UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Submillimetre observations of the Subaru Deep Field Coppin, Kristen Erin Kathryn 2003

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

Item Metadata

Download

Media
831-ubc_2003-0542.pdf [ 6.17MB ]
Metadata
JSON: 831-1.0091401.json
JSON-LD: 831-1.0091401-ld.json
RDF/XML (Pretty): 831-1.0091401-rdf.xml
RDF/JSON: 831-1.0091401-rdf.json
Turtle: 831-1.0091401-turtle.txt
N-Triples: 831-1.0091401-rdf-ntriples.txt
Original Record: 831-1.0091401-source.json
Full Text
831-1.0091401-fulltext.txt
Citation
831-1.0091401.ris

Full Text

Submillimetre Observations of the Subaru Deep Field by Kristen Erin Kathryn Coppin B.Sc. (Physics and Astronomy) University of V i c t o r i a , 2000 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L M E N T O F THE REQUIREMENTS FORTHEDEGREE OF MASTER OF SCIENCE in T H E F A C U L T Y O F G R A D U A T E STUDIES (Department of Physics and Astronomy)  We accept this thesis as conforming to the require^ standard  T H E UNIVERSITY O F BRITISH C O L U M B I A October 10, 2003 © K r i s t e n E r i n K a t h r y n C o p p i n , 2003  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of B r i t i s h C o l u m b i a , I agree that the L i b r a r y 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  T h e University O f B r i t i s h C o l u m b i a Vancouver, C a n a d a  11  ABSTRACT We have measured the submillimetre wavelength continuum emission from the Subaru Deep F i e l d ( S D F ) at 450 and 850 (j,m w i t h the Submillimetre C o m m o n - U s e r Bolometer A r r a y ( S C U B A ) detector on the James Clerk M a x w e l l Telescope ( J C M T ) . T h e S D F is as deep a near-IR image as is available and contains four 'hyper extremely red objects' ( H E R O s ) . These d a t a have allowed us to test the connection between 'extremely red objects' ( E R O s ) found in I R surveys and the population of bright submillimetre sources found w i t h S C U B A . T h i s is an important building block in our understanding of the star formation history of the universe. We examine the entire submillimetre m a p of the S D F region and perform correlation analyses of a /^'-selected catalogue of galaxies i n the S D F w i t h our submillimetre S D F map. We find that there is no clear correlation between the near-Infrared positions and submillimetre flux. We also present upper limits to the fluxes of the four H E R O s , or a weak measurement of the average flux for the four of them. O u r data are consistent w i t h the H E R O s being Arp220-like objects out to z ~ 1.6. However, they would need to be one and a half times as luminous as Arp220 and extincted in the J - b a n d by about 1 magnitude; a plausible scenario since only modest adjustments are made to Arp220's luminosity and dust content.  iii  CONTENTS Abstract  ii  Contents  iii v  List of Tables  vi  List of Figures .  1  1 Introduction  2  1.1  G a l a x y Formation.and Evolution  1  1.2  Submillimetre Galaxies  2  1.2.1  T h e O r i g i n of Submillimetre Emission  2  1.2.2  Inhabitants of the High-Redshift Universe  3  1.2.3  Submillimetre Emission as a High-Redshift P r o b e  3  1.2.4  EROS  7  1.2.5  Hyper-Extremely R e d Objects  8  1.3  T h e Subaru Deep F i e l d  10  1.4  G u i d e to T h i s Thesis  15  The Instrument and Data  17  2.1  T h e J C M T and S C U B A  17  2.1.1  18  2.2  2.3  Observing Modes  Observations  21  2.2.1  24  A d d e d Noise i n the 2003 D a t a  D a t a Reduction 2.3.1  P r e l i m i n a r y Reduction w i t h S U R F  30 .  30  iv  3  4  5  2.3.2  M a k i n g the M a p s  31  2.3.3  Calibration  34  2.3.4  F l u x Measurements  35  HEROs  40  3.1  Cirrus Contribution  40  3.2  H E R O I R Colours and 850 / m i Fluxes  41  3.3  U s i n g Arp220 as a Template  41  3.4  U s i n g a N G C 3 9 3 8 as a Template  51  3.5  Results  51  3.6  A r e the Models Feasible?  51  3.6.1  Number Counts of Submillimetre Sources  54  3.6.2  C o u l d the H E R O s be Starburst Galaxies?  54  3.6.3  C o u l d the H E R O s be N o r m a l Spiral Galaxies?  61  S D F Sources  68  4.1  Correlating the I R Galaxies w i t h the Submillimetre D a t a  68  4.1.1  91  Detections  Conclusions  96  5.1  98  Future Work  Bibliography  99  V  LIST O F T A B L E S 1.1  S u m m a r y of I R information on the H E R O s  13  2.1  Summary of S C U B A observations  23  2.2  Summary of F l u x Conversion Factors  36  2.3  Summary of S C U B A data for the H E R O s w i t h upper l i m i t flux estimates and photometric redshifts  39  3.1  Submillimetre sources w i t h red I R colours and known redshifts  49  3.2  S u m m a r y of S C U B A data for the H E R O s w i t h upper l i m i t flux estimates  4.1  and photometric redshifts  53  Detected sources  95  vi  LIST O F F I G U R E S 1.1  T h e extragalactic background  4  1.2  Negative K-correction  6  1.3  I — KAB versus KAB  9  1.4  T h e Subaru Deep F i e l d i f ' - b a n d image  11  1.5  J and i f ' - b a n d images of the H E R O s  12  1.6  T h e 850 pm cumulative source counts  16  2.1  S C U B A wideband filter profiles  19  2.2  T h e S C U B A bolometer arrays  20  2.3  F F T s of jiggle map data  2.4  F F T of the average of 2003 jiggle map data  2.5  Standard deviation of 10 integrations for 64 positions of the central bolometer 28  2.6  Contour plots of the standard deviation of 10 integrations for 64 positions  .  25 27  of the central bolometer  29  2.7  C h o p p i n g strategy for photometry observations of the S D F  33  2.8  850 am S D F S N R map  37  2.9  450 pm S D F S N R map . .  38  3.1  Is there a colour-flux correlation for the H E R O s ?  42  3.2  S E D of Arp220  44  3.3  Comparison of 3 different luminosity S E D s  45  3.4  Effective wavelengths and relative extinction values from Schlegel, Finkbeiner & Davis [53]  46  3.5  Creating an extinction law  47  3.6  Testing our model against sources w i t h known redshifts  50  vii 3.7  SEDofNGC3938  52  3.8  Expected 850 / i m versus redshift using Arp220  55  3.9  /(''-band brightness versus redshift using Arp220  56  3.10 Required J a n d I f ' - b a n d dust extinction using Arp220  58  3.11 T h e T o t a n i et al. [63] model of a typical elliptical galaxy  60  3.12 M o d e l of H R 1 0 based on an Arp220 S E D  62  3.13 Expected 850 /tin versus redshift using N G C 3 9 3 8  63  3.14 i T - b a n d brightness versus redshift using N G C 3 9 3 8  64  3.15 Required J and i f ' - b a n d dust extinction using N G C 3 9 3 8  . .  4.1  Histogram of number of pixels at each flux level for the 8 5 0 / i m map . . .  4.2  T h e difference of the flux bins and the flux bins reflected about 0 m J y for the 850 (ira. map  4.3  Histogram of number of pixels at each flux level for the 450 /tm map . . .  4.4  T h e difference of the flux bins and the flux bins reflected about 0 m J y for the 450 / i m map  65 69  70 71  72  4.5  Histogram of the mean of the 20 brightest i f ' - b a n d objects per redshift b i n 74  4.6  Histogram of observed average 850 /tin flux per redshift bin  75  4.7  Histogram of observed average 450/tm flux per redshift bin  76  4.8  D i s t r i b u t i o n of correlation coefficients obtained from a set of M o n t e C a r l o simulations of stacked 850 /zm flux in redshift bins  4.9  78  Frequency of obtaining a certain correlation coefficient from a set of Monte C a r l o simulations of stacked 850 /tm flux in redshift bins  79  4.10 D i s t r i b u t i o n of correlation coefficients obtained from a set of M o n t e C a r l o simulations of stacked 450 /tm flux in redshift bins  80  4.11 Frequency of obtaining a certain correlation coefficient from a set of Monte C a r l o simulations of stacked 450 /tm flux in redshift bins  81  4.12 Average 850 /tm flux per redshift bin w i t h an equal number of objects per bin  83  viii 4.13 A plot of observed average 450 am flux per redshift b i n w i t h an equal number of objects per bin  84  4.14 D i s t r i b u t i o n of correlation coefficients obtained from a set of Monte C a r l o simulations of stacked 850 um flux w i t h an equal number of objects per redshift bin  85  4.15 Frequency of obtaining a certain correlation coefficient f r o m a set of M o n t e C a r l o simulations of stacked 850 /um flux w i t h an equal number of objects per redshift bin  86  4.16 D i s t r i b u t i o n of correlation coefficients obtained from a set of Monte C a r l o simulations of stacked 450 /xm flux w i t h an equal number of objects per redshift b i n  87  4.17 Frequency of obtaining a certain correlation coefficient from a set of Monte C a r l o simulations of stacked 450 am flux w i t h an equal number of objects per redshift bin  88  4.18 Stacked 850 jum flux measured in K magnitude bins  89  4.19 Stacked 450 pm flux measured in K magnitude bins  90  4.20 Scatter-plots of 850 am flux for F - b a n d detected objects  92  4.21 Stacked 850 pm flux measured in M  v  bins  93  4.22 Stacked 850 / / m flux measured in M  v  bins for objects w i t h z > 2 and z > 4 94  ix  Acknowledgements I would like to thank my research supervisor, D r . M a r k H a l p e r n , for his expertise, guidance, patience, enthusiasm and financial support through N S E R C . A d d i t i o n a l funding was received from the National Research C o u n c i l of C a n a d a and was used for trips to the J C M T to collect data for this project. M u c h thanks also goes out to the staff of the J C M T for their assistance w i t h the S C U B A observations. T h e James Clerk M a x w e l l Telescope is operated on behalf of the Particle Physics and Astronomy Research C o u n c i l of the U n i t e d K i n g d o m , the Netherlands Organisation for Scientific Research, and the N a t i o n a l R e search C o u n c i l of C a n a d a . T h i s research has made use of the N A S A / I P A C Extragalactic Database ( N E D ) which is operated by the Jet Propulsion Laboratory, C a l i f o r n i a Institute of Technology, under contract w i t h the National Aeronautics and Space A d m i n i s t r a t i o n . We acknowledge the use of N A S A ' s Sky View facility (http://skyview.gsfc.nasa.gov) located at N A S A G o d d a r d Space Flight Center. In addition, G u i l l a i n e Lagache supplied her group's starburst and normal galaxy templates and an unpublished version of a paper describing these galaxy templates which was much appreciated. I would also like to thank A n n a Sajina for help using the templates and for getting me started w i t h I D L . M y gratitude goes out to C o l i n Borys who helped me w i t h using his map-making codes and for generally useful advice. Douglas Scott and James Dunlop also provided many useful suggestions and comments about the work contained i n this thesis.  1  CHAPTER 1 INTRODUCTION 1.1  Galaxy Formation and Evolution  T w o fundamental unsolved problems in modern cosmology are: how and when did  the  first galaxies form; and how long did their initial starbursts last? There are m a i n l y two theories depicting how galaxies may have formed: the monolithic collapse and hierarchical merging scenarios. T h e current consensus is that massive galaxies formed v i a hierarchical structure formation, but there is possible contrary evidence suggesting that a monolithic collapse scenario may still play a role in some instances. Based on the motions of old stars in our Galaxy, Eggen, L y n d e n - B e l l &: Sandage [27] conceived the monolithic collapse scenario as a theory of galaxy formation, involving a single rapid collapse of stars at high redshift, producing a violent and short-duration burst of star-formation followed by a quiescent evolution of the stellar population. T h i s theory was questioned by Searle & Z i n n [54] as new and better data became available and they suggested that star formation i n our G a l a x y was a more prolonged chaotic process. New theories were subsequently developed (see Davis et al. [16]) describing galaxy formation as a "bottom-up" process, i.e. hierarchical merging, or the gradual formation of a galaxy by the merging of smaller collections of stars at moderate redshifts of z < 1.5 producing moderate and continuous star-formation (e.g. W h i t e & Frenk [66], Kauffmann [44]). We see evidence of hierarchical merging as our neighbouring satellite, the Sagittarius dwarf galaxy, is being tidally disrupted by our G a l a x y ' s gravitational field (see Ibata, G i l m o r e &; Irwin [39]).  But  older stellar populations of galactic spheroids seem to be better explained by a monolithic collapse scenario. So we see evidence of both processes throughout the history of the Universe but we do not know how much each process plays into the entire galaxy formation picture.  1.  CHAPTER  INTRODUCTION  2  1.2 Submillimetre Galaxies Very luminous high-redshift galaxies have recently been discovered at submillimetre wavelengths and may be significant players in deducing the history of galaxy formation and evolution. Far-Infrared and submillimetre galaxies are responsible for approximately 2 / 3 of the total power output of all known galaxies and they are excellent tracers of massive regions of star-formation at high redshifts (Gispert, Lagache & Puget [33]). M o s t of the star formation history of the Universe is "hidden" from optical sight, but remains transparent in the Infrared (IR) and submillimetre regimes. Careful submillimetre surveys of distant galaxies should be able to shed some light on the relative importance of each formation process since submillimetre emission can indicate galaxies undergoing hightened rates of star-formation. ( B l a i n et al. [3]).  