The K-correction term is (1 + z)−1+α.The comoving distance dc and luminosity distance dL = (1 + z)dc arecalculated from the redshift. In the very local Universe, these two distancesconverge and are proportional to the redshift (Hubble’s Law). The exactrelationship of redshift and distance depends on the cosmological model.The general expression of comoving distance is expressed in a numericalintegration from the Friedmann equation:dc(z) =cH0∫ z0(ΩM (1 + z′3) + Ωk(1 + z′2) + ΩΛ)− 12dz′. (5.3)H0 is the Hubble constant; the value used is 50 kms−1Mpc−1. The Ωi arerelative density to today’s critical densities ρc =3H028piG . For calculation con-705.1. Brief description of the Wall, Pearson & Longair (WPL) techniquevenience and to check agreement with WPL, the cosmological model chosenin this chapter is the matter-dominated flat universe with no cosmologicalconstant: ΩM = Ω = 1,Ωk = 0,ΩΛ = 0. In this case, the comoving distancecan be obtained analytically:dc(z) =2cH0Ω2(1 + z)[Ωz + (Ω− 2)(√1 + Ωz − 1)]. (5.4)With the distance Equation 5.4, differential comoving volume is subse-quently obtained and will be utilized in later steps:dV (z) = 4pidc(z)2dz,∆V (z) = 4pidc(z)2∆z. (5.5)From the calculation of radio luminosities from the sample with flux andredshift data, a discrete histogram distribution of radio luminosity can bebuilt. To set up a continuous luminosity distribution instead, each sourceis then convolved with a narrow Gaussian, centred at its own luminosity(WPL).All these Gaussians are superposed and divided by the sample angulararea sky coverage to form a master luminosity distribution N(P ). HereN(P )dP represents the number of radio sources per steradian with radiopower between P to P+dP and redshift z = 0 to zmax(P, S0) (the maximumredshift at which a source with power P could be detected above flux limitS0). Due to the large range of radio power, the luminosity distribution isconstructed with logarithmic values of P .The main feature of the WPL scheme is the construction of a masterluminosity distribution. The only other input is a low-frequency sourcecount. The aim is to find a parametrized evolution function to reproducethe source count.The luminosity function may be shown to be given by:ρ0(P )dP =4piN(P )dP∫ zmax(P,S0)0 F (P, z)dV (z). (5.6)The key equation is derived from the reverse calculation of the number ofradio sources per steradian in a small range of redshift and luminosity fromz to z + dz and P to P + dP :n(P, z)dz =14piF (P, z)ρ0(P )dV (z). (5.7)Again for computational convenience, discrete summation is utilized insteadof numerical integration. The (P, z) evolution plane mapping F (P, z), is715.1. Brief description of the Wall, Pearson & Longair (WPL) techniquepixelated with 100 × 100 cells. The range of the radio power is set to befrom 20 to 30 in log luminosity, in unit of WHz−1sr−1. The redshift rangein log(1 + z) is chosen to be from 0 to 1.With the discretization, the number of sources per sr between Pi to Pi+1and zj to zj+1 is nown(Pi, zj) =14piF (Pi, zj)ρ0(Pi)∆P∆V (zj). (5.8)The discrete version of formulating the local radio luminosity function (LRLF)becomes:ρ0(Pi)∆P =4piN(Pi)∆P∑j(Pi,S0)1 F (Pi, zj)∆V (zj). (5.9)The next and final step is modelling the unspecified F (P, z) function andfitting to the observed radio source count.5.1.3 WPL procedure: fitting the source countSince the master luminosity distribution is fixed by the input data, theLRLF depends exclusively on the form of the evolution function F (P, z).In the WPL calculation, the procedure is to guess a parametrized form forF (P, z) and optimize the parameters. This optimization is done by fittingthe calculated radio source count to an observed source count.The cumulative radio source countN(>S) is the number of sources abovethe flux density Sν per steradian, at a given radio frequency. More oftenused is the differential radio source count dNdS . The differential source countis approximated by ∆N∆S in specific bin sizes in a radio sample. It has theadvantage that counts in each bin are statistically independent of each other,which also allows for the construction of a radio source count by combiningdifferent radio surveys reaching different flux levels.Within the framework of the previous section, the modelled radio sourcecount can be calculated by a double summation in the luminosity-redshiftspace:N(> S) =14pi100∑i=1j(Pi,S)∑j=1ρ0(Pi)F (Pi, zj)∆P∆V (zj), (5.10)where the unique flux of pixel (i, j) must satisfy the inequalityS≥Pi/dc(zj)2/(1 + zj)1+α. (5.11)725.1. Brief description of the Wall, Pearson & Longair (WPL) techniqueThe model differential source count can then be calculated by subtractingtwo cumulative radio source count values obtained at two consecutive fluxdensity levels: (dNdS)i=N(∆Si)∆Si=N(Si+1)−N(Si)Si+1 − Si . (5.12)The choices of the flux levels should be identical to the observed radio sourcecount bin data in order to perform a goodness-of-fit test.WPL chose the classical Pearson chi-square χ2 score to measure thequality of the fit. For each source count bin ∆Si, the model differentialsource count N(∆Si) is compared with the data Nobs(∆Si), normalizingthe calculated source count to match the survey area Ai of the bin. Theuncertainty of the observed count is the square root value of itself, so that:χ2 =∑i(N(∆Si)Ai −Nobs(∆Si))2Nobs(∆Si)(5.13)Going from the master luminosity distribution towards the χ2 value, everystep is governed by the parameters of the evolution function F (P, z). Using~α to denote the set of model parameters, the radio source count and the χ2are then both functions of ~α:N(≥ S, ~α) = 14pi∫ ∞0dP∫ zmax(S,P )0ρ0(P )F (P, z, ~α)dV (z); (5.14)χ2(~α) =∑ (N(∆Si, ~α)Ai −Nobs(∆Si))2Nobs(∆Si). (5.15)With the straightforward dependence of ~α, the downhill simplex algo-rithm AMOEBA (Press et al., 1992) was used in this application to minimizethe value of χ2(~α) for the source count fitting. This is the best algorithmamong the popular optimization methods, since downhill simplex only in-volves functional evaluation rather than derivative information, which is notpractical in this WPL calculation. Also downhill simplex itself is efficient inexploring multi-parameter spaces.5.1.4 The method of MCMC and estimation of parameteruncertaintiesCompared with downhill simplex, the MCMC (Lewis & Bridle, 2002) methodcan not only perform parameter optimization, but also sample the posteriorprobability around the optimum location of each parameter.735.1. Brief description of the Wall, Pearson & Longair (WPL) techniqueThe MCMC algorithm introduces a chain of parameter values, movingsemi-randomly in the parameter space. Each step of the chain might resultin a move to a new set of parameters or just stay at the current position.After a sufficient number of steps, the chain value distribution convergesto the stationary posterior probability of the model parameter. Thereafter,direct statistical quantities can be derived from the chain of the parametervalues for each parameter.There are several versions of step strategy in using the MCMC tech-nique. In this WPL scheme analysis, the Metropolis-Hasting algorithm(HASTINGS, 1970) was selected to generate the random walk of modelparameters.If the evolution function model to be investigated is F (P, z, ~α), the priorprobability distribution p(~α) of model parameter ~α must be specified first.Normally the functional form of the prior is assumed as a uniform distribu-tion p(~α) ∝ const or the Jeffreys prior: p(~α) ∝ 1αi .Then supposing ~αi is the i’th value in the parameter chain, a pro-posal function pi( ~αi+1|~αi) is required to move to the next set of multi-parameter values in the chain. The proposal function has to be symmetric:pi( ~αi+1|~αi) = pi(~αi| ~αi+1). The usual choice of proposal function is a multi-variate Gaussian.Given the value of the parameters ~α, the likelihood or the conditionalprobability of the source count data based on the parameter value ~α isp(Data|~α), related to the χ2 fitting value byp(Data|~α) ∝ (χ2(~α)) d2−1e−χ2(~α)2 , (5.16)where d is the number of degrees of the freedom (d.o.f) of the fitting modeldetermined by the number of source count bins N and the number of pa-rameters of the model n, d = N − n+ 1.Therefore the posterior probability of the model parameters is then derivedby Bayes’ theorems,p(~α|Data) ∝ p(~α)(χ2(~α)) d2−1e−χ2(~α)2 . (5.17)In the Metropolis-Hasting algorithm, the posterior probability ratio r =p( ~αi+1|Data)/p(~αi|Data) is calculated. If the ratio is greater than 1, the745.1. Brief description of the Wall, Pearson & Longair (WPL) techniquenew position ~αi+1 will be accepted. If not, the acceptance probability ~αi+1is r, determined by a random variable drawn from a uniform distributionfrom 0 to 1. If the new ~αi+1 is rejected, then ~αi+1 will be set equal to ~αi,preparing for the next proposal chain value.Given the parameter chain, a normalized histogram distribution of eachparameter can be constructed and this behaves as an approximate posteriorprobability distribution. The parameter expectation value and uncertaintyare simply the mean and standard deviation of the chain value. Meanwhilethe optimal parameter set is achieved. Furthermore, the joint