Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Quasar detection in the UBC-NASA multi-narrowband survey Braglia, Katia 2002

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

Item Metadata

Download

Media
831-ubc_2002-0345.pdf [ 4.88MB ]
Metadata
JSON: 831-1.0085131.json
JSON-LD: 831-1.0085131-ld.json
RDF/XML (Pretty): 831-1.0085131-rdf.xml
RDF/JSON: 831-1.0085131-rdf.json
Turtle: 831-1.0085131-turtle.txt
N-Triples: 831-1.0085131-rdf-ntriples.txt
Original Record: 831-1.0085131-source.json
Full Text
831-1.0085131-fulltext.txt
Citation
831-1.0085131.ris

Full Text

Quasar detection in the UBC-NASA multi-narrowband survey. by K A T I A B R A G L I A Laurea in Astronomia, Universita ' degli Studi di Bologna, Italy, 1999 A T H E S I S S U B M I T T E D I N P A R T I A L F U L F I L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R O F S C I E N C E in T H E F A C U L T Y O F G R A D U A T E S T U D I E S (Department of Physics and Astronomy) We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A August 11, 2002 © K A T I A B R A G L I A , 2002 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of Br i t i sh Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Physics and Astronomy The University Of Br i t i sh Columbia Vancouver, Canada ABSTRACT i i A B S T R A C T The goal of this thesis is to select quasars by applying a novel analy-sis to the U B C - N A S A Mul t i -Narrow band Survey ( U N M S 1 ) catalog. The database consists of drift-scan observations taken wi th the 3-m N A S A L i q -uid Mir ror Telescope ( L M T ) in 1996-1997 and in 1999, using 35 narrow band filters, from 4500 to 9500 A , and 4 broad bands (B, V, R, I), necessary for calibration purposes. The method presented here is based on the comparison between U N M S 1 catalog sources and stellar, galaxy, and quasar templates through a x2 m m -imizat ion procedure combined with Bayesian analysis. The x2 parameter is useful to determine which model is the best fit for an observed Spectral E n -ergy Distr ibut ion (SED); the odds ratio parameter, from the Bayes' theorem, is necessary in order to know the most likely category the source belongs, and it involves information such as the number of models for each category and the surface densities of stars, galaxies and quasars at a given magnitude. After the method was applied to templates, treated as test-objects, it is concluded that more than 85% of quasar candidates, selected with at least 30 filter measurements, are correctly classified: the misclassification is due mostly to the similarity between stellar models and quasars when important spectral features are missed. O f al l the 39040 selected sources, 3056 quasar candidates were identified: ABSTRACT i i i most of them have typical redshift (z ~ 0.3 - 4), apparent V magnitude (V ~ 16 - 20) and spectral index ( a ~ -2 - 0.8), but there is also a non-negligible number of objects localized well outside these ranges. Analysing the position of these sources in the redshift - V magnitude plane, redshift - spectral in-dex plane and V magnitude - spectral index plane, it is possible to identify misclassified quasars and remove them away from the sample. The final list of quasar candidates includes 2294 objects, among which 1 is in common with Veron's quasar catalog. CONTENTS iv C O N T E N T S A B S T R A C T i i C O N T E N T S iv L I S T O F T A B L E S v i i L I S T O F F I G U R E S ix I THESIS 1 1 I N T R O D U C T I O N 2 1.1 R E C E N T S U R V E Y S A N D M E T H O D S T O F I N D Q U A S A R S . 5 2 L I Q U I D M I R R O R T E L E S C O P E S 13 2.1 L I Q U I D M I R R O R T E L E S C O P E S I N T H E P A S T 13 2.2 T H E N O D O L I Q U I D M I R R O R T E L E S C O P E 15 2.3 D A T A C O L L E C T I O N 17 2.4 L M T A D V A N T A G E S 19 2.5 L M T D I S A D V A N T A G E S 20 2.6 L M T T E C H N I C A L ISSUES 22 2.7 D A T A A N A L Y S I S 23 2.7.1 O B J E C T D E T E C T I O N A N D P H O T O M E T R Y 23 CONTENTS v 2.7.2 A S T R O M E T R I C A N D P H O T O M E T R I C C A L I B R A -T I O N 25 2.7.3 T H E O B J E C T C A T A L O G 26 2.8 C H A R A C T E R I S T I C S O F T H E S U R V E Y 27 2.8.1 P H O T O M E T R I C S E L E C T I O N C R I T E R I A 30 3 Q U A S I S T E L L A R O B J E C T S 33 3.1 Q S O ' S F U N D A M E N T A L C H A R A C T E R I S T I C S 33 3.2 Q S O S T R U C T U R E 38 4 T H E U N M S Q U A S A R S U R V E Y 41 4.1 T H E U N M S 1 C A T A L O G 41 4.2 M E T H O D 44 4.3 T E M P L A T E S 48 4.3.1 S T E L L A R T E M P L A T E S 48 4.3.2 G A L A X Y T E M P L A T E S 50 4.3.3 Q U A S A R T E M P L A T E S 51 4.4 M E T H O D C A L I B R A T I O N F R O M T E M P L A T E S 58 4.4.1 A F I R S T A N A L Y S I S 60 4.4.2 x2 M I N I M I Z A T I O N M E T H O D A N D O D D S R A T I O P R O C E D U R E A P P L I E D T O O R I G I N A L T E M P L A T E S . 66 4.4.3 x2 A N D O D D S R A T I O P R O C E D U R E S A P P L I E D T O A L T E R E D T E M P L A T E S 77 5 Q U A S A R S I N T H E U N M S l C A T A L O G 98 5.1 R E S U L T S 98 5.1.1 S T E L L A R C A N D I D A T E S I N T H E U N M S S U R V E Y . . 99 CONTENTS v i 5.1.2 G A L A X Y C A N D I D A T E S I N T H E U N M S S U R V E Y . . 100 5.2 P R O P E R T I E S O F Q U A S A R C A N D I D A T E S 107 5.2.1 T H E N U M B E R - R E D S H I F T D I S T R I B U T I O N . . . . 107 5.2.2 T H E N U M B E R - A P P A R E N T V M A G N I T U D E DIS-T R I B U T I O N 110 5.2.3 T H E N U M B E R - S P E C T R A L I N D E X a D I S T R I B U -T I O N 112 5.3 D I S C U S S I O N 114 5.4 Q U A S A R C A N D I D A T E S 119 5.5 I D E N T I F I C A T I O N 123 6 S U M M A R Y A N D C O N C L U S I O N S 124 B I B L I O G R A P H Y 127 LIST OF TABLES v i i LIST O F T A B L E S 1.1 Area covered, epoch, magnitude, A A and number of expected or detected quasar candidates for the most recent optical Q S O surveys 6 2.1 Fi l ter specifications, (a) mean wavelength (nm), from trans-mission curve; (b) bandwidth (nm): equivalent width/central transmission; (c) log of central frequency (Hz): c/mean wave-length; (d) log frequency bandwidth: 0.434 x bandwidth/mean wavelength; (e) central transmission; (f) equivalent width (nm): integral of transmission curve 29 3.1 Inventory of emission lines [23] 37 4.1 Parameters in the U N M S 1 catalog 42 4.2 F l u x corrections by Remi Cabanac [priv. comm.] 43 4.3 Surface densities for stars, galaxies and quasars in the broad bands: the units are num/deg 2 for the B and / filters, and num/deg 2 0.5 mag in the R and V bands. The "a" case is for B < 19.5 and the "b" is for B > 19.5 47 4.4 Parameters in Pickles' stellar library 49 4.5 Spectral templates from Kinney & Calzet t i galaxies' library. . 51 4.6 Fi t ted power-law index 54 LIST OF TABLES v i i i 4.7 Power-law index in frequency and wavelength for quasar tem-plates in the U V and optical-near-infrared (Optical-IR) region. 58 4.8 Template sequence for the x2 analysis 69 5.1 Star, galaxy and quasar identifications 98 5.2 Conversion table between stellar type and number 102 5.3 Logari thmic star counts [# / 10 4pc 3] by stellar type and lumi-nosity class from data published in Al l en (1973) [1] 102 5.4 Logari thmic star counts [# / 10 4pc 3] for candidates found in the L M T survey. 103 5.5 Conversion table between galaxy type and number 103 5.6 Fraction of ellipticals (E), SO galaxies, Sa+Sb, Sc+Sd and Sm spirals, Irregular (Im) galaxies, Starburst (SB1-6) and U V h o t galaxies in the C f A l , CfA2 and L M T galaxy samples 105 5.7 Equatorial coordinates ( R A , D E C ) , apparent V magnitude and redshift z for the quasar common to Veron's catalog and U N M S 1 quasar catalog 123 6.1 Distr ibut ion of star, galaxy and quasar candidates 125 LIST OF FIGURES ix LIST O F F I G U R E S 1.1 Evolut ion of telescope aperture over time 4 1.2 Two-colour diagram: quasars (points) are well separated by main sequence stars (line with stellar types BO - MO) (ht tp: / /ned www.ipac.caltech.ca/Cambridge/Cambridge l_3_3.html). . . . 9 2.1 Components for a l iquid mirror 16 2.2 Basic optical layout for the U B C - L a v a l 2.7-m l iquid mirror telescope 18 2.3 Fi l ter transmission curves 31 3.1 General layout for a quasar spectrum 35 3.2 Structure of A G N ' s inner region (ht tp: / / nedwww.ipac.caltech. edu / leve l5 /Urry l /Ur ryPl .h tml ) 39 4.1 Corrections wi th the best fit curve 44 4.2 Kinney-Calzet t i ' s galaxy templates normalized at 4000 A. . . . 52 4.3 Composite spectrum of 101 quasars, binned to 2 A 53 4.4 Composite spectrum: power law fits to the estimated flux are shown 55 4.5 First part of the composite spectrum in the 300-3400 A range: it can be noted that the L y a (1215 A) and C I V (1549 A) are among the strongest emission lines 56 LIST OF FIGURES x 4.6 Second part of the composite "flat" spectrum in the 3400-8550 A range. Ha (6563 A ) and [OIII] (4363 A ) are among the strongest emission lines 57 4.7 Quasar templates in the rest frame: lines I, II, III and I V represent respectively template I, II, III and I V 59 4.8 Galactic (blue line) and quasar (red line) templates in the rest frame with Supergiant star templates (green line) 61 4.9 Galactic (blue line) and quasar (red line) templates in the rest frame with bright giant templates (green line) 62 4.10 Galactic (blue line) and quasar (red line) templates in the rest frame wi th different giant star templates (green line) 63 4.11 Galactic (blue line) and quasar (red line) templates in the rest frame wi th different subgiant star templates (green line). . . . 64 4.12 Galactic (blue line) and quasar (red line) templates in the rest frame wi th different main sequence star templates (green line). 65 4.13 Galactic (blue line) and quasar (red line) templates at redshift z—1.2 with supergiant star templates (green line) 66 4.14 Galactic (blue line) and quasar (red line) templates at redshift z=1.2 with bright giant star templates (green line) 67 4.15 Galactic (blue line) and quasar (red line) templates at redshift z=1.2 with giant star templates (green line) 70 4.16 Galactic (blue line) and quasar (red line) templates at redshift z=1.2 with sub-giant star templates (green line) 71 4.17 Galactic (blue line) and quasar (red line) templates at redshift z=1.2 with main sequence stars (green line) 72 LIST OF FIGURES x i 4.18 x2 distribution for 4 QSO N.14 at z=0.3: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates 73 4.19 x2 distribution for QSO N.15 at z=1.4: line I, II and III come from the comparison respectively to galactic, quasar and stel-lar templates 74 4.20 x2 distribution for QSO N.16 at z=3.5: line I, II and III come from the comparison respectively to galactic, quasar and stel-lar templates 75 4.21 x2 distribution for QSO N.17 at z=5.6: line I, II and III come from the comparison respectively to galactic, quasar and stel-lar templates 76 4.22 x2 distribution for the altered QSO N.14 at z=0.3: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates 78 4.23 x2 distribution for the altered QSO N.15 at z=1.4: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates 79 4.24 x 2 distribution for the altered QSO N.16 at z=3.5: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates 80 4.25 x2 distribution for the altered QSO N.17 at z=5.6: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates 81 LIST OF FIGURES x i i 4.26 Distr ibut ion of stellar (hexagon), galaxian (cross) and quasar (triangle) template-objects as classified according to the odds ratio method 83 4.27 Normalized distribution for correctly classified stars (hexagon) and for objects wrongly classified as stars (cross) 84 4.28 Normalized distribution in the redshift range z=0-0.6 for cor-rectly classified galaxies (hexagon) and for objects wrongly classified as galaxies (cross) 85 4.29 Normalized distribution in the redshift range z=0-0.6 for cor-rectly classified quasars (hexagon) and for objects wrongly classified as quasars (cross) 86 4.30 Normalized distribution in the redshift range z=0-0.6 for cor-rectly classified galaxies (hexagon) and for objects wrongly classified as galaxies (cross) 87 4.31 Normalized distribution in the redshift range z=0.6-1.2 for correctly classified quasars (hexagon) and for objects wrongly classified as quasars (cross) 88 4.32 Normalized distribution in the redshift range z=l.2-4.1 for correctly classified quasars (hexagon) and for objects wrongly classified as quasars (cross) 89 4.33 Normalized distribution in the redshift range z=4.1-7 for cor-rectly classified quasars (hexagon) and for objects wrongly classified as quasars (cross) 90 LIST OF FIGURES x i i i 4.34 Normalized odds ratio distribution for a Sa spiral galaxy (N.4) at redshift 0.2, 0.6, 0.8 and 1.2 in favor of stellar (hexagon), galactic (croos) and quasar (triangle) template 92 4.35 Normalized odds ratio distribution for a starburst galaxy (N.7) at redshift 0.2, 0.6, 0.8 and 1.2. in favour of stellar (hexagon), galactic (cross) and quasar (triangle) template 93 4.36 Normalized odds ratio distribution for Q S O N.14 at redshift 0.3, 2.1, 4.2 and 7.in favour of stellar (hexagon), galactic (tri-angle) and quasar (cross) template 94 4.37 Normalized odds ratio distribution for Q S O N.15 at redshift 0.3, 2.1, 4.2 and 7. in favour of stellar (hexagon), galactic (triangle) and quasar (cross) template 95 4.38 Normalized odds ratio distribution for Q S O N.16 at redshift 0.3, 2.1, 4.2 and 7.in favour of stellar (hexagon), galactic (tri-angle) and quasar (cross) template 96 4.39 Normalized odds ratio distribution for Q S O N.17 at redshift 0.3, 2.1, 4.2 and 7.in favour of stellar (hexagon), galactic (tri-angle) and quasar (cross) template 97 5.1 Star counts by spectral types 99 5.2 Galaxy distribution by the morphological type 104 5.3 Distr ibut ion of galaxy counts with spectral type and redshift. . 106 5.4 Redshift distribution of quasar candidates 108 5.5 The redshift distribution for quasars in the Hewett&Burbidge catalog (top) and in the Sloan Sky Survey catalog (bottom). . 109 5.6 The V magnitude distribution for quasar candidates I l l LIST OF FIGURES x iv 5.7 The V magnitude distribution from the Hewett&Burbidge cat-alog 112 5.8 Spectral index distribution for quasar candidates in the U N M S l catalog 114 5.9 The spectral index distribution by Warren et al . (top picture, upper line) and by Francis et al . (bottom) 115 5.10 Quasar candidates in the redshift - V magnitude plane 117 5.11 Quasar candidates in the redshift - spectral index plane. . . . 118 5.12 Quasar candidates in the spectral index - V magnitude plane. 119 5.13 The number - redshift distribution for the "cleaned" set of quasars 120 5.14 The number - V magnitude distribution for the "cleaned" set of quasars 121 5.15 The number - a distribution for the "cleaned" set of quasars. . 122 P A R T I Thesis CHAPTER 1. INTRODUCTION 2 C H A P T E R 1 I N T R O D U C T I O N Quasars were first discovered in 1963, when several faint radio sources inside the 3rd Cambridge Catalogue of Discrete Radio Sources (3C) were identified as optical objects undistinguishible from stars [13]. Their diame-ters, smaller than 1 arc sec, indicated an unusually high surface brightness in the optical and radio spectral regions. Their positions, away from the galactic equator, suggested an extragalactic nature. The name quasar (quasi stellar radio source) was coined because of their stellar appearance and the fact that their radio emission is comparable to that of Cygnus A . The first discovered quasar was 3C 273 [62], a star-like source identified as the optical counterpart of a radio galaxy whose position was known accu-rately. First considered a stellar object, its real extragalactic nature became clear only after A l a n Sandage saw broad emission lines at positions unusual for stars and M . Schmidt identified a continuum and strong emission features with redshift z=0.158, implying a distance around 950 M p c [73]. In 1965 Sandage and Veron [61], working on the identification of radio sources, realized that there were many galaxies very similar to stars on pho-tographic plates, having a compact structure, a high surface brightness and weak radio emission or no emission at a l l . It was therefore necessary to dis-tinguish between QSRs (Quasi Stellar Radio sources), wi th radio emission, CHAPTER 1. INTRODUCTION 3 and QSOs (Quasi Stellar Objects), wi th little or no emission [13]. It was eventually realized how important these "new" objects were, not only as peculiar sources (as described in §3, they radiate an amount of en-ergy around 10 5 3 - 10 5 5 Joule from regions as small as the Solar System), but also for their role in cosmology and for what could be understood about host galaxies, and the nature and properties of the material between the observer and the source. The small angular size and faintness of QSOs is a challenge for their detec-tion: the faintest objects can be investigated only by the largest telescopes, characterized by a greater light-gathering power and a superior resolution. The light-gathering power of a telescope depends on the objective's area: the bigger it is, the more light is collected in a given time yielding bright images of even distant objects. The telescope's resolving power is the abili ty to reveal fine details: ac-cording to the formula 0 = 1 . 