1.2.1  The Origin of Submillimetre Emission  There are two different mechanisms responsible for submillimetre emission from galaxies: continuum thermal emission and line emission. Supernovae are thought to be the most likely dominant mechanism of dust production at high redshifts (Dunne et a l . [25]). These micron-sized interstellar dust grains, primarily composed of silicates and polycyclic aromatic hydrocarbons ( P A H s ) , absorb h a r d - U V photons from regions of intense high mass star-formation or from A c t i v e Galactic Nuclei ( A G N ) accretion disks and are heated to temperatures between 20-200 K . T h e obscuring dust grains re-radiate this absorbed radiation as thermal continuum emission, peaking i n the Far-Infrared ( F I R ) . T h i s emission becomes visible at submillimetre wavelengths when the cosmological expansion of the Universe redshifts the F I R peak into this regime. T h e other mechanism involves atomic and molecular transitions i n the interstellar gas resulting i n line emission. A p p r o x i m a t e l y 99 per cent of the power output of galaxies at submillimetre and far-IR wavelengths is produced by continuum thermal emission and the rest comes from line emission. ( B l a i n et al. [3]).  1.  CHAPTER  1.2.2  INTRODUCTION  3  Inhabitants of the High-Redshift Universe  In the 1980's, a ~ U y sensitive all-sky survey at 12, 25, 60 and 100 / i m by the Infrared A s t r o n o m i c a l Satellite  (IRAS), revealed a population of previously unseen optically-faint  galaxies out to a redshift of ~ 0.3 and that the amount dust reprocessing increases w i t h star production and cannot be measured solely w i t h optical or U V data (see Sanders & M i r a b e l [52]). After the dominant C M B emission is accounted for, about half of the remaining energy budget of the extragalactic background is contained i n the Far-Infrared Backgound ( F I R B ) (see F i g . 1.1). T h e Infrared Space Observatory (ISO), sensitive only to the low redshift region, resolved about 10 per cent of the F I R B into discrete galaxies lying at redshifts less than about 1 (Dole et a l . [20]). The discovery of a submillimetre extragalactic background by COBE ( C O s m i c Background Explorer) was an amazing find and requires the existence of far-IR  emitting  galaxies i n the early throes of their evolution or elliptical galaxies bursting w i t h star formation.  T h i s submillimetre background implies a strong cosmological evolution, or  that early type galaxies i n the distant Universe must be very different from galaxies i n the local Universe which we know to be predominantly composed of relatively o l d stars. T h i s progression greatly affects the star formation history, as a large part of the stars being formed are most likely hidden i n dust-enshrouded galaxies, eluding optical surveys. What are the objects responsible for producing this background emission and what is their role in galaxy formation? T h e S C U B A camera on the J C M T is able to detect rest-frame far-IR radiation from high-redshift galaxies and has revealed a galaxy population w i t h star-formation rates of l O O M Q y r " , similar to that expected from dusty protogalaxies 1  (Steidel [60]). There are relatively few secure identifications and redshifts unfortunately, but most S C U B A galaxies are thought to be at redshifts greater than 1.  1.2.3  Submillimetre Emission as a High-Redshift Probe  Submillimetre emission may be used as a probe of the high-redshift Universe since i t c a n penetrate dust, and because of another powerful effect called the negative K-correction  CHAPTER 1.  INTRODUCTION  4  energy (keV)  10  3  1  lO" lO" wavelength (mm) 3  6  10"  9  10-  12  Figure 1 . 1 : T h i s figure depicts the energetics of the extragalactic background which is comprised of six m a i n components: the Cosmic Microwave Background ( C M B ) , the Cosmic Infrared Background ( C I B ) or Far-Infrared Background ( F I B ) , the Near-Infrared ( N I R ) , the O p t i c a l - U V , the X - R a y Background ( X R B ) and the G a m m a - R a y Background ( G R B ) . Use of this figure courtesy of Scott et al. [55].  CHAPTER  1.  INTRODUCTION  5  which facilitates submillimetre observations of galaxies out to the distant reaches of the Universe.  T h e negative K-correction is strongest for wavelengths longer t h a n about  250/xm, being the most pronounced at millimetre wavelengths as seen i n F i g . 1.2, and causes a galaxy's flux density to remain about constant as it is moved to higher redshifts (between 1 < z < 10). Resolved submillimetre-wave observations exist only for the lowest redshift galaxies and it is usual practice to use the results of these observations as templates for galaxies of the more distant Universe (which may or may not be a reasonable assumption given the lack of understanding of high-redshift galaxies and of the feedback processes involved in star formation).  IRAS  uncovered a distinct class of extremely luminous objects of  1 0 L , dubbed L u m i n o u s Infrared Galaxies or L I R G s . Not-so-distant cousins to L I R G s , 1 1  o  U l t r a Luminous Infrared Galaxies ( U L I R G s ) are probably more similar to galaxies i n the high-redshift world w i t h respect to their luminosities and physical properties. U L I R G s are a small class of special objects, believed to be interacting or merging pairs of spiral galaxies.  T h e y make up less than 0.1 per cent of galaxies i n the local Universe but  were more important i n the past (Sanders [51]). T h e y represent the most luminous of galaxies i n the local Universe w i t h bolometric luminosities i n excess of 1 0 L © and the 12  dust emission dominated peak of their spectral energy distribution ( S E D ) lies in the F I R region of the spectrum. T h i s peak gets redshifted into the submillimetre regime, m a k i n g submillimetre observations an ideal way to see the highest redshift dusty U L I R G s . U L I R G s have irregular morphologies and are bursting w i t h star-formation, most likely indicating recent galaxy mergers. U L I R G s predominantly display elliptical galaxy surface brightness profiles (e.g. Zheng et al. [69]), hinting that U L I R G formation could be just the way to form elliptical galaxies at high redshifts. Submillimetre-selected galaxies are an important piece of the galaxy formation and evolution puzzle but are difficult to study, as they are usually very faint i n the complementary wavelength regimes. L i m i t e d progress has been made i n determining the relationship of the submillimetre population to other high-redshift galaxy populations. O f all the classes of galaxies, Extremely R e d Objects ( E R O s ) have emerged to be one  1.  CHAPTER  6  INTRODUCTION  1000.00  1  1  1  1  1  1 1  1  1  1 1 |  1  1  1  1  1  1 1 1-  \ 100.00 b  r:  \  \  — \  -  \  \^  \  " N.  £  -  \  -  -  \  -  10.00b  :  c Q  >  \  \  \  \^  - \  '.^  1.00  0) OT  \  \ S s  s.  \  \  ~  \  \  \  \ \  o  :  \  :  \  \ \  \ \  0.10 — :  >v  \ \  \ \  \ \  \  \ \  M i l  1.0  \ \  \\  0.1  : ;  \ \  \  0.01  \ \  \  \  \ \  \  l  i  i  i  i  i iil  10.0  \ \  \  i . i \  i  i  i ii  100.0  Redshift  Figure 1.2: T h i s figure depicts the effect of negative K-correction on the predicted flux density of A r p 2 2 0 , a dusty U L I R G galaxy, as a function of redshift.  Note  the powerful K-correction at submillimetre wavelengths of 850 / i m (solid line) and 4 5 0 / i m (dashed line), where the flux density is almost independent of redshift past z « 1. T h e sudden drop i n flux at high redshift occurs because we begin to sample the S E D on the other side of the thermal F I R peak. The i f - b a n d does not benefit from this effect and is shown for comparison (dot-dashed line).  1.  CHAPTER  INTRODUCTION  7  of the most promising and important tools for studying the history of galaxy formation and star formation i n the universe (Dickinson et a l . [18], Y a h a t a et a l . [68]), since dusty objects at high redshifts are expected to be strongly reddened.  1.2.4  EROS  The recent development of large near-IR imaging cameras led to the discovery of a faint class of extremely red objects, or E R O s .  T h e E R O class contains objects w i t h very  red optical to near-IR colours, typically R — K  > 5 to 6 or J — i f  > 4. Suggested  explanations for the very red spectra include the possibility that they might be dusty galaxies at z < 1, extraordinarily high-redshift 'J-dropout' galaxies, or that they are old elliptical galaxies containing low-mass cool stars seen at z « 1 or 2, passively evolving after their i n i t i a l burst of star formation, and that line emission might be involved. T h e E R O class may very well contain several types of object. Dickinson et a l . [18]. T h i s selection criterion was originally designed to pick out old (> 1 G y r ) passively evolving elliptical galaxies at redshifts between 1 and 2. B u t a small fraction of all E R O s selected this way seem to be very dusty ellipticals undergoing their i n i t i a l burst of star formation.  It is important to find the dominant population amongst the E R O s since  the high-redshift old ellipticals would indicate a "monolithic collapse" scenario of galaxy formation whereas the "hierarchical" scenario would be better supported by evidence of many dusty starbursts living at intermediate redshifts. M a n y of the most luminous submillimetre sources discovered have been identified as convincing counterparts to E R O s .  For example, C U D S S 1 4 A (Gear et a l . [32]), S M M  J00266+1708 (Frayer et al. [31]), W - M M D 1 1 (Chapman et al. [12]), S M M J09429+4658, and S M M J04431+0210 (Smail et al. [58]) have a l l been identified w i t h E R O s . These submillimetre sources have fluxes i n the range of 7 to 20 m J y .  Because of the high  frequency w i t h which very bright submillimetre sources are identified w i t h E R O s , S m a i l et a l . [58] suggest that submillimetre sources comprise the majority of the reddest E R O s . T o t a n i et a l . [63] discovered four objects in the Subaru Deep F i e l d that were even more  CHAPTER  1.  INTRODUCTION  8  red than the reddest known E R O s . So the U B C submillimetre group collaborated w i t h the Subaru team and applied for time to search for these rare objects at submillimetre wavelengths, to test the conjecture that very red objects are submillimetre-bright. T h e search for dusty starbursts at high redshifts is a key step to unraveling the mystery of galaxy formation at early epochs.  1.2.5  Hyper-Extremely Red Objects  A m o n g the E R O s which are faintest i n A'-band, K > 22, an increasing fraction are extraordinarily red, even compared to other E R O s , w i t h J — K > 3 or 4. A ' - b a n d sources which are very faint or undetected i n optical data hold a strong possibility of being S C U B A bright. F i g . 1.3 is the near-IR colour-magnitude diagram for the C F R S (CanadaFrance Redshift Survey) and demonstrates that U L I R G s make up a large fraction of the reddest objects i n this field.  B u t does a redder colour necessarily guarantee that an  object w i l l be bright i n the submillimetre? Tota n i et a l . [63] cleverly dubbed this very distinct population of galaxies HyperExtremely R e d Objects, or H E R O s , and i n detailed modeling find that they are much too red t o be passively evolving elliptical galaxies at z < 2. These objects are most likely to be either very dusty elliptical galaxies which formed at redshift 2 « 4 - 7 and are still undergoing rapid star formation when seen at redshift z > 2, or they are clean, L y m a n break galaxies seen at extraordinary redshifts of z > 10. R e d optical colours sometimes suggest that dust extinction is present, and this is precisely what submillimetre instruments can see. K n o w i n g the submillimetre flux of these H E R O s would distinguish between the galaxy formation models and could provide very valuable information for understanding when galaxies first formed and how long their initial starbursts lasted.  CHAPTER  1.  9  INTRODUCTION  Figure 1.3: T h e near-IR colour-magnitude diagram for the identifications the C F R S 14h and 3h fields (see Clements et al., in preparation).  T h e solid circles corre-  spond to radio-detected objects, and the solid diamonds denote radio nondetections. T h e open circles show the best identification for the remaining objects. For comparison, the magnitudes and colours of all the galaxies i n the C F H T I R C U D S S + 1 4 image are also shown (plus signs). N o t included i n this plot are the possible L B G identifications (whose colors are i n the grz filter system). T h e solid, diagonal black line denotes the I detection l i m i t of the C F R S . T h e dashed lines are tracks of I — KAB, KAB w i t h redshift for sources 14.1, 14.3, 14.18, 3.10, and 3.15. T h e three solid lines show the predicted colors for the three U L I R G s studied by Trentham, K o r m e n d y & Sanders [64]. These have been scaled to M  K a b  = 24.4 (approximately M* — 2). T h e tracks  begin at z = 0.5 and are marked (small triangles) every 0.5 step i n redshift. T h i s plot illustrates that U L I R G s comprise most of the reddest populations. Use of this plot and caption courtesy of Webb et a l . [65].  CHAPTER 1. INTRODUCTION  1.3  10  The Subaru Deep Field  T h e Subaru Telescope, atop M a u n a K e a , has been used to produce deep images in J-band A = 1.16 — 1.32 pm and i n i f ' - b a n d A = 1.96 — 2.30 / i m of a 'blank' 2 arcmin x 2 arcmin 1  field near the north Galactic pole (Maihara et a l . [46]) (see F i g . 