distributionfor each pair of parameters is also available in order to study the correlationsof model parameters.Figure 5.1: Master luminosity distribution from the Robertson all-skycatalogue (Robertson, 1973), 87 sources included with S408 > 10 Jy. Thehistogram of the sources is binned in intervals of 0.5 in logP . The curve isthe superposition of all the individual source Gaussian curves; it representsa continuous-function estimate of the master luminosity distribution.755.1. Brief description of the Wall, Pearson & Longair (WPL) technique5.1.5 Data and radio samplesThe WPL numerical procedure requires two sets of radio source data asinput for the entire calculation: a flux-density-limited complete sample toproduce the master luminosity distribution; and a set of radio source countdata to be fit by the model calculation. The survey frequency must beidentical for each. If not, the individual radio luminosities of the luminositydistribution has to be transposed by spectral index calculation Lν ∝ ν−α tothe same frequency (and this only works if the frequencies are closer thanabout a factor of 2).The data used here are the same as in Wall et al. (1980) at 408 MHz.The flux-limited complete sample is constructed from the Robertson all-sky catalogue (Robertson, 1973) and the 3CR sample (Laing et al., 1983).There are 87 bright radio sources in total including QSOs, radio galaxiesand Seyfert galaxies. The flux limit S0 is 10 Jy at 408 MHz and survey areais 5.86 sr.The master luminosity distribution is plotted in Figure 5.1. The half-width (standard deviation) of the smoothing curve applied in summing eachsource is 0.4 in the logP axis.The 408-MHz radio-source count data are a combination of three surveys,including the Robertson all sky catalogue (Robertson, 1973), the 5C survey(Pooley & Kenderdine, 1968, Pearson, 1975) and the Bologna B2 survey(Colla et al., 1973). The source count data can be found in Wall et al.(1980). For this entire chapter, the source count will be plotted in theEuclidean normalization convention: ∆Ni∆N0 where N0 ∝ S−1.5. For a uniformconstant space density distribution, N(>S) ∝ S−1.5.765.2. Analysis of the evolution function model5.2 Analysis of the evolution function modelThe aim of re-using the data from the 1980s is to test whether the newstatistical methods reach the same conclusions and to set up such a techniqueto use the vastly improved modern data to find improved models for differentpopulations.5.2.1 Investigating the models used in WPLFigure 5.2: Source count fit of the non-evolving model with the optimalscale factor 2.65. The blue and green crosses are the normalized 408 MHzsource count data compiled by Wall et al. (1980). The normalization N0 istaken to be 1200S−1.5, and fixed as such for the rest of this chapter.Non-evolving model The non-evolving model (model 1) assumes thereis no cosmic evolution of extragalactic radio sources, i.e. F (P, z) = 1.0.Although there is no parameter in F (P, z), an arbitrary scale factor k couldbe created to adjust the overall normalization of the source count:775.2. Analysis of the evolution function modelχ2(~α, k) =∑ (k ∗N(∆Si, ~α)Ai −Nobs(∆Si))2Nobs(∆Si)(5.18)Treating this scale factor as the single parameter of the non-evolvingmodel, the optimization procedure can reduce the χ2 value of the sourcecount fit significantly. The optimal value of the scale factor is 2.65 with aminimum χ2 of 700 (22 degrees of freedom), rather than 1900 if the scalingfactor is fixed at 1.0. We note that the expected error in the scale factor is1√87≈ ±0.1.The source count plot of the non-evolving model is shown in Figure 5.2.The source count fit and the χ2 value indicate that it is extremely unlikelythat the extragalactic radio sources do not show any degree of cosmic evo-lution, as WPL concluded.Figure 5.3: MCMC output of the WPL non-evolution model. The param-eter chain length is 5000. The optimum value and uncertainty of the onlyparameter, the scale factor, is k = 2.649± 0.065.The MCMC method is implemented in this simple model as it is veryefficient to generate chains with a small number of parameters.The posteriordistribution of the scale factor is shown in Figure 5.3.WPL model 4 (the number does not have any special meaning but isjust a label of models in the original paper Wall et al., 1980). Both model785.2. Analysis of the evolution function model4 and model 5 are were categorized as successful due to their satisfactorysource count fitting. The modelling details are duplicated in this thesis.The functional form of the exponential model 4 isF (P, z) = exp[M(P )τ(t)],where M(P ) encodes the characteristic that the degree of the radio sourceevolution depends on its radio power in the form that less powerful sources(P < P1) undergo mild evolution, while intermediate sources (P1 < P < P2)and powerful sources (P > P2) evolve exponentially in space density withincreasing radio luminosity:M(P ) = 0 for P < P1 ;M(P ) = MmaxlogP−logP1logP2−logP1 for P1 < P < P2 ;M(P ) = Mmax for P > P2 .Here τ(t) is the cosmic look-back time defined as the difference between thetime when a distant source emits its light at te and until it reaches us at thecurrent the age of the Universe t0:τ(t) = (1− tt0 ) = 1H0∫ z0d(z′)(1+z′)√(ΩM (1+z′3)+Ωk(1+z′2)+ΩΛ).If the universe is dominated by matter, the look-back time is simplified to:τ(t) = [1− (1 + z)−1.5] .in units of Hubble time.The constraint on the redshift cutoff parameter zc can be added asIf z > zc, F (P, z) = 0.Therefore there are four parameters P1, P2,Mmax and zc in WPL model 4.The optimal parameter set was obtained using AMOEBA. Optimized pa-rameter values of model 4 are logP1 = 24.44, logP2 = 28.10, M = 15.37 andzc = 2.21. The minimum χ2 value is down to 18 (with 19 degrees of free-dom). This successfully reproduces the initial increase and rapid turnovertowards fainter fluxes of the source count. The source count fit of model 4is shown in Figure 5.4, together with the local radio luminosity function asa by-product of the WPL-type calculation.2The MCMC result for model 4 and other multi-parameter models wouldbe a next step for the analysis of this radio source evolution model. For nowwe find that the parameter results are close to those of WPL, and producea markedly better source-count fit.2The WPL optimized parameter set was (25.0, 27.3, 11.0, 3.5).795.2. Analysis of the evolution function modelFigure 5.4: Source count fit and the local radio luminosity function of WPLmodel 4 at 408 MHz, using the optimal parameters found in this work.805.2. Analysis of the evolution function modelThe space density enhancement plot, or simply the F (P, z) plane formodel 4 is presented in Figure 5.5.Figure 5.5: Space density enhancement of WPL model 4, using optimal pa-rameters found in this work. The colour bar represents the logF (P, z) value.The maximum space density for powerful sources in this model reaches fiveorders of magnitude in amplification.Alternative model 4 as adopted by Jackson & Wall (1999). The suc-cessful WPL model 4 was slighted adjusted into a more physically plausiblemodel with the adoption of a more gradual redshift cutoff:If z > zc2 , F (P, z) = F (P, z − zc2 ) ;If z > zc, F (P, z) = 0.With the above seteup, the maximum radio source epoch could be in-terpreted as zc2 . The optimal parameter values are logP1 = 24.63, logP2 =27.53, M = 13.29 and zc = 3.83. The calculated source count and LRLF815.2. Analysis of the evolution function modeldiffer very little from those shown in Figure 5.4 and the form of the evolutionfunction is shown in Figure 5.6.Figure 5.6: Space density enhancement of an alternative WPL model 4,using optimal parameters found in this work.. With the modification of zc,a trend of space density turnover can be traced from the peak radio sourceepoch (z = zc2 ) to the redshift cutoff (z = zc).WPL model 5 Evolution model 5 defines a transition power Pt to dis-tinguish the evolution of weak and powerful sources and a more “gentle”transition. For weak radio sources, there is mild evolution so that F (P, z)is close to 1 while at high radio power, F (P, z) is dominated by the factorφ(z). The model does not rely on the existence of a redshift cutoff. It isdefined byF (P, z) = X1(P ) + φ(z)X2(P ) ,X1(P ) = (PtP )n/[1 + (PtP )n] ,X2(P ) = 1/(1 + [PtP )n] .The transition power Pt is further parametrized as a monotonically increas-825.2. Analysis of the evolution function modeling function of redshift:Pt = a log(z) + b .The expression for φ(z) is the same as model 4’s exponential form:φ(z) = exp(M(t)) = exp[M(1− (1 + z)−1.5)] .Figure 5.7: Source count fit for WPL model 5. As for model 4, it success-fully captures the shape of the observed source count.There are again four parameters in WPL model 5: M , a, b and n. Theoptimized set of values is M = 10.84, a = 2.16, b = 26.70 and n = 1.22. Thesource count fit is shown in Figure 5.7. The minimum χ2 obtained is 20.0(d.o.f = 19).Fig 5.8 presents a plot of the local radio luminosity function and spacedensity enhancement from WPL model 5. The enhancement reaches a factorof 104.835.2. Analysis of the evolution function modelFigure 5.8: Local radio luminosity function and space density enhancementfor WPL model 