2 2 ^ ( i . i ) where 8 is the angular distance between two bodies, A is the observation wavelength and D is the primary mirror diameter, the resolving power can be improved either by increasing the diameter D or reducing the wavelength A. Since the optical wavelength range of the electromagnetic spectrum spans from 4500 A to 9500 A, progress has come from larger diameters. F i g 1.1 shows the evolution of telescope's aperture over time: the diameter has increased slowly with time due to the difficulty of producing accurate optics (both in terms of casting the primary mirror substrate and of polishing CHAPTER 1. INTRODUCTION 4 1 5 0 0 1 6 0 0 Figure 1.1: Evolut ion of telescope aperture over time. i t) . Sophisticated technology has overcome this problem, but more time has been necessary to improve the ground-based optical telescope resolution, pro-portionate to the diameter of the primary lens or mirror. Images produced by CHAPTER 1. INTRODUCTION 5 large telescopes suffer various problems because of the Earth 's atmosphere. Its turbulent nature causes density changes over small distances, creating regions where the light is refracted in nearly random directions; in this situ-ation the image of a point source is effectively blurred. Correctors to remove optical aberrations and adaptive optics to counter atmospheric turbulence are now functioning on most modern telescopes. Though the l iquid mirror telescope ( L M T ) technology st i l l needs develop-ments, the optics production and other factors (such as the mount, required to support telescopes often heavy and to move them with accuracy) are not an issue and this fact and other advantages, such as the low cost wi th respect to a conventional glass telescopes, make it a really interesting and alternative technological project, described in detail in chapter §2. 1.1 Recent surveys and methods to find quasars. The scientific goal of this thesis is to find quasars using a novel method which is applied to a database of objects observed with the N A S A L M T . Beside a detailed description of the method, it is also useful to compare this survey to those which characterized the last decade or earlier, wi th particular attention to the Sloan, and discuss briefly how they selected quasar candidates. Table 1.1 shows the area, the epoch, the l imi t ing magnitude, the spectral coverage and the number of objects found in the most important optical surveys. The L M T - N O D O survey is explained in detail in §2; nevertheless some CHAPTER 1. INTRODUCTION 6 Survey Ref. Epoch Area magn A A quasars (deg 2) A M B Q S [52] 1978-1981 109 B17.6 3500-7000 32 P B Q S [63] 1981 10714 B16.2 U , B 114 P T Q S [65] 1985-1989 61 R22 4400-7500 232 L B Q S [34] 1986-1989 454 B19 3400-5100 1058 H N Q S [32] 1988-1998 14000 R18.6 3600-6500 376 H B Q S [18] 1989 153 B18.8 U , B , V , R 284 A A T [11] 1989-1990 1.6 B21 3500-6400? 420 E Q S [27] 1989-1997 330 B18.5 U , B , V , R , I 224 A Q S [28] 1990 14019 R15.4 B , V , R 46 H S Q S [31] 1990-1994 10000 B17.5 3200-5400 160 F Q S [12] 1991 0.9 B22 U ^ ^ r / . I 66 A P M 1 [71] 1991-1995 2500 B19 B , R , I 31 SA94 [17] 1992 10 B19.9 U , B , V 200 L M T [39] 1996-1999 20 B21.5 4500-9500 ~ 3000 T Q S [64] 1997 20 B21.3 U , B , V 368 2dF [68] 1997-2002 740 B20.9 U , B j , R 25000 SDSS [79] 1998-... 10225 g23.3 4000-7500 10 5 A P M 2 [72] 1999-2000 5500 B20.8 B j , R , I 31 Table 1.1: Area covered, epoch, magnitude, A A and number of expected or detected quasar candidates for the most recent optical Q S O surveys. CHAPTER 1. INTRODUCTION 7 points can be made here. Compared to these surveys, the l imi t ing magnitude reached by the U B C -N A S A Mul t iband Survey (UNMS1) catalog (21-22 mag in the most sensitive bands) is inferior only to those of the Palomar Transit (22 mag), Faint (22.3 mag) and Sloan Digi ta l (23 mag) surveys. This means these surveys include fainter objects than those observed by the l iquid mirror telescope. Considering the sky area covered, regions observed by the L M T - N O D O survey overlap those studied by the Sloan and the Hamburg Nor th surveys. Therefore, observations taken with the l iquid mirror telescope are matched only by those of the Sloan survey. This fact, which could turn out useful since it can provide an independent confirmation of quasar candidates se-lected wi th the method described in §4, is not surprising: the Sloan Digi ta l Sky Survey (SDSS) [79] is in fact one of the biggest surveys and quasar study (involving mainly quasar clustering effects, Q S O evolution and associ-ation with galaxies ) represents only one of the many fields of investigation (www.ast ro .pr inceton.edu/PBOOK/ science/quasars.htm). Inside the group of 113 QSOs found so far, 27 have redshift 3.5 < z < 4.5 and magnitude 18.55 <m; < 20.97, 4 have redshift z > 4.95; more than 100000 quasars are expected to be detected when the survey is completed. The method adopted to select quasars varies according to the redshift: at low redshifts ( z < 2.2), the lack of a detectable Balmer jump in spectra helps to distinguish between QSOs and stars; at higher redshifts, the presence of the strong emission line Lya and the absorption by the L y forest cause the broadband colours of quasars to become increasingly redder with redshift. The Hamburg sur-vey (HQS) [32] for the Northern sky and the first A P M U K S T colour survey CHAPTER 1. INTRODUCTION 8 ( A P M 1 ) [71] base their selection process on these absorption systems as well. A t 3.5 < z < 5, quasars and stars occupy different zones of the g'r'i'z' colour space (modified Thuan-Gunn system, see transmission curves in F i g . 1 of [29]) but at z > 5 the quasar track approaches the red end of the stellar locus in the r'i'z' diagram and new discriminators are required. The selection method is then based on three regions in SDSS colour space: • r' - V > 1.35 and i' - z' < 0.3; • r' - i' > 2 and V - z' < 0.7; • z-band detection only, i.e. z' < 20.8 and the detection in the other bands is below 5<7. In addition, an object must be classified as a point source by the SDSS pro-cessing software and have z' < 20.8. Surveys listed in Table 1.1 use other methods to select quasar candidates. The A S I A G O - E S O / R A S S Q S O survey (AQS) [28], derived by merging the R A S S (Rosat-All-Sky-Survey) [76], G S C (Guide Star Catalogue) [47], U S N O (United States National Observatory) [54] and DSS (Digitized Sky Survey: http://arch-http.hq.eso.org/dss/dss) catalogs, found quasars accord-ing to their point appearance and a magnitude criterion: 11 < VGSC < 14.5 and 13.5 < RUSNO < 15.4. Common optical selection criteria are based on colours: in the two-colour diagram showed in F i g . 1.2, quasars are always much brighter in the ultravi-olet than main sequence stars with the same B-V colour and the threshold JJ-B=-0.4 is used to separate these two categories. CHAPTER 1. INTRODUCTION 9 Figure 1.2: Two-colour diagram: quasars (points) are well separated by main sequence stars (line wi th stellar types BO - MO) (http:/ /ned www.ipac.caltech.ca/Cambridge/Cambridge l_3_3.html). This method, used in the Palomar Bright Q S O survey ( P B S Q ) [63], the Med ium Bright Quasar survey ( M B Q S ) [52], the Homogeneous bright Q S O survey (HBQS) [18], the S A 94 Q S O survey (SA94) [17] and the Anglo-Australian-Telescope survey ( D U R H A M / A A T ) [11], must be used carefully as that quasar region inside the two-colour diagram is also well populated by white dwarfs that, at B magnitude smaller than 16, are much more numerous than quasars. Other surveys rely on the red excess: the spectroscopic survey of faint QSOs (FQS) [12] and the second A P M U K S T colour survey ( A P M 2 ) [72] CHAPTER 1. INTRODUCTION 10 (for which only sources with Bj — R > 2.5 are considered quasar candidates), are an example. Techniques with more colours can be more efficient: besides U-B and B- V colours, U-J and J-F (where J is at 12500 A and F is at 6250 A) are used. Quasars wi th redshift higher than 2.2 have a colour J-F bluer than stars wi th the same U-J colour. This method is adequate to find objects no more distant than z ~ 3.2. Mult icolour techniques with broader bands, already mentioned for the SDSS, are efficient to identify more distant quasars: this is the way in which the 2dF Q S O Redshift survey (2dF) [68] (http://www.2dfquasar.org) oper-ated as well. Other research methods are based on broad lines: the spectrum of a quasar, wi th strong emission features, is very different, at low resolution, from that of red stars and galaxies. A Gr i sm (grating-lens system), charac-terized by a prism objective associated to a diffraction grating, allow simul-taneous measurements of low resolution spectra for many objects. This led to the discovery of quasars at redshift higher than 4 [22]. This is the case of the Large Bright Q S O survey ( L B Q S ) [34] and the Palomar Transit Gr i sm survey ( P T G S ) [65]. Other possible methods to identify quasars could rely on their variabili ty or on the absence of proper motions. These methods require long exposures on different time scales to separate possible candidates: the Tautenburg sur-vey (TQS) [64] worked in fact on Schmidt plates taken over the last three decades. Recently, the Pr incipal Component Analysis approach ( P C A ) has resulted CHAPTER 1. INTRODUCTION 11 in successful classification of objects such as stars, galaxies, QSOs etc. from multi-band photometry, using mock catalogs which match as closely as pos-sible the observations of the Large Zenith Telescope ( L Z T ) , mentioned in §2 [15]. The P C A is a non-parametric approach already employed with mult i -colour photometry (generally fewer than 10 colour bins) or wi th medium to high-resolution spectroscopy, but it was not tested on spectral energy distri-bution (SED) with R ~ 40. The P C A methodology can be summarized in this way: from a set of N vectors "f" wi th M elements (where M is the number of filters) and each of them normalized to have unit scalar product, the P C A derives a set of orthogonal eigenvectors e\j , using criteria of decreasing maximun variance of the spectra when projected onto the eigenvectors. Then, vectors fxi are the linear combinations of eigenvectors e\j (where e\ is the mean vector over f\i, e 2 lies in the direction of the highest variance orthogonal to d , etc.) and eigencomponents which are the weight of the "jth" eigenvector in the "«**" vector. The main advantage of this approach is that when the vectors f\i are correlated (as it is for astronomical SEDs) , most of the classification power of the linear combination mentioned above is carried by the first 10 eigen-vectors (ei, ... eio) when describing a set of objects of different classes and at different redshift intervals. In order to use the P C A for spectral classification, one must relate the internal correlations outlined by the P C A with the physical properties of the objects: it is necessary then to apply the P C A to mock catalogs of templates and extract important information for classying objects and measuring their CHAPTER 1. INTRODUCTION 12 redshifts. This was the approach chosen by [15] : once the principal eigen-components of the catalog are measured, the first 10 eigenvectors are used to defined a 10-dimensional eigenbasis, in which each class of objects occupies a given locus, defined by its spectral type among its class, and its redshift for galaxies and quasars. For a median signal-to-noise ratio of 6, 98% of stars, 100% of galaxies and 93% of quasars are correctly classified. This technique, tested only on simulated observations so far, is going to be calibrated on the same data studied in this thesis. CHAPTER 2. LIQUID MIRROR TELESCOPES. 13 C H A P T E R 2 LIQUID M I R R O R T E L E S C O P E S . 2.1 Liquid mirror telescopes in the past. The idea of a l iquid mirror telescope dates back to 1850, when a scientist at the Italian Observatory of Naples, Ernesto Capocci , thought to use for astronomical observations a well known physical concept according to which the equilibrium configuration of a l iquid surface, which rotates uniformly, assumes just a paraboloid shape with a focal length / where co is the rotation angular frequency, g is the gravitational constant. Capocci realized that a highly-dense l iquid, such as mercury, might be used as a reflecting surface. The New Zealander Henry Skey, from the observatory of Dunedin, was the first to bui ld such an instrument in 1872. Unfortunately, the fact it couldn't point and track objects, and complications related to a l iquid such as mercury, prevented this k ind of telescope from being useful for astronomical research unti l a century later. Here is a brief list of the most important l iquid mirror telescopes built in the past or s t i l l in the development phase [25]: (2.1) CHAPTER 2. LIQUID MIRROR TELESCOPES. 14 • in 1992, E . Borra created the first rotating mercury mirror, 1.5 m in diameter, that was able to produce diffraction l imited images: it was followed by a second one with a diameter of 2.5 m; • in March 1994, a 2.7-m telescope designed and built by P. Hickson was used for astronomical observations. Its resolution was l imited only by the atmospheric seeing. It was equipped with a C C D operating in T D I mode [36], [37], [38]; • in the Spring of 1996, the 3-m Nasa Orbi ta l Debris Observatory ( N O D O ) built by M . Mulrooney and P. Hickson began operation; • in 1997, a 3.7-m l iquid mirror was completed by Borra [9]: its optical testing is on its way; • the 6-m Large Zenith Telescope ( L Z T ) was recently buil t by P. Hickson near Vancouver, Br i t i sh Columbia: it wi l l be operational before the end of 2002. One of the most ambitious projects for future L M T investigations is L A M A , the Large-Aperture Mir ror Array: it is an optical interferometry project based on an array of 18 fixed 10-meter liquid-mirror telescopes, covering a circle of 60-m diameter and wi th an effective aperture ~ 42m. Likely located either at an altitude of 5000 meters in Cerro Chajnantor, in Northern Chile or at an altitude of 2800 meters in the Sacramento Mountains of New Mexico, this instrumentation wi l l produce distortion-free diffraction-limited images over a 1 arcmin instantaneous field of view, with pointing and tracking as far as 4° from the zenith (h t tp : / /www.as t ro .ubc .ca /LMT/lama/ index .h tml) . CHAPTER 2. LIQUID MIRROR TELESCOPES. 