1.4). T h e Subaru Deep F i e l d ( S D F ) is a blank survey region and was chosen according to the following criteria: 1. to be i n a different location than the Hubble Deep F i e l d ( H D F ) since the Universe may have different characteristics depending on the direction you look (the H D F also has a higher airmass than the S D F from M a u n a K e a ) , 2. to have a reference star nearby for future adaptive optics ( A O ) observations, 3. to have low G a l a c t i c HI column density, no close-by stars (except for the A O guide star), galaxies or clusters of galaxies. Subaru's large aperture and  0.3 arcsec seeing have allowed extremely deep images  which rival any other images at these wavelengths (5<r = 25.1 and 23.5 i n J and K' respectively). These images are providing us w i t h v i t a l information for understanding physical conditions during the earliest epochs of galaxy formation, just as studying the Hubble Deep F i e l d has done . Near-IR observations measure the starlight directly from 2  older stellar populations or redshifted o p t i c a l - U V emission. T h e A"-band is about 10 times less obscured than optical bands and is therefore a good measure of intrinsic luminosity. So deep near-IR imaging has the potential to reveal a significant population of galaxies at high redshifts. A m o n g the ~ 350 galaxies detected in the AT'-band (K'<  23.5) i n the S D F , there are  four H E R O s whose K' magnitudes and J — K' colours are derived i n M a i h a r a et a l . [46] and are given in Table 1.1. T h i s sample of objects has a weighted average K' and J — K  1  x  magnitude  colour of 21.57 and 2.99, respectively.  The if'-band is centred at a wavelength of 2.13 fim compared to the xY-band being centred at a  wavelength of 2.2 /xm. 2  The Hubble Deep Field (HDF) revealed that faint, irregular and smaller galaxies are more abundant,  perhaps a characteristic of the infant high-redshift Universe (Ferguson, Dickinson & Williams [29]).  CHAPTER  1.  INTRODUCTION  11  Figure 1.4: T h e Subaru Deep F i e l d A " - b a n d image. Circles are drawn around the four H E R O s shown in more detail in F i g . 1.5. T h e numbers simply indicate the order of presentation i n F i g . 1.5. T h e circles, drawn to guide the eye, are roughly 10 arcsec in diameter, ~ 2 / 3 of the S C U B A 8 5 0 / / m full-width halfm a x i m u m ( F W H M ) beam size.  CHAPTER  1.  INTRODUCTION  12  Figure 1.5: T h e four H E R O s identified in the S D F survey are shown i n the J-band (upper row of images) and in i f ' - b a n d (lower row). Notice that S D F 1 appears to be a merging system and S D F 3 appears to be an interacting pair of galaxies (although they could just appear to be close neighbours through projection effects — see text).  CHAPTER 1. INTRODUCTION ID  13  Position (2000.0)  K'  J -  K'  SDF1  13 24 22?38  +27°29'49'.'5  20.91 ( ± 0.05)  2.97 ( ± 0.14)  SDF2  13 24 22?39  +27°29'01'.'9  22.03 ( ± 0.09)  3.65 ( ± 0.40)  SDF3  13 24 21?16  +27°29'01'.'9  21.99 ( ± 0.05)  2.81 ( ± 0.20)  SDF4  13 24 22?84  +27°30'08'.'4  22.31 ( ± 0.14)  4.12 ( ± 1.04)  21.57  2.99  h  m  h  m  h  m  h  m  Mean  Table 1.1: T h e positions, K' magnitude, and colour of the H E R O s from M a i h a r a et a l . [46]. Astrometry is derived from the coordinates of a H u b b l e Space Telescope ( H S T ) guide star in the flanking field. T h e estimated positional accuracy is ±0715. One of the bright S D F H E R O s may be a merging system and another may be part of an apparent interacting pair of galaxies (see F i g . 1.5), which could be exciting since mergers have been known to play a role i n models of U L I R G s . To evaluate the likelihood of these objects appearing as close neighbours v i a chance projection effects, we use the P statistic (see Downes et al. [22]). For a given surface density of objects JV, the probability that an object lies w i t h i n a distance d of another object at random is given by:  p = l U s i n g N = 10 d e g 5  - 2  down to  e  - ^  N  .  (1.1)  22 — 23 ( M a i h a r a et a l . [46]) as the t o t a l galaxy counts  i n S D F , and an estimate that one of the H E R O s lies w i t h i n 1.5 arcsec of another galaxy, we calculate that there is only a 5.3 per cent chance that two galaxies would appear this close together at random. Using simple Poisson statistics and the previous result, we determine that there is an 80 per cent chance ((1.0 — 0.053) ) of not seeing one galaxy 4  pair (within 1.5 arcsec) among four galaxies. We consider how unlikely it would be to see a galaxy pair amongst the rare H E R O s using a surface density of 3.5 x 1 0  - 4  arcsec  -2  CHAPTER  1.  INTRODUCTION  14  (5 H E R O s i n the 2 arcmin x 2 arcmin S D F image ). T h e chance that two H E R O s i n the 3  S D F lie w i t h i n 1.5 arcsec of each other at random is 0.25 per cent. It is therefore extremely unlikely to see a pair of H E R O s 1.5 arcsec apart if they were not interacting. Presumably, at least S D F 1 could be an interacting pair and be well-described by a U L I R G - t y p e galaxy. Since we see one pair of H E R O s i n the S D F , the surface density of paired interacting H E R O s is 1/(4 arcmin ) or 900 d e g . How bright do we expect a pair of H E R O s be i n 2  - 2  the submillimetre? A s s u m i n g that we can select bright submillimetre objects by finding H E R O s at I R wavelengths first, and supposing that we w i l l be lucky to see a bright submillimetre source only 10 per cent of time, the surface density S C U B A - b r i g h t paired H E R O s becomes 9 0 d e g . We locate this surface density on the 850pm number counts - 2  plot ( c f . F i g . 1.6), and we find that this surface density predicts that the interacting pair of S C U B A - b r i g h t H E R O s could have a flux level anywhere from 11 — 19 m J y , depending on how "lucky" we are to see bright submillimetre sources. G i v e n the high frequency w i t h which S C U B A sources w i t h luminosity i n the range of 5 to 2 0 m J y are identified w i t h very red E R O s , and the number densities of both types of sources, it is reasonable to expect that these four sources might have measurable submillimetre  flux.  Totani et a l . [63] deduced that S C U B A sources are likely to have  the same origin as the H E R O s i n the S D F by comparing the surface density of H E R O s to S C U B A sources. B y estimating that H E R O s make up 3 per cent of the t o t a l galaxy counts i n the S D F , they estimate the surface density of H E R O s to be « 3 x 10 d e g . 3  - 2  T h i s is similar to the surface density of 2 m J y S C U B A sources found by Borys [6] (see F i g . 1.6 for the 8 5 0 / i m source count model). T h e hypothesis that S C U B A sources are likely to have the same origin as H E R O s is reasonable, and a submillimetre detection would strongly suggest that they are dusty starburst galaxies at reasonably high redshifts. 3  For these purposes, we consider SDF1 to be a merger of two HEROs since they appear to be 2  separate objects in the iiT'-band.  CHAPTER  1.4  1.  INTRODUCTION  15  Guide to This Thesis  In this thesis, fluxes for the four H E R O s are determined from submillimetre observations, and we compare them to a model based on redshifting a galaxy template, and varying the overall luminosity and amount of dust reddening i n the Infrared. W e investigate 2 templates: Arp220 (a starburst galaxy) and N G C 3 9 3 8 (a normal cold spiral galaxy). Chapter 2 describes the instrument and data reduction, Chapter 3 addresses the H E R O s , Chapter 4 examines all of the S D F sources, and Chapter 5 states the conclusions that have been drawn from this work. Standard cosmological parameter values of H JIM = 0.3 are assumed for this work.  0  = 75kms  - 1  Mpc  - 1  , Q\  = 0.7, and  CHAPTER  1.  INTRODUCTION  T  i  1  16  1  1—i—i—r—i—|  0.1  !—i  1  1  1—i—i—i  i |  10  S/mJy  Figure 1.6: T h e 850 / i m cumulative source counts. T h e solid diamonds and yellow jagged line denote results from Borys [6]. O p e n squares and circles represent source counts from cluster studies by C h a p m a n et a l . [14] and S m a i l et a l . [59], respectively. T h e Scott et a l . [56] U K 8 m J y survey counts are the open triangles. Stars mark the counts from Hughes et a l . [38], and the open diamonds mark the counts from Borys et al. [8]. A two power-law model (solid line) is overlaid, as well as two predictions based on galaxy evolution models of Rowan-Robinson [49]. T h e dashed line is a £2 A = 0.7, £2M — 1-0 universe, and the dotted line is a £2A = 0.7, £2 = 0.3 universe. T h e dot-dashed line M  near the bottom of the graph is the count prediction from extrapolating the IRAS 60 /tm counts w i t h no evolution. We use this plot to derive the number of sources we should see at the flux level of the H E R O s . Use of this courtesy of Borys [6].  figure  CHAPTER  2.  THE INSTRUMENT  AND  DATA  17  CHAPTER 2 T H E  2.1  I N S T R U M E N T  A N D  D A T A  The J C M T and S C U B A  T h e 15-metre diameter James Clerk M a x w e l l Telescope ( J C M T ) is the largest s u b m i l limetre telescope i n the world and is situated atop M a u n a K e a on the B i g Island of Hawaii. A t this height (approximately 13,400 feet or 4092 metres) we are above most of the water vapour in our atmosphere.  It is very challenging to observe astronomi-  cal objects of interest in the submillimetre wavelength regime since there are very few "windows" where emission can penetrate the water-laden atmosphere, the largest opacity source i n the submillimetre regime. T h e Submillimetre C o m m o n - U s e r Bolometer A r r a y ( S C U B A ) (Holland et al. [36]), a continuum detector mounted at the left N a s m y t h focus of the J C M T , was built w i t h special filters i n these windows. Figure 2.1 shows the observed transmission of the S C U B A niters. S C U B A has been i n operation since M a y 1997 and is still the premier detector operating i n the submillimetre regime and is the most highly cited instrument to date. T h e S C U B A detector has a field of view of 2.3 arcminutes and consists of hexagonal arrays of 37 and 91 bolometers operating at 850 and 4 5 0 / / m , respectively (see Figure 2.2). It is cooled to about 70 m K , making it background-noise limited. Light collected by the telecope is reflected off of a series of mirrors, decreasing the f-ratio to f/4 where it is then split by a dichroic beam-splitter w i t h a 97 per cent transmission efficiency. T h e 1  beamsizes for the arrays are telescope diffraction-limited and the 850 jum beam is well fit by a Gaussian w i t h a F W H M of 14.7 arcseconds. T h e 450 pm beam has a F W H M of 7.5 arcseconds, although the beamshape is not so well-defined (it has significant sidelobes), is more prone to surface inaccuracies, and is characteristically less stable over time. 1  T h i s allows the arrays to operate simultaneously  CHAPTER  2.  THE INSTRUMENT  AND  DATA  18  T h e Caltech Submillimetre Observatory ( C S O ) has a radiometer operating at 225 G H z which monitors the changing conditions in the atmosphere by performing skydips approximately every 10 minutes. We use the relations for the post-upgrade wide band filters derived by A r c h i b a l d , W a g g & Jenness [1] to translate the C S O r atmospheric opacity values to the wavelengths of interest i n order to correct our d a t a for sky extinction:  7850 = 4.02TCSO + 0.001  (2.1)  and T  4 5 0  = 26.2rcso + 0.014.  (2.2)  Often, the C S O r meter is looking at a completely different part of the sky w i t h a different airmass and it would be more desirable to measure the atmospheric opacity term at the same time and airmass as the astronomical source. In response to this need, since the summer of 2001 the J C M T has been using a cabin-mounted water vapour monitor ( W V M ) which looks at the 183 G H z water vapour line once every 6 seconds using a three channel double side band receiver. T h i s instrument is particularly useful during unstable weather conditions, where the C S O tau meter may be giving unreliable readings.  2.1.1  Observing Modes  Sky emission dominates the astronomical signal being measured and must be removed. T h e secondary mirror "chops", or nutates, at a default frequency of 7.8125 H z w i t h a userdefined chop "throw", or amplitude, and position angle. T h i s produces a differential measurement of the signal from the reference position and another point on the sky, effectively cancelling any rapid atmospheric backgound variations. T h e telescope also "nods" on and off the source (or "beam-switches") every 10-20 seconds while chopping in order to compensate for the gradual change of the sky brightness over the timescale of an observation. S C U B A has four standard modes of operation: photometry, jiggle-mapping, scanmapping, and polarimetry.  We w i l l briefly describe the first two since they were the  CHAPTER  2.  THE INSTRUMENT  AND  19  DATA  Measured SCUBA filter profiles  250  350  450  550  650  750  850  950  Wavelength (microns) Figure 2.1: T h i s figure shows the wideband filter profiles measured in 2000 w i t h the U n i versity of Lethbridge Fourier Transform Spectrometer. T h e y are plotted over the old narrowband filter profiles for comparison. T h e t h i n black line (1mm P W V ) traces the submillimetre transmission function assuming a precipitable water vapour content of 1mm. T h i s figure was created by D a v i d Naylor and Wayne Holland and is from the J C M T website.  CHAPTER  2. THE INSTRUMENT  AND  DATA  SHORT WAVE ARRAY (91  20  LONG WAVE ARRAY  detectors)  (37 detectors)  1.3mm  2 mm  1.1mm (photometric pixel) 2.3 arcminutes  |iiii|iiii|iiii|ini|iiii|  0  5 10 15 20 25 mm  Figure 2.2: T h i s figure displays the arrangement of bolometers i n the short wave a n d long wave S C U B A arrays, which operate simultaneously at 450 and 850 pm, respectively. T h i s figure is from the J C M T website.  