5, using optimal parameters found in this work.845.2. Analysis of the evolution function model5.2.2 New but less successful modelsIn this section, new evolution models are investigated. Some models mightonly be conceptually established with as yet imperfect fitting of the sourcecount.Model A: Pure density evolution model with scale factor The formof the famous density evolution model F (P, z) = (1 + z)β has already beendiscussed in the WPL paper, and in earlier literature, such as Longair (1966).Here the source count scale factor k, which was previously utilized in thenon-evolving model, participates as the second parameter in order to betterfit the radio source count.The best pair of parameters are k = 2.134 (again we noted that k ≈1.0±0.1) and β = 2.019, with minimum χ2 value 686 (21 degrees of freedom).The best count of model A is shown in Figure 5.9.Figure 5.9: Source count fit of model A: The model optimal source count fitcannot recover the shape of the observed one, even using the best parametersfound here.The probability distribution of the two parameters is successfully found855.2. Analysis of the evolution function modelby the method of MCMC (Figure 5.10). With a chain length of 10000, theoptimized parameter values with uncertainties from MCMC are k = 2.134 ±0.064 and β = 2.015 ± 0.050, in good agreement with the AMOEBA result.It can be seen that both the scale factor and the evolution exponentprobabilities are distributed as Gaussians. Moreover, they are strongly anti-correlated, with a correlation coefficient −0.665. We would expect this anti-correlation, since more evolution means higher source count.Figure 5.10: MCMC results of Model A. The diagonal plots are the nor-malized posterior distribution of model parameters. The off-diagonal plotsare the joint distribution of the pair of parameters.865.2. Analysis of the evolution function modelModel B: model A + transition power + redshift cutoff On top ofmodel A, a transition radio power Pt and an abrupt redshift cutoff parameterzc are added into the evolution model:If P < Pt , F (P, z) = 1 ;If z > zc , F (P, z) = 0 ;else, F (P, z) = (1 + z)β, identical to model A.Figure 5.11: Source count fit of model B, using optimal parameters foundhere. The model optimal source count cannot recover the observed shape.The optimized parameter sets is k = 1.52, β = 5.04, logPt = 26.68 andzc = 3.62 with minimum χ2 value 151 (19 degrees of freedom). Both Pt andzc lie in the range consistent with successful models in section 5.2.1. Figure5.11 shows the source count of model B.875.3. A new approach to evolution modelling5.3 A new approach to evolution modellingWith the visualization of the F (P, z) evolutions function in previous models,it is clear that the space density of radio sources varies rapidly at differentradio luminosities or epochs. Though additional parametric forms of evolu-tion can equally well fit the source count, there are several trends commonto all models.These shared properties can be summarized as the effect of redshift cutoffbehaviour and cosmic down-sizing (Cowie et al., 1999). The redshift cutoffhas been introduced in Section 3.8.3. The cosmic down-sizing refers to thephenomenon that the redshift where peak space density of the powerfulradio sources is reached decreases as a function of source luminosity, whileincluding less evolution for less powerful sources.The ‘grid model’ has been developed to address the question: can mod-ern data + computing power + statistical techniques obtain the form ofthe evolution with no assumptions? Motivated by the visual inspection ofthe space densities of parametric models above, the grid model assumesthat a certain patch or mega-pixel of the luminosity-redshift plane evolvesuniformly.Therefore the grid model is simply a two-dimensional step-wise func-tion. In the P–z plane, horizontal and vertical dividing lines are set to cutthe plane into mega-pixels being rectangular regions. In each rectangle, aconstant value is assigned to the evolution function as the degree of spacedensity enhancement. The location of dividing lines can also be modelledto be optimized in the calculation or taken from prior knowledge about theradio source population. Here the second approach is taken.The 2 × 2 grid model Work on the grid model starts with one horizontaland vertical dividing line, i.e. four mega-pixels. The vertical line plays therole of redshift cutoff while the horizontal line would act as a transitionpower. The initial vertical line is fixed at z = 3 and logP = 26.5 for thehorizontal line. These values are chosen for illustrative purpose only, asbroadly consistent with prior knowledge.An additional source count scale factor is involved in order to find thebest fit. Therefore there are five parameters in total in this grid model.The calculation follows the WPL technique, with the same master lumi-nosity distribution and source count at 408 MHz. The modelling results arepresented in the Figure 5.12 and 5.13. The minimum χ2 is 104.25 (d.o.f =18) which cannot be regarded as a success. The model source count does itspoorest job at the higher flux levels.885.3. A new approach to evolution modellingFigure 5.12: Source count fit and local radio luminosity function of the2×2 grid model, using the optimal parameters found here.895.3. A new approach to evolution modellingThe optimal scale factor value for the 2×2 grid model is 0.932, ratherclose to 1. The log values of the optimized grid parameter set are(0.269 2.7000.050 1.857)corresponding to location in the P–z plane (Figure 5.13). A direct limita-tion is that the 2×2 grid model cannot capture the redshift cutoff, becausethere is so much evolution happening before z = 3 that a single constantvalue cannot represent it. Therefore the last model, the 3×3 grid model,is implemented to try to reveal the cosmic history before the space densityturnover.Figure 5.13: Space density enhancement of the 2×2 grid model, using theoptimal parameters found in this work with the WPL formulation.The 3×3 grid model Will the situation improve with increased numberof mega-pixels? To answer this question, in the last model investigated thereare two dividing lines are for each direction, located at z1 = 1.5, z2 = 3.0,905.3. A new approach to evolution modellinglogP1 = 24.0 and logP2 = 26.0. There will be nine parameters in the 3×3grid model, referred to as a11 a12 a13a21 a22 a23a31 a32 a33Figure 5.14: Source count fit, 3×3 grid model, using the optimal parametersfound in this work with the the WPL formulation.From prior knowledge about the existence of a redshift cutoff, cosmic down-sizing and differential evolution with radio power, it is reasonable to putconstraints on adjacent patches. For example, reading upwards, each entryin the middle column should be the maximum value in its row; this couldreveal both the radio-source peak epoch and the cutoff effect at higher red-shift. Prior knowledge also says that within each column, the degree ofevolution should increase upwards.During the optimization process, it was found that the parameters au-tomatically satisfy all constraints except that a12 < a13. This is not entirelysurprising since the most powerful radio sources may not fully reach their915.3. A new approach to evolution modellingmaximum density at z = 3; hence a13 could further increase to some degree.From another perspective, this evolution is further compatible with the ef-fect of cosmic downsizing, since powerful sources peak at earlier epochs.Meanwhile three patches a31, a33 and a23 show almost no evolution.Figure 5.14 shows the source count for the 3×3 grid model. The mini-mum χ2 value is 70.7 (12 degrees of freedom). Except at intermediate fluxes,the model over predicts the source count at both ends. This situation mightbe improved by using a scale factor to pull down the overall model sourcecount.Figure 5.15: The local radio luminosity function, in the 3×3 grid model,using the optimal parameters found in this work.In the local radio luminosity function of the 3×3 grid model (Figure5.15), the shape develops kinks compared to previous parametric models.This is due to the non-smooth nature of the evolution function. The valueof F (P, z) changes discontinuously at the dividing lines.The optimized value set of the parameters is presented in the matrix925.4. Gridding continuedrepresentation corresponding to their position in the P–z plane.0.235 2.047 2.5580.713 1.893 0.0000.000 1.882 0.000The positions match with the enhancement plot in Figure 5.16. The max-imum density increase is only 2.5 orders of magnitude, less than the para-metric models of the previous section.Figure 5.16: Space density enhancement, 3×3 grid model, using the optimalparameters found in this work.5.4 Gridding continuedDespite the far from perfect χ2 value and non-smooth luminosity function,the overall shape of this 3×3 grid model is consistent with prior informationas discussed above. It will be intriguing to further explore the grid model.One possible measure will be resetting the positions of the dividing lines tobe variables. Another elaboration would be dropping the constant value of935.4. Gridding continuedF (P, z) for each patch and obtaining the