15 2.2 The NODO liquid mirror telescope. The data analyzed and discussed in this thesis were obtained from the N O D O liquid mirror telescope in the state of New Mexico, in the Sacramento mountains, at 2756 meters of altitude. This 3-m mirror with a 4.5-m focal length was built by P.Hickson and M . Mulrooney for N A S A . The principal objective was to characterize orbital debris, wi th a second goal of astronom-ical research [57]. The most important component of the entire structure is the mirror sup-port: a styrofoam core enclosed by Kevlar . This material was chosen because of its light weight, high stiffness, and good dampening characteristics. The Kevlar surface is covered by a layer of polyurethane, created by spincasting. The final surface, which is parabolic to wi thin a fraction of a millimeter, is then covered by 1.5 m m of spinning mercury. A thin oxide layer forms over the surface after the mercury is exposed for a few hours to air and it has the advantages of dampening surface waves and eliminating the evaporation. The reflectivity of M g is in the range 70 - 80%. The structure is able to support 170 kg of mercury without vibration and wi th a maximum deviation from a paraboloid around 0.1 mm. The deviation of the mercury surface from a perfect parabola is approximately A/20, where A is the wavelength. To achieve this performance, there are two fundamental requirements: a stable rotational velocity and the rotation axis alignment parallel to the gravitational field of the Ear th . The first requirement is obtained using a synchronous motor which drives the mirror support: the rotation period is approximately 6 seconds. The second requirement is achieved by employing a precise air bearing and a three-point mount with which the rotation axis CHAPTER 2. LIQUID MIRROR TELESCOPES. 16 can be aligned to within 0.25 arcseconds [9]. The basic mirror setup is shown in Fig. 2.1. Container A i r compresser + Drier Figure 2.1: Components for a liquid mirror. The telescope structure consists of a steel tripod 5.8 meters high which supports correcting lenses, filters, an alignment system, focus mechanism and a C C D camera. The corrector contains 4 elements which reduce the optical aberrations (especially field distortions) to less than 1 arcsec over a 0.5° field at infrared CHAPTER 2. LIQUID MIRROR TELESCOPES. 17 and optical wavelengths. The N O D O L M T ' s basic layout is similar to that shown F i g 2.2 for the U B C - L a v a l 2.7-m l iquid mirror telescope: main differences are in the absence of the brace at half the height and of the drive belt. The drive system includes a motor stator (which is mounted directly on the air-bearing base), a rotor (which is attached to the rotating spindle) and an optical encoder (which senses the angular velocity). 2.3 Data collection. A front i l luminated 2048x2048 pixel Lora l C C D (for 1996-1997 observa-tions) and a back il luminated 1024x1024 S I T E C C D (for 1999 observations) are the detectors used for this survey. The first chip, for which the light enters through thin polysilicon gates of the parallel register, is transparent to long wavelength and opaque for wavelengths shorter than 4000 A. The second device, for which the light enters through the thinned backside of the C C D register where there is no gate structure, exhibits high sensitivity to light up to the near-infrared regions of the spectrum due to its thicknesss reduced to 10 fxm. Bo th devices don't show a good response in the blue. Each pixel has a size of 15 pm. wi th an image scale of 0.6 arcsec/pixel for the first detector, and a size of 24 fj,m wi th an image scale of 0.96 arcsec/pixel for the second detector, in right ascension and declination. The C C D is situated inside a Dewar and is cooled thermically to a temperature of -30°C, reducing the dark current level. Dur ing observations, the C C D is scanned continuously in the time-delay CHAPTER 2. LIQUID MIRROR TELESCOPES. 18 air bearing stand Figure 2.2: Basic optical layout for the U B C - L a v a l 2.7-m l iquid mirror tele-scope. CHAPTER 2. LIQUID MIRROR TELESCOPES. 19 integrate (TDI) mode [51]. In this operation, the charges accumulated by the C C D are moved across it at the siderial rate. The effective integration time is around 97 seconds for the 2048x2048 C C D and 78 seconds for the 1024x 1024 C C D . This represents an object's cross in time along the chip due to our planet's rotation. Data are accumulated continuously at a rate of ~ 50 Kb/sec for a total of 1.8 G b each night. Software written by P. Hickson supplies interactive control of data acquisition parameters and a continuous display of image data; zoom and contrast controls on the display allow one to monitor the focus and to check image quality during observations. The Lora l C C D is connected to a photometric controller which, being de-signed for high-speed operation (necessary for satellite debris observations), has a relatively-high read noise (28e~); for the S I T E C C D the read noise is of l i e - . This fact represents a l imit for the detector, particularly at short wavelengths where the sky is not so luminous and the C C D does not have a good response [39]. This fact could influence the study of quasars in this thesis since the selection process is based on the comparison of the observed source and templates: if features in the blue are missed or not very accurate, it is likely to misclassify the source. 2.4 L M T advantages. For clear reasons of construction, the L M T is a non-steerable instru-ment and it observes objects inside a narrow strip of sky passing through the zenith. Here we can find the first advantage: only zenith observations minimize the atmospheric extinction and refraction, allowing one to observe CHAPTER 2. LIQUID MIRROR TELESCOPES. 20 wi th the best seeing. The increasing interest in L M T s is justified by numerous practical ad-vantages wi th respect to conventional telescopes: a l l the challenges related to the mounting, which supports most of the telescope weight, and to the trai l ing of celestial bodies, are here eliminated. The simple general design of the entire instrument simplifies maintenance. Dust can be removed from the mirror surface in less than a hour loosing a negligible amount of mercury: this allows one to use the telescope contin-uously with the opportunity of observing the variabili ty of certain sources night by night. L M T s are well-suited for many different kinds of surveys, from large-scale structure to galactic evolution, galaxy luminosity function evolution, Q S O observations, supernovae rates, gravitational lenses and cosmology. A l l these research topics would be difficult to study continuously wi th conventional telescopes. Most of the success of this type of instrument is due also to its low cost: an order of magnitude lower than a conventional telescope wi th the same aperture. 2.5 L M T disadvantages. Since the l iquid mirror rotates around a vertical axis, the field is l imited to ~ 0.5° square near the zenith and, thanks to the terrestrial rotation, the survey studies a narrow strip of sky, whose extent depends on the integration time. CHAPTER 2. LIQUID MIRROR TELESCOPES. 21 The mechanical disadvantages, such as the lack of tracking and directed pointing, might be overcome in the near future using correctors with movable optics [42], correctors with movable mirrors [8], introducing magnetic parti-cles into the l iquid and deforming its configuration through magnetic fields or using a viscous liquids, covered by a very thin reflecting film so that the instrument can point a few tens of degrees off the zenith without altering its parabolic shape [10]. Star-trail curvature, which affects observations away from the celestial equator, can be distinguished from other problems since these trails are lightly concave, to North direction, about an amount which depends on the size of the C C D and is practically constant over the entire field. A t N O D O latitude, the peak-to-edge curvature is around 3.3 arcsec. This problem could be avoided by using a corrector having ti l ted and decentreted lenses [40]. For this telescope, the curvature produced by the rotation of the field is smaller than the average seeing, around 2 arcsec. The field distorsion, experienced by the N O D O telescope wi th its original 3-elements corrector (later replaced by a low distortion 4-element design), is a more serious issue: it causes an increasing degree of image smearing from the center of the field to the extremites Nor th and South (up to 2 degrees). Just for this reason the survey is l imited to 75% of the field, starting from the center [39]. Different factors must be accounted for the image degradation: seeing, instrumental aberrations, and the T D I observation mode. Aberrations of a parabolic mirror depend.on the field angle: using the corrector, they are smaller than 1 arcsec, negligible with respect to the see-CHAPTER 2. LIQUID MIRROR TELESCOPES. 22 ing. The final contribution to the image aberration comes from T D I : this observation mode is necessary since the instrument does only zenith obser-vations and then it is fundamental that images of al l the sources drift at the same rate on parallel linear tracks, otherwise there wi l l be an image spread. This requirement is challenged by the star-trail curvature mentioned above and the siderial rate varying with declination: there appear errors dependent on the latitude and increasing proportionally to the square of the angular field of view. Since the magnitude of the effect is typically 1-2 arcsec, it is important to use 4-optical-element correctors at the prime focus [40]. 2.6 L M T technical issues. A s already mentioned, the layout of this telescope is very simple, but this doesn't mean that the telescope is free from technical difficulties, some of them already solved and other in search of a solution. Besides the technical problem to give a uniform rotation to the mirror, today solved with a modern drive system, a real difficulty is the formation of surface waves, which degrades the image quality. Such waves can result from two causes: vibration transmission from the floor to the mirror; inaccuracy in leveling the mercury (avoided wi th a better alignment between the mirror rotation axis and the gravitational field) and an unstable rotation speed. Some optical tests carried out at Laval University on the quality of a mercury surface [3] showed that the total amplitude of ripples is between A/10 and A/15 at 5000 A. CHAPTER 2. LIQUID MIRROR TELESCOPES. 23 Vibrat ions can be caused by wind as well: the mercury surface doesn't tolerate turbulent local air velocities, stronger than 12 m/s . The telescope enclosure must therefore provide wind protection. 2.7 Data analysis. 2.7.1 Object detection and photometry. Before data analysis, it is necessary to subtract the dark current and the sky, and to correct for response variations (flat field). The dark current is corrected using data accumulated in T D I mode wi th the C C D covered; the flat field is obtained from observations on nights wi th moon-brightened cloudy skies. Cosmic rays and stellar images are removed by median filtering. Once these corrections are done, consecutive C C D rows are put together in blocks of 2048 overlapping lines: the sky subtraction is then obtained by subtracting the mode (the most common value inside the distribution) of each line and row. After this procedure, since the good uniformity of C C D images, systematic background variations are lower than 0.02%. The object detection and the photometry are done on individual data blocks: in order to diminish the noise, the object identification is conducted on a smoothed image copy, obtained using a 5 x 5 pixel boxcar filter which provides a five-fold reduction in background noise. In each block, the back-ground mean standard deviation is determined eliminating iteratively the stellar images row by row: then, al l the pixels with a value exceeding the mean by 2.5 standard deviation are marked, areas of contiguous pixels are CHAPTER 2. LIQUID MIRROR TELESCOPES. 24 identified and the outline of each of them is considered the object isophote. For sources inside the image, the first three moments about the intensity distribution are determined within the isophote considered: - the zero-order moment, the isophotal flux; - the first-order moment, from which the image coordinates can be obtained; - the second-order moment, from which the inertia tensor is determined and whose eigenvalues give the major axis, the minor axis and the position angle of the intensity distribution. A l l the measurements are done on the un-smoothed image and inside the area of the object isophote determined from the smoothed image. Since raw isophotal magnitudes are not realistic for d im objects (an in-creasing fraction of light falls outside the isophote and adding flux from outside the same isophote increases the measurement noise), a correction is applied to these magnitudes depending on the flux and on the mean intensity inside the isophote registered. This correction was calculated assuming the relationship between the intensity and the flux is the same as for a Gaussian intensity profile, which is a good approximation to seeing-altered profiles of the faintest galaxies in the image. In order to identify and separate objects whose images overlap, the de-tection algorithm is repeated with increasing detection thresholds of surface brightness: at each step, the area inside the isophotal outline of an object previously detected is examined. The isophote level is increased unt i l either the object no longer reaches the threshold, or it breaks into two parts. In the latter case, the total flux and the isophotal flux of the " ini t ia l" source is divided between the two parts, proportionally to their isophotal fluxes. CHAPTER 2. LIQUID MIRROR TELESCOPES. 25 For data obtained with this telescope, the fraction of objects which suffer blending is around 1%. The photometry program produces a list of al l the sources recorded in one night of observation: instrumental magnitudes, positions, errors, image parameters and the full-width at half-maximum intensity ( F W H M ) . Instrumental coordinates were obtained by applying corrections for aber-ration and nutation of the coordinates of the image centroids, and then re-ferring these value to the standard epoch (J2000). 2.7.2 Astrometric and photometric calibration. For the astrometric calibration process the U S N O A l stellar catalog, pro-duced by the U S Naval Observatory for the Precise Measurement Machine Project [55], was used. The first stage is to find a number of astrometric reference stars from the astrometry file and then to fit the astrometric data using the nearest calibration star. Astrometric fitting equations in right ascension and declination are based on 7 parameters including offsets and linear and quadratic scale factors for both the coordinates. The resulting coordinates have typical errors around 0.3 arcsec, both in right ascension and declination. The photometric calibration was conducted using spectro-photometric stars inside the survey field. These were calibrated by observations con-ducted at the Kitt Peak National Observatory: 22 stars were chosen to cross the telescope field every 30 minutes, producing a calibration accurate to wi thin ~ 5 % at every wavelength [41]. For each band, the product between the filter transmission curve and the CHAPTER 2. LIQUID MIRROR TELESCOPES. 26 specific flux for the standard star is integrated to give the magnitude zero point in that band. To account for possible sky transparency variations during the observa-tion, a second order polynomial is fit to the magnitude zero points from the standard stars and applied to the instrumental magnitudes to give the cali-brated values for al l the objects. A B magnitudes are used, defined in this way: m„ = -56.10 - 2.5log{fv) (2.2) where f„ is the average specific flux for the filter passband in W m~2 Hz'1 [39]. 2.7.3 The object catalog. In order to obtain the final catalogue of objects, it is necessary to merge the photometry files of al l the nights for which the same filter was used. They were identified as objects al l those detected on more than one night (in order to reject cosmic rays and spurious detection of noise), having a difference of magnitude smaller than 1 and the same position to wi thin an error smaller than 3.5 arcsec. For each night the magnitude error is estimated from the object flux, the isophotal area and the background variance. The average magnitude is de-termined weighing the magnitude for each night by the reciprocal variances. The same weights are then used when the average values for other pho-tometric and astrometric parameters are calculated. The final photometric errors are calculated easily: for each source, the CHAPTER 2. LIQUID MIRROR TELESCOPES. 