CHAPTER 2. THE INSTRUMENT AND DATA  21  modes used for this work.  Photometry P h o t o m e t r y is used for small point sources and is done by pointing the central bolometer at the target and then making a very small 9-point jiggle around the source position, resulting i n an undersampled map. T h i s is the most efficient mode for getting down to the desired signal-to-noise level for a point source w i t h accurately known coordinates. E a c h integration takes approximately 18 seconds including the nod.  Jiggle-Map T h e jiggle-mapping mode is ideally suited for sources larger than a beam, but less than 2.3 arcminutes i n extent. T h e secondary mirror makes a 16-point jiggle pattern of offset positions from the source's central pointing position. T h i s produces a fully sampled image at 850 fim. In order to get a fully sampled image at 450 pm, the telescope must make a 64-point jiggle pattern, as 3 arcsecond spacing between points is required as opposed to the 6 arcsecond spacing required at 850 pan. E a c h integration takes approximately 128 seconds including the nod.  2.2  Observations  T h e S D F was observed w i t h a resolution of 14.7 arcsec and 7.5 arcsec at 850 and 450 /zm respectively w i t h S C U B A on the J C M T during M a y 2001, M a y 2002, February and M a r c h 2003. T h e centre of the field lies at an approximate R A and Dec (J2000) of 13 24 21?20 h  m  and +27°29'25'.'0. 64-point jiggle map data of the S D F were obtained, providing measurements of the continuum at both wavelengths simultaneously w i t h S C U B A . In addition, 9-point photometry data were obtained i n 2002 and 2003 using the central bolometer on the S C U B A array for each of the four H E R O s and one suspected redshift 7 source . 2  2  The estimated photometric redshift of z = 7 was derived for one of the SDF galaxies by Kashikawa  et al. [43].  CHAPTER 2. THE INSTRUMENT  AND DATA  22  T h e atmospheric zenith opacity at 225 G H z was monitored w i t h the C S O tau monitor i n 2001 and 2002 and w i t h the J C M T water vapour monitor at 183 G H z i n 2003. Table 2.1 summarizes the observations, telescope parameters, and weather conditions. T h e secondary mirror was chopped at the standard frequency w i t h alternating position angles of 0 and 90 degrees and chop thows ranging from 30 to 86 arcsec. P o i n t i n g checks were performed hourly on blazars and planets, and offsets from the pointing centre were nominal i n all cases.  Observations were co-added to create maps w i t h t o t a l  integration times of 13.5 hours at 450 pm (excludes poor-weather photometry-mode d a t a from 2002) and 16.5 hours at 850 /zm, over a variety of weather conditions. T h e maps have estimated rms values of about 15 and 2 m J y , at 450 and 850 /um, respectively.  CHAPTER  I—1  CP  B  2.  CO  d  OS eo  THE INSTRUMENT  I—1  o  00  CN  i—i  o o  AND  CM  O  o  o  DATA  lO OS  00 o  23  00 ©  00 o  CO CO  CP  • f—(  SH  SH  co i  O  cd  o  CP  <  O  00  XS  CO  "3)  SH  a  co  Xi  O  d  CP CO  co  o  OS  o"  o Oi  o  OS  o OS  o  o  os os  o Oi  o  o  OS  o  os os  o o  d o  o  OS  o  CP CP s-i faO CP  +3  CP  CO  PH  CO  O SH  E-I OH O  Xi  o  o o o co © " © " ©" co eo CO  o CP  co  o o" co  SH  CM IC CM <M CO m <© os oco co_ oo oo o o CM" in" in" c o" co" co" in co co CO CO CO CO CO  MH  CT3  PH  .2 a SH  faO CP  O  o  o < < < < CS3  O  N  tS3  XI  OH  o  N  < < <  Xl  in CM  O  co CM  O lI  O  o  o o O  i—1  CM CM  C5  ©  T—1  I  ti-  CM ©  ii  I CM  o  i—1 o  o  CM  Q  CM  >>  CS  OH  SH  o  CP  -a  o  cc3  CP 'hb faO  CO  o o  CM <M~ s->  CO  o o  CM  cS-lo" cc3  CO  o o  CM  "*" l-i cc3  o  CO  o o  CM OO" (-I  a  os i—i to o  C N  I  oo O  c o o o  <M  o o  CM OS  X fo  X fo  +3  CP  a o +3  o  Xi PM  i-H  I  O  O I T}H  o  I CO  CO  1—1  oo  ©  CM  cc3  I  CP  I CM  i-H  CM CO  CO  >> S-l  c o o  oo oo  O o  CN  co"  o  os" o  fo S-l CP  a  o +3  O  XS  ©  CO  co CM  X3  H  <i  T—1  >>  o  X  fo  fo  SH  SH  PQ  2003  1 1  CM  , 2003  C5  SH  CM  2003  TP  , 2003  d cc3  i—i  T—1  7, 2002  CP faO  O  in"  X3  GO  "cp  fo p SH  CP  -1-3  ao a aoo  OH  o o  CO  a  X! fo  X5  cu  SH  OH OH  -d o CO SH  HH  CP SH  o  a  OH  cS  P CO  fo  CM fo  CO  GO  Q  P  CO  fo  fo  GO  GO  Q  P  SH  CP  XS O  X!  o OH  co CP SH  X!  CP  -u  cc3  H J  O  SH  O  H-3  O  "73  • r-t  -4-3 SH  CP CO X!  13  CM  X  CO CP XI  fcuO Cl  • 1—I >  SH  CP CO  X  O SH  CP  X3  O  CO Cl  faO Cl  "8, OH O  XI  o 'rt d  a • -H  Cl  O  SI  O  ct3  CP  +s CO  -c|  bO  a  SH  ^H  CP SH  SI  • 1-4  T3  d  d o  • i—i  £ SH 0)  CO  X  O  d  X3  T3  o .a  •'—:  X!  fo  O CP  CT3  CP XI SH CP  CP  O  -d  ce Cl  CP bO  cc3  o cu  O  X  SH  cc3  CO  CP  +=  CO  O  cu co  CP  .a o a  N  •fH  SH  T3  O  SH  d o CP  CP  a  CP  ^3  • -H  * I—I  o  SH  >  CP  CP SH  CO  CH • I—I  O O O  O  o  Xl  cc3 PI  "(3  X -u  a  CP  d cc3  CP CO SH  d o o CP  XI  -u fcuO  d  • -H  SH  d -cc? o SH SH  Cfi  CHAPTER 2. THE INSTRUMENT AND DATA  2.2.1  24  Added Noise in the 2003 Data  In June 2003, C o l i n Borys, a former U B C P h D . student, discovered a large power spike at «  1/16 H z coincident w i t h the frequency of the secondary chopper i n de-nodded,  flatfielded,  and extinction corrected 64-point jiggle map data. T h e telescope perfoms  1.024 second integrations on each of the 16-point jiggle pattern positions and nods to one side before repeating this procedure and nodding to the opposite side. T h i s pattern is repeated two more times to complete a 64-point jiggle p a t t e r n . T h u s , the telescope 3  is moving by a large amount every 16 seconds (1 nod every 16 points x 1.024s/point w 1 6 n o d s / s ) . We hypothesize that there may be some microphonic noise pickup when the telescope moves during the n o d , corresponding to the noise appearing stronger i n the bolometers at a frequency of « 1/16 H z . 300 jiggle map observations between 1998 and 2003 were analyzed by C o l i n B o r y s and Remo Tilanus of the Joint Astronomy Centre ( J A C ) , and only observations taken between December 2002 and M a r c h 2003 seem to be affected by this anomaly (see F i g . 2.3). T h e S C U B A fridge cycle at the end of M a r c h is likely responsible for fixing the problem (Borys 2003, private communication). U p o n analysis of the fourier spectrum of our data, we discovered that this power spike occurs on the long-wave array i n about two-thirds of the bolometers, while the remaining t h i r d had nominal signals, similar to what B o r y s found. T h e effect is also present in the short-wave array, but to a lesser extent. O n l y about 10 per cent of the bolometers show this noise spike. Usually this effect would be automatically removed after subtracting off the sky signal if all of the bolometers had this same spurious signal present i n their frequency spectra. B u t since only a fraction of the bolometers have this signal, the effect of removing the mean sky background actually imposes this signal on the non-affected bolometers, corrupting them as well. T h e concern is that the presence of a periodic noise  fluctuation  on a similar timescale to the jiggle pattern may be affecting the data, thereby imposing 3  When S U R F de-nods the data, each data point it reports is actually the sum of two 1.024 second  measurements taken 16 seconds apart.  CHAPTER  2.  THE INSTRUMENT  AND  DATA  25  0.10  0.15  0.20  0.25  0.10  0.15  0.20  0.25  Frequency (Hz)  1.2x1 0 " ' 4  1.0x10-  4  >  8.0X10-  5  o  6.0x10  -  - 5  4.0x10~  5  2.0X10-  5  Frequency (Hz)  Figure 2.3: T h e top plot is the F F T of the timestream for bolometer G 1 5 (or number 12) of the long-wave array, from 64-point jiggle map data taken of the S D F i n M a y 2002. T h e bottom plot shows the F F T of the timestream for the same bolometer from 64-point jiggle map data taken of the same field i n M a r c h 2003. Note the power contained in the large noise spike at about 1/16 H z i n the lowermost plot.  CHAPTER  2.  THE INSTRUMENT  AND  DATA  26  unwanted structure on the map data to some degree. (Borys 2003, private communication). T h e data were examined i n a variety of ways. F i r s t , we tried three different methods of separating the data: averaging all of the 2003 data together, separating the d a t a according to chop throw and position angle and taking the average, and concatenating all of the timestream data for one night. U p o n examination of the fourier spectrum of each of these new averaged data-sets we observed that the signal persisted. In the 2003 averaged data-set, the signal decreased by about an order of magnitude, but so d i d the overall noise level, resulting in a constant signal-to-noise ratio for the spike. See F i g . 2.4 for an example. We first investigate if and how this additional noise affects the d a t a i n the time domain (i.e. the bolometer timestream). We trace how the standard deviation (a) of the signal varies w i t h time. One jiggle observation consists of 10 integrations of a 64-point jiggle pattern, adding up to 640 timestream elements for each bolometer. For the central bolometer H 7 (number 19), we calculate a noise of o = 0.410 V over the course of one tenintegration observation. For a single bolometer, we calculate the mean and error of each 64-point jiggle map integration and see no trend for the error (noise) either increasing or decreasing over the set of ten integrations. We therefore conclude that the noise is randomly  fluctuating,  as it does not vary significantly i n the bolometer timestream over  the course of an observation. Next, we investigate if and how the noise spike affects the data i n the spatial domain (i.e. the map structure). Obviously, one might worry that unwanted structure is being imposed on the map. We trace how the signal varies across the 64-point jiggle pattern array. A bolometer w i l l hit 64 different positions in the sky during the course of one 64-point jiggle integration. Since each observation is made up of ten integrations, each position w i l l be hit 10 times during the course of an observation. We add up the contribution of each integration at each jiggle position, calculate the mean, a and the error i n cr, and look for a trend i n the noise. We do not see any trends of the noise either increasing or decreasing. For an error of a — 0.410 V (found above), we determine the typical error  CHAPTER  2.  THE  INSTRUMENT  AND  27  DATA  > 1.OX10-  4  5.OX10-  5  0.00  0.05  0.10  0.15  0.20  0.25  0.20  0.25  Frequency ( H z ) 4X10"  5  3xl0" E5  > o  2x10'  X10-  5  0.00  0.05  0.10  0..15  Frequency ( H z )  Figure 2.4: T h e top plot is the F F T of the timestream for bolometer G 1 5 (or number 12) of the long-wave array, from 64-point jiggle m a p d a t a taken of the S D F in M a r c h 2003. T h e bottom plot shows the F F T of the timestream for the same bolometer from the average of a l l the 2003 64-point jiggle m a p data taken of the same field. Note the factor of « 10 decrease i n the power of the noise spike at about 1/16 H z and i n the overall noise level i n the lowermost plot from the one above.  CHAPTER  2.  THE INSTRUMENT  AND  DATA  Jiggle m a p  28  position  Figure 2.5: T h i s plot shows the standard deviation of ten points (ten integrations) for each of the 64 different jiggle positions for the central bolometer, H 7 . T h e solid line represents the mean standard deviation value. T h e dashed lines represent the la error i n the standard deviation (~  ±l/y/2N).  CHAPTER  2.  THE INSTRUMENT  0  B o l o m e t e r  19  2  10  4  6 8 X grid  B o l o m e t e r  0  AND  2  4  6 8 X grid  29  B o l o m e t e r  12  2 0  10  DATA  0  2  4  6 8 X grid  B o l o m e t e r  12  1 3  10  12  2 6  0  12  Figure 2.6: T h e 64-pt jiggle pattern is observed i n 4 sections of 16 points (solid lines). For 4 different bolometers on the array, we plot the standard deviation of ten measurements (ten integrations) at each position in the 64-point jiggle pattern using contours to specify the noise level. Contours begin at a = 0.0 and increase i n steps of 0.05 up to a m a x i m u m of a = 1.0. It is evident that the noise varies randomly across the array for any given bolometer.  CHAPTER  2.  THE INSTRUMENT  AND  DATA  30  i n a to be about 0.0920 V . Figs. 2.5 and 2.6 illustrate that the noise appears to vary randomly over the course of 64-point jiggle pattern for a given bolometer. We conclude that the noise is randomly  fluctuating,  does not depend on position, and therefore does  not impose any structure on the map. C o l i n Borys postulates that there was some k i n d of vibration-induced microphonic pickup when the telescope suddenly moves at the end of the 16 point jiggle pattern to nod, m a k i n g the noise on the array stonger at a frequency of 1/16 H z . It may also be due to a noisy atmosphere or to a temperature instability which makes the bolometer gains fluctuate  i n a random way (which would have been fixed by the S C U B A fridge cycle).  In any case, we conclude that the large power spike at 1/16 H z is just introducing more random noise  2.3 2.3.1  l / \ / 2 ) into the data and not imposing any structure on our maps.  