F (P, z) value by 2d interpolationfrom the surrounding pixels.The MCMC method for multi-parameter evolution models needs to befully implemented. Knowing the uncertainty of the model parameters andtheir correlations would be a major step forward and a completion of theproof of concept.More systematic work needs to be done to scrutinize the validity of theevolution model proposed. For instance, how the entire luminosity functionchanges magnitude and shape with respect to redshift needs to be examined.Further steps would then be to generalize to many more pixels, and to theuse of more data sets, source counts, and luminosity distributions at otherfrequencies and flux levels.For the WPL model in this thesis and original paper, the radio sourcepopulation is described only by its power during the calculation. This is acrude approximation. Chapters 3 and 4 showed how radio sources can bedistinguished by morphological appearance and other physical quantities,since unified models predict different population characteristic would havedifferent evolution properties. Larger radio samples with classified popula-tions are needed for this purpose.The limitation of ‘one’ WPL scheme is that it only uses radio data atone single frequency with a single spectral index. An ambitious plan wouldbe collecting WPL modelling at multiple radio frequencies and making surethat at each frequency the WPL modelling works effectively. Then it couldbe attempted to connect these (source count fitting or radio luminosity func-tion) by: either (a) their spectral index; or (b) more ambitiously via a unifiedmodel. We then might approach a full description of the space density evo-lution of radio AGN.94Chapter 6Direct construction of partof the local radio luminosityfunction from the MWAGLEAM 4-Jy sampleFigure 6.1: An MWA antenna tile with dipole elementsLow radio-frequency surveys from specially designed observatories in-95Chapter 6. MWA GLEAM local radio luminosity functionclude Cambridge 3C, 4C, 6C and 7C surveys, Low Frequency Array (LO-FAR), Square Kilometre Array (SKA) and Murchison Widefield Array (MWA).The supervisor of the student, Professor Jasper V. Wall is currently collab-orating on an MWA-related science project, which is expected to lead tofuture PhD thesis work. In this chapter, a first analysis of the 4-Jy MWAGLEAM subsample is introduced as the start of a developing MWA collab-oration.The MWA (Bowman et al., 2013) is located at the Murchison Radio-Astronomy Observatory in Western Australia. It applies radio interferom-etry to perform observations at low radio frequencies from 73 MHz to 300MHz in five frequency bands (bandwidth = 30.72 MHz). Traditional diffi-culties of low-frequency radio surveys, such as poor surface brightness sensi-tivity and inaccurate position estimates, are tremendously reduced by MWAinstrumentation and confusion analysis (Franzen et al., 2016). The MWAis located in an ‘ionization trough’ in the hopes of less ‘smearing’ by ionclouds. The MWA consists of more than 8000 individual dipole antennas,arranged in 512 ‘sub-tiles’. Each tile contains a 4 × 4 antenna array. Mosttiles are distributed within a circular area of 3 km in diameter.The scientific goals of MWA involve: studying the Universe at the epochof re-ionization through the redshifted 21cm line; solar, heliospheric andionospheric research; detection of transient radio events; pulsar discovery;and deep sky surveys. Detailed information of MWA science and design isdescribed in Tingay et al. (2013) and Bowman et al. (2013).A GaLactic and Extragalactic All-sky survey with the MWA(GLEAM) (Wayth et al., 2015) is designed to catalogue the brightest ex-tragalactic radio sources in the southern hemisphere. The next section willintroduce a subsample of the GLEAM survey of bright (>few Jy) sourcesThe bright end of the radio source count could provide important impli-cations for the luminosity distribution of radio sources and their evolutionbehaviour (Longair, 1966, Wall et al., 1980).966.1. The MWA GLEAM 4-Jy sample6.1 The MWA GLEAM 4-Jy sampleFigure 6.2: The sky distribution of MWA GLEAM bright sources (S150 >4 Jy). The horizontal axis is Right Ascension and the vertical axis is Dec-lination in degrees. The hollow U-shape is the zone of avoidance for theGalactic Plane (Galactic latitude |b| > 10◦).There are more than 300,000 extragalactic sources detected in the GLEAMsurvey in its first year of operation using drift scanning. With the large fieldof view, GLEAM is able to cover a large sky area to provide a statisticallymeaningful radio sample with bright and rare sources. With MWA’s sensi-tivity, local faint sources and more distant strong radio sources can both bedetected, making it possible to explore the not well-defined regions of theradio luminosity function.The survey covers the sky from declinations +25◦ to −80◦. The fre-quency is centred at 150 MHz and ranges from 72 to 231 MHz. A detailedsummary of GLEAM can be found in Jackson et al. (2016).From the entire GLEAM sample, sources with flux density S greaterthan 4 Jy at 151 MHz are selected, as these are fully complete in most ofthe survey area (except the Magellanic Clouds and low Galactic latitudes976.1. The MWA GLEAM 4-Jy sample|b| < 10◦). The total area of completeness to 4 Jy is 7.4 steradians. TheGLEAM 4-Jy sample used here was acquired from Dr.Tomas Franzen andProf. Carole Jackson by private communication. There are 2130 sources intotal, ten times larger than the size of the 3CR sample (Laing et al., 1983).Figure 6.3: Normalized source count from the MWA GLEAM 4-Jy sub-sample.Strictly speaking, a portion of the ‘sources’ in the 4-Jy sample actu-ally represent components of extended radio emitting sources. With carefulmatching, only 1873 sources are confirmed to be genuine isolated sources.This issue is not considered in this thesis; each source with a flux value isassumed to be an individual source for this initial analysis.In order to determine more accurate source positions, the 4-Jy subsamplewas further examined with higher frequency data (Jackson et al., 2016). Thepositional cross-match procedure used the Sydney University Molonglo SkySurvey (SUMSS) at 843 MHz (Bock et al., 1999) and NVSS at 1.4 GHz(Condon et al., 1998), both of which have better resolution.Though the issue of component sources exists, the 4-Jy bright samplesource count is shown in Figure 6.3 as a first approximation. There is ageneral decrease of the source count towards higher flux density. The binsare equally divided in log radio flux intervals, except that sources above986.2. The subsample for the local radio luminosity function calculation30 Jy are placed in one bin. The source count data are listed in Table 6.1.Bin start Bin end Number of Euclidean normalized counts(Jy) (Jy) sources Jy3/2 sr−14.00 5.00 675 3847.38 ± 148.095.00 6.26 484 3859.67 ± 175.446.26 7.83 296 3302.48 ± 191.957.83 9.79 227 3543.38 ± 235.189.79 12.25 156 3406.91 ±272.7712.25 15.33 101 3086.04± 307.0715.33 19.17 58 2479.43 ± 325.5619.17 23.98 49 2930.64 ± 418.6623.98 30.00 32 2677.69 ± 473.3530.00 +∞ 37 1765.98 ± 290.33Table 6.1: The MWA GLEAM 4-Jy source count data6.2 The subsample for the local radio luminosityfunction calculationCross match with 6dFGS To proceed to the local luminosity function,redshift data are required to estimate the distance and radio luminosity ofthe 4-Jy sample. Cross-matching with existing low-redshift extragalacticsky surveys is a fast method to obtain the redshift information. Hencethe 6 degree-Field Galaxy Survey (6dFGS) (Jones et al., 2004, 2009) wasemployed for the cross matching, since 6dFGS estimates redshift for brightlow-redshift galaxies (median redshift: 0.053) over the whole southern skyexcept the Galactic plane (17046 deg2). This overlaps most of the sameregion covered by GLEAM.The positional cross-matching was done using the software TOPCATwith a matching radius of 50 arcsec. The GLEAM subsample’s positionsare those of the NVSS and SUMSS data. The number of resultant matchesis 165 from 124,647 6dFGS galaxies and 2130 GLEAM 4-Jy sources.There are three factors producing lack of matching for the rest of theGLEAM 4 Jy sources. Firstly, 6dFGS has a K-band limiting magnitude of12.75. Secondly, 6dFGS and GLEAM do not overlap in the declination re-gion +25◦ to 0◦ and −80◦ to −90◦. Lastly, 6dFGS is a low-redshift (z ∼ 0.15)996.2. The subsample for the local radio luminosity function calculationextragalactic survey; it will not match GLEAM sources of larger redshift.The final sources, systematically identified, are likely to fall into one of thefollowing categories: weak radio AGN; star-forming galaxies; or starburstgalaxies.Redshift constraint Here a cutoff value 0.1 was manually adopted forthe redshift, when determining the sources considered in the local radioluminosity function. It turns out that 98 of the 165 sources of the abovesubsample have redshift z < 0.1.Cross-match with 2MASS Since the 6dFGS sources have a limitingK-band magnitude (mK = 12.75) because they were selected from the TwoMicron All Sky Survey 2MASS (Skrutskie et al., 2006), knowing the K-band(wavelength λ = 2.2µm) magnitude of the sources with the redshift is veryimportant