27 variance of the mean is found from the magnitude variances and this allowed the estimation of the random noise. The total error, the sum of random and systematic one, is estimated by calculating directly the variance in the magnitude obtained on different nights. The highest of this estimation is kept for final record. The entire procedure gives a single photometric file for each band and then, in order to obtain the S E D (Spectral Energy Distr ibution), the final files correspondent to al l the bands must be grouped. 2.8 Characteristics of the survey. The survey was conducted in different narrow bands covering 20.13° square of sky, centered at the declination of +33° and at right ascension from lOhr to 18hr. The observation in a band is done by inserting a glass filter between the corrector lenses and the C C D window. A filter is positioned at the beginning of the observation and used for al l the night. This survey is based on observations taken in 1996, 1997 and 1999 using 35 narrow band-filters covering 454 - 993 nm, and 4 broad band-filters: 5(440 nm), 7(550 nm), #(700 nm), 7(900 nm). These filters, being broader than the narrow ones, reach a fainter l imi t ing magnitude and they make easier the comparison with other broad-band photometries. Table 2.1 shows the characteristics for narrow filters. F i g 2.3 displays their transmission curves. The fact blue filters have lower transmission curves (a lower percentage of photons is received at wave-CHAPTER 2. LIQUID MIRROR TELESCOPES. 28 ID \a A o A A6 log{v)c Alog(u)d te la wf 948 947.7 39.08 14.5003 0.019 0.933 36.43 925 924.5 40.04 14.5111 0.019 0.928 36.96 906 906.3 35.32 14.5198 0.018 0.900 31.71 883 883.1 41.28 14.5311 0.021 0.924 38.10 868 867.9 35.10 14.5388 0.018 0.952 33.38 844 843.8 35.58 14.5509 0.019 0.932 33.09 825 824.8 33.67 14.5608 0.018 0.950 31.96 806 805.9 34.62 14.5709 0.019 0.936 32.27 788 787.5 33.31 14.5809 0.019 0.927 30.85 770 769.6 31.86 14.5910 0.018 0.937 29.79 752 752.4 33.25 14.6008 0.019 0.955 31.72 735 734.7 32.17 14.6111 0.019 0.940 30.17 719 718.7 30.54 14.6208 0.019 0.954 29.13 704 704.4 29.88 14.6293 0.019 0.930 27.78 688 688.0 29.20 14.6397 0.019 0.936 27.30 671 671.3 29.08 14.6503 0.019 0.933 27.10 655 654.6 27.99 14.6612 0.019 0.930 26.03 641 641.1 23.98 14.6705 0.016 0.919 21.99 629 628.7 26.39 14.6789 0.018 0.952 25.10 614 613.7 23.62 14.6893 0.018 0.910 21.45 598 597.6 24.31 14.7010 0.018 0.717 17.41 586 585.6 23.10 14.7099 0.018 0.720 16.63 571 571.1 21.71 14.7207 0.017 0.750 16.26 Table 2.1: Continue. CHAPTER 2. LIQUID MIRROR TELESCOPES. 29 ID \a Ao A A 6 log(v)c Alog(v)d te lo wf 557 557.0 21.35 14.7314 0.017 0.707 15.05 545 545.1 21.00 14.7409 0.017 0.726 15.22 533 532.7 22.76 14.7505 0.019 0.730 16.59 519 519.0 22.72 14.7609 0.022 0.679 15.38 510 510.2 22.36 14.7698 0.019 0.689 15.39 498 498.1 21.91 14.7798 0.019 0.670 14.66 486 486.0 20.22 14.7904 0.019 0.752 15.18 476 475.6 19.30 14.7998 0.018 0.690 13.31 466 465.9 18.48 14.8090 0.018 0.673 12.42 455 454.5 17.67 14.8196 0.018 0.632 11.17 Table 2.1: Fi l ter specifications, (a) mean wavelength (nm), from transmis-sion curve; (b) bandwidth (nm): equivalent width/central trans-mission; (c) log of central frequency (Hz): c/mean wavelength; (d) log frequency bandwidth: 0.434 x bandwidth/mean wavelength; (e) central transmission; (f) equivalent width (nm): integral of transmission curve. lenghts shorter than 600 nm) and the exposure time is the same for each filter, influence signal-to-noise ratios in such a way that, for blue bands, the signal-to-noise ratios result lower than those of red filters. Each filter was used on more than one night in order to recognize cosmic rays, to create an independent estimate of the photometric accuracy and to improve the object signal-to-noise ratio. The sequence of the filter observa-tions depended on lunar phase. In order to minimize the effect of the moon CHAPTER 2. LIQUID MIRROR TELESCOPES. 30 on the sky brightness, the red and infrared bands were used when the moon was bright and the blue filter were used when the moon was new. 2.8.1 Photometric selection criteria. The recorded objects are l imited by the surface brightness, by the ap-parent magnitude and by the angular separation. In order to distinguish an object, the surface brightness must exceed the detection threshold for a minimum number of connected pixels corresponding to a min imum area of Am = 1.788 arcsec2. Then, i f / a n d i are respectively the flux and the median intensity inside the isophote, the first selection cr i-terion is f/i > Am. (2.3) To record the object ~i > im (2.4) where im is the detection intensity threshold: to maximize the number of de-tected objects, im is set to the lowest possible level allowed by the background noise, which is generally dominated by the sky light and varies according to the lunar phase. The formula 2.4 represents the second criterion. The thi rd criterion requires a minimum signal-to-noise ratio £. The noise is due to both the image's Poissonian noise and the background noise, then / > C2(g + °2l~i) (2.5) CHAPTER 2. LIQUID MIRROR TELESCOPES. 3 1 j i i I i i i l _ i i i l i i i l i i i 8 0 9 0 K 0 Z d uo L s s i uusue J + CHAPTER 2. LIQUID MIRROR TELESCOPES. 3 2 where a is the background noise variation and g is the system gain (signal produced by a single photo-electron). The l imi t ing magnitude for each block of survey data can be estimated from the surface brightness, by the minimum area required to find the object and by the seeing F W H M . This function can change on short time scales since it depends on atmospheric conditions. CHAPTER 3. QUASI STELLAR OBJECTS. 33 C H A P T E R 3 QUASI S T E L L A R OBJECTS. 3.1 QSO's fundamental characteristics. Schmidt described the optical properties of quasars in 1964 in the following way [13]: - starlike objects identified with radio sources; - variable light; - large UV lines in spectra; - broad emission lines in spectra; - large redshift of the spectrum. Since then, many surveys have been completed at different frequencies and much has been learnt about quasars. Now we know that even though quasars were first found as radio sources, only 10% of those optically selected are radio loud, with a power higher than 1026 W Hz'1. Their radio structure is not so different from that of giant ellipticals. Their X emission is very strong as well, Lx ~ 1021 —1023 W Hz'1 [22]. Since the liquid mirror telescope works in the optical range, it is useful to describe quasar optical properties in detail. In general the optical luminosity is extremely high: typical values are around Lopt ~ 1022 — 1024 watt Hz'1 which corresponds to an absolute CHAPTER 3. QUASI STELLAR OBJECTS. 34 magnitude M o p < ~ -11 - -28 (to make a comparison, the absolute magnitude of giant ellipticals is -22 - -23). The optical variability, from weeks to a few months, is of few tenths of magnitude. From that, it is possible to calculate the linear scale implied by the duration of this variation through this formula: where r 0 and r are respectively the time variation in the quasar's and in the observer's frame. It is found that al l the energy comes from a region smaller than 1016 — 1017 cm in diameter; this situation can be explained only by means of relativistic effects. The general trend for a quasar spectrum beyond the optical range is shown in F i g . 3.1. It is characterized by a continuum plus strong absorption and emission lines. The continuum is well approximated by a power law where / „ is the monochromatic flux at frequency v and a is the spectral index: different values of this parameter were found from surveys, but the most common range is [-2, 0.8] as we describe in §4. In the optical region the continuum tends to flatten and it shows a more or less pronounced " U V Bump" or " B i g Blue Bump" , a real deviation from the power law. It has been proposed by Shields [69], that the blue bump might be due to thermal emission from the surface of an optically thick but geometrically thin accre-tion disc which surrounds the central source. The emitted spectrum is then the superposition of black bodies of temperatures decreasing from the inner D < c r 0 = c r / ( l + z) (3-1) fv<xua (3.2) CHAPTER 3. QUASI STELLAR OBJECTS. 35 WAVELENGTH (MICRONS) 10.0 3.0 1.0 0.3 J : LA i L _ 13.5 14.0 14.5 15.0 LOG F R E Q U E N C Y (Hz) Figure 3.1: General layout for a quasar spectrum. radius to the outer radius of the disc. Many features of this emission could be explained by accretion disc models which are based on possible matter accretion wi th some angular momentum, provided that an adequate source of viscosity is available to transport this angular momentum towards the outer regions of the nucleus [69]. Table 3.1 lists the most important quasar emission lines in the U V and CHAPTER 3. QUASI STELLAR OBJECTS. 36 optical range. These lines are so important to identify quasars that i f a few of them are missed, since the method is based on the comparison between the observed S E D and models, it is likely to misclassify objects. Many lines which fall in the U V range are detectable from optical instru-ments only at high redshift, according to the relation Kbs = A e m ( l + z) (3.3) where A 0 ; , s is the observation wavelength, A e m is the rest frame wavelength and z is the redshift. Besides numerous and strong emission lines, absorption features are also observed. Since they are caused by material between the observer and the luminous source along the line of sight, the absorbing material has a redshift smaller the the quasar. For the same object, three categories of absorption lines can be found: - metallic systems: groups of two or three lines at wavelengths very close to each other and wi th the same redshift. The most common are CIVA1548/1551 A and M g IIA 2796/2803 A but also C , N , 0 , S i , S, A l and Zn systems can be present. - the La forest: at emission wavelengths lower than 1260 A there are many weak and narrow absorption lines which have the dramatic effect of strongly reducing the quasar emission intensity. In this case, each line has its own redshift. - broad absorption lines: very large features which begin from the blue part of strong emission lines and extend for many thousands of k m sec~l. CHAPTER 3. QUASI STELLAR OBJECTS. 37 line A Ly/3 1025.7 0 V I 1035 L y a 1215.7 N V 1241.5 0 I + [S II] 1305 Si I V + 0 I V 1400.0 C I V 1549 He II 1640 A l III 1859 C III 1909 M g II 2800 Ne V 3426 [0 II] 3728 Ne III 3869 U6 4101.7 H 7 4340.5 [0 III] 4363 He II 4686.5 H/? 4861 He I 5875.6 H a 6563 Table 3.1: Inventory of emission lines [23]. CHAPTER 3. QUASI STELLAR OBJECTS. 38 Their connection to strong emission lines is perhaps due to fact they originate close to the quasar. 3.2 QSO structure. Quasars are the most powerful sources among A G N s , a heterogeneous class of objects characterized by a great manifestation of energy which can-not be explained by ordinary stellar processes. There is a growing convergence of opinions that A G N s are in reality the same type of object: it might be possible to develop a model common to most A G N s which could account for factors such as variability, evolution and orientation effects. The variabili ty is a significant problem: since many objects show a vari-able emission over few years, it is easy to put them into the wrong category. Fairal l 9, a quasar wi th a S y l spectrum, became a Sy2 galaxy decreasing its magnitude by a factor of 3 in 5 years, or a B L Lac, PKS0521-36, became a S y l galaxy in less than 6 years. The evolution factor could be important i f the variabili ty time was very long and systematic: for example, since quasars radio loud show an opti-c a l / U V / X luminosity which decreases slowly and systematically to the point to be negligible, we could think of this transformation from quasars to radio-galaxies as a sort of evolution. Recently, attention has focussed on models that explain A G N s ' obser-vational characteristics as due to different orientations wi th which objects are seen from Ear th , with different accretion rates and masses of the cen-CHAPTER 3. QUASI STELLAR OBJECTS. 39 tral black hole and with something which prevents the radiation from being isotropic. According to the most recent model, a funnel-shaped thin shell outflow creates all these features (fig. 3.3). Figure 3.2: Structure of AGN's inner region (http:// nedwww.ipac.caltech. edu/level5/Urryl/UrryPl.html). In more detail, a torus of opaque material such as dust, with a radius of few parsecs, surrounds a nucleus which is likely a 101 7 cm black hole, from which a thermal and not-thermal continuum is emitted. Inside the torus there is a Broad Line Region (BLR), dense and highly mobile clouds, and outside the torus there is a Narrow Line Region (NLR), less dense and slow clouds. Sometimes a relativistic jet can emerge from the nucleus and relativistic effects will be seen if the line of sight is close to the CHAPTER 3. QUASI STELLAR OBJECTS. 40 jet direction. Looking at directly into the jet, we should see O V V (Optical Violent Variable, characterized by an exceptional optical variability) and B L Lac objects (with agreat activity and variability from the radio to the X-par t of the spectrum). If the B L R material is absent or the boosting doppler in the optical continuum is very strong, the B L R s shouldn't be visible. A t a larger angle, radio quasars should appear. When the central source is obscured but the B L R s are partially covered, broad line radio galaxies ( B L R G ) could be observed. Increasing the angle, with the central source and the B L R s now obscured, only the narrow lines are visible: this is a N L R G , narrow line radio galaxy. When the radio jet is very weak or absent, it is likely to have a radio quiet quasar or Seyfert 1 (mainly associated to spiral galaxies having strong emission continuum with broad and forbidden narrow emission lines) when the central source is not covered, or a Seyfert 2 (with broad and narrow emission lines and a non-negligible X-ray emission) when the the nucleus is obscured. Even without a collimated jet, one can expect a strong wind from the nucleus, able to blow away part of the torus or a B L R cloud: in this case a Broad Absorption Line quasar ( B A L ) could be the end result. CHAPTER 4. THE UNMS QUASAR SURVEY. 41 C H A P T E R 4 T H E U N M S Q U A S A R S U R V E Y . 4.1 The U N M S 1 catalog. The U N M S 1 catalog [Hickson, P., priv. comm.] includes more than 2000000 objects: Table 4.1 displays parameters provided for each object, be-sides their spectral energy distribution (SED) in as many as 35 narrow bands and 4 broad bands. Parameters "nband" and "pa" represent respectively the number of bands in which an object was observed and the angle between the image semi-major axis and the line North-South direction, measured from East to West starting from North . Details about filters were given in §2. As explained in §4.4.3, only objects with reliable flux measurements in 30 or more filters were used in this study: the selected sample contains 39040 objects. Recently Remi Cabanac (ESO) , using Pr incipal Component Analysis on bright stars, discovered that U N M S 1 SEDs show a systematic deviation from spectra published by Pickles [56]. The U N M S l SEDs tend to be brighter at long and short wavelengths. This could result from an error in the absolute calibration derived from K P N O observations. To make the U N M S l catalog more consistent with that of Pickles, the original fluxes Forig(fUteri) were corrected according to the relation CHAPTER 4. THE UNMS QUASAR SURVEY. 42 parameter abbr. units Right Ascension ra hh:mm:ss Declination dec deg: arcmin: arcsec band number nband semimajor axis a arcsec semiminor axis b arcsec position angle pa grad broad bands B - V - R - I mag narrow bands 14.48 - 14.82 mag Table 4.1: Parameters in the U N M S 1 catalog. Fcorr (f Uteri) = Forig{f ilter i)/C(f UterA (4.1) where Fcorr(f Uteri) is the corrected flux in the aith" filter and C(filteri) is the correction correspondent to this filter. This formula can be directly applied to narrow bands since the largest bandwidth is around 40 A; for broad bands the most accurate way to proceed is to convolve the transmission curves wi th the best fit curve of the discrete set of corrections, according to the formula C(fUteri) = ^ ZiXk (4-2) where C\k is the value from the correction curve and W{ is the transmission curve value for the "ith" filter at wavelength A*.. Table 4.2 shows the flux corrections provided by Remi Cabanac. CHAPTER 4. THE UNMS QUASAR SURVEY. 