Data Reduction Preliminary Reduction with SURF  T h e S U R F ( S C U B A User Reduction Facility; Jenness & Lightfoot [40]) reduction packages were used for nod compensation extinction correction  (extinction)  (reduce_switch),  of the data. T h e  flat-fielding  reduce.switch  (flatf ield)  and  c o m m a n d subtracts  the off-position from the on-position and splits the raw data array into separate components. Next, the  flatf ield  command "flatfields" or corrects the gain of the array  by m u l t i p l y i n g each bolometer by the volume flatfield value relative to a reference pixel, usually the central pixel of the array. T h e  extinction  command corrects the d a t a for  source airmass (elevation) and sky opacity, r. A great deal of care was taken to eliminate any spikes in the C S O r readings using the polynomial fits provided by the J A C , and a constant r value was used over the duration of each observation. T h e e x t i n c t i o n command has the capability to interpolate between two specified r values (before and after the observation) but the r did not vary significantly during the course of any one observation so this feature was not used. T h i s last command also splits up the d a t a into  2.  CHAPTER  THE INSTRUMENT  AND  DATA  31  the two different wavelength components, namely 850 and 450 pm. T h e work produced in this thesis made use of locally developed code written i n C + + by C o l i n B o r y s (Borys [6]) for his U B C doctoral dissertation. We converted the observation files into the more versatile " F I T S " format using the program SCUBA2FITS, m a k i n g it readable for the next series of locally developed programs. T h e files all now contain a header, and an array of three vectors containing: d a t a timesteps, bolometer number, and the data. T h e data vector itself contains the signal i n Volts and three R A / D e c offset positions corresponding to the three positions on the sky that the bolometer looked at while taking the difference measurement.  2.3.2  Making the Maps  A program called  BS_SCAN_REDUCE  was then used to 4 a despike and rebin a l l d a t a types  (photometry and jiggle maps) into a single map, producing three useful d a t a products (similar to those that can be produced by S U R F ) : a signal map, a noise map (the standard deviation or a), and a signal-to-noise ratio ( S N R ) map. B a d pixels were essentially eliminated by weighting each pixel by its timestream inverse variance relative to the central pixel. T h e data were despiked by means of calculating the root-mean-square (rms) for each bolometer and rejecting any data 4 times greater than this value. Sky subtraction was performed by calculating the mean sky at each timestep using the median of all of the bolometers save the noisiest bolometers and subtracting this result from the data. A bolometer was deemed too noisy to include in the sky subtraction by calculating its variance and cutting out the bolometers that were above a certain threshold (usually a > 0.005). T h e flux from the negative off-beams were then "folded i n " to increase the overall sensitivity of the 850 pm map. For each sample in the bolometer timestream, we add -0.5 times the measured flux to the position of the off-beam. Since a bolometer spends only half as much time i n each of the off-positions (reference positions that do not lie on the target) as i n the on-position during the course of a single observation, we assign the  CHAPTER  2. THE INSTRUMENT  AND  DATA  32  on-position a weight of it; = 1 and the off-positions weights of w — —0.5 each . Following 4  the treatment by Sajina [50], the folding can be represented by:  =  and (2-4) where /x is the mean, a is the standard deviation, and x is the voltage measured. It should be noted that the beam is assumed to be Gaussian i n shape which is usually the case at 850 fim but not at 450 fim, where the beam is often more severely distorted and non-Gaussian i n shape. Therefore we do not fold i n the off-beams to produce the 450 /zm map.  In this way, folding i n the off-beams w i l l improve the sensitivity of the m a p by  about a factor of:  (2-5)  2  V 0 . 5 + 1.0 + 0.5 2  2  V 3  The "-f" option of the B S _ S C A N _ R E B I N program was used to produce this result. T h i s step is essentially equivalent to a single iteration i n the iterative deconvolution technique described i n Borys [6]. T h i s method is only valid if sources are far enough apart a n d are not being chopped onto one another. W e have taken great care to use a chopping strategy that avoids the H E R O s being chopped onto each other by selecting appropriate chop throws i n azimuthal coordinates, allowing the chop position to rotate o n the sky during the observation (see F i g . 2.7). T h i s procedure of folding i n the off-beams was used successfully by Borys [6] to produce the Hubble Deep F i e l d Supermap and by Eales et al. [26] to produce the 14 Hour F i e l d , for example. The 850 /tm data were then regridded onto a rectangular grid of 3 arcsec pixels and smoothed using the program B S J 3 M O O T H F I T to a final F W H M of 14.7 arcsec (the size 4  This can also be seen in that for an observation, both values vary as N"  1  (WotKES^)4  since /t  =  CHAPTER  2.  THE INSTRUMENT  SDF-  AND  33  DATA  SDF-2  1  •  SDF-4  • SDF-3  ^^^^^^^ JlP^^illlk  •  00  •  9  £r^^k>  Ik  ^^mw  H  M  Figure 2.7: We plot the chopping strategies for photometry observations of the four HEROs.  For each plot, the shaded source is the H E R O that the central  bolometer is pointed at during photometry, and the other small circles (scaled to the represent the beam-size at 850 fxm) represent the other H E R O s i n the field. We also plot the path that the bolometer w i l l chop on the sky as the sky rotates during the night. We label the chop throws i n arcseconds on each path.  CHAPTER  2.  THE INSTRUMENT  AND  DATA  34  of the J C M T beam at 850 /mi) and the 450 /zm data were smoothed to a F W H M of 7.5 (the beamsize at 4 5 0 / i m ) . B y rebinning the 850 / i m map using 3 arcsec pixels, the m a p w i l l have pixels which are correlated since the beam is 4.93 of these new pixels wide. Another reason for smoothing is that the 3 arcsecond pixel grid gives an artificially small resolution which is not the true resolution of the image (given by the beam size) and w i l l have the effect of making the map appear more noisy than it truly is. Neglecting the undersampled and hence noisier edges of the m a p , we get a mean signal of —0.047 (±0.03) m J y from the central 3425 pixels of the 850 am map. T h i s is i n agreement w i t h the expected map average of 0 m J y for differential measurements. A s a check, we examine the rms of the map without the off-beams folded i n and compare it to the rms of the m a p w i t h the off-beams folded i n . T h e rms before off-beam folding is 2.26 m J y and the rms post off-beam folding is 1.80 m J y , which is entirely consistent w i t h an estimated resulting rms of 2.26 x  = 1.85. Similarly, we measure the mean signal of  the central 1662 pixels of the 450 /zm map to be 0.32 (±0.24) m J y . W e measure an rms for the 450 / i m m a p of 9.7mJy.  2.3.3  Calibration  Because the map is output i n Volts and we wish to measure flux i n an astronomicallymeaningful units (i.e. m J y ) we must calibrate the data-set by observing objects w i t h known submillimetre fluxes. For consistency, the calibrators were reduced i n the same way as the S D F m a p and the mean of the peak pixel values of the calibration sources were used to convert the measurements into units of flux. See table 2.2 for the list of average F C F s used. W e adopt the average value of all of the flux conversion factors ( F C F s ) obtained over one night of observing and use this value as the F C F for every observation on the same night. T h e average values and standard deviations of all of the F C F s we obtained are 224(±27) m J y at 850 / i m and 339(±120) m J y at 450 / i m . These flux conversion factors are consistent w i t h the standard gains (within the uncertainties) at 450 a n d 850 / i m of  CHAPTER  2.  THE INSTRUMENT  AND  DATA  35  308(±109) a n d 219(±21) m J y , respectively, from the J C M T website. F o r photometry mode, the standard gains are slightly different: 384(±82) and 197(±13) m J y / V o l t at 450 and 850 fj.m. R m s uncertainties i n the calibrations are approximately 10 per cent for the 850 fim jiggle m a p a n d photometry data and 30 per cent for the 450 /zm jiggle m a p data-set. These uncertainties are mainly caused by the variability of the atmospheric transmission and changes i n the dish surface accuracy induced by temperature  fluctuations  (Jenness  et al. [41]).  2.3.4  Flux Measurements  S C U B A operates at wavelengths appropriate,to detect thermal emission from dust w i t h temperatures of 3 - 3 0 K . A t higher temperatures (excluding high-redshift objects), dust radiates mostly i n the far-IR, so high-redshift far-IR emitters w i l l have this emission redshifted into the submillimetre (Holland et a l . [36]). Since the angular extent of any of these sources is likely to be much smaller than the 3 arcsec pixel size (for example a l k p c source at a distance of z = 3 would appear to be 0.14 arcsec i n the sky), we simply measure the flux of each source i n the pixel corresponding to the near-IR source coordinates on the beam-convolved map. T h e best estimate of the noise associated w i t h each pixel is given i n the noise map. We ignore the calibration uncertainty, which is unimportant w i t h these low signal-to-noise ratio data. T h e flux measurements include contributions from a l l jiggle m a p a n d photometry data which have a l l been inversevariance weighted i n the co-addition process. T h e final maps are displayed i n Figs. 2.8 and 2.9. T h e results for the H E R O s are summarized i n Table 2.3. It is interesting to note that the 3 reddest objects (J — K' > 3) a l l have positive submillimetre flux greater than the least red H E R O , S D F 3 .  CHAPTER  2.  AND  DATA  Observing  Number of  FCF  Mode  Calibrators  M a y 27, 2001  Jiggle M a p  M a y 16, 2002  Date  THE INSTRUMENT  36  Number of  FCF  (mJy/Volt)  Calibrators  (mJy/Vo  5  325  8  235  Photometry  -  -  0  *197  M a y 17, 2002  Photometry  -  -  0  *197  Feb 6, 2003  Photometry  2  324  2  201  Feb 9, 2003  Photometry  0  *384  0  *197  Feb 10, 2003  Photometry  2  404  2  233  M a r 2, 2003  Jiggle M a p  2  292  2  200  M a r 3, 2003  Jiggle M a p  3  402  3  232  M a r 4, 2003  Jiggle M a p  1  318  1  219  M a r 5, 2003  Photometry  3  417  3  230  M a r 8, 2003  Jiggle M a p  4  260  5  215  450  850  Table 2.2: For each night of data, a different F l u x Conversion Factor ( F C F ) , or gain, is derived from the average value of calibration observations taken near to the observations. In cases where the calibration source was too extended to give a useful measurement or where too few calibration observations were made (denoted by an asterisk), we just use the "standard" gain value from the J A C website.  T h e 450 fxm data from M a y 2002 were not used because  the weather was very poor (see Table 2.1). T h e standard gains are 308(±109) and 219(±21) m J y / V o l t at 450 and 850 pm respectively i n mapping-mode and 384(±82) and 197(±13) m J y / V o l t in photometry mode (cf. S C U B A calibration webpage).  CHAPTER  2.  THE INSTRUMENT  AND  DATA  37  Figure 2.8: T h e Subaru Deep F i e l d 850 pm signal-to-noise ratio ( S N R ) map. T h e brightest points have a S N R of w 3 — 4 and the lowest points have a S N R of w —3. T h e circles are roughly 14.7 arcsec in diameter, the size of the S C U B A 850 / i m F W H M beam size. T h e circles w i t h crosses inside mark the location of the four H E R O s , and the other circle marks a > 3 a detection.  CHAPTER  2.  THE INSTRUMENT  AND  DATA  38  Figure 2.9: T h e Subaru Deep F i e l d 450 /zm signal-to-noise ratio ( S N R ) map. T h e brightest points have a S N R of w 3 — 4 and the lowest points have a S N R of « —3. T h e circles are roughly 7.5 arcsec in diameter, the size of the S C U B A 450 /zm F W H M beam size. T h e circles w i t h crosses inside mark the location of the four H E R O s , and the other two circles mark > 3 a detections.  CHAPTER  ID  2.  THE INSTRUMENT  S450  5g5o  (mJy)  (mJy)  AND  DATA  39  Upper L i m i t s g  8 5 0  SDF1  i:47 ( ± 5.94)  2.15 ( ± 0.92)  < 3.46  SDF2  - 7 . 8 4 ( ± 6.36)  1.01 ( ± 0.94)  < 2.64  SDF3  6.10 ( ± 6.48)  - 0 . 2 2 ( ± 0.88)  < 1.63  SDF4  1.73 ( ± 5.76)  1.85 ( ± 0.98)  < 3.45  Mean  0.42 (±3.06)  1.15 (±0.46)  < 1.76  Table 2.3: Measured flux densities w i t h errors and predicted redshift upper limits for the four S D F H E R O sources. T h e fourth column lists the 95 per cent Bayesian upper limits (in m J y ) at 8 5 0 / i m . T h e bottom row lists the error-weighted mean of the 850 / i m fluxes, and the 95 per cent Bayesian upper confidence l i m i t to this mean flux.  CHAPTER  3.  HEROS  40  CHAPTER 3 HEROS In the following chapter, we estimate the cirrus contribution to our 850 //m map. We also briefly discuss the relation between H E R O colour and submillimetre flux. W e then create two template S E D s , a starburst galaxy and a normal galaxy, and use them to facilitate estimates of the redshifts of the H E R O s . We compare and evaluate the models to determine how realistic these templates are at describing the H E R O s .  3.1  Cirrus Contribution  We need to estimate the amount of cirrus contribution to our map i n case it is providing a large fraction of the flux, which would mean we are just measuring galactic emission and not extragalactic sources at all. We use a portion of the Schlegel, Finkbeiner & Davis [53] full sky 100 am map, a reprocessed composite image of the (Diffuse InfraRed Background Experiment) and IRAS/ISSA  (IRAS  COBE/DIRBE  Sky Survey Atlas)  maps. These authors carefully removed zodiacal foreground emission, artefacts from the IRAS scan pattern, and confirmed point sources, resulting i n a map w i t h calibration and IRAS  DIRBE-qu&lity  resolution. F r o m this map we obtain 0 . 9 M J y / s t e r a d i a n i n the  direction of the S D F pointing centre coordinates. Franceschini [30] presents a plot of the S E D of the diffuse dust due to cirrus for a nearby prototype starburst galaxy, M 8 2 .  