if a local radio luminosity function is constructed by the V/Vmaxmethod. The value of the K-band magnitude was obtained by anothercross-match with 2MASS for the remaining 98 GLEAM sources.The cross-match with 2MASS resulted in 88 sources with direct K-bandinformation. For the other 10 sources, each was searched for in NASA Ex-tragalactic Database (NED). Seven of the remaining 10 sources do have a K-magnitude in the database, but two of them have K-band magnitudes above12.75. Therefore the final subsample size of GLEAM local radio sources was93.Final GLEAM local radio sub-sample The final subsample of 93sources is listed in Table 6.2.Table 6.2: The MWA GLEAM local bright radio source sub-sample.MWA name 6dFGS name S151MHz mK z zmax zmax zmax(Jy) (mag) radio K-bandJ185043-630428 g1850490-630428 4.193 11.614 0.016 0.016 0.027 0.016J004600-633354 g0046003-633352 5.139 11.852 0.075 0.085 0.117 0.085J005734-012321 g0057349-012328 21.084 10.549 0.045 0.100 0.133 0.100J021645-474842 g0216451-474909 11.689 10.541 0.064 0.107 0.189 0.100J031002-301930 g0310015-301940 4.433 12.342 0.068 0.071 0.083 0.071J031757-441418 g0317577-441417 7.988 10.465 0.076 0.106 0.231 0.100Continued on next page1006.2. The subsample for the local radio luminosity function calculationTable 6.2 – continued from previous pageMWA name 6dFGS name S151MHz mK z zmax zmax zmax(Jy) (mag) radio K-bandJ034631-342238 g0346306-342246 18.338 12.047 0.053 0.110 0.075 0.075J034844-041252 g0348430-041313 4.845 12.689 0.068 0.075 0.070 0.070J035139-274353 g0351358-274435 29.989 12.539 0.066 0.171 0.073 0.073J044829-203214 g0448306-203214 4.796 12.271 0.074 0.081 0.094 0.081J050122-033219 g0501201-033239 4.396 12.531 0.088 0.092 0.098 0.092J050452-101452 g0504531-101453 10.3 11.198 0.04 0.063 0.086 0.063J051823-561403 g0518264-561413 7.47 11.602 0.095 0.128 0.167 0.100J052524-324154 g0525272-324216 5.643 10.895 0.077 0.091 0.191 0.091J081631-703917 g0816228-703940 5.764 10.883 0.07 0.083 0.175 0.083J112554-352319 g1125529-352340 12.437 9.701 0.033 0.057 0.147 0.057J121418-415950 g1214188-415954 5.951 12.698 0.069 0.083 0.071 0.071J124849-411842 g1248493-411840 23.912 7.142 0.01 0.024 0.150 0.024J130100-322622 g1301008-322629 5.067 8.507 0.017 0.019 0.134 0.019J130527-492802 g1305273-492805 14.908 4.483 0.002 0.004 0.101 0.004J133709-295152 g1337005-295160 10.272 4.619 0.002 0.003 0.094 0.003J163856-031327 g1638569-031411 7.01 12.757 0.06 0.079 0.060 0.060J192531-425726 g1925297-425711 4.34 12.538 0.077 0.080 0.086 0.080J195816-550935 g1958185-550930 27.518 11.813 0.058 0.145 0.092 0.092J201126-564321 g2011275-564407 4.984 12.056 0.055 0.061 0.077 0.061J202435-051632 g2024353-051641 6.505 11.634 0.082 0.103 0.142 0.100J205202-570406 g2052023-570408 5.705 8.75 0.012 0.014 0.082 0.014J210139-280147 g2101377-280154 26.748 10.346 0.039 0.097 0.127 0.097J220114-374641 g2201171-374624 8.828 10.856 0.033 0.049 0.083 0.049J223911-172052 g2239114-172028 10.2 11.362 0.074 0.115 0.147 0.100J231358-424337 g2313586-424339 4.459 10.587 0.056 0.059 0.162 0.059J235701-344531 g2357007-344533 24.988 9.817 0.049 0.118 0.206 0.100J000310-544451 g0003112-544458 4.409 10.176 0.033 0.035 0.117 0.035J000313-355629 g0003130-355614 8.006 10.477 0.05 0.070 0.153 0.070J002113-191038 g0021075-191006 6.332 12.036 0.096 0.119 0.137 0.100J002530-330334 g0025292-330354 9.533 12.379 0.051 0.078 0.061 0.061J003703-010907 g0037041-010908 22.498 11.915 0.073 0.165 0.110 0.100J004615-420740 g0046178-420752 33.612 12.659 0.053 0.146 0.055 0.055J010241-215222 g0102418-215256 13.966 10.512 0.057 0.104 0.171 0.100J011109-135741 g0111091-135740 4.573 10.475 0.052 0.055 0.159 0.055J011815-255143 g0118147-255200 4.717 12.614 0.053 0.057 0.057 0.057Continued on next page1016.2. The subsample for the local radio luminosity function calculationTable 6.2 – continued from previous pageMWA name 6dFGS name S151MHz mK z zmax zmax zmax(Jy) (mag) radio K-bandJ013716-091140 g0137153-091152 4.165 10.462 0.041 0.042 0.126 0.042J023136-204022 g0231370-204022 5.04 12.034 0.09 0.100 0.128 0.100J024240-000045 g0242407-000048 20.291 5.788 0.004 0.009 0.111 0.009J025926-394040 g0259267-394038 4.162 11.527 0.066 0.067 0.121 0.067J031638-435126 g0316393-435117 4.509 10.906 0.063 0.067 0.156 0.067J033403-385911 g0334067-385933 6.807 11.473 0.061 0.079 0.115 0.079J040845-133551 g0408441-133525 4.231 12.563 0.089 0.091 0.098 0.091J042904-534922 g0429082-534940 23.715 9.761 0.038 0.090 0.165 0.090J043636-222614 g0436354-222639 4.896 10.964 0.069 0.076 0.166 0.076J051104-131645 g0511048-131731 5.929 10.51 0.043 0.052 0.130 0.052J062140-524109 g0621433-524133 15.7 9.801 0.051 0.098 0.216 0.098J062622-534122 g0626221-534126 53.786 10.385 0.054 0.184 0.173 0.100J062646-543234 g0626496-543234 22.783 10.784 0.052 0.119 0.137 0.100J062707-352906 g0627067-352915 18.536 10.509 0.055 0.114 0.166 0.100J064425-434359 g0644251-434349 7.593 11.106 0.061 0.083 0.137 0.083J065021-554929 g0650215-554933 6.256 10.397 0.05 0.062 0.159 0.062J071706-362143 g0717081-362159 13.499 10.326 0.031 0.056 0.102 0.056J080536-005813 g0805378-005818 9.834 10.703 0.09 0.138 0.242 0.100J100439-321638 g1004400-321642 5.566 12.164 0.088 0.103 0.118 0.100J101503-235708 g1015007-235728 5.024 12.653 0.032 0.036 0.034 0.034J105533-283128 g1055334-283134 6.37 11.292 0.061 0.076 0.125 0.076J105854-362045 g1058548-361921 8.207 12.05 0.07 0.099 0.099 0.099J110611-244443 g1106121-244444 9.475 12.359 0.05 0.076 0.061 0.061J122343-423529 g1223434-423532 8.076 12.335 0.027 0.038 0.033 0.033J125441-291338 g1254410-291340 13.385 9.885 0.057 0.102 0.230 0.100J125530-133843 g1255294-133823 4.018 11.767 0.097 0.097 0.157 0.097J125722-302140 g1257219-302149 7.024 10.313 0.054 0.071 0.179 0.071J133224-330821 g1332253-330816 8.754 11.857 0.049 0.072 0.076 0.072J141028-424649 g1410289-424656 10.717 11.015 0.051 0.082 0.120 0.082J142003-493539 g1420037-493542 14.816 11.666 0.091 0.168 0.155 0.100J142152-392346 g1421554-392333 7.667 12.395 0.083 0.113 0.099 0.099J145510-053912 g1455090-053927 6.997 11.485 0.04 0.053 0.074 0.053J145509-365540 g1455096-365508 7.156 11.332 0.095 0.125 0.190 0.100J164418-771517 g1644161-771549 22.428 11.2 0.043 0.099 0.092 0.092J170242-774143 g1702410-774157 6.661 12.505 0.095 0.121 0.107 0.100Continued on next page1026.3. The GLEAM local radio luminosity functionTable 6.2 – continued from previous pageMWA name 6dFGS name S151MHz mK z zmax zmax zmax(Jy) (mag) radio K-bandJ180958-455239 g1809579-455241 4.109 12.012 0.07 0.071 0.101 0.071J192605-574007 g1926057-574017 6.662 11.242 0.061 0.078 0.128 0.078J192819-293141 g1928170-293144 8.01 9.555 0.024 0.034 0.114 0.034J193138-335437 g1931382-335442 8.513 11.863 0.098 0.140 0.152 0.100J201758-553909 g2018013-553932 7.933 11.638 0.061 0.085 0.106 0.085J203444-354849 g2034447-354902 8.812 11.792 0.089 0.129 0.143 0.100J204343-263259 g2043457-263301 10.378 10.298 0.041 0.065 0.137 0.065J214014-441237 g2140141-441228 5.073 12.391 0.071 0.080 0.085 0.080J214358-563746 g2143592-563721 6.928 11.165 0.082 0.106 0.178 0.100J215707-694120 g2157060-694124 104.471 10.049 0.028 0.135 0.105 0.100J222751-303341 g2227507-303343 4.765 12.323 0.056 0.061 0.069 0.061J232519-120728 g2325197-120727 16.056 12.096 0.082 0.158 0.113 0.100J021046-510059 g0210462-510102 6.369 11.977 0.032 0.040 0.047 0.040J052257-362729 g0522580-362731 61.993 11.342 0.057 0.207 0.114 0.100J155701-791347 g1556589-791404 6.256 12.466 0.059 0.073 0.068 0.068J222350-020640 g2223495-020613 34.813 11.602 0.056 0.156 0.099 0.099J234740-280835 g2347433-280837 5.786 12.241 0.027 0.032 0.034 0.032Column 1 and 2 list the source ID in GLEAM and 6dFGS. Column 3 is theradio flux measured by MWA at 151 MHz. Column 4 is the 2MASS K-bandmagnitude of the source. Column 5 is the redshift of source from 6dFGScross?matching. Column 6 and Column 7 are the maximum redshifts if thesource was at the respective flux limits. Column 8 is the final maximumredshift taking the upper limit z = 0.1 into consideration (See Section 6.3).6.3 The GLEAM local radio luminosity functionWith the final GLEAM subsample, it is now possible to construct the localradio luminosity function (LRLF). The LRLF provides information of howradio sources are distributed in radio power in the local Universe. It alsoperforms as a calibrated starting point for radio source evolution.The 93 sources all lie in the overlapping region of MWA and NVSS withan area of 17046 deg2. Their radio luminosity is calculated from Equation5.2. The cosmology adopted here is the Lambda Cold Dark Matter universe1036.3. The GLEAM local radio luminosity functionwith H0 = 70 km s−1Mpc−1, ΩM = 0.3 , ΩΛ = 0.7, for the distance calcu-lation in Equation 5.3. Figure 6.4 shows the luminosity-redshift diagram ofthe subsample.Figure 6.4: Luminosity-redshift diagram of the GLEAM local subsample.The majority of the sources populate logP from 25.0 to 26.5 (log WHz−1).The sources are then binned in uniform log luminosity intervals ∆ logP .Following the 1/Vmax method, with the conventions of Schmidt (1968), Con-don (1989) and Best & Heckman (2012), the equation of direct constructionof luminosity function (space density) for N sources in each luminosity bin[logP , logP + ∆ logP ] isρ(P ) =N∑i=1( 1Vm,i), (6.1)The uncertainty of the luminosity function in each bin isσρ(P ) =[ N∑i=1( 1Vm,i)2]1/2. (6.2)where the Vm,i is the maximum comoving accessible volume of each sourcein the subsample. The Vm,i value is calculated by the maximum redshift1046.4. Determining Vmax of each sourcezmax at which the source could be marginally detected above the flux limit(optical and radio), selecting the minimum of the optical and radio zmax.6.4 Determining Vmax of each sourceVmax is determined as followsRadio limit For each source with given 6dFGS redshift z, the absoluteradio luminosity Pν is calculated from Equation 5.2, assuming a spectralindex value of -0.7 for the radio K-correction. The maximum radio limitedredshift corresponds to the maximum redshift at which the source could bedetected above the GLEAM bright source limit S0 = 4 Jy:Pν = dc(zmax)2S0(1 + zmax)1+α. (6.3)The maximum radio redshift can then be inversely determined from Equa-tion 6.3.Optical (K-band) limit Similarly, in order to find the maximum opticalredshift, the absolute K-band magnitude mK is calculated firstly bymK = mK + 5 log(dL(z)/10pc)−K(z). (6.4)The term K(z) is the K-band K-correction. The form of the K-band K-correction is adopted from Glazebrook et al. (1995). Then the K-band max-imum redshift is found by inverse calculation from the absolute magnitudeMK and the K-band limit, using the maximum magnitude mKlim = 12.75.MK = mK lim + 5 log(dL(zmax)/10pc)−K(zmax). (6.5)Both the radio redshift limit and K-band redshift limit are recorded in Table6.2. Each source must survive above both limits in order to be detected.The true value of the maximum redshift is the minimum of these two andthe manual redshift cut at z = 0.1. The final maximum redshift for eachsource is also listed at the last column of the GLEAM data table.Obtaining comoving volume V is obtainec as by V = A4pi4pi3 dc(z)3, thesame for obtaining Vmax by zmax.1056.5. Checks and comparisons6.5 Checks and comparisonsV/Vmax check The V/Vmax method (Schmidt, 1968) is a quick way to checkwhether the radio sources have cosmic evolution in a flux-limited completesample. The V/Vmax distribution is expected to be uniform in [0, 1] if there isno evolution. A mean value of V/Vmax greater than 0.5 will indicate positiveevolution towards larger redshift.Figure 6.5: The V/Vmax distribution of the GLEAM local radio subsample.The V/Vmax distribution is shown in Figure 6.5. The calculated valueof 〈V/Vmax〉 is 0.540 ± 0.030. Although the sample suffers small numberstatistics at the low radio luminosity end (logP < 25.0), the overall dis-tribution is broadly consistent with a non-evolving scenario in the redshiftrange z < 0.1. The sample standard deviation is 0.287, matching closely tothe standard deviation of a uniform distribution ( 112)0.5 = 0.289.The GLEAM local radio luminosity function For compilation of theGLEAM LRLF, the luminosity interval ∆ logP was chosen to be 0.4, asin Condon (1989), Sadler et al. (2002) and Mauch & Sadler (2007). This1066.5. Checks and comparisonscorresponds to one magnitude difference in a magnitude system as ∆m =m2−m1 = 2.5 log10(P1/P2). Also the choice of 0.4 ensures enough bins andavoids small number statistics in most of the bins.Figure 6.6: The GLEAM LRLF. The LRLFs from Mauch & Sadler (2007),Best & Heckman (2012) are also shown.The LRLF is plotted in Figure 6.6. Calculation of each bin is listedin Table 6.3. The GLEAM bright local radio luminosity function is com-pared with 6dFGS (Mauch & Sadler, 2007) and SDSS (Best & Heckman,2012). Mauch & Sadler (2007) calculated their LRLF with 6667 sourcesfrom cross-matching the NVSS at 1.4 GHz and 6dFGS, while Best & Heck-man (2012) cross-matched NVSS and SDSS to obtain a sample of 9048radio sources. Both LRLFs are shifted horizontally here by P151MHz =P1.4GHz(151/1400)−α, assuming α = 0.7.1076.5. Checks and comparisonsBin start Bin end Number of log(ρ/logP (W Hz−1) logP (W Hz−1) sources Mpc−3mag−1)22.8 23.2 2 -3.4323.2 23.6 1 -4.9723.6 24.0 1 -5.5824.0 24.4 1 -5.7524.4 24.8 2 -5.7824.8 25.2 7 -5.8925.2 25.6 13 -6.2525.6 26.0 37 -6.1926.0 26.4 23 -6.6226.4 26.8 6 -7.21Table 6.3: The MWA GLEAM local radio luminosity function dataComparison of the GLEAM LRLF with other LRLFs After con-verting to the same low radio frequency of 151 MHz, the GLEAM LRLF hasa lower estimation of space density than the other two below logP = 25.0W Hz−1. In the luminosity range less than logP = 24.0W Hz−1, smallnumber statistics of GLEAM local sample minimizes the reliability of thespace density calculation and should probably not be considered. No errorbars were drawn in three bins with only one source for each.Between 24.0 < logP < 25.0, the new estimation is less than 1 order ofmagnitude below the other two. This could be accounted for by a couple offacts. The sample size of Mauch & Sadler (2007), Best & Heckman (2012)are about 100 times greater than GLEAM subsample. However, the manuallocal redshift cut used in Best & Heckman (2012) is z < 0.3. This willreduce each individual value of 1Vmax resulting in a decrease of the spacedensity estimation.Furthermore a 1.4-GHz radio survey will be biased towards flat-spectrumsources (Wall, 1994). Those flat spectrum radio sources will fall below thedetection limit of low frequency surveys. But this factor is trivial in thissituation, since the NVSS sample limit is set at 2.8 mJy, deep enough forGLEAM sources to be observed at GHz frequencies (but not above 4 Jy).Nevertheless the variation of radio source nature among different fre-quency domains affects the credibility of the comparison, even though radiopowers are properly adjusted by the spectral index.1086.6. Future work on the GLEAM 4-Jy bright-source sample6.6 Future work on the GLEAM 4-Jybright-source sampleThe GLEAM bright radio sub-sample has already shown the importance ofGLEAM by direct estimation of the local radio luminosity function. Moreinteresting projects could be performed on the full GLEAM 4-Jy sample.Several projects could be started immediately.(1) The discrepancy between the LRLF derived from GLEAM and NVSShas not been fully understood. Figuring out reasonable explanations is thefirst thing to check for the GLEAM sample.(2) The detailed morphological study of the 4 Jy sources: identifyingthe ‘sources’ that are components of extended radio objects in order todetermine the true sample size of isolated radio sources.(3) Spectroscopic identification of all the GLEAM bright sources: The 4-Jy sample is intriguing in that it is complete over the entire Southern sky, toa much deeper level than 3CR or the MRC. It will be statistically powerfulonce it becomes redshift-complete. So far the cross-match with 6dFGS onlyproduced redshift information on 165 sources out of 2130.109Chapter 7ConclusionAlthough the modern framework of both radio-loud and radio-quiet AGNunified models has been firmly established for over twenty years (Antonucci,1993, Urry & Padovani, 1995, Chapter 2 and 3), there are concerns aboutunexplained issues and complications. For radio loud unification, the prob-lem of projected linear size (Singal, 1993b, DiPompeo et al., 2013, Chapter4) is still not clearly understood.To analyze in an unbiased way the issues in radio-loud unified modelssuch as projected linear size, the effects of strong radio source evolutionare unavoidable, since otherwise selection effects would bias the statisticalanalysis. To quantitatively capture the radio source evolution, grid modelsbased on a classical modelling technique (WPL Wall et al., 1980) have beenimplemented. The 3×3 grid model reveals radio source evolution consistentwith prior knowledge (Chapter 5) and suggests a way forward to encompassmodern data sets.Apart from the new modelling strategy of radio source evolution, themodern generation of low-frequency radio data can also provide direct testsfor the unified models with fewer primary selection effects. A tentativelocal radio luminosity function constructed from the MWA GLEAM sample(Jackson et al., 2016) has already shown the value of the radio data (Chapter6). It shows clearly how to extract more science by progressive completionof the GLEAM data reduction process.The ultimate aim is to provide better theoretical understanding of AGN-related physics such as galaxy formation and black-hole AGN co-evolutionthrough observational evidence plus statistical analysis. It is anticipatedthat the framework of unified models will provide a solid contribution to thesubject.110BibliographyAllington-Smith, J. R. 1982, MNRAS, 199, 611Andrew, B. H., MacLeod, J. M., Harvey, G. A., & Medd, W. J. 1978, AJ,83, 863Angel, J. R. P. & Stockman, H. S. 1980, Ann Rev Astron Astrophys, 18,321Antonucci, R. 1993, Ann Rev Astron Astrophys, 31, 473Antonucci, R. 2012, Astronomical and Astrophysical Transactions, 27, 557Antonucci, R. R. J. 1983, Nature, 303, 158Antonucci, R. R. J. 1984, ApJ, 278, 499Antonucci, R. R. J. & Miller, J. S. 1985, ApJ, 297, 621Antonucci, R. R. J. & Ulvestad, J. S. 1985, ApJ, 294, 158Arshakian, T. G. 2005, A&A, 436, 817Axon, D. J. 2001, Evidence for the Torus, Past, Future, and Elsewhere: TheUnified Theory of Active Galactic NucleiBailey, J., Sparks, W. B., Hough, J. H., & Axon, D. J. 1986, Nature, 322,150Baldi, R. D., Capetti, A., & Giovannini, G. 2015, A&A, 576, A38Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5Barthel, P. D. 1989, ApJ, 336, 606Barthel, P. D. 1994, in Astronomical Society of the Pacific Conference Series,Vol. 54, The Physics of Active Galaxies, ed. G. V. Bicknell, M. A. Dopita,& P. J. Quinn, 175111BibliographyBarthel, P. D., Hooimeyer, J. R., Schilizzi, R. T., Miley, G. K., & Preuss,E. 1989, ApJ, 336, 601Baum, S. A., Zirbel, E. L., & O’Dea, C. P. 1995, ApJ, 451, 88Becker, R. H., White, R. L., & Helfand, D. J. 1995, ApJ, 450, 559Beckmann, V. & Shrader, C. R. 2012, Active Galactic NucleiBegelman, M. C., Blandford, R. D., & Rees, M. J. 1984, Reviews of ModernPhysics, 56, 255Bennert, N., Jungwiert, B., Komossa, S., Haas, M., & Chini, R. 2006, A&A,459, 55Best, P. N. & Heckman, T. M. 2012, MNRAS, 421, 1569Best, P. N., Rottgering, H. J. A., & Lehnert, M. D. 1999, MNRAS, 310, 223Bianchi, S., Maiolino, R., & Risaliti, G. 2012a, Advances in Astronomy,2012, 782030Bianchi, S., Panessa, F., Barcons, X., et al. 2012b, MNRAS, 426, 3225Bicknell, G. V. 1994, ApJ, 422, 542Bicknell, G. V., Dopita, M. A., & O’Dea, C. P. O. 1997, ApJ, 485, 112Blain, A. W., Smail, I., Ivison, R. J., Kneib, J.