43 A correction A correction 4541.0 1.180418 6873.0 0.925380 4865.0 1.059962 7033.0 0.923201 4979.0 0.992161 7197.0 0.915377 5095.0 1.055990 7364.0 0.908164 5213.0 1.031924 7536.0 0.948763 5335.0 0.967368 7711.0 0.941986 5459.0 0.990099 7891.0 0.950050 5586.0 0.997617 8075.0 1.010598 5716.0 0.997977 8263.0 1.015200 5850.0 0.958713 8455.0 1.033491 5986.0 0.996552 8652.0 1.087448 6125.0 0.937052 8854.0 1.064739 6268.0 0.964520 9060.0 1.109830 6414.0 0.903055 9271.0 1.204436 6563.0 0.893435 9487.0 1.224466 Table 4.2: F l u x corrections by Remi Cabanac [priv. comm.]. The best fitting curve, found from a x2 minimizat ion procedure, is Cx = 2.893959 - 5.80649710 _ 4A + 4.26502710 _ 8 A 2 (4.3) and it is graphically represented by the line in F i g . 4.2 wi th the discrete set of corrections. CHAPTER 4. THE UNMS QUASAR SURVEY. 44 5000 6000 7000 8000 9000 lambda Figure 4 . 1 : Corrections wi th the best fit curve. 4.2 Method. The method used to separate quasars from stars and galaxies is based on the x2 minimizat ion procedure combined with a Bayesian approach. According to the Bayesian theory, given two models and a data set, it CHAPTER 4. THE UNMS QUASAR SURVEY. 45 is often more useful to consider the ratio of model probabilities than the probabilities directly [30]. The so-called odds ratio in favor of the model M j over the model Mj is defined as 0.. = pmp,i) (44) where p(Mi\D, I) is the probability of "ith" model Mi given the data D and the prior information I. According to Bayes' theorem, the posterior proba-bi l i ty of the model M j based on data and the prior information is pmDJ)=mnmBdi (4.5) where p(Mi\I) is the probability of the model M j given the prior information I, p(D\Mi,I) is the global likehood for the model M j (data influence the choice of the model) and p(D\I) is the probability of data given the prior information (this parameter is independent of the model). Substituting these terms in (4.7), the odds ratio becomes: = p(Mi\I)p(D\MhI) =P(Mj\I) U i j piMillMDlM^I) piMAlf l4-°j where is the Bayesian factor. If p(9\Mi,I) is the prior probability of each sub-model of that specific category M j , characterized by the parameter 9 (it is reasonable to take this probability equal to 1/Niy where A j^ is the total number of sub-models of that category) and p(D\9, Mi, I) is the posterior probability of the template (the likehood function of that template), the global likehood of the model M j is defined as CHAPTER 4. THE UNMS QUASAR SURVEY. 46 p(D\Mi,I) = jpWMiJWD^MiJ)^ = ^Jp{D\e,Mhi)r\e = ^ • E p W . J W i , / ) (4.7) iV* Oi=l For one template of the category Miy the maximun likehood for " N " detected bands is just the product of the individual probability functions for the observed fluxes and the model fluxes in the same filter: assuming a Gaussian distribution, the maximum likehood is then [4] N -1 I N p _ A pT P{D\euMui) = m — T H e ^ [ - o E I " "H (4-8) n=l °ViV lK 1 „ = i °V» where F n and a n are the observed flux and its error in the unth" filter, F* is the template flux in the "nth" filter and A is the normalization constant. The x2 parameter is defined as N J? — Ah?T X 2 = E[ n? (4-9) n=l therefore p(D|(9 i,M1,/)=n[^=][c-*xa] (4-10) n=l CnV^TT Maximiz ing the likehood function involves the minimizat ion of the ex-ponent: that involves the x2 minimizat ion through the right choice of the normalization constant, whcih can be found from the condition S\2 CHAPTER 4. THE UNMS QUASAR SURVEY. 47 The final formula for the odds ratio is p{Mi\I) Nj[e-^] O p ( M J - | J ) ^ [ e - h ? ] ' (4.12) stars galaxies quasars B 1 Q ( 0 . 1 6 2 B [ J ] - 0 . 5 1 6 ) l Q ( 0 . 2 9 * B [ i ] - 2 . 6 1 ) A . - | _ Q ( 0 . 8 6 * B [ i ] - 1 5 . 8 0 ) b: -10.95 + 15.61 io° - 2 8 *( B W- 1 9 - 1 5 ) V ^Q(0.15*V[i]+0.05) same as for B filter same as for B filter R l Q ( 0 . 4 1 * i i [ i ] - 5 . 5 2 ) •j_Q(0.11*fl[i]+0.15) same as for B filter I 1 Q ( 0 . 3 8 * / [ » ] - 4 . 4 2 ) 1 0 ( 0 . 1 * / [ i ] + 0 . 5 ) same as for the B filter Table 4.3: Surface densities for stars, galaxies and quasars in the broad bands: the units are num/deg 2 for the B and I filters, and num/deg 2 0.5 mag in the R and V bands. The "a" case is for B < 19.5 and the "b" is for B > 19.5. The quantity p(Mi\I) is related to the surface density of quasars, stars and galaxies given the magnitude in a filter. Table 4.3 shows these quantities, extrapolated from other surveys, for the broad bands B, V, R, I. A l l the con-sulted quasar surveys report surface densities in the B band: this situation, which explains why V, R, I bands have the same surface density as the B, is due to the fact that these surveys studied quasars at small and medium redshift ( 0.3 - 3) for which important spectral features, such as the Lya, are st i l l in the B. For high redshift quasars (z > 3), the observed flux in the B band is reduced dramatically by the presence of the increasingly preva-lent Lya forest lines and of the higher column density metal-line absorption CHAPTER 4. THE UNMS QUASAR SURVEY. 48 systems. For surveys of high redshift quasars, working in the R band or at longer wavelenths is a fundamental prerequisite to select this category of sources: these kind of studies have just started. The unaccuracy to have the same surface density for different broad bands doesn't influence the object classification: the amount of quasars is a really small percentage of that of stars and galaxies in every bands. The identification is done through a code which, for each object, cal-culates the odds ratio for every pair of categories (star-galaxy, star-quasar, galaxy-star, galaxy-quasar, quasar-star and quasar-galaxy, six combinations in total), and chooses the highest one: for example, i f the highest odds ratio is 0Sjq between a star and a quasar template, the object in question is iden-tified as a star. 4.3 Templates. 4.3.1 Stellar templates. Stellar templates were taken from Pickles' database [56]. The stellar spectral l ibrary includes 131 flux-calibrated spectra encompassing al l normal spectral types and luminosity classes at solar abundances, metal-weak and metal-rich F - K dwarfs and G - K giant components. The library has complete spectral coverage from 1150-10620A, in steps of 5A, up to 25000A for about half of them (mainly later stellar types). Monochromatic fluxes F\ are tab-ulated and normalized to unity at 5556 A. Table 4.4 shows how these templates are organized inside the library: the CHAPTER 4. THE UNMS QUASAR SURVEY. 49 first column gives the library number, the second column the stellar type, and the thi rd column the luminosity class. The first group includes 5 metal-weak and 5 metal-rich G - K V spectra, the third group includes 7 metal-weak and 7 metal-rich G - K III spectra. l ibrary number stellar type luminosity class 1-45 0 5 - M 6 V 46-59 B 2 - K 3 I V 60-105 O8-M10 III 106-113 B 2 - M 3 II 114-131 B 0 - M 2 I Table 4.4: Parameters in Pickles' stellar library. In order to be compared to the observed spectral energy distributions, these templates must first be convolved wi th the transmission curves of those filters used in this survey. For the " T " 1 " template, the monochromatic flux integrated over the "ith" filter, is given by (4.13) F T ^f=lFlWJkQXk nuwikQXk where Fjk is the template's flux at wavelength \ k , W{k is the "ith" filter transmission value at wavelength \ k and Q\k is the system response. Since U N M S 1 spectra are given in frequency, Fv versus v, monochromatic fluxes in wavelength are converted to fluxes in frequency according to the formula CHAPTER 4. THE UNMS QUASAR SURVEY. 50 Fu = -^Fx (4.14) The system response Q\k is approximated by the C C D Quantum Effi-ciency (QE) . C C D s used in this survey don't show any response for wave-lengths higher than 1 pm. Therefore, the sum in (4.4) is carried only as far as A = 1000 nm: this involves those filters whose transmission curves extend further than this threshold, the broad-band I and the narrow bands log(u) = 14.48 and log(v) = 14.49 (these narrow filters were used in 1999 only). The procedure was applied to galaxy and quasar templates as well. 4.3.2 Galaxy templates. Most of the galaxy SEDs were taken from Kinney & Calzetti 's l ibrary [44] and integrated with few others from Rocca-Volmerange's database [59] in order to have a complete set of templates. The first group of spectra are given as F\ versus A, wi th the flux in erg c m - 2 s _ 1 A - 1 and steps of 1.5 A. The galactic types used are elliptical, bulge, SO, Sa, Sb and Starburst galaxies from 1200 A to 8000 A, in some cases up to 10000 A. Table 4.5 summarizes the template types. A l l these spectra, normalized at 4000 A, are shown in F i g . 4.2. The second group of templates includes Sc and U V - h o t E / S 0 models, 12.08 G y r old, extended from fa r -UV (1220 A) to 1 /im wi th a spectral resolution of 10A. A l l the templates were convolved with filter transmission curves in the same way as the stellar templates. Since models provided are built in the rest frame, they were redshifted from z=0.00 to z=1.20, wi th step 5z=0.01; CHAPTER 4. THE UNMS QUASAR SURVEY. 51 library name galactic type colour excess SB1 starburst galaxy E ( B - V ) < 0.10 SB2 starburst galaxy 0.11 < E ( B - V ) < 0.21 SB3 starburst galaxy 0.25 < E ( B - V ) < 0.35 SB4 starburst galaxy 0.39 < E ( B - V ) < 0.50 SB5 starburst galaxy 0.51 < E ( B - V ) < 0.60 SB6 starburst galaxy 0.61 < E ( B - V ) < 0.70 SO SO galaxy Sa Sa galaxy Sb Sb galaxy bulge bulges tempi. elliptical elliptical tempi. Table 4.5: Spectral templates from Kinney & Calzett i galaxies' library, in this way, 1573 templates were obtained. 4.3.3 Quasar templates. The rest-frame template of quasars is the result of merging two templates taken from different authors, in order to cover the largest possible wavelength range (330-8550 A). The first S E D model comes from Zheng et al . [78] and was constructed using 284 H S T F O S spectra of 101 quasars with redshifts z > 0.33. The uncalibrated spectrum is given in F i g . 4.3 and it covers wavelengths between 350 and 3000 A in the rest frame. CHAPTER 4. THE UNMS QUASAR SURVEY. 52 CHAPTER 4. THE UNMS QUASAR SURVEY. 53 Lya+N V 10 8 4 h FOS Quasar Spectrum i 0.8 400 600 800 1000 2000 3000 Rest—frame Wavelength (A) Figure 4.3: Composite spectrum of 101 quasars, binned to 2 A. It is clear there is a significant steepening of the continuum slope around 1050 A. Table 4.6 shows continuum fit parameters using radio-loud quasars, radio-quiet quasars and both types together: power law indexes av = -1.96 for the range 350-1050 A and av = -0.99 for the range 1050-2200 A characterize the original template. The second model, in F i g 4.4, was taken from Vanden Berk et al . [74]. The composite spectrum, with resolution of 1 A, is the result of a homogeneous data set based on more than 2200 spectra from the Sloan Dig i ta l Sky Survey. The median composite template covers a range from 800 to 8555 A. A double power law with av = -0.46 for A < 4700 and av - -1.58 for A > 4700 A is the CHAPTER 4. THE UNMS QUASAR SURVEY. 54 Sample A l l Quasars Radio-loud Radio-quiet Number of objects 101 60 41 Mean redshift 0.93 0.87 0.95 Wavelength range (A): 1050-2200 -0.99 ± 0.01 -1.02 ± 0.01 -0.86 ± 0.01 600-1050 -2.02 ± 0.05 -2.45 ± 0.05 -1.83 ± 0.03 350-1050 -1.96 ± 0.02 -2.16 ± 0.03 -1.77 ± 0.03 Table 4.6: F i t ted power-law index. best fit. The spectral region blueward of the L y a emission line was ignored when calculating the flux density scaling, since the L y a forest flux density varies greatly from spectrum to spectrum. This is another reason which explains why this region was then replaced by the first template. The first step was to integrate these two models. This was done making them completely flat (spectral features were reproduced proportionally to their reciprocal intensities on a null-slope spectral continuum), calculating the normalization constant in the wavelength range they overlap and applying this constant to obtain a unique template. F i g . 4.5 and F i g . 4.6 show this unique "flat" model respectively for the range 300-3400 A and the range 3400-8550 A. Many important features such as the L y a , Ly/3, H a and OIII emission lines are well pronounced. The second step was to bui ld different types of templates introducing various slopes for the spectral continuum. The idea was to look for the largest possible range of the spectral index a „ cited in papers, and then to divide this range into equal intervals. For the U V part of the template to be CHAPTER 4. THE UNMS QUASAR SURVEY. 55 1 1 1 1 1 1 1 I 1 1 1 1 1 1 1 1 1 1 1 1—1—I—I | I I I ( ! I I I I | i Ly a 1000 2000 4000 6000 8000 Rest Wavelength, X (A) Figure 4.4: Composite spectrum: power law fits to the estimated flux are shown. constructed, only one value of slope was used, av= -1.96, which is the average between the lowest U V slope in literature, -2.16 from Zheng et. al [78], and the highest value, au~ -1.76 found from Hatziminaoglou et. al [33]. The same idea was used for the optical-near infrared (Opt-IR) range: the lowest slope value is -2.5 from Richstone & Schmidt [58] and the highest value is 1 from Hatziminaoglou et. al [33]. Div id ing this range into three equal intervals, I obtained those values of ctv used in the Opt - IR part: -2.5, -1.33, -0.16 and 1.01. Table 4.7 shows al l the slopes in frequency and wavelength used for the CHAPTER 4. THE UNMS QUASAR SURVEY. 56 Figure 4.5: First part of the composite spectrum in the 300-3400 A range: it can be noted that the L y a (1215 A) and C I V (1549 A) are among the strongest emission lines. quasar templates considering that the relationship between the power law index in frequency and that in wavelength is ct\ = —(au + 2). CHAPTER 4. THE UNMS QUASAR SURVEY. 57 o a *4 _L J I I I I I l_ _L 4000 5000 6000 7000 lambda(A) 8000 9000 Figure 4.6: Second part of the composite "flat" spectrum in the 3400-8550 A range. H a (6563 A) and [OIII] (4363 A) are among the strongest emission lines. In F i g 4.7 al l the quasar templates (in the rest frame) are shown. Each of these models was then redshifted to z=7.00 wi th step 5z=0.02, CHAPTER 4. THE UNMS QUASAR SURVEY. 58 a\ template U V -1.96 -0.04 Opt - IR -2.5 0.5 I -1.33 -0.67 II -0.16 -1.84 III 1 -3 I V Table 4.7: Power-law index in frequency and wavelength for quasar templates in the U V and optical-near-infrared (Optical-IR) region. but only those with z > 0.30 were compared to U N M S 1 catalog objects: in total 1344 templates were used. The choice of z=7.00 is due to the presence of the Lya emission line: as long as it appears inside the optical range, it is a useful tool to identify quasars. 4.4 Method calibration from templates. So far, we have not calibrated the reliability of quasar detections and how their identification is contaminated from other astronomical objects, especially stars. A good way to test this method is to apply it to something we already know and to see if results are consistent wi th this "a priori" knowledge. The idea is to work with templates of each category at different redshifts and to consider them as "test objects", called also "template-objects". Results from a pure minimizat ion x2 procedure and from the odds ratio procedure are important in order to see if new conditions should be considered when CHAPTER 4. THE UNMS QUASAR SURVEY. 59 1 , , 1 , , , 1 1 1 1 1 1 I I | _ J I I I I I I I I I I I I L 2000 4000 6000 8000 LAMBDA Figure 4.7: Quasar templates in the rest frame: lines I, II, III and I V repre-sent respectively template I, II, III and I V . the method is applied to real sources. CHAPTER 4. THE UNMS QUASAR SURVEY. 60 4.4.1 A first analysis. As already discussed, the database includes 131 stellar templates, 13 galactic templates spread from redshift 2=0.00 to 2=1.20 and step Az=0.01 (1573 galaxies in total) and 4 quasar types from 2=0.30 to 2=7.00 and step A2=0.02 (1344 in total). For an appropriate analysis, models were normalized to the flux in the filter "671" ( l o g f » = 1 4 . 6 5 , A=6716 A , see Table 2.1) and plotted in few diagrams. In this way it is possible to recognize peculiar features that may be im-portant to separate quasars from stars and galaxies. F i g . 