F r o m this plot, we take the ratio of the  cirrus contribution at 100 / m i to the contribution at 850 / m i and scale the 100 / m i cirrus measurement by this ratio i n order to estimate the 850 / m i cirrus contribution. We obtain a cirrus flux contribution of 5 x 1 0 ~ M J y / s t e r a d at 850 / m i . G i v e n that 4  there are about 60,000 14.7 arcsecond S C U B A beams i n 1 square degree of sky, the flux contribution becomes 2 x 10~ m J y per S C U B A beam at 850 / m i . T h i s is much less than 3  CHAPTER  3. HEROS  the confusion l i m i t of our map  41 l m J y ) and is extremely negligible. A t 4 5 0 / i m we  obtain a cirrus flux contribution of 3 x 1 0 is 4 x 1 0  - 3  - 3  M J y / s t e r a d . T h e 450 /zm cirrus contribution  m J y due to the smaller beam size of 7.5 arcseconds F W H M . We can therefore  safely ignore the neglible contribution of cirrus in our measurements.  3.2  HERO IR Colours and 850 jam Fluxes  S C U B A galaxies are often found to be associated w i t h E R O s but we have shown that h y p e r - E R O s are not always necessarily associated w i t h S C U B A - b r i g h t galaxies. We do not find that the H E R O s are S C U B A - b r i g h t , and we also do not find any evidence for their 850 / i m submillimetre flux increasing w i t h J — K' colour. T h e latter finding is illustrated in F i g . 3.1, where we plot 850 /tm against J — K' colour and get a slope of 0.81 ± 0.89 and a x  2  value of 3.3 for the best-fit line through the data. G i v e n that extremely red  objects associated w i t h S C U B A sources tend to be dusty starburst galaxies, we might expect an overlap between the H E R O s and the U L I R G population at high redshifts. We use the S E D of A r p 2 2 0 , constrained to fit the near-IR and submillimetre observations, to investigate if this is a reasonable assumption. W e compare the results against a normal galaxy template as well.  3.3  Using Arp220 as a Template  We created a template S E D using public photometric data available for A r p 220 from the N A S A / I P A C Extragalactic Database ( N E D ) . A r p 220 is the most luminous object in the nearby universe (at z = 0.018) and is the most well-studied example of a U L I R G . It is necessary to use this low-redshift analogue of a S C U B A source since higher redshift sources have more sparsely sampled S E D s , due to the difficulty of observing them i n multiple wavelength regions. We use the Lagache, Dole & Puget [45] U V to radio S E D models of a typical starburst galaxy to fill i n the gaps i n the data for Arp220.  T h e Lagache, Dole & Puget [45]  CHAPTER  3.  HEROS  42  E =4. o m oo  .j  i  i i_  J - K  Figure 3.1: Is there a colour-flux correlation for the H E R O s ? We plot 850 / / m flux versus J — K' colour. T h e best-fit line through these d a t a has a slope of 0.81 ± 0 . 8 9 and a goodness-of-fit x  2  value of 3.3, suggesting that there is no correlation  between the measured 850 fim submillimetre flux and I R colour i n the case of the S D F H E R O s .  CHAPTER  3.  HEROS  43  templates evolve (become hotter) w i t h luminosity over a range of luminosity levels from about L— 10 — 1 O L . 9  These templates cover the range of bolometric luminosities  1 3  0  of galaxies that comprise the Cosmic Infrared Background up to z < 2. For example, F I R B A C K (Far InfraRed B A C K g r o u n d ) sources comprise a cold nearby population w i t h L=  10 — l O L 9  n  0  and a more distant (z < 1.2) cold (or warm and highly luminous)  population w i t h L= 1 0 L . We employ their model number 30 (albeit heavily modified 1 2  Q  in terms of its dust content i n the I R portion of the spectrum) in order to draw a smooth line through Arp220 (see F i g . 3.2). F i g . 3.3 shows that for the i f ' - b a n d and at 850 / u n , the flux does not vary by more than about 10 per cent from the Lagache, Dole & Puget [45] template number 41, the most luminous template available. We have constructed a three parameter model based on the S E D of A r p 2 2 0 , constrained only by the K' magnitude and the J — K' colour. We effectively redshift, dust extinct, and brighten the S E D by known amounts and then read off the flux at 850 / o n . After choosing a redshift, we find the J — K' colour and add dust to fix the colour to the observed value for each H E R O . T h e model we have employed fits known sources even though it is very simple (one-parameter), and this gives it an advantage over more complicated multi-parameter models. We add dust by multiplying our Arp220 template by an extinction function which varies w i t h wavelength. We determine such an extinction law by constructing a powerlaw fit to the relative extinction values in the Landolt V , R, I and U K I R T J, H,  K  and V bandpasses versus effective wavelength from Schlegel, Finkbeiner & Davis [53] (see Figs. 3.4 and 3.5). T h e y employ a 'diffuse I S M ' mean value of R  v  = 3.1 for the  extinction laws of Cardelli, C l a y t o n & M a t h i s [10] and O ' D o n n e l l [47]. We note that the difference between the extinction curves of the M i l k y Way, the Magellanic Clouds and starburst galaxies is almost negligible at wavelengths longer than w 0.26 /tm (see C a l z e t t i , K i n n e y & Storchi-Bergmann [9], Cardelli, C l a y t o n & M a t h i s [10]). Therefore, using a different extinction law should not affect our results since only galaxies at redshifts of z < 3 are considered, corresponding to a rest wavelength of 0.31 u-m being redshifted into the J-band.  CHAPTER  3.  HEROS  44  Figure 3.2: T h e model spectral energy distribution of a nearby U L I R G : the Lagache, Dole & Puget [45] starburst template (dashed line represents template number 30) fit at z — 0.018 to Arp220 photometric data (solid line), which are marked w i t h the plus signs. T h e energy output is dominated by a modified blackbody and the m i d - I R emission features seen here are attributed to P A H molecules. T h e slope of the drop at « 1 / m i i n the original S E D has been extrapolated to shorter wavelengths and fits an Arp220 ROSAT i n the plot) w i t h i n its error bar.  X - r a y d a t a point (not shown  CHAPTER  3.  HEROS  10  2 7  45  f  m22r 0.1  ,  ,  ,  ,  1  1.0  10.0  100.0  1000.0  10000.0  Wavelength  [^m]  Figure 3.3: T h e upper panel shows 3 different luminosity templates from Lagache, Dole & Puget [45]. T h e lowest line is the template used for Arp220 (number 30), the middle line is template number 35, and the highest line is the brightest template available (number 41). T h e lower panel gives plots of the ratio of the Arp220 template w i t h the other two higher luminosity templates. Note how the template S E D s actually change intrinsically in shape w i t h luminosity. Vertical lines indicate the portion of the spectrum that we w i l l be sampling over a redshift range 0 < z < 3 at wavelengths of 850 fim and K'.  For an  observed band, the ratio of the AT'-band to 850 fj,m flux does not vary by more than 10 per cent across the models chosen.  CHAPTER  3.  HEROS  46  TABLE 6 RELATIVE EXTINCTION FOR SHLECTED BANDPASSBS Filter Landolt V Landolt B Landolt V , , Landolt R Landolt I CTIO U CTIO B CTIO V CTIO R CTIO/ UKIRT J UKIRT H UKIRT K UKTRTE Gunn i Gunn z APMb,  (A)  A/A(V)  3372 4404 5428 6509 8090 3683 4393 5519 6602 8046 12660 16732 22152 38079 5244 6707 7985 9055 6993 4690  1.664 1.321 1.015 0.819 0.594 1.521 1.324 0.992 0.807 0.601 0.276 0.176 0.112 0.047 1.065 0.793 0.610 0.472 0.755 1.236  A*t A/E(B-V) 5.434 4315 3.315 2.673 1540 4568 4.325 3.240 2.634 1.962 0.902 0.576 0.367 0.153 3.476 Z590 1.991 1.540 2.467 4.035  Filter Stromgren u Stromgren ft Stromgren y Sloan f WFPC2 F300W WFPC2 F450W WFPC2 F555W WFPC2 F606W WFPC2 F702W WFPC2 F814W DSS-II g DSS-II r DSS-II/  A  A/A(V)  A/E(B-V)  3502 4676 4127 4861 5479 3546 4925 6335 7799 9294 3047 4711 5498 6042 7068 8066 4814 6571 8183  1.602 1.240 1.394 1.182 1.004 1.579 1.161 0.843 0.639 0.453 1.791 1.229 0596 0.885 0.746 0.597 1.197 0.811 0.580  5.231 4.049 4.552 3.858 3.277 5.155 3.793 2.751 2.086 1.479 5.849 4.015 3.252 2.889 1435 1.948 3.907 2.649 1.893  NOTB.—Magnitudes of extinction evaluated in different passbands using the R = 3.1 extinction laws of Cardelli et al. 1989 and ODonneB 1994. Thefinalcolumn normalizes the extinction to photoelectric measurements of E(B—V). r  F i g u r e 3.4: W e use the effective wavelengths (A ff column) a n d relative extinction values e  (A/A(V)  column) for the L a n d o l t V, R, I a n d U K I R T J , H, K a n d 11  bandpasses i n order to construct an extinction law. T h i s table has been extracted from Schlegel, F i n k b e i n e r & Davis [53], a n d they employ a 'diffuse I S M ' mean value of R  v  = 3.1 for the extinction laws of C a r d e l l i , C l a y t o n &  M a t h i s [10] a n d O ' D o n n e l l [47].  CHAPTER  3.  HEROS  47  Figure 3.5: W e plot the magnitudes of extinction relative to the F - b a n d (A/A (V)) for the Landolt (triangles) and U K I R T (squares) bandpasses (see F i g . 3.4 for the complete table). W e fit 2 . 5 1 o g ( l / ( l +  (1^3)2.6)0.2x)  w  i  t  h  x  =  1  0  ^  e  2  .51og  of E q n . 3.1) to the data points (solid line). T h e values i n the function were chosen arbitrarily to give a reasonable fit to the data points: 1.23 places the line at the correct vertical position relative to the points, 2.6 gives the correct line steepness to fit the data, and 0.2 places the graph i n the correct horizontal location relative to the data points and also adjusts the line steepness to a lesser degree.  CHAPTER  3.  HEROS  48  We redden the original template by a factor of: 1  (l + (  (3.1)  L|3)2.6)0.2a:  where x is our free parameter which adjusts the reddening (i.e. J — K' colour) and A is wavelength in microns.  T h e function was fabricated to fit the relative extinction  values of the L a n d o l t and U K I R T bandpasses (which are similar enough to the Subaru bandpasses), and the values in the function were chosen arbitrarily to give the best fit to the data points (see caption of F i g . 3.5 for details). Choosing template number 30, redshifted to z = 0.018 and using a value of x = 2.5 i n the equation above gives a good fit to the A r p 2 2 0 photometric data (see F i g . 3.2 solid line). We alter the luminosity of the template to fit the observed K' flux, as determined from the observed K' magnitude following Skinner [57]: = 10 -0.4m  F where m is the magnitude and F  0  zero-point for the i f ' - b a n d is F  0  (3.2)  is the zero-point flux for the photometric band. T h e  = 718.903 J y and was extracted from the interpolation  of the broad-band flux of Vega (Motohara 2002, private communication).  where d  L  is the luminosity distance. Following C a r r o l l , Press & Turner [11], the l u m i -  nosity distance, <ii, i n a flat Universe is derived from: (3.4) with H(Z)  = H {(I  + ) n z  0  z  M  +  n y/ , 2  A  (3.5)  and  Q = 1 - Cl A  (3.6)  CHAPTER  3.  HEROS  49  ID  Opt ical  5850  •^spec  •^model  (mJy) 2 2 . 5 , 7 - K = 4.0  4.4 ± 1.2  1.013  1.2 ± 0 . 3  -  2 2 . 3 , 7 - K = 3.5  3.0 ± 0 . 8  1.219  1.3 ± 0 . 3  I  =  2 8 . 7 , 7 - K = 5.2  7.0 ± 0 . 5  4.1 ± 0 . 5  3.4 ± 0 . 1  I  =  2 4 . 9 , 7 - K - 6.5  4.9 ± 0 . 7  1.44  0.82 ± 0 . 1  2 3 . 9 6 , 7 - -K •= 4.23  6.3 ± 1 . 4  0.449  1.8 ± 0 . 2  SMMJ123628+621048  7  SMMJ123635+621239  I  SMMJ123652+621227 SMMJ164502+4626.4  F I R B A C K F N 1 40 ( J l ) I  Table 3.1: Submillimetre sources w i t h red I R colours and known redshifts.  Model-  predicted redshifts and errors (estimated using the errors i n the 850 fim fluxes) are given i n the last column. We test this model on a number of red objects w i t h measured I R colour, 850 fim flux and spectroscopic redshift to test whether we can reproduce these redshifts photometrically. We use an E R O and a V R O (very red object), two red objects from the Hubble Deep F i e l d ( H D F ) w i t h measured spectroscopic redshifts from the work of (Borys [6]). A l s o from the H D F , H D F 8 5 0 . 1 , detected by Hughes et a l . [38] i n the submillimetre, is a S C U B A - b r i g h t IR-faint E R O w i t h a redshift deduced by Monte C a r l o photometric redshift techniques from D u n l o p et a l . [24]. is also included i n this test.  W e also i n -  clude H R 1 0 (or E R O J164502+4626.4, Dey et al. [17]), the first E R O discovered to be associated w i t h a U L I R G , and F N 1 40 ( J l ) , discovered i n the F I R B A C K (Far InfraRed Background) survey (Dole [19]). See Table 3.1 for the I R colours, submillimetre fluxes, spectroscopic redshifts and predicted redshifts. T h e mean difference between the spectroscopic redshifts and model-estimated redshifts  (^  m o  dei ~  z ) spec  is +0.06 ± 0.8. F i g . 3.6  illustrates how accurately the model predicts source redshifts by comparing model results to spectroscopically measured redshifts. We are therefore confident that this simple model is suitable for estimating redshifts of the reddest population of objects when the IR-colour and 850 fim flux are known.  CHAPTER 3. HEROS  0  50  1  2 spectroscopic  3  4  5  redshift  Figure 3.6: We plot model-estimated redshifts (with errors) against spectroscopic redshifts to see how well our model does for sources w i t h known redshifts. Points falling on the line indicate a good correlation between the spectroscopic redshift and model-predicted redshift. O n average, the mean difference between the spectroscopic redshifts and model-estimated redshifts ( z d e i m0  +0.06 w i t h a spread of ± a = 0.8.  —  ^spec) is  CHAPTER  3.4  3.  