-P., & Frayer, D. T. 2002,physrep, 369, 111Blandford, R. D. 1999, in Astronomical Society of the Pacific ConferenceSeries, Vol. 160, Astrophysical Discs - an EC Summer School, ed. J. A.Sellwood & J. Goodman, 265Blandford, R. D. & Ko¨nigl, A. 1979, ApJ, 232, 34Blandford, R. D., Netzer, H., Woltjer, L., Courvoisier, T. J.-L., & Mayor,M., eds. 1990, Active Galactic Nuclei, 97Blandford, R. D. & Rees, M. J. 1974, MNRAS, 169, 395Blandford, R. D. & Rees, M. J. 1978, physscr, 17, 265Bock, D. C.-J., Large, M. I., & Sadler, E. M. 1999, AJ, 117, 1578Bolton, J. G., Stanley, G. J., & Slee, O. B. 1949, Nature, 164, 101112BibliographyBoroson, T. A. 2011, in Bulletin of the American Astronomical Society,Vol. 43, American Astronomical Society Meeting Abstracts 217, 142.22Boroson, T. A. & Green, R. F. 1992, ApJS, 80, 109Bowman, J. D., Cairns, I., Kaplan, D. L., et al. 2013, pasa, 30, e031Bregman, J. N. 1990, aapr, 2, 125Bridle, A. H., Hough, D. H., Lonsdale, C. J., Burns, J. O., & Laing, R. A.1994, AJ, 108, 766Bridle, A. H. & Perley, R. A. 1984, Ann Rev Astron Astrophys, 22, 319Browne, I. W. A. 1983, MNRAS, 204, 23PBrowne, I. W. A. 1989, in Lecture Notes in Physics, Berlin Springer Verlag,Vol. 334, BL Lac Objects, ed. L. Maraschi, T. Maccacaro, & M.-H. Ulrich,401Browne, I. W. A., Clark, R. R., Moore, P. K., et al. 1982, Nature, 299, 788Browne, I. W. A. & Murphy, D. W. 1987, MNRAS, 226, 601Browne, I. W. A. & Perley, R. A. 1986, MNRAS, 222, 149Burbidge, G. & Hewitt, A. 1992, in Variability of Blazars, ed. E. Valtaoja& M. Valtonen, 4Burn, B. J. 1966, MNRAS, 133, 67Capetti, A., Macchetto, F., Axon, D. J., Sparks, W. B., & Boksenberg, A.1995, ApJL, 452, L87Cara, M. & Lister, M. L. 2008, ApJ, 674, 111Chiaberge, M. & Marconi, A. 2011, MNRAS, 416, 917Code, A. D., Meade, M. R., Anderson, C. M., et al. 1993, ApJL, 403, L63Cohen, M. H., Cannon, W., Purcell, G. H., et al. 1971, ApJ, 170, 207Cohen, M. H., Linfield, R. P., Moffet, A. T., et al. 1977, Nature, 268, 405Cohen, M. H., Lister, M. L., Homan, D. C., et al. 2007, ApJ, 658, 232113BibliographyCohen, M. H., Ogle, P. M., Tran, H. D., Goodrich, R. W., & Miller, J. S.1999, AJ, 118, 1963Cohen, M. H. & Unwin, S. C. 1984, in IAU Symposium, Vol. 110, VLBI andCompact Radio Sources, ed. R. Fanti, K. I. Kellermann, & G. Setti, 95Colla, G., Fanti, C., Fanti, R., et al. 1973, A&A, 11, 291Condon, J. J. 1984a, ApJ, 287, 461Condon, J. J. 1984b, ApJ, 284, 44Condon, J. J. 1989, ApJ, 338, 13Condon, J. J., Cotton, W. D., Greisen, E. W., et al. 1998, AJ, 115, 1693Condon, J. J. & Ledden, J. E. 1982, AJ, 87, 219Cotton, W. D., Wittels, J. J., Shapiro, I. I., et al. 1980, ApJL, 238, L123Cowie, L. L., Songaila, A., & Barger, A. J. 1999, AJ, 118, 603de Zotti, G., Massardi, M., Negrello, M., & Wall, J. 2010, aapr, 18, 1della Ceca, R., Lamorani, G., Maccacaro, T., et al. 1994, ApJ, 430, 533di Serego Alighieri, S., Courvoisier, T. J.-L., Fosbury, R. A. E., Tadhunter,C. N., & Binette, L. 1988, Nature, 334, 591DiPompeo, M. A., Runnoe, J. C., Myers, A. D., & Boroson, T. A. 2013,ApJ, 774, 24Donoso, E., Best, P. N., & Kauffmann, G. 2009, MNRAS, 392, 617Dopita, M. A. 1997, pasa, 14, 230Dunlop, J. S. & Peacock, J. A. 1990, MNRAS, 247, 19Efstathiou, A. & Rowan-Robinson, M. 1995, MNRAS, 273, 649Eilek, J. A., Burns, J. O., O’Dea, C. P., & Owen, F. N. 1984, ApJ, 278, 37Elitzur, M. 2012, ApJL, 747, L33Elvis, M., Wilkes, B. J., McDowell, J. C., et al. 1994, ApJS, 95, 1Fabbiano, G., Trinchieri, G., Elvis, M., Miller, L., & Longair, M. 1984, ApJ,277, 115114BibliographyFalcke, H., Gopal-Krishna, & Biermann, P. L. 1995, A&A, 298, 395Falcke, H., Ko¨rding, E., & Markoff, S. 2004, A&A, 414, 895Falomo, R., Pesce, J. E., & Treves, A. 1993, ApJL, 411, L63Fanaroff, B. L. & Riley, J. M. 1974, MNRAS, 167, 31PFanti, C., Fanti, R., Parma, P., Schilizzi, R. T., & van Breugel, W. J. M.1985, A&A, 143, 292Fanti, R., Kellermann, K. I., & Setti, G., eds. 1984, IAU Symposium, Vol.110, VLBI and compact radio sources; Proceedings of the Symposium,Bologna, Italy, June 27-July 1, 1983Ferrarese, L., Ford, H. C., & Jaffe, W. 1996, ApJ, 470, 444Ferrarese, L. & Merritt, D. 2000, ApJL, 539, L9Franzen, T. M. O., Jackson, C. A., Offringa, A. R., et al. 2016, MNRAS,459, 3314Garrington, S. T., Conway, R. G., & Leahy, J. P. 1991, MNRAS, 250, 171Ghisellini, G. & Celotti, A. 2001, A&A, 379, L1Ghosh, H., Pogge, R. W., Mathur, S., Martini, P., & Shields, J. C. 2007,ApJ, 656, 105Glazebrook, K., Peacock, J. A., Miller, L., & Collins, C. A. 1995, MNRAS,275, 169Goodrich, R. W., Veilleux, S., & Hill, G. J. 1994, ApJ, 422, 521Gopal-Krishna, Kulkarni, V. K., & Mangalam, A. V. 1994, MNRAS, 268,459Gopal-Krishna, Kulkarni, V. K., & Wiita, P. J. 1996, ApJL, 463, L1Gregg, M. D., Becker, R. H., White, R. L., et al. 1996, AJ, 112, 407Grimes, J. A., Rawlings, S., & Willott, C. J. 2004, MNRAS, 349, 503Haas, M., Klaas, U., Mu¨ller, S. A. H., et al. 2003, A&A, 402, 87Hales, S. E. G., Baldwin, J. E., & Warner, P. J. 1988, MNRAS, 234, 919115BibliographyHASTINGS, W. K. 1970, Biometrika, 57, 97Heckman, T. M. 1980, A&A, 87, 152Heckman, T. M. & Best, P. N. 2014, Ann Rev Astron Astrophys, 52, 589Heckman, T. M., Chambers, K. C., & Postman, M. 1992, ApJ, 391, 39Hill, G. J., Goodrich, R. W., & Depoy, D. L. 1996, ApJ, 462, 163Hine, R. G. & Longair, M. S. 1979, MNRAS, 188, 111Hine, R. G. & Scheuer, P. A. G. 1980, MNRAS, 193, 285Ho, L. C. 2008, Ann Rev Astron Astrophys, 46, 475Hough, D. H. & Readhead, A. C. S. 1987, ApJL, 321, L11Jackson, C., Franzen, T. O., Seymour, N., et al. 2016, ArXiv e-prints1604.04041Jackson, C. A. & Wall, J. V. 1999, MNRAS, 304, 160Jackson, C. A., Wall, J. V., Shaver, P. A., Kellermann, K. I., & Hook, I. M.2003, 300, 267Jaffe, W., Ford, H. C., Ferrarese, L., van den Bosch, F., & O’Connell, R. W.1993, Nature, 364, 213Jenkins, C. J., Pooley, G. G., & Riley, J. M. 1977, memras, 84, 61Jones, D. H., Read, M. A., Saunders, W., et al. 2009, MNRAS, 399, 683Jones, D. H., Saunders, W., Colless, M., et al. 2004, MNRAS, 355, 747Jones, M. H., Lambourne, R. J. A., & Serjeant, S. 2015, An Introduction toGalaxies and CosmologyKapahi, V. K. 1981, aaps, 43, 381Kapahi, V. K. & Kulkarni, V. K. 1986, in IAU Symposium, Vol. 119,Quasars, ed. G. Swarup & V. K. Kapahi, 207–210Kapahi, V. K. & Saikia, D. J. 1982, Journal of Astrophysics and Astronomy,3, 465116BibliographyKauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003, MNRAS, 346,1055Kellermann, K. I. & Pauliny-Toth, I. I. K. 1969, ApJL, 155, L71Kellermann, K. I. & Pauliny-Toth, I. I. K. 1981, Ann Rev Astron Astrophys,19, 373Kellermann, K. I., Sramek, R., Schmidt, M., Shaffer, D. B., & Green, R.1989, AJ, 98, 1195Kellermann, K. I. & Wall, J. V. 1987, in IAU Symposium, Vol. 124, Obser-vational Cosmology, ed. A. Hewitt, G. Burbidge, & L. Z. Fang, 545–562Kembhavi, A. K. & Narlikar, J. V. 1999, Quasars and active galactic nuclei: an introductionKewley, L. J., Groves, B., Kauffmann, G., & Heckman, T. 2006, MNRAS,372, 961Kharb, P., Lister, M. L., & Cooper, N. J. 2010, ApJ, 710, 764Kinney, A. L., Antonucci, R. R. J., Ward, M. J., Wilson, A. S., & Whittle,M. 1991, ApJ, 377, 100Kovalev, Y. Y., Kardashev, N. S., Kellermann, K. I., et al. 2016, ApJL, 820,L9Krichbaum, T. P., Alef, W., Witzel, A., et al. 1998, A&A, 329, 873Krolik, J. H. 1998, Active Galactic Nuclei: From the Central Black Hole tothe Galactic EnvironmentKrolik, J. H. & Begelman, M. C. 1988, ApJ, 329, 702Laing, R. A. 1988, Nature, 331, 149Laing, R. A. & Bridle, A. H. 1987, MNRAS, 228, 557Laing, R. A., Jenkins, C. R., Wall, J. V., & Unger, S. W. 1994, in Astro-nomical Society of the Pacific Conference Series, Vol. 54, The Physics ofActive Galaxies, ed. G. V. Bicknell, M. A. Dopita, & P. J. Quinn, 201Laing, R. A. & Peacock, J. A. 1980, MNRAS, 190, 903Laing, R. A., Riley, J. M., & Longair, M. S. 1983, MNRAS, 204, 151117BibliographyLawrence, A. 1987, PASP, 99, 309Lawrence, A. 1991, MNRAS, 252, 586Lawrence, A. & Elvis, M. 1982, ApJ, 256, 410Lewis, A. & Bridle, S. 2002, Phys Rev D, 66, 103511Liu, Y. & Zhang, S. N. 2007, ApJ, 667, 724Longair, M. S. 1966, MNRAS, 133, 421Lynden-Bell, D. 1969, Nature, 223, 690Maiolino, R., Ruiz, M., Rieke, G. H., & Keller, L. D. 1995, ApJ, 446, 561Maiolino, R., Salvati, M., Bassani, L., et al. 1998, A&A, 338, 781Marscher, A. P. 1995, Proceedings of the National Academy of Science, 92,11439Mateos, S., Carrera, F. J., Alonso-Herrero, A., et al. 2016, ArXiv e-prints1601.04439Mauch, T. & Sadler, E. M. 2007, MNRAS, 375, 931McCarthy, P. J. 1991, AJ, 102, 518McCarthy, P. J., Kapahi, V. K., van Breugel, W., et al. 1998, VizieR OnlineData Catalog, 210McCarthy, P. J., van Breughel, W., Kapahi, V. K., & Subrahmanya, C. R.1991, AJ, 102, 522Meier, D. L. 2002, New Astron Rev, 46, 247Miley, G. 1980, araa, 18, 165Miller, H. R., Carini, M. T., & Goodrich, B. D. 1989, Nature, 337, 627Miller, J. S. & Antonucci, R. R. J. 1983, ApJL, 271, L7Miller, J. S. & Goodrich, R. W. 1990, ApJ, 355, 456Miller, J. S., Goodrich, R. W., & Mathews, W. G. 1991, ApJ, 378, 47Miller, L., Peacock, J. A., & Mead, A. R. G. 1990, MNRAS, 244, 207118BibliographyMorganti, R., Fosbury, R. A. E., Hook, R. N., Robinson, A., & Tsvetanov,Z. 1992, MNRAS, 256, 1PMorris, S. L., Stocke, J. T., Gioia, I. M., et al. 1991, ApJ, 380, 49Murphy, D. W., Browne, I. W. A., & Perley, R. A. 1993, MNRAS, 264, 298Nenkova, M., Ivezi’c, v., & Elitzur, M. 2002, ApJL, 570, L9Netzer, H. 2006, in Lecture Notes in Physics, Berlin Springer Verlag, Vol.693, Physics of Active Galactic Nuclei at all Scales, ed. D. Alloin, 1Netzer, H. 2015, Ann Rev Astron Astrophys, 53, 365O’Dea, C. P. 1998, PASP, 110, 493Ogle, P., Whysong, D., & Antonucci, R. 2006, ApJ, 647, 161Orr, M. J. L. & Browne, I. W. A. 1982, MNRAS, 200, 1067Osterbrock, D. E. 1978, Proceedings of the National Academy of Science,75, 540Osterbrock, D. E. 1989, Astrophysics of gaseous nebulae and active galacticnucleiOsterbrock, D. E. 1991, Reports on Progress in Physics, 54, 579Owen, F. N. & Ledlow, M. J. 1994, 54, 319Owen, F. N., Ledlow, M. J., & Keel, W. C. 1996, AJ, 111, 53Padovani, P. 1997, memsai, 68, 47Padovani, P. 1999, 65, 159Padovani, P., Bonzini, M., Miller, N., et al. 2014, 304, 79Padovani, P. & Urry, C. M. 1990, ApJ, 356, 75Padovani, P. & Urry, C. M. 1992, ApJ, 387, 449Peacock, J. A. 1985, MNRAS, 217, 601Peacock, J. A. 1987, in NATO Advanced Science Institutes (ASI) SeriesC, Vol. 208, NATO Advanced Science Institutes (ASI) Series C, ed.W. Kundt, 185–196119BibliographyPeacock, J. A., Miller, L., & Longair, M. S. 1986, MNRAS, 218, 265Pearson, T. J. 1975, MNRAS, 171, 475Pearson, T. J., Unwin, S. C., Cohen, M. H., et al. 1981, Nature, 290, 365Penston, M. V. & Cannon, R. D. 1970, Royal Greenwich Observatory Bul-letins, 159, 83Perley, R. A., Dreher, J. W., & Cowan, J. J. 1984, ApJL, 285, L35Peterson, B. M. 1997, An Introduction to Active Galactic NucleiPier, E. A. & Krolik, J. H. 1992, ApJ, 401, 99Pogge, R. W. 1988, ApJ, 328, 519Pooley, G. G. & Kenderdine, S. 1968, MNRAS, 139, 529Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1992,Numerical Recipes in C (2Nd Ed.): The Art of Scientific Computing (NewYork, NY, USA: Cambridge University Press)Ramos Almeida, C., Levenson, N. A., Alonso-Herrero, A., et al. 2011, ApJ,731, 92Readhead, A. C. S. & Wilkinson, P. N. 1978, ApJ, 223, 25Rees, M. J. 1966, Nature, 211, 468Rees, M. J. & Simon, M. 1968, ApJL, 152, L145Reyes, R., Zakamska, N. L., Strauss, M. A., et al. 2008, AJ, 136, 2373Richards, G. T., Strauss, M. A., Fan, X., et al. 2006, AJ, 131, 2766Rigby, E. E., Best, P. N., Brookes, M. H., et al. 2011, MNRAS, 416, 1900Risaliti, G. & Elvis, M. 2004, in Astrophysics and Space Science Library,Vol. 308, Supermassive Black Holes in the Distant Universe, ed. A. J.Barger, 187Risaliti, G., Maiolino, R., & Salvati, M. 1999, ApJ, 522, 157Robertson, J. G. 1973, Australian Journal of Physics, 26, 403Robson, I. 1996, Active galactic nuclei120BibliographyRoche, P. F., Aitken, D. K., Smith, C. H., & Ward, M. J. 1991, MNRAS,248, 606Rowan-Robinson, M. 1995, MNRAS, 272, 737Rybicki, G. B. & Lightman, A. P. 1979, Radiative processes in astrophysicsRyle, M. & Clarke, R. W. 1961, MNRAS, 122, 349Ryle, M. & Hewish, A. 1950, MNRAS, 110, 381Sadler, E. M., Cannon, R. D., Mauch, T., et al. 2007, MNRAS, 381, 211Sadler, E. M., Jackson, C. A., Cannon, R. D., et al. 2002, MNRAS, 329, 227Saikia, D. J. 1995, Proceedings of the National Academy of Science, 92,11417Sandage, A. 1972, ApJ, 178, 25Sandage, A., Ve´ron, P., & Wyndham, J. D. 1965, ApJ, 142, 1307Saripalli, L. 2012, AJ, 144, 85Scheuer, P. A. G. 1974, MNRAS, 166, 513Scheuer, P. A. G. 1987a, in Superluminal Radio Sources, ed. J. A. Zensus &T. J. Pearson, 104–113Scheuer, P. A. G. 1987b, in NATO Advanced Science Institutes (ASI) Se-ries C, Vol. 208, NATO Advanced Science Institutes (ASI) Series C, ed.W. Kundt, 129–136Scheuer, P. A. G. & Readhead, A. C. S. 1979, Nature, 277, 182Schmidt, M. 1963, Nature, 197, 1040Schmidt, M. 1968, ApJ, 151, 393Schmidt, M. 1969, Ann Rev Astron Astrophys, 7, 527Schmidt, M. 1970, ApJ, 162, 371Schmitt, J. L. 1968, Nature, 218, 663Schraml, J., Pauliny-Toth, I. I. K., Witzel, A., et al. 1981, ApJL, 251, L57121BibliographySeyfert, C. K. 1943, ApJ, 97, 28Shaver, P. A., Hook, I. M., Jackson, C. A., Wall, J. V., & Kellermann, K. I.1999, 156, 163Shaver, P. A., Wall, J. V., Kellermann, K. I., Jackson, C. A., & Hawkins,M. R. S. 1996, Nature, 384, 439Sikora, M., Stawarz,  L., & Lasota, J.-P. 2007, ApJ, 658, 815Singal, A. K. 1993a, MNRAS, 263, 139Singal, A. K. 1993b, MNRAS, 262, L27Singal, A. K. 1996, MNRAS, 278, 1069Singal, A. K. 2014, AJ, 148, 16Singal, A. K. & Laxmi Singh, R. 2013, ApJ, 766, 37Singal, A. K. & Singh, R. L. 2013, MNRAS, 435, L38Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163Smith, D. A. & Done, C. 1996, MNRAS, 280, 355Snellen, I. A. G. 2008, ArXiv e-prints 0802.1976Stickel, M., Padovani, P., Urry, C. M., Fried, J. W., & Kuehr, H. 1991, ApJ,374, 431Stocke, J. T., Morris, S. L., Weymann, R. J., & Foltz, C. B. 1992, ApJ, 396,487Strittmatter, P. A., Hill, P., Pauliny-Toth, I. I. K., Steppe, H., & Witzel, A.1980, A&A, 88, L12Tadhunter, C. 2008, New Astron Rev, 52, 227Tadhunter, C. & Tsvetanov, Z. 1989, Nature, 341, 422Tadhunter, C. N., Scarrott, S. M., Draper, P., & Rolph, C. 1992, MNRAS,256, 53PTingay, S. J., Goeke, R., Bowman, J. D., et al. 2013, pasa, 30, e007Tran, H. D. 2001, ApJL, 554, L19122BibliographyTran, H. D., Cohen, M. H., & Goodrich, R. W. 1995, AJ, 110, 2597Tran, H. D., Cohen, M. H., Ogle, P. M., Goodrich, R. W., & di SeregoAlighieri, S. 1998, ApJ, 500, 660Tran, H. D., Miller, J. S., & Kay, L. E. 1992, apj, 397, 452Trump, J. R., Impey, C. D., Kelly, B. C., et al. 2011, ApJ, 733, 60Ulvestad, J., Johnston, K., Perley, R., & Fomalont, E. 1981, AJ, 86, 1010Unwin, S. C., Cohen, M. H., Biretta, J. A., et al. 1985, ApJ, 289, 109Urry, C. M. & Padovani, P. 1994, 54, 215Urry, C. M. & Padovani, P. 1995, Publ.Astron.Soc.Pac., 107, 803Urry, C. M., Padovani, P., & Stickel, M. 1991, ApJ, 382, 501Urry, C. M. & Shafer, R. A. 1984, ApJ, 280, 569Urry, M. 2007, in The Nuclear Region, Host galaxy and Environment ofActive Galaxiesvan der Wolk, G., Barthel, P. D., Peletier, R. F., & Pel, J. W. 2010, A&A,511, A64Veilleux, S., Rupke, D. S. N., Kim, D.-C., et al. 2009, ApJS, 182, 628Venturi, T., Castaldini, C., Feretti, L., et al. 1994, in Compact ExtragalacticRadio Sources, ed. J. A. Zensus & K. I. Kellermann, 67Vermeulen, R. C. & Cohen, M. H. 1994, ApJ, 430, 467Vermeulen, R. C., Ogle, P. M., Tran, H. D., et al. 1995, ApJL, 452, L5Veron-Cetty, M. P. & Veron, P. 2000, aapr, 10, 81Verschuur, G. L. & Kellermann, K. I. 1988, Galactic and extra-galactic radioastronomyWall, J. V. 1994, Australian Journal of Physics, 47, 625Wall, J. V. & Jackson, C. A. 1997, MNRAS, 290, L17Wall, J. V. & Jackson, C. A. 1999, astro-ph/9907441123BibliographyWall, J. V., Jackson, C. A., Shaver, P. A., Hook, I. M., & Kellermann, K. I.2005, A&A, 434, 133Wall, J. V. & Peacock, J. A. 1985, MNRAS, 216, 173Wall, J. V., Pearson, T. J., & Longair, M. S. 1980, MNRAS, 193, 683Wardle, J. F. C., Moore, R. L., & Angel, J. R. P. 1984, ApJ, 279, 93Warren, S. J., Hewett, P. C., & Osmer, P. S. 1988, in Astronomical Societyof the Pacific Conference Series, Vol. 2, Optical Surveys for Quasars, ed.P. Osmer, M. M. Phillips, R. Green, & C. Foltz, 96Wayth, R. B., Lenc, E., Bell, M. E., et al. 2015, pasa, 32, e025Weedman, D. W. 1977, Ann Rev Astron Astrophys, 15, 69Wehrle, A. E., Pian, E., Urry, C. M., et al. 1998, ApJ, 497, 178Weiler, K. W. & Johnston, K. J. 1980, MNRAS, 190, 269Wilkes, B. J. & Elvis, M. 1987, ApJ, 323, 243Willott, C. J., Rawlings, S., Blundell, K. M., & Lacy, M. 2000, MNRAS,316, 449Wills, B. J. 1999, 162, 101Wilson, A. S., Braatz, J. A., Heckman, T. M., Krolik, J. H., & Miley, G. K.1993, ApJL, 419, L61Wilson, A. S. & Colbert, E. J. M. 1995, ApJ, 438, 62Wilson, A. S. & Tsvetanov, Z. I. 1994, AJ, 107, 1227Windhorst, R. A. 1984, PhD thesis, Ph. D. thesis, University of Leiden(1984)Wolter, A., Caccianiga, A., della Ceca, R., & Maccacaro, T. 1994, ApJ, 433,29Woltjer, L. 1990, in Active Galactic Nuclei, ed. R. D. Blandford, H. Netzer,L. Woltjer, T. J.-L. Courvoisier, & M. Mayor, 1–55124


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items