4.8 - 4.12 show all the galactic (blue line) and quasar (red line) tem-plates in the rest frame wi th different stellar templates (green line): main sequence stars, giant stars, supergiant stars, subgiant stars and bright giant stars. To avoid crowded diagrams, labels were added only to indicate the galactic and quasar models. Peculiar features at log(y) = 14.49, log(i/) = 14.63, \og(u) = 14.73 and log(^) = 14.85 correspond to the broad bands J, R, V and B. Quasar templates are truncated at log(i/)=14.56 since the original high-resolution model was not complete enough to cover the entire optical range. This inconvenient doesn't influence results from the odds ratio method since only quasar models at redshift higher than 0.3 are used. The Q S O plots show two evident lines: the H a line at log(i/)=14.66 and the H/? at log(^)=14.79. In any of these diagrams, the range \og{y)= 14.55-14.65 is characterized by a particular feature, common only to quasars and starburst galaxies. If this structure persisted at any redshift, it could be an interesting tool firstly CHAPTER 4. THE UNMS QUASAR SURVEY. 61 I I I I I I I I I I I I I I I I L 1 4 . 5 1 4 . 6 1 4 . 7 1 4 . 8 log(fre) Figure 4.8: Galactic (dashed line) and quasar (dashed-dot line) templates in the rest frame with Supergiant star templates (solid l ine). to separate stars from extragalactic objects, and secondly to find quasars using both the Bayesian approach and the % 2 procedures. In fact if a source is already identified as a quasar by the Bayesian approach and a quasar tern-CHAPTER 4. THE UNMS QUASAR SURVEY. 62 14.5 14.6 14.7 14.8 log(fre) Figure 4.9: Galactic (dashed line) and quasar (dashed-dot line) templates in the rest frame with bright giant templates (solid line). plate is considered the best fit when the pure x2 is calculated only in the region 14.55-14.65, then that object can reasonably be considered a quasar candidate. CHAPTER 4. THE UNMS QUASAR SURVEY. 63 14.5 14.6 14.7 14.8 log(fre) Figure 4.10: Galactic (dashed line) and quasar (dashed-dot line) templates in the rest frame with different giant star templates (solid line). F i g . 4.13-4.17 show all the galactic and quasars templates at redshift z=1.2 with stellar templates. The structure has disappeared and, in another region, the M g l l line is now prominent for only two quasar templates. Be-CHAPTER 4. THE UNMS QUASAR SURVEY. 64 Figure 4.11: Galactic (dashed line) and quasar (dashed-dot line) templates in the rest frame with different subgiant star templates (solid line). cause these characteristics are a function of redshift, they are not useful for the separation quasars. CHAPTER 4. THE UNMS QUASAR SURVEY. 65 — I 1 1 1 1 1 1 i i i I i i i i L_ 14.5 14.6 14.7 14.8 log(fre) Figure 4.12: Galactic (dashed line) and quasar (dashed-dot line) templates in the rest frame with different main sequence star templates (solid line). CHAPTER 4. THE UNMS QUASAR SURVEY. 66 I I I I I I I I I I I I I I I I I L 14.5 14.6 14.7 14.8 log(fre) Figure 4.13: Galactic (dashed line) and quasar (dashed-dot line) templates at redshift z=1.2 with supergiant star templates (solid line). 4.4.2 x 2 minimization method and odds ratio procedure applied to original templates. The first test is to compare each template-object to al l the models avail-able, finding not only the one giving the lowest x2 a n d the highest odds ratio CHAPTER 4. THE UNMS QUASAR SURVEY. 67 I i i i i i i i i i i i 14.5 14.6 14.7 14.8 log(fre) Figure 4.14: Galactic (dashed line) and quasar (dashed-dot line) templates at redshift z=1.2 with bright giant star templates (solid line). but also the x2 distribution from all the templates. These were organized firstly according to the stellar and galactic type, and secondly according to the luminosity class for stellar models. Table 4.8 displays this sequence. CHAPTER 4. THE UNMS QUASAR SURVEY. 68 N . T E M P L . N . T E M P L . N . T E M P L . N . T E M P L . 1 elliptical 38 b l v 75 gOiv 112 k3i 2 bulge 39 b2ii 76 gOv 113 k34ii 3 SO 40 b2iv 77 wgOv 114 k3i i i 4 Sa 41 b3i 78 rgOv 115 rk3i i i 5 Sb 42 b3ii i 79 g2i 116 wk3i i i 6 Sc 43 b3v 80 g2iv 117 k3iv 7 SB1 44 b5-7v 81 g2v 118 k3v 8 SB2 45 b5i 82 g5i 119 k4i 9 SB3 46 b5ii 83 g5ii 120 k4i i i 10 SB4 47 b5ii i 84 g5iii 121 wk4i i i 11 SB5 48 b6iv 85 wg5iii 122 rk4i i i 12 SB6 49 b8i 86 rg5iii 123 k4v 13 U V h o t 50 b8v 87 g5iv 124 k5i i i 14 Q S O l 51 b9ii i 88 g5v 125 rk5i i i 15 Q S 0 2 52 b9v 89 wg5v 126 k5v 16 Q S 0 3 53 fOi 90 rg5v 127 k7v 17 Q S 0 4 54 fOii 91 g8i 128 mOiii 18 o8iii 55 fOiii 92 g8iii 129 mOv 19 o5v 56 f0-2iv 93 wg8iii 130 m l i i i 20 o9v 57 fOv 94 g8iv 131 m l v 21 aOi 58 f2ii 95 g8v 132 m2i 22 aOiii 59 f2iii 96 k O l i i 133 m2i i i 23 aOiv 60 f2v 97 kOiii 134 m2v 24 aOv 61 f5i 98 wkOiii 135 m2.5v Table 4.8: Continue. CHAPTER 4. THE UNMS QUASAR SURVEY. 69 N . T E M P L . N . T E M P L . N . T E M P L . N . T E M P L . 25 a2i 62 f5iii 99 rkOiii 136 m3ii 26 a2v 63 f5iv 100 kOiv 137 m3i i i 27 a3iii 64 f5v 101 kOv 138 m3v 28 a3v 65 wf5v 102 rkOv 139 m4i i i 29 a47iv 66 f6v 103 k l i i i 140 m4v 30 a5iii 67 rf6v 104 w k l i i i 141 m5i i i 31 a5v 68 f8i 105 r k l i i i 142 m5v 32 a7iii 69 f8iv 106 k l i v 143 m6i i i 33 a7v 70 f8v 107 k2i 144 m6v 34 bOi 71 wf8v 108 k2i i i 145 m7i i i 35 bOv 72 rf8v 109 wk2i i i 146 m8i i i 36 b l i 73 gOi 110 rk2i i i 147 m9i i i 37 b l - 2 i i i 74 gOiii 111 k2v 148 mlOi i i Table 4.8: Template sequence for the x2 analysis. CHAPTER 4. THE UNMS QUASAR SURVEY. 70 I i i i i I i i i i i i i i ' i i 14.5 14.6 14.7 14.8 log(fre) Figure 4.15: Galactic (dashed line) and quasar (dashed-dot line) templates at redshift z=1.2 wi th giant star templates (solid line). F i g . 4.18-4.21 show the \ 2 distribution for several quasar template-objects when compared to al l the 3048 models. This is useful to understand which templates have similar % 2 and confuse the object identification (this CHAPTER 4. THE UNMS QUASAR SURVEY. 71 Figure 4.16: Galactic (dashed line) and quasar (dashed-dot line) templates at redshift z=1.2 with sub-giant star templates (solid line). explains why these plots show only the bottom region and not the complete range). The blue line represents the comparison with galactic templates, the red line quasar templates and the green line stellar template. CHAPTER 4. THE UNMS QUASAR SURVEY. 72 14.5 14.6 14.7 14.8 log(fre) Figure 4.17: Galactic (dashed line) and quasar (dashed-dot line) templates at redshift z=1.2 with main sequence stars (solid line). When the template-object is just one of the models, it is logical that the lowest x2 comes from the original template. From these plots it is important to see is that there are some stellar and galactic models wi th low % 2 and close CHAPTER 4. THE UNMS QUASAR SURVEY. 73 QSO templ.14z=0.3 50 100 150 TEMPLATE Figure 4.18: x2 distribution for 4 Q S O N.14 at z=0.3: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. to the object in question. For example, quasar N.14 is close to the first six galactic models at low redshift (mostly ellipticals and spirals) and to K and CHAPTER 4. THE UNMS QUASAR SURVEY. 74 Figure 4.19: x2 distribution for Q S O N.15 at z=1.4: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. M stellar templates; quasar N.15 is very close to late spirals and starburst galaxies and to G and K stars, quasar N.16 is mostly near F stars and finally CHAPTER 4. THE UNMS QUASAR SURVEY. 75 QSO templ.16z=3.5 _ J i i i i i i i i i i 50 100 150 TEMPLATE Figure 4.20: x 2 distribution for Q S O N.16 at z=3.5: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. quasar N.17 is very close to starburst galaxies . This situation is confirmed by the odds ratio's results: 40 objects, al l from CHAPTER 4. THE UNMS QUASAR SURVEY. 76 Figure 4.21: x2 distribution for Q S O N.17 at z=5.6: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. the quasar category, are infact misidentified. In particular N.14 quasars in the redshift range 1.30-1.46 are mostly identified as U V h o t and spiral galaxies at CHAPTER 4. THE UNMS QUASAR SURVEY. 77 low and medium redshift. Quasars N.15, N.16 and N.17 are mostly confused wi th F and B stars. The 40 misclassified elements are just a small portion of the entire sample which involves more than 3000 test objects. Since the final goal is to find quasars in a catalog of real objects, it is interesting to analyze the results from the same test applied to altered models, as we explain below. 4.4.3 x 2 a n d odds ratio procedures applied to altered templates. For the second test, template-objects were modified to simulate realistic conditions. Noise was added to them simulating 39 measurements, from the standard Gaussian distribution [4]: with mean 0 and standard deviation 1. The "ith" deviated flux is given by where Fi and 8Fi are the flux with its error in the "ith" filter. F i g . 4.22-4.25 show the x2 distribution found for the same objects cited in the previous paragraph, this time wi th noise added: as expected, the x2 between the smeared template-object and the original model is no longer null but is of the same order as those of other templates. This k ind of ambiguity is fatal for ~500 template-objects for which the X2 method fails to recover the correct model. (4.15) Fi,dev — Fi -f- SFiOj (4.16) CHAPTER 4. THE UNMS QUASAR SURVEY. 78 Figure 4.22: x2 distribution for the altered Q S O N.14 at z=0.3: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. From the odds ratio approach, 1082 template-objects of 3048 are misclas-sified. In particular, N.14 quasars are confused wi th U V h o t , ell iptical and CHAPTER 4. THE UNMS QUASAR SURVEY. 79 Figure 4.23: x2 distribution for the altered Q S O N.15 at z=1.4: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. spiral galaxies, N.15 quasars are mainly classified as starburst and late spiral galaxies, and G and F stellar types, N.16 quasars is mostly seen as F type CHAPTER 4. THE UNMS QUASAR SURVEY. 80 Figure 4.24: x2 distribution for the altered Q S O N.16 at z=3.5: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. stars and quasars N.17 with B type stars. U n t i l now all the results are based on template-objects wi th 39 flux mea-CHAPTER 4. THE UNMS QUASAR SURVEY. 81 Figure 4.25: x 2 distribution for the altered Q S O N.17 at z=5.6: line I, II and III come from the comparison respectively to galactic, quasar and stellar templates. surements in the optical range. This situation is not very realistic: in the U N M S l catalog, most of the sources have 10-20 filters and only 4% of the CHAPTER 4. THE UNMS QUASAR SURVEY. 82 entire database has more than 35 magnitudes. So, it is important not only to add noise to these template-objects, but also to analyze the results when more and more filters are randomly neglected (jackknife approach). F i g . 4.26 shows the distribution of stars (green line), galaxies (blue line) and quasars (red line) as classified by the program against the number of fil-ters: each distribution is normalized to the object number available for each category (131 stars, 1573 galaxies and 1344 quasars). Note that the num-ber of stars and quasars increases and decreases respectively when more and more filters are randomly neglected, while the galaxy distribution is nearly unaffected. 10-15 point quasar SEDs (with some emission lines missed) can be easily confused wi th star SEDs and this suggests the importance of using a criterion on the minimum filter number when the odds ratio method is applied. The fact the galaxy distribution is practically flat means galaxies are not easily confused with other types of objects and the number of filters is not so important for this category. It is interesting to split the distribution of stars, galaxies and quasars into 4 redshift ranges and include the fraction of template-objects correctly classified (coloured line) and the fraction of models incorrectly classified in that category (black line). In this way it is possible to understand how this misclassification between stars and quasars changes wi th redshift. In order to do that, 4 redshift ranges are considered: • range I at 2=0-0.6 which involves 131 stars, 793 galaxies and 64 quasar (Fig. 4.27-4.29) ; • range II at 2=0.6-1.2 which includes 780 galaxies and 120 quasars (Fig. 4.30-4.31); CHAPTER 4. THE UNMS QUASAR SURVEY. 84 _i i i i i i i i i i i i i i i i 10 20 30 40 FILTERS Figure 4.27: Normalized distribution for correctly classified stars (hexagon) and for objects wrongly classified as stars (cross). As usual, these distributions were normalized to the object number available in each redshift range. For galaxies the distribution is almost flat in every range. When S E D s CHAPTER 4. THE UNMS QUASAR SURVEY. 85 z > 3 o O 0.5 10 20 30 40 FILTERS Figure 4.28: Normalized distribution in the redshift range z=0-0.6 for cor-rectly classified galaxies (hexagon) and for objects wrongly clas-sified as galaxies (cross). are constructed with only 10 flux points, 85% of galaxies are s t i l l correctly classified. CHAPTER 4. THE UNMS QUASAR SURVEY. 86 z w > P '2" o o FILTERS Figure 4.29: Normalized distribution in the redshift range z=0-0.6 for cor-rectly classified quasars (hexagonal) and for objects wrongly classified as quasars (cross). For quasars, as the number of filters decreases, the distribution drops very quickly, first of al l in the second redshift range. This behavior is consistent CHAPTER 4. THE UNMS QUASAR SURVEY. 87 z > o o o FILTERS Figure 4.30: Normalized distribution in the redshift range z=0-0.6 for cor-rectly classified galaxies (hexagon) and for objects wrongly clas-sified as galaxies (cross). with that of stars uncorrectly classified in the first range: the black line in-creases when fewer filters are used and this opposite trend implies that most CHAPTER 4. THE UNMS QUASAR SURVEY. 88 z > XI o x> o FILTERS Figure 4.31: Normalized distribution in the redshift range z=0.6-1.2 for cor-rectly classified quasars (hexagon) and for objects wrongly clas-sified as quasars (cross). of them are in real terms QSOs. This could be due to the presence of lines in the quasar S E D which resemble those in the stellar spectrum or, since CHAPTER 4. THE UNMS QUASAR SURVEY. 89 2 > £1 O XI O FILTERS Figure 4.32: Normalized distribution in the redshift range z=l.2-4.1 for cor-rectly classified quasars (hexagon) and for objects wrongly clas-sified as quasars (cross). filters are neglected randomly, some emission features are not considered and the quasar's continuum is easily confused to a stellar "spectrum". Another CHAPTER 4. THE UNMS QUASAR SURVEY. 90 Figure 4.33: Normalized distribution in the redshift range z=4.1-7 for cor-rectly classified quasars (hexagon) and for objects wrongly clas-sified as quasars (cross). interesting consideration is that the black line, which represents the number of sources misclassified as quasars, does not appear in any plot for quasars. CHAPTER 4. THE UNMS QUASAR SURVEY. 91 This means that when a test object is classified as quasar, it is really so. We can conclude that in order to have at least 80% of well classified quasars, the method should be applied only to sources with 30 or more magnitudes. The situation is improved for galaxies: in this case the threshold is closer to 20 magnitudes. To confirm the validity of this criterion, fig. 4.34-4.39 show the distri-bution of the odds ratio's logarithm normalized to the one from the correct category and calculated with 39 filters, for 4 quasars and 2 galaxies at dif-ferent redshift (black line). The odds ratio can be used to determine which of two models is the pre-ferred one considering the ratio of their probabilities. In each plot, the black line represents the odds ratio distribution in favour of the correct category, and the red, green and blue lines are the odds ratio distributions in favour of quasar, star and galaxy models as the best fit of the template-object anal-ysed. When a line exceeds the other ones in a plot, the related category of models is the preferred one. For quasar test objects, in most cases the odds ratio in favour of stars and galaxies (wrong classification) is lower than the one in favour of quasars (correct classification) for more than 25 filters at least. For galaxies fewer filters are required to avoid an eventual misclassifi-cation: it is reasonable to expect that only 20 filters are required. CHAPTER 4. THE UNMS QUASAR SURVEY. 