HEROS  51  Using a NGC3938 as a Template  For comparison, we also created a normal spiral galaxy S E D using N E D public photometric d a t a for N G C 3938. N G C 3938 is a well-studied, multiple-armed, early luminosity class Sc(s) face-on galaxy 10.8 M p c away. We utilize the Lagache, Dole &: Puget [45] template of a normal cold spiral galaxy to fill in the gaps in the photometric d a t a i n the same way as for Arp220. Unlike the starburst galaxy, the normal galaxy template of Lagache, Dole & Puget [45] does not evolve w i t h luminosity, since the lack of data limits the contraints that can be placed on the model (Lagache 2003, private communication). We scale the template evenly across all wavelengths and add a small amount of dust (x = 0.1, c f . E q n 3.1) to the I R region of the spectrum to match the photometric d a t a as closely as possible (see F i g . 3.7).  3.5  Results  We present the model-estimated redshifts derived from our analysis using the Arp220 and N G C 3 9 3 8 templates i n Table 3.2.  3.6  Are the Models Feasible?.  Here we perform a quick check to see that the flux level we measure for the H E R O s is consistent w i t h the number of H E R O s found in the S D F . We then examine the model parameters such as dust content and luminosity that were chosen to satisfy the colour and K magnitude constraints at each redshift step (usually A 2 = 0.2). If either of the starburst or normal galaxies are reasonable analogues of the H E R O s then we would not need to add in an extraordinary amount of dust (i.e. more than a magnitude or 2 i n J-band) nor brighten the template by an unreasonable amount (i.e. more than a factor of a few) at the redshift where the expected 850 pm flux matches the upper l i m i t to the 850 fim observed flux.  CHAPTER  3.  HEROS  52  Figure 3.7: T h e model spectral energy distribution of a nearby normal spiral galaxy: the Lagache, Dole & Puget [45] normal galaxy template (dashed line represents template w i t h no added dust) fit at z = 0.002699 to N G C 3 9 3 8 photometric data (solid line), which are marked w i t h the plus signs. A g a i n , the model S E D has been extrapolated to shorter wavelengths.  CHAPTER  ID  3.  53  HEROS  5450  £850  (mJy)  (mJy)  Upper L i m i t s 5850  •^burst  •^spiral  SDF1  1.47 ( ± 5.94)  2.15 ( ± 0.92)  < 3.46  < 1.65  < 1.79  SDF2  - 7 . 8 4 ( ± 6.36)  1.01 ( ± 0.94)  < 2.64  < 1.65  < 1.77  SDF3  6.10 ( ± 6.48)  - 0 . 2 2 ( ± 0.88)  < 1.63  < 1.83  < 1.97  SDF4  1.73 ( ± 5.76)  1.85 ( ± 0.98)  < 3.45  < 1.66  < 1.76  Mean  0.42 (±3.06)  1.15 (±0.46)  < 1.76  < 1.61  < 1.74  Table 3.2: Measured flux densities w i t h errors and predicted redshift upper limits for the four S D F H E R O sources. The fourth column lists the 95 per cent Bayesian upper limits (in m J y ) at 850 (im.  T h e last two columns list the starburst  galaxy and spiral galaxy model-estimated redshifts, respectively, based on the upper l i m i t flux of each source. T h e b o t t o m row lists the error-weighted mean of the 8 5 0 / i m fluxes, the 95 per cent Bayesian upper confidence l i m i t to this mean flux, and the model-estimated redshift based on this upper l i m i t .  CHAPTER  3.6.1  3.  HEROS  54  Number Counts of Submillimetre Sources  B y looking at number counts of submillimetre sources above some threshold flux level we can predict how many sources we would see at the measured flux level of the H E R O s . Based on interpolations from 850 i/m cumulative source counts (see F i g . 1.6) we expect to find « 4 ± 2 sources brighter than some flux threshold S, which we observe to be S > 1.76 m J y (the 95 per cent upper l i m i t to the average measured flux of the H E R O s ) . So the flux level we measure for the H E R O s is consistent w i t h the number of H E R O s i n the S D F .  3.6.2  Could the H E R O s be Starburst Galaxies?  We use the modified starburst template to create F i g . 3.8, a plot of expected 850 /tm flux as a function of redshift for a H E R O which is representative of the whole population w i t h the average colour and magnitude of the four-source sample. Since we cannot claim detection above 3 a of any of our faint sources, Bayesian 95 per cent upper confidence l i m i t flux estimates can be calculated for each source by integrating over the non-negative flux regions of a normalized Gaussian probability function. These upper flux limits were provided earlier i n Table 3.2. C o m b i n i n g the results from the four sources allows us to obtain an average object flux for our sample of H E R O s .  T h e error-weighted average flux density for the whole  sample at 850 am is 1.15 ± 0.46 m J y , and has a 95 per cent upper l i m i t of 1.76 m J y . T h e average flux density at 450 m J y is 0.42 ± 3.06 m J y . W h e n we force our template S E D to match this same 850 fj,m flux, constrained by a faint K' magnitude and very red J — K' colour, we estimate a corresponding redshift of 1.6. T h i s result is consistent w i t h the median redshift estimates for S C U B A sources (see e.g. D u n l o p [23]). Based on the assumption that hyper-extremely red objects are well-represented by an S E D that fits Arp220, then they are at a redshift less than ~ 1.6. F i g . 3.9 investigates how much we need to brighten the S E D to get the correct K' magnitude after adding i n an amount of dust to the Arp220 S E D to get the correct  CHAPTER  3.  HEROS  10.00  F  0.01 I 0.0  55  1  1  i  1  i  I  1  ,  i  1  |  0.5  1  ,  1  i  1  i  — r — i — i — i — i — | — ' — i — i — ' — q  ,  |  i  1.0  ,  i  ,i l  1.5  i!  .  >  ,  I  2.0  Redshift z  Figure 3.8: A plot of expected 850 pm versus redshift for our average H E R O . Calculated data points are marked w i t h plus signs and the asterisks represent where we read off a redshift, based on the mean and upper l i m i t fluxes. W e expect this semilog plot to have a linear shape since we add dust to the SED by applying a power-law function to the template. We expect a range of models to occupy a " b a n d " around this line, predicting different flux levels for a given redshift.  CHAPTER  3.  HEROS  56  1 00.00 F  1 0 . 0 0 b-  o  1.00  0.10  0.01 0.0 Redshift z  Figure 3.9: A plot of the brightness i n the A " - b a n d (giving the correct K'  magnitude)  compared to the brightness of Arp220 (after adding i n the right amount of extra dust extinction included to obtain the observed J — K' colour) versus redshift for our average H E R O . Based on the assumption that this model describes the H E R O s , then they are approximately as luminous as A r p 2 2 0 .  CHAPTER  3. HEROS  57  J — K' colour. We notice that we need the K' flux to be dimmer than that of this 'dustenriched' Arp220 until z ~ 1.5, where the H E R O starts becoming intrinsically brighter and more luminous than the dusty template. For z ~ 1.6, we need to m u l t i p l y the S E D by a factor of ~ 1.5 to get the correct K' magnitude for the H E R O sample, which is reasonable (i.e. not brighter than a factor of a few) since we believe that more distant galaxies w i l l be more intrinsically luminous ( B l a i n et al. [3]). E v e n the brightest H E R O in this sample, S D F 1 , only needs to be brightened by a factor of ~ 3 at its upper redshift l i m i t of z ~ 1.65 F i g . 3.10 shows that the amount of dust reddening applied to the A r p 2 2 0 S E D i n order to match J — K' decreases w i t h redshift. (Recall that our A r p 2 2 0 template S E D is comprised of Lagache et al.'s (2002) starburst template number 30 w i t h dust added v i a equation (3.1) w i t h m = 0.5). After z ~ 1, the amount of required dust decreases very slowly but never reaches zero. T h e curves show an asymptotic behaviour, illustrating that some amount of dust extinction must always be applied to the S E D . A t z ~ 1.6, we add i n only  0.4 magnitudes i n K' and « 1.2 magnitudes i n J. For the reddest  H E R O , S D F 4 , we need to add « 1.8 magnitudes of dust i n J a t the upper redshift l i m i t of z = 1.66. T h e H E R O s could be about the same luminosity as A r p 2 2 0 and only slightly more dusty. A g a i n , we stress that the model possesses an advantage over more complicated m u l t i parameter models i n that it fits known sources well, despite its single-parameter simplicity. In F i g . 3.11, we compare our S E D w i t h that of Totani et a l . [63] to get an idea of the amount of uncertainties i n our model and which way a more sophisticated model would affect the results. We note that we must add about 1.80 magnitudes i n J and 0.57 magnitudes i n K' i n order to obtain the correct colour of 4.12 at a redshift of z = 2.3. T h i s results i n a predicted 850 /zm flux of ~ 0.5 m J y , close to predicted value of 1 m J y that T o t a n i et al. [63] determined. In comparison, when we use the A r p 2 2 0 S E D and place it at this same redshift of z = 2.3 we predict a 850 /zm flux of « 20 m J y . T h i s demonstrates how different redshifts will be inferred, given the large difference in the predicted 850 / i m flux, depending entirely on the type of S E D that is chosen. T h e result  CHAPTER  3.  HEROS  58  Figure 3.10: T h e amount of dust extinction that required to redden our we our template S E D to match the data versus increasing inferred redshift at the J (solid line) and K' (dotted line) rest wavelengths. T h e curves display an asymptotic behaviour to values above zero, illustrating that some amount of dust extinction must always be applied to the S E D .  CHAPTER  3.  HEROS  59  of 0.5 m J y from the typical elliptical S E D is plausible given our measured average H E R O flux.  T h e H E R O s could i n fact be typical dusty starburst galaxies at z ~ 2.3 which  formed at z ~ 3 and are the progenitors of present-day giant ellipticals, as T o t a n i et a l . [63] suggest. We make note here of some caveats w i t h this model and address some issues. We have assumed that the far-IR luminosity does not increase or decrease as we alter the amount of dust or extinction i n the optical/near-IR portion of the S E D v i a equation 3.1. Performing numerical integration over the S E D of F i g . 3.2 reveals that the total power absorbed in reddening the starburst galaxy template is only about 3 — 5 per cent of the t o t a l galactic emission. T h e small increase i n the far-IR luminosity can be neglected as )  the reddening is altered. We also scale the galaxy brightness uniformly across the spectrum, which inherently assumes that intrinsic properties such as temperature and density remain constant, which essentially has the effect of simply changing the physical size of the galaxy. T h e simple one-to-one relation between dust temperature and I R luminosity which is widely used empirically is also assumed here and is likely to break down for high-redshift galaxies from a theoretical standpoint (see Totani & Takeuchi [62]). Dust grains in a real galaxy always span a range of temperatures (20-200K), but in reality a large amount of it is cold (i.e. at the lower end of the 20-200K range). Therefore, the underlying assumption of a constant temperature seems reasonable here. T h e redshift upper l i m i t would be weakened if the dust temperature was allowed to increase over a modest range. A s alluded to earlier, high-redshift source S E D s may be systematically different, since p r i m o r d i a l galaxies w i l l contain smaller-sized and more chemically simpler dust which w i l l heat u p to higher temperatures due to the higher reflectivity and grain physics involved (see for e.g. Takeuchi et al. [61]). Since the H E R O s are extreme examples of E R O s , it is useful to consider how the results would change by fitting an ERO-specific S E D to the data aswell. H R 1 0 was detected by H u & Ridgway [37], has subsequently been well-studied at many wavelengths, and was the first E R O discovered to be associated w i t h a U L I R G . N o t suprisingly, the  CHAPTER  3.  HEROS  1 1  60  i  i  i 111111 i  n—i i 111111  1 — i i 1 1 1 i|i |  1—i—i 111111  Observed Wavelength  1—i i 1 1 1 1 1  [fim]  Figure 3.11: W e scale and plot the Totani et a l . [63] model of a typical elliptical galaxy w i t h present-day absolute magnitude of M  B  = —20, J — K' colour of 4, and  formation redshift z? = 3 at a redshift of z = 2.3. T h e model contains emission from the direct stellar light that survived absorption by dust, as well as emission from heated dust. We scale the brightness to match the A'-band brightness of S D F 4 and adjust the J — K' colour slightly to that of S D F 4 (4.12) using our extinction law ( c f . E q n . 3.1) (solid line). For comparison, we plot Arp220 (dashed line) at z = 2.3 and adjust the brightness a n d . colour to match that of S D F 4 . T h e three vertical lines represent (from left . to right), the J-band, the A^'-band, and 850 / m i . T h i s plot demonstrates how different redshifts will be inferred, given the large difference i n the predicted 850 / a n flux, depending entirely on the type of S E D that is chosen.  CHAPTER  3.  HEROS  61  S E D for H R 1 0 matches the S E D shape of Arp220, a nearby U L I R G , but is brighter (i.e. more intrinsically luminous) by a factor of 3.8 (see F i g . 3.12 and E l b a z et al. [28]). H R 1 0 lies at a redshift of z — 1.44 and is thought to be no more than a distant clone of Arp220. If we examine this template specific to E R O s , and fit it to the H E R O s , we find that the brightness factor is reduced to the H E R O s being 0.34 times at bright as H R 1 0 as opposed to being 1.3 times as bright as Arp220 at a redshift of 1.6. M a i h a r a et a l . [46] state that the H E R O s are likely to be remote galaxies, based on their "stellarity" indices and their apparent spatial extent. F r o m the S D F A ' ' - b a n d image, we estimate the spatial extent of the H E R O s to be between 1.2 to 2.0 arcsec i n diameter. Based on this estimate from the image, if the H E R O s are at z ~ 1.6, then this would make them ~ 4 — 6.5 kpc i n size. T h i s is consistent w i t h the known range of galaxy sizes of 1 kpc up to 50 kpc.  3.6.3  Could the HEROs be Normal Spiral Galaxies?  