92 Figure 4.34: Normalized odds ratio distribution for a Sa spiral galaxy (N.4) at redshift 0.2, 0.6, 0.8 and 1.2 in favor of stellar (hexagon), galactic (croos) and quasar (triangle) template. CHAPTER 4. THE UNMS QUASAR SURVEY. 93 Figure 4.35: Normalized odds ratio distribution for a starburst galaxy (N.7) at redshift 0.2, 0.6, 0.8 and 1.2. in favour of stellar (hexagon), galactic (cross) and quasar (triangle) template. CHAPTER 4. THE UNMS QUASAR SURVEY. 94 Figure 4.36: Normalized odds ratio distribution for Q S O N.14 at redshift 0.3, 2.1, 4.2 and 7.in favour of stellar (hexagon), galactic (triangle) and quasar (cross) template. CHAPTER 4. THE UNMS QUASAR SURVEY. 95 Figure 4.37: Normalized odds ratio distribution for Q S O N.15 at redshift 0.3, 2.1, 4.2 and 7. in favour of stellar (hexagon), galactic (triangle) and quasar (cross) template. CHAPTER 4. THE UNMS QUASAR SURVEY. 96 Figure 4.38: Normalized odds ratio distribution for Q S O N.16 at redshift 0.3, 2.1, 4.2 and 7.in favour of stellar (hexagon), galactic (triangle) and quasar (cross) template. CHAPTER 4. THE UNMS QUASAR SURVEY. 97 Figure 4.39: Normalized odds ratio distribution for Q S O N.17 at redshift 0.3, 2.1, 4.2 and 7.in favour of stellar (hexagon), galactic (triangle) and quasar (cross) template. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 98 C H A P T E R 5 Q U A S A R S IN T H E U N M S l C A T A L O G . 5.1 Results. A s shown in the previous chapter, QSOs can be found and identified wi th a completeness of ~ 85% when the method described in §4 is applied to the U N M S l catalog for objects satisfying two criteria: a min imum number of 30 flux measurements and a minimum separation radius of 3 arcsecs. Based on these criteria, our algorithm selects 39040 sources out of the 20.13-deg2 field. Table 5.1 summarizes the number of objects. Category # S T A R S G A L A X I E S Q U A S A R S 25193 10791 3056 Table 5.1: Star, galaxy and quasar identifications. We first describe star and galaxy candidates; quasar candidates wi l l be discussed in §5.2. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 99 Figure 5.1: Star counts by spectral types. 5.1.1 Stellar candidates in the UNMS survey. F i g . 5.1 shows the distribution of stars, organized according to the order presented in Table 5.2. Two major structures appear: from the 46th to the 99th stellar type and CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 100 from the 108th to the 131th stellar type. The first group corresponds to F, G, K and M stars of any luminosity class; the second group correspond to metal-weak and metal-rich F, G, K and M stars, mainly main sequence and giant stars. The most common templates are main sequence G stars and giant G stars. A and B stars (1-32) and O stars (105-108) represent just a small fraction of the stellar distribution. This result is in agreement with the star counts by spectral type commonly known. Table 5.3 shows the number density logarithm of stars [ # / 104pc3] by spectral types from data published in Allen (1973) [6] [1]: for every luminosity class, F-G-K-M stars are the most frequent. Table 5.4 is similar to Table 5.3 but shows the results from the LMT survey: it is useful to see if the method can recognize stars not only of different type but also of different luminosity classes. The comparison of these two tables confirms that the method seperate stars of different stellar types: many F, G, k and M stars are found with respect to other types; the only discrepancy is in large offsets of O stars in the LMT sample. The star distribution according to the luminosity class, is not good: the number of giants and supergiants is too high with respect to that of main sequence stars. 5.1.2 Galaxy candidates in the UNMS survey. Fig. 5.2 shows the galaxy counts by morphological type according to the order described in Table 5.5: the term "Bulge" refers to galaxies with a pro-nounced bulge (spheroidal central region); the term "SB" refers to starburst CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 101 N U M . T Y P E N U M . T Y P E N U M . T Y P E N U M . T Y P E 1 aOi 34 fOi 67 kOiv 100 m5v 2 aOiii 35 fOii 68 kOv 101 m6ii i 3 aOiv 36 fOiii 69 k l i i i 102 m6v 4 aOv 37 fOv 70 k l i v 103 m7i i i 5 a2i 38 f2ii 71 k2i 104 m8i i i 6 a2v 39 f2iii 72 k2i i i 105 m9i i i 7 a3iii 40 f2v 73 k2v 106 o5v 8 a3v 41 f5i 74 k34ii 107 o8iii 9 a47iv 42 f5iii 75 k3i 108 o9v 10 a5iii 43 f5iv 76 k3i i i 109 rf6v 11 a5v 44 f5v 77 k3iv 110 rf8v 12 a7iii 45 f6v 78 k3v 111 rgOv 13 a7v 46 f8i 79 k4i 112 rg5iii 14 bOi 47 f8iv 80 k4i i i 113 rg5v 15 bOv 48 f8v 81 k4v 114 rkOiii 16 b l 2 i i i 49 gOi 82 k5i i i 115 rkOv 17 b l i 50 gOiii 83 k5v 116 r k l i i i 18 b l v 51 gOiv 84 k7v 117 rk2i i i 19 b2ii 52 gOv 85 mOiii 118 rk3i i i 20 b2iv 53 g2i 86 mOv 119 rk4i i i 21 b3i 54 g2iv 87 mlOi i i 120 rk5i i i 22 b3ii i 55 g2v 88 m l i i i 121 wf5v 23 b3v 56 g5i 89 m l v 122 wf8v Table 5.2: Continue. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 102 N U M . T Y P E N U M . T Y P E N U M . T Y P E N U M . T Y P E 24 b57v 57 g5ii 90 m2.5v 123 wgOv 25 b5i 58 g5iii 91 m2i 124 wg5iii 26 b5ii 59 g5iv 92 m2i i i 125 wg5v 27 b5ii i 60 g5v 93 m2v 126 wg8iii 28 b6iv 61 g8i 94 m3ii 127 wkOiii 29 b8i 62 g8iii 95 m3i i i 128 w k l i i i 30 b8v 63 g8iv 96 m3v 129 wk2i i i 31 b9ii i 64 g8v 97 m4i i i 130 wk3i i i 32 b9v 65 k O l i i 98 m4v 131 wk4i i i 33 f02iv 66 kOiii 99 m5i i i Table 5.2: Conversion table between stellar type and number. 0 B A F G K M Giants and Supergiants -0.3 0.2 0.6 -0.6 M a i n Sequence -3.6 0.0 0.7 1.4 1.8 2.0 2.8 Table 5.3: Logarithmic star counts [# / 10 4pc 3] by stellar type and luminos-ity class from data published in Al l en (1973) [1]. galaxies of different types as shown in Table 4.5. Each histogram bin in F i g . 5.2 includes galaxies of the same type in-tegrated over al l redshifts (the redshift ranges from 0 to 1.2, wi th step 5z — 0.01. There are three pronounced peaks which correspond to SO and Sc galaxies, and starburst galaxies. A s explained later, the fact to have many CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 103 0 B A F G K M Giants and Supergiants 1.9 3 2.9 2.1 M a i n Sequence 1.0 0.0 0.64 2.2 2.8 2.9 2.6 Table 5.4: Logarithmic star counts [# / 10 4pc 3] for candidates found in the L M T survey. N U M . T Y P E N U M . T Y P E N U M . T Y P E 1 Bulge 6 Sc 11 S B 5 2 E 7 S B 1 12 S B 6 3 SO 8 S B 2 13 U V h o t 4 Sa 9 S B 3 5 Sb 10 SB 4 Table 5.5: Conversion table between galaxy type and number. spirals is reasonable i f compared wi th other works on galactic types [6]. It is useful to check the galaxy distribution with that found in the C f A l - 2 Redshift Surveys [50]. The C f A l sample includes 2397 galaxies (ellipticals, SO, Sa, Sb, Sc, Sd and Sm spirals, Irregular galaxies) distributed over 2.7 steradians; the CfA2 survey has 1862 galaxies over 0.42 steradians. Table 5.6 shows the galaxy percentage distributed according to the morphological type for C f A l - 2 and L M T surveys. Discrepancies between the C f A l and CfA2 surveys are due to the fact they sample different structures: the mor-phological mix varies with local density [19]. For the U N M S survey, the fraction of Sc galaxies does not include Sd spirals. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 104 1500 ~i 1 r-1000 o o o 500 _l I I l_ _l I I [_ 10 galactic type Figure 5.2: Galaxy distribution by the morphological type. Considering only common galactic types, what appears evident in each survey is the low fraction of ellipticals: at this point, this result seems rea-sonable. The major discrepancy is the smaller number of spiral candidates found CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 105 C f A l C fA2 L M T % % % bulge 4.1 E 11.1 5.3 4.7 SO 25.3 34.4 11 Sa+sb 33.2 40.3 17.3 Sc+Sd 22.1 12.9 12.6 Sm+Im 25.3 5.1 SB1-6 38.6 U V h o t 6.1 Table 5.6: Fraction of ellipticals (E), SO galaxies, Sa+Sb, Sc+Sd and Sm spi-rals, Irregular (Im) galaxies, Starburst (SB1-6) and U V h o t galax-ies in the C f A l , CfA2 and L M T galaxy samples. in the L M T sample: this spiral fraction is not in fact completely consistent with the luminosity function of Binggeli, Sandage and Tammann [5] where, considering the galaxy distribution in the local field, the fraction of spirals wi th respect to ellipticals seems higher than that obtained using the method described in §4. Among galaxy candidates, there is a big number of starburts with re-spect to normal galaxies. This th i rd result doesn't agree wi th the luminosity function of Sanders & Mirabe l (Fig. 1) [2]: for bolometric luminosity lower than 1 O 1 1 L 0 , the luminosity function of normal galaxies is well above that of starbusts; for higher luminosities, it seems there are more irregular galaxies than normal galaxies but the trend of starburst luminosity function is not CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 106 1000 CO E-CJ Cd 3 O 500 I ' ' ' I I 1 1 1 I 1 1 1 I 1 BULGE I SPIRALS II 0 0.2 0.4 0.6 0.8 1 1.2 1000 H 500 h ' 1 I 1 ' 1 I 1 1 1 I 1 1 1 I 1 1 1 I 1 1 1 J| STARBURSTS _ L I i i i I t i i I i i i I 0 0.2 0.4 0.6 0.8 1 1.2 1000 c o E-U H CQ O 500 h _ L ' I 1 1 1 I 1 1 1 I 1 1 1 I 1 1 1 J| ELLIPTICALS _L 0 0.2 0.4 0.6 0.8 Redshift Z 1 1.2 1000 h 500 h 0 0.2 0.4 0.6 0.8 Redshift Z 1 1.2 Figure 5.3: Distr ibut ion of galaxy counts with spectral type and redshift. well defined. Since starbursts are characterized by a great stellar activity and their spectrum show emission lines, it is possible that some stars were misidentified as starburts. F i g . 5.3 shows the galaxy distribution in redshift according to the galactic CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 107 type. El l ipt icals are more concentrated at low redshifts: this is a selection effect due to the difficulty to observe non-emission like objects at high red-shifts. Spirals are mainly found in the range z = 0-0.6: this is consistent with the fact that, though dwarf spheroidal ellipticals are the closest galaxies to the M i l k y Way, they are not as easily detected as spirals, the most luminous field galaxies close to our Galaxy. Starburst candidates are detected at al l redshifts: although they are mainly concentrated in the range z = 0 - 0.3, there are pronounced peaks at z ~ 0.5 and z ~ 0.9. Many of the features such as tails and ringlike struc-tures visible in starburst pictures can be convincingly explained in terms of colliding and/or merging galaxies [67]. The fact that galaxies need time to form and, first of al l , to come so close to the point to interact, could explain the great number of starburts found at low redshift. Another explanation for the peak in the range z = 0 - 0.3 could be that some of these local starburts are misidentified stars. 5.2 Properties of quasar candidates. 5.2.1 The number - redshift distribution. F i g . 5.4 shows the number of quasar candidates in each redshift range 5z = 0.3, from z = 0.3 to z = 7.0. Most sources fall in two regions: 0.5 - 1.5 and 3.5 - 5.5. It is useful to check this result with other redshift distributions such as the Hewett & Burbidge catalog (7110 quasars) [35] and the Sloan Sky CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 108 400 Redshift z Figure 5.4: Redshift distribution of quasar candidates. Survey catalog (3814 quasars) [66] (see F i g . 5.5). In the first work, quasars are mostly concentrated in the range 0 - 2 . 3 wi th a ta i l extending up to z = 4. The second study shows many quasars in the range 0.25 - 2 but there are s t i l l objects at z = 5. In this last case, the peak at z ~ 0.25 is caused by CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 109 Figure 5.5: The redshift distribution for quasars in the Hewett&Burbidge catalog (top) and in the Sloan Sky Survey catalog (bottom). CHAPTER 5. QUASARS IN THE UNMSl CATALOG. . 110 Seyfert galaxies. Comparing a l l these distributions, two discrepancies can be noted: the L M T sample shows a dir th of detections for z ~ 2 and an unusual number of objects at redshift z > 4. The L y a emission line could explain both these two effects. A t z ~ 2 the Lya is not visible in the optical part and this, besides having an incomplete S E D , could contribute to the misclassification of some quasars as stars: this could be the reason behind the first discrepancy. The second effect could be explained in the following way: the L y a emission line enters into the optical spectrum (e.i. it can be observed in quasar templates at wavelenghts A > 4500 A ) at redshift higher than 3. If a galaxy spectrum or, more likely a stellar spectrum, shows a strong line but it misses other important flux measurements, this source can be misclassified as quasar. This misclassification was in part expected: the method is supposed to separate quasars from galaxies and stars with an accuracy of 85%. Since a lot more than 15% quasars seem misidentified, it is then necessary to analyse more carefully the quasar sample in order to catch unlikely quasar candidates. 5.2.2 The number - apparent V magnitude distribution. F i g . 5.6 shows the object distribution according to the apparent V magnitude: quasar candidates are peaked around magnitude 19, with a long and thick ta i l extending to magnitude 12. Even though the brighest quasar (3C 273, h t tp : / /www. seds.org/~spider/spider/Misc/3c273.! has a visual brightness 12.8, it is unlikely that al l the objects brighter than magnitude 16 are correctly classified. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. I l l Figure 5.6: The V magnitude distribution for quasar candidates. The magnitude distribution is compared to that of Hewett & Burbidge (Fig. 5.7): both distributions show a pronounced peak at magnitude 19 and the fact our sample misses many faint objects at V > 20 could be due to the lower l imi t ing magnitude reached by the N O D O - L M T (21 mag) and also CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 112 1500 i 1 1 1 1 1 1 1 [ 1 1 1 r V (magnitudes) Figure 5.7: The V magnitude distribution from the Hewett&Burbidge cata-log. to the smaller number of quasar candidates counted in the U N M S l sample (30% of that of Hewett & Burbidge). The real discrepancy between these two distributions is the bigger number of bright sources in the L M T sample: it is possible that many quasar candidates with magnitude brighter than 16 are misclassified stars. 5.2.3 The number - spectral index a distribution. The continuum part of a quasar spectrum can be modeled by a simple power law Fv oc ua, where F„ is the flux in frequency and a is the spectral index. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 113 To determine the continuum, one must exclude emission lines. This is done in steps. First , the S E D is fitted with a power law considering al l magnitudes. Second, al l the points for which \Pobs,v -PuQ\ >25Fobs,v (5.1) are rejected as possible lines. Here FobStV and 6Fobs>I/ are the observed flux and its error at frequency u, a and (3 are the characteristic parameters of the best power law fit. The third and final step is to fit a second time the S E D with the restricted set of magnitudes. F i g . 5.8 shows the object distribution according to the spectral index a. A prominent peak at a ~ -1 and a second peak for a ~ -2 are the most interesting characteristics. F i g . 5.9 shows spectral index distributions by Warren et al . [77] and by Francis et al . [24]. The first study is a multicolour survey for high redshift quasars and it includes 144 sources; the second work uses quasars taken from the Large Bright Quasar Survey to create a high signal-to-noise composite spectrum (718 sources). Bo th of them show a peak around a ~ -0.5, a bit shifted from the peak at a ~ -1 observed in this survey. The main difference between the U N M S a distribution and those found in these works is the second peak, centered at a ~ -2 and extending up to a ~ -5. Since typical values of a cited in literature (§4.3.3) are in the range [-2.5, 1], it is likely that some misclassified stars are included in this second peak, but it is also possible that the elimination of some points (no more than 9 points according to the criterion used) in spectra has contributed to make quasar SEDs steeper if the flux measurements eliminated are at the edges of the optical range. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 114 spectral index Figure 5.8: Spectral index distribution for quasar candidates in the U N M S l catalog. 5.3 Discussion. In §4 we found that it is possible to identify quasars wi th a completeness of 85%. Yet, result diagrams, shown in the previous section, are not fully CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 115 Figure 5.9: The spectral index distribution by Warren et al . (top picture, upper line) and by Francis et al . (bottom). CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 116 consistent with the results obtained from other surveys. Therefore, objects outside expected ranges of redshift [0.5, 4], magnitude [16, 23] and spectral index [-3, 1] were analysed more completely. To do that, three more diagrams were considered: object distribution in the redshift - V magnitude plane (Fig. 5.10), in the redshift - spectral index plane (Fig. 5.11) and in the spectral index - V magnitude plane (Fig. 5.12). F i g . 5.10 shows that, faint sources are equally distributed at high and low redshifts, but quasar candidates brighter than the 16th magnitude are mostly localized in the redshift range 3-6. This last result is not acceptable because those quasars would have to have exceptionally high luminosities. The second diagram (Fig. 5.11) shows that sources are mainly grouped around a ~ -1. Most of those having spectral index lower than -3, have redshift higher than 4. They represent just a small group (78 sources) and neglecting them would have little effect. The next diagram (Fig. 5.12) shows the source distribution in the spectral index - V magnitude plane: only few of the bright objects have unusual spectral indexes lower than -3. These considerations suggest that, in order to reduce misclassification, criteria should be based on the first diagram (Fig. 5.10). Though we wish to eliminate objects that satisfy one or more of the following criteria: • V mag < 16 and z > 3.5; • V mag < 16 and a < -3; • z > 3.5 and a < -3. The fact that we chose a = -3 as the minimum value for the spectral index CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 117 Figure 5.10: Quasar candidates in the redshift - V magnitude plane. instead of -2.5, cited in §5.2.3, is justified by the desire to retain as good quasar candidates as possible. Each object has a min imim number of 30 flux measurements over a total of 39: i f an observed S E D shows the flux of an emission line and no flux in the filters right beside it, the continuum part of CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 118 Figure 5.11: Quasar candidates in the redshift - spectral index plane, the spectrum would appear steeper (a < -2.5). CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 119 2 0 - 2 - 4 -6 spectral index Figure 5.12: Quasar candidates in the spectral index - V magnitude plane. 5.4 Quasar candidates. Weeding out misclassified quasar candidates which satisfy at least one of the criteria mentioned above, the final list now includes 2294 objects, 6% of the entire group of sources analysed with this method. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 120 According to the surface densities mentioned in §4, the distribution of stars, galaxies and quasars should be respectively 49.5%, 49.5% and 1% of the object sample: it is clear that the number of quasars found is s t i l l higher than 400 Redshift z Figure 5.13: The number - redshift distribution for the "cleaned" set of quasars. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 121 Figure 5.14: The number - V magnitude distribution for the "cleaned" set of quasars. expected and it is likely this list s t i l l includes misclassified objects. Diagrams in F i g . 5.13, 5.14 and 5.15 show the distribution in redshift, magnitude and spectral index using the new quasar candidate sample: these plots are now CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 122 0 -2 spectral index Figure 5.15: The number - a distribution for the "cleaned" set of quasars. more consistent wi th the knowledge that we have about quasars and no other criteria seem possible to further reduce the quasar candidate list. CHAPTER 5. QUASARS IN THE UNMSl CATALOG. 123 5.5 Identification. As another way to test the reliability of quasar candidates, the U N M S l quasar catalog was cross-correlated wi th Veron's database [75], which, re-cently updated after the first release of the 2dF quasar catalogue (Croom et al.), now includes 23760 quasars, 608 B L Lac objects and 5751 active galaxies (of which 2765 are Seyfertls). In the region observed by the L M T Survey (12h <a< 1SH, 32.5° < 5 < 33.5°), Veron's catalog has 100 quasars among which 1 is in common wi th the quasar candidate catalog found in this thesis. Since the unavailability of optical SEDs , the cross-correlation was done using the object position and visual brightness. Table 5.7 shows coordinates, V magnitude and redshift of this object. V E R O N U N M S l N A M E R X J1730.0+3301 R A 17:30:4 17:30:4 D E C 33:01:02 33:01:05 V 17.9 16.43 z 0.632 0.62 Table 5.7: Equatorial coordinates ( R A , D E C ) , apparent V magnitude and redshift z for the quasar common to Veron's catalog and U N M S l quasar catalog. CHAPTER 6. SUMMARY AND CONCLUSIONS. 124 C H A P T E R 6 S U M M A R Y A N D CONCLUSIONS. This thesis has attempted to develop a method to identify quasars from stars and galaxies included in the U N M S l catalog, compiled from observa-tions taken in 1996-1997 and 1999 with the N O D O Liqu id Mi r ro r Telescope. The method proposed in this thesis is based on a x2 minimizat ion pro-cedure combined with a Bayesian approach. The x2 parameter (eq. 4.12), between the theoretical and observed S E D , estimates how well a S E D model fits the data. The odds ratio (eq. 4.15), from Bayes' theorem, includes other important considerations: it accounts for the number of models, which differs from category to category (131 stellar templates, 1573 galaxy templates and 1344 quasar templates), and our prior knowledge of the number of quasars to be expected in 1 degree square at a particular magnitude (surface densities are generally obtained from other surveys). The method was first tested on templates: each was treated as test object, modified by adding noise, resampling the flux measurements in a jackknife approach and analysed in terms of x2 and odds ratio. It was found that 1082 template-objects of 3048 are misclassified: using the notation of Table 4.7, type-I quasars are confused with UVho t , elliptical and spiral galaxies, type-II quasars are mainly classified as starburst and late spiral galaxies, and G and F stellar types, type-I l l quasars are mostly seen as F type stars and type-IV CHAPTER 6. SUMMARY AND CONCLUSIONS. 125 quasars as B type stars. A n important step was to realize how important the number of reliable magnitudes is on classifying objects correctly : in the U N M S l catalog most of the sources have 10 - 20 niters and only 35% of the entire database have more than 35 flux measurements. If important spectral features are missed, it is easy to misclassify objects. The final test was to analyse method's results when more and more filters are randomly neglected. From the distribution of stars, galaxies and quasars (Fig. 4.16), two features appear: the number of stars increases and the number of quasars decreases when more and more filters are randomly neglected. However, the galaxy distribution is almost flat. This situation and further studies described in §4 leads us the conclu-sion that, in order to classify galaxies properly, wi th an accuracy of 85% or higher, 20 flux measurements or more are required. To classify stars and quasars, it is necessary to have 30 or more flux measurements. The method, applied to the U N M S l sources satisfying the above criterion and more separated by more than 3 arcsec, produced the results summarized in Table 6.1: Category Object N U M . S T A R 25193 G A L A X Y 10791 Q U A S A R S 3056 T O T . 39040 Table 6.1: Distr ibut ion of star, galaxy and quasar candidates. Comparing surface density values cited in §4 with these results, there are CHAPTER 6. SUMMARY AND CONCLUSIONS. 126 more quasars than what are expected. Useful results are obtained from the quasar candidate distribution in red-shift, magnitude and spectral index: there are many objects brighter than magnitude 16, localized at redshift higher than 4, and wi th the spectral index a lower than -3.5. Considering that bright objects cannot be too distant (otherwise they would emit an extremely large amount of energy) and that, generally, quasar spectra can be well fitted by a power law with spectral index in the range [-2.5, 1], sources satisfying the following criteria are not considered reasonable quasar candidates: • V mag < 16 and z > 3.5; • V mag < 16 and a < -3; • z > 3.5 and a < -3. After this "cleaning procedure", the new quasar sample includes 2294 objects: it is likely there are st i l l misclassified quasars, but we cannot identify any new selection criteria to further refine our catalog. From the cross-correlation of our quasar list wi th Veron's quasar catalog, which has 100 objects in the sky region observed with the N O D O - L M T , 1 object was identified; basic parameters are shown in Table 5.7. BIBLIOGRAPHY 127 B I B L I O G R A P H Y [I] Al l en , C . W , 1973, Astrophysical Quantities, 3rd ed. Athlone Press, London. [2] Bachiller, R . et al . , 1996, Annual Review in Astronomy and Astrophysics, V o l . 34. [3] Beauchemin, M . , 1985, M.Sc . Thesis, Laval University. [4] Bevington,P. R. , Robinson, D . K . , 1992, Data reduction and error anal-ysis for the physical sciences, ed. S. J . Tubb and J . M . Morr is , U S A . [5] Binggeli , B . , Sandage, A . , Tammann, G . A . , 1988, Annu . Rev. Astron. Astrophys., 26, 509. [6] Binney, J . , Merrifield, M . , 1998, Galactic Astronomy, ed. Princeton University Press, Princeton, New Jersey. [7] Borra , E . F . , astro-ph/950319. [8] Borra , E . F . , Moretto, G . , Wang, M . , 1995, A & A , 109, 563. [9] Borra , E . F . , astro-ph/9706098. [10] Borra, E . F . , Ritcey, A . , Ar t igau, E . , astro-ph/9901295. [II] Boyle, B . J . , Fong, R. , Shanks, T. , Peterson, B . A . , 1990, M N R A S , 243, I B . BIBLIOGRAPHY 128 [12] Boyle, B . J . , Shanks, T. , Jones, L . R . , 1991, M N R A S , 251, 482B. [13] Burbidge, Geoffrey & Margaret 1967, Quasi-Stellar Object, ed. W.H.Freeman and Company, San Francisco. [14] Cabanac, R . A . , astro-ph/9706156. [15] Cabanac, R . A . , de Lapparent, V . , Hickson, P., astro-ph/0206062 [16] Carrol l , B . W , Ostlie, D . A . , 1996, An Introduction to Modern Astro-physics, ed. Addison-Wesley Publishing Company, United States of America . [17] Cris t iani , S., Barbieri , C , Iovino, A . , La-Franca, F . , 1989, A & A , 177, 161C. [18] Cris t iani , S. et a l , 1995, A & A , 112, 347C. [19] Dressier, A . , 1980, A p J , 236, 51. [20] Fairal l , A . , 1998. Large-scale structures in the Universe, ed.John Wi ley &; Sons, New York. [21] Fan, X . et al . , 2001, A J , 122, 2833F. [22] Fanti , Car la & Roberto, 2000, Lezioni di Radioastronomia, Dipar t i -mento di Astronomia, Bologna. [23] Foster, K . , Green, P . J . , k Aldcroft, T . L . , Vestergaard, M . , Foltz, C . B . , Hewett, P . C . , 2001, A p J S , 134, 35F. BIBLIOGRAPHY 129 [24] Francis, P . J . , Hewett, P . C . , Foltz, C . B . , Chaffee, F . H . , Weymann, R . J . , Morris , S.J., 1991, A p J , 373, 465F. [25 [26 [27; [28 [29 [30 [31 [32 [33 [34; [35 [36 Gibson, B . K . , 1991, J R A S C , 85, 158. Gi lmozz i , R. , Science with 100 m telescope, 2000, S P I E 4005. Goldschmidt, P. et al . , 1991, The Space Distribution of Quasars, proc. 93G. Grazian, A . , Cris t iani , S., D'Odorico, V . , Omizzolo, A . , Pizzel la , A . , 2000, 119, 2540-2555. Grebel, E . K . , astro-ph/0105138. Gregory, P . C . , 2002, Statistical Inference in Science: Bayesian and Frequentist Perspectives, P H Y S 5 0 9 notes. Groote, D . , Reimers, D . , Wohler, T. , Wisotzki , L . , 1996, A & A , 115, 227W. Hagen, H - J . , Engels, D . , Reimers, D . , 1999, A & A , 134, 483. Hatziminaoglou, E . , Mathez, G . , Pello ' , R. , 2000, A & A , 359, 9H. Hewett, P . C . , Foltz, C B , Chaffee, F . H . , 1995, A J , 109, 1498H. Hewitt , A . , Burbidge, G . , 1993, A p J S , 87, 451H. Hickson, P., Gibson, B . K . , Borra, E . F . , Content, R. , 1990, project description. BIBLIOGRAPHY 130 [37] Hickson, P., Walker, G . A . H . , Borra , E . F . , Cabanac, R . A . , 1994, S P I E , 2199, 922. [38] Hickson, P., Borra , E . F . , Cabanac, R . A . , Content, R. , Gibson, B . K . , Walker, G . A . H . , 1994, A p J , 436, 201H. [39] Hickson, P., Mulrooney, M . K . , 1998, A p J S , 115, 35. [40] Hickson, P., Richardson, E . H . , 1998, P A S P , 110, 108 . [41] Hickson, P., Mulrooney, M . K . , 1998, A p J , 506, 191. [42] Hickson, P., 2002, M N R A S , 330, 540H. [43] Jenkner, H . , Lasker, B . M , Sturch, C . R . , McLean , B . J . , Shara, M . M . , Russel, J . L . , 1990, A J , 99, 2082. [44] Kinney, A . L . , Calzet t i , D . , Bohl in , R . C . , McQuade, K . , Storchi-Bergmann, T. , Schmitt, H .R . , 1996, A p J , 467, 38K. [45] La-Franca, F . , Cris t iani , S., Barbieri , C , 1992, A J , 103, 1062L. [46] La-Franca, F . , Cris t iani , S., 1997, A J , 113, 1517L. [47] Lasker, B . M . et al . 2000, 119, 2540-2555. [48] Lasker, B . M . , Sturch, C .R . , McLean , B . J . , Russel, J . L . , Jenkner, H . , Shara M . M . , 1990, A J , 99, 2019. [49] Malkan, M . A . , Sargent, W . L . W . , 1982, A p J , 254, 22M. [50] Marzke, R. , Geller, M . , Huchra, J.P., Corwin , H . G . J . , 1994, A p J , 108, 4 3 7 M M . BIBLIOGRAPHY 131 [51] McGraw, J .T. , Angel , J .R .P . , Sargent T . A . , 1980, In Application of Dig-ital Image Processing to Astronomy, ed. El io t D . A . (Bellingham: SPIE) , 20. [52 [53 [54 [55 [56 [57; [58 [59 [60 [61 [62 [63 [64; [65 [66 Mitchel l , K . J . , Warnoch, A . , Usher, P .D . , 1984, A p J , 287L, 3 M . Mitsouras, D . , Brandenberger, R. , Hickson, P., astro-ph/9806360. Monet, D . , 1996, A & A , 188, 5404. Monet, D . et a l , 1997, U S N O l , C , 0 M . Pickles, A . J . , 1998, A J , P A P S 110, 863. Potter, A . E . , Mulrooney, M . , 1997, A d S p R , 19, 213P. Richstone, D . O . , Schmidt, M . , 1980, A p J , 235, 361R. Rocca-Volmerange, B . , Guiderdoni, B . , 1988, A & A S , 75, 93R. Russel, J . L , Lasker, B . M . , McLean , B . J . , Sturch, C . R . , Jenkner, H . , 1990, A J , 99, 2059. Sandage, A . R . , Veron, P., 1965, A p J , 142, 412. Schmidt, M . , 1963, Nature, 197, 1040. Schmidt, M . , Green, R. , 1983, A J , 269, 352. Scholz, R . D . , Meusinger, H . , Irwin, M . , 1997, A & A , 315, 417S. Schneider, D .P . , Schmidt, M . , Gunn J . E . , 1994, A J , 107, 1245S. Schneider, D .P . et al . , 2002, A J , 123, 567. BIBLIOGRAPHY 132 [67] Schweizer, F . , 1990, Dynamics and Interactions of Galaxies, ed. Wielen, R. , (Berlin: Springer), p.60, Observational phenomena associated wi th mergers. [68] Shanks, T . et a l , astro-ph/0003206. [69] Shields, G . A . , 1978, Nat, 272, 706. [70] Skey, H . , 1872, Trans. New Zealand. Inst., 5, 119. [71] Storrie-Lombardi, L . J . , McMahon , R . G . , Irwin, M . J . , Hazard, C . , 1996, A p J , 468, 121S. [72] Storrie-Lombardi, L . J . , Irwin, M . J . , M c M a h o n , R . G . , Hook I . M . , astro-ph/0012446. [73] Unsold, A . , Baschek, B . , 1991, The New Cosmos., 4th ed. Springer-Verlag, Ber l in . [74] Vanden-Berk, D . E . et a l , 2001, A J , 122, 549V. [75] Veron-Cetty, M . P . , Veron, P., 2001, A & A , 374, 92V. [76] Voges, W . , 1992, Environment Observation and Climate Modelling Through International Space Projects: Space Sciences with Particular Emphasis on High Energy Astrophysics., (Noordwijk:ESA), 9-19(RASS). [77] Warren, S.J., Hewett, P . C . , Osmer, P.S., 1991, A p J S , 76, 23W. [78] Zheng, W . , Kriss , Z . G . A . , Telfer, R . C . , Grimes, J.P., Davidsen, A . F . , 1997, A p J , 475, 469Z. BIBLIOGRAPHY 133 [79] Zheng, W . , astro-ph/0005247. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0085131/manifest

Comment

Related Items