We use the modified normal spiral galaxy template to create F i g . 3.8, a plot of expected 850 //m flux as a function of redshift for a H E R O which is representative of the whole population w i t h the average colour and magnitude of the four-source sample. A s before, we use the Bayesian 95 per cent upper confidence l i m i t flux estimates to derive the upper redshift limits listed i n the last column of Table 3.2. W h e n we force our template S E D to match the 95 per cent upper l i m i t to the 850 fim flux of 1.76 m J y , constrained by a faint K' magnitude and very red J — K' colour, we estimate a corresponding redshift of 1.74. Based on the assumption that hyper-extremely red objects are well-represented by an S E D that fits N G C 3 9 3 8 , then they are at a redshift less than ~ 1.74. F i g . 3.14 investigates how much we need to brighten the S E D to get the correct K' magnitude after adding in an amount of dust to the N G C 3 9 3 8 S E D to get the correct J — K' colour. We notice that we need the K' flux to be dimmer than that of this 'dustenriched' N G C 3 9 3 8 until z ~ 0.6, where the H E R O starts becoming intrinsically brighter  CHAPTER  3.  HEROS  10-  1  62  10°  10  1  10  2  Observed Wavelength  10  3  10  4  10  5  [/i.m]  Figure 3.12: T h i s is a model of H R 1 0 based on an Arp220 S E D brightened by a factor of 3.8 to fit the S E D of H R 1 0 at a redshift of z = 1.44. N o dust has been added to this model. T h e crosses mark photometric data points and the arrows denote upper limit flux estimates from N E D .  CHAPTER  3.  HEROS  10.00 F  0.01 I 0.0  63  1  1  i  1  i  I  1  i  i  I 0.5  1  1  ,  1  ,  1  i  I  i  1  I 1.0  Redshift  I— — — — —q  1  1  ,  i  i  ,  1  L_1J 1.5  1  1  U_J  i _ l 2.0  z  Figure 3.13: A plot of expected 8 5 0 / i m versus redshift for our average H E R O , based on the normal spiral galaxy template. T o reach a flux level of 1.76 m J y at 850/tm, the template must be moved out to z ~ 1.74. U s i n g the Arp220 S E D , the template only needed to be moved out to z ~ 1.61 i n order to reach the same 850 am flux. Calculated data points are marked w i t h plus signs and the asterisks represent where we read off a redshift, based on the mean and upper l i m i t fluxes.  CHAPTER  3.  HEROS  64  Figure 3.14: A plot of the brightness i n the i f ' - b a n d (giving the correct K' magnitude) compared to the brightness of spiral galaxy N G C 3 9 3 8 (after adding i n the right amount of extra dust extinction included to obtain the observed J — K' colour) versus redshift for our average H E R O .  CHAPTER  3.  HEROS  65  Figure 3.15: T h e amount of dust extinction that we must add to our normal galaxy S E D template for increasing redshift at the J (solid line) and K'  (dotted line)  rest wavelengths. and more luminous than the dusty template. For z ~ 1.7, we need to m u l t i p l y the S E D by a factor of ~ 30 to get the correct K' magnitude for the H E R O sample. T h e brightest H E R O , S D F 1 , requires the S E D to be brightened by a factor of 60 at its upper redshift l i m i t of z = 1.79. F i g . 3.7 reveals the energetics of the normal cold spiral galaxy, and it is clear that there is just as much power in the near-IR peak as i n the far-IR peak. T h i s was not the case in the starburst galaxy template, where the far-IR peak dominated the energy output of the galaxy. W i t h this in m i n d , it is clear that for such galaxies we cannot  CHAPTER  3.  66  HEROS  ignore the effect of dust extinction on the total power of the galaxy. T h e power that would be "lost" due to reddening the spiral galaxy must be added back into the far-IR region of the spectrum in order to conserve energy. We calculate the t o t a l power of the galaxy before adding dust, and calculate the power in both the far-IR and near-IR peaks after adding some dust to the galaxy. We subtract the power i n the near-IR peak from the total power "before dust" to find the amount of energy that must be elsewhere i n the spectrum i n order to conserve energy. We achieve this by first splitting the spectrum into two different components (the near-IR bump and the far-IR bump) and m u l t i p l y i n g each b u m p by a smooth function of the form:  (3.7)  A(s) = e<£>" for a decreasing exponential, or f (x) 2  = 1 - e<£>  n  (3.8)  for an increasing exponential. A* is the cut-off wavelength where the function drops off steeply and n controls the steepness of the function's drop-off (we choose a large enough value, n = 20, to get sufficient steepness). Subsequently, we add the two components back together to create one unified template. We take the ratio of the "missing power" (i.e. the power not i n the near-IR peak) to the amount of energy i n the far-IR to find the far-IR "boosting" factor. M u l t i p l y i n g the far-IR bump by this factor effectively moves the entire far-IR vF  v  spectrum up by the same amount. B y conserving energy i n this  way, the total power absorbed i n reddening the galaxy is extremely negligible.  Even  though this scaling method is rather crude, it avoids changing the temperature and other intrinsic physics of the galaxy. F i g . 3.15 shows that the amount of dust reddening applied to the N G C 3 9 3 8 S E D in order to match J — K' decreases w i t h redshift. (Recall that our N G C 3 9 3 8 template S E D is comprised of Lagache et al.'s (2002) normal cold spiral galaxy template w i t h dust added v i a E q n (3.1) w i t h x = 0.1). After z ~ 1, the amount of required dust decreases very slowly. A t z = 1.74, we add in only « 0.7 magnitudes in K' and « 1.2 magnitudes  CHAPTER  3.  HEROS  67  i n J. For the reddest H E R O , S D F 4 , we need to add « 1.8 magnitudes of dust i n J at the upper redshift l i m i t of z = 1.76 N G C 3 9 3 8 is not a reasonable analogue for the H E R O s , since we need to brighten the galaxy by an extraordinary amount to fit the entire S E D (i.e. more t h a n a factor of a few). A n 8 5 0 / i m flux of 1.76mJy indicates that the galaxy lies at z = 1.74, the galaxy would need to be intrinsically brighter by a factor of about 30 compared w i t h a normal spiral galaxy, not a very realistic scenario for this type of galaxy. T h e possibility of the H E R O s being normal spiral galaxies is therefore eliminated. Even if N G C 3 9 3 8 was at a lower redshift and would therefore not need to be brightened by such a large factor, we would need to add i n more dust than would be reasonable (i.e. more than one or two magnitudes of extinction i n J, c f . F i g . 3.15). If the flux is just above the confusion l i m i t at ~ 0.6 m J y then the redshift is about z ~ 1.3 and the galaxy would be still need to be brighter than N G C 3 9 3 8 by a factor of 8 and contain about 1.3 magnitudes of extinction i n J-band.  CHAPTER  4.  SDF  SOURCES  68  CHAPTER 4 SDF SOURCES A n image is "confused" when multiple unresolved faint objects cluster i n one beam-size. A s we look fainter and fainter, d i m objects become more numerous, superimposing their signals on each other, until the confusion l i m i t is reached. We investigate the number of pixels w i t h i n a certain flux level in F i g . 4.1, a histogram of the number of pixels at each flux level i n the 850 / i m S D F map. T h e positively skewed non-Gaussian single peak distribution of flux demonstrates that a population of submillimetre sources lies i n the positive t a i l of the distribution, below the confusion l i m i t of these d a t a (Condon [15]). If we subtract the flux bins reflected about 0 m J y from the original flux bins, this effect becomes apparent (see F i g . 4.2). We can therefore infer that a population of sources lies just below the confusion l i m i t of the 850 / i m S D F map. Figs. 4.3 and 4.4 demonstrate this also for the 450 / i m map. T h i s result hints that we are detecting submillimetre emission from the S D F and that it could be correlated w i t h S D F data i n another wavelength region.  4.1 Correlating the IR Galaxies with the Submillimetre Data Kashikawa et a l . [43] have constructed a deep i f ' - b a n d selected B,V,R,I,z',J,K' colour sample of 439 galaxies (K'<  multi-  24.0) in the Subaru Deep F i e l d and estimated a  photometric redshift for each galaxy using the public domain HYPERZ code written by Bolzonella, Miralles & Pello [5]. HYPERZ finds the redshift of a galaxy using a standard S E D fitting procedure, i.e. comparing the observed magnitudes w i t h those computed from template S E D s .  Throughout this thesis we use the more complete, unpublished  526 if'-selected galaxy catalogue based on the criterion of K{  <  sophot  24.6 (Kashikawa  2002, private communication). In F i g . 4.5, we plot the mean of the 20 brightest i f ' - b a n d  CHAPTER  4.  SDF  250  SOURCES  T  1  69  1  1  1  r  Figure 4.1: We plot the number of pixels at each brightness level i n the 850 pm map as a histogram w i t h 100 bins of w i d t h 0.2 m J y (solid jagged line). T h e overlayed dashed histogram represents the flux bins reflected about 0 m J y , demonstrating that there is clearly more positive flux i n the map than negative flux. T h e flux distribution is clearly non-Gaussian and has enhanced high and low flux tails when compared to the overlaid Gaussian. T h e Gaussian fit to the data (smooth solid line) has a height of 186 ± 0.27, center of - 0 . 1 2 ± 2.3, and a half-width half-maximum of 2.26 ± 0.004.  CHAPTER  4.  SDF  SOURCES  70  Figure 4.2: We plot the difference of the flux bins and the flux bins reflected about 0 m J y . If we take the difference between a positively skewed single peaked distribution and its reflection about O m J y , we reproduce this result. We can therefore infer a population of sources lying just below the confusion l i m i t of the 850 / i m map.  CHAPTER  4.  SDF  1 20 I  SOURCES  1  1  1  I  1  71  1  1  I  1  1  1  Flux  1  1  1  1  1  1  1  '  1  1  '  r  Bins ( m J y )  Figure 4.3: We plot the number of pixels at each brightness level i n the 450 pm map as a histogram w i t h 120 bins of w i d t h 1.0 m J y (solid jagged line). T h e overlayed dashed histogram represents the flux bins reflected about 0 m J y , demonstrating that there is clearly more positive flux i n the map t h a n negative flux. T h e flux distribution is clearly non-Gaussian and has enhanced high and low flux tails when compared to the overlaid Gaussian. T h e Gaussian fit to the data (smooth solid line) has a height of 93 ± 0.27, center of - 0 . 7 3 ± 0.04, and a half-width half-maximum of 11.4 ± 0.04.  CHAPTER  4.  SDF  SOURCES  72  c C 'D >  o  I  "D (D  x> o o  Q XI  c o  Q c in  CQ  O  _  -60  -40  -20 Flux  0 20 Bins (mJy)  40  60  Figure 4.4: We plot the difference of the flux bins and the flux bins reflected about 0 m J y . If we take the difference between a positively skewed single peaked distribution and its reflection about 0 m J y , we reproduce this result. W e can therefore infer a population of sources lying just below the confusion l i m i t of the 450 fim map.  CHAPTER  4. SDF  SOURCES  73  objects i n redshift bins of 0.5 w i d t h , and observe that the faintest i f ' - b a n d objects are 1  also at the highest redshifts. It is interesting to investigate the measured submillimetre emission of the A"'-selected galaxies as a function of redshift, since i t c a n potentially reveal if amounts of hidden star formation evolve w i t h redshift. Peacock et a l . [48] have performed a similar analysis w i t h ultraviolet positional information for the Hubble Deep F i e l d and found little correlation at z < 1, but a definite positive signal spread out over higher redshift bins (z > 2.5). T h i s demonstrates that their submillimetre m a p of the H D F receives emission from galaxies over a wide range of redshifts. A A ' - b a n d source which is very faint or undetected i n optical data (i.e. J or / - b a n d ) holds a strong possibility of being S C U B A bright - or at least we know the converse to be true. W e currently do not know how t o predict which E R O s w i l l be S C U B A - b r i g h t . W e would like to test if faint near-IR emission is correlated w i t h submillimetre emission. W e might expect higher redshifted near-IR objects to be stronger submillimetre sources, since the far-IR peak of the S E D w i l l be significantly shifted into the submillimetre regime. T h e best way t o see if we are detecting IR-galaxy emission statistically i n the 850 /mi map is to measure the flux density at the location of each galaxy and look for a correlation. We take the value from the map at the position of each I R source as the best estimate of its 850 /im flux density. These flux densities are then averaged or "stacked" into redshift bins (see F i g . 4.6). W e also perform the same analysis on the 450/mi map (see F i g 4.7). B y dividing the maps up into redshift slices, we can now see if one redshift band tends to dominate more i n the submillimetre than another. In order to quantify the degree of correlation we use the linear Pearson correlation coefficient, a number between —1 and +1 which measures the degree to which two variables are linearly related using a least-squares fit. A value of 0 indicates no correlation and values of —1 and +1 indicate a strong correlation (inverse proportionality and direct proportionality, respectively). W e weight the mean fluxes by a factor pf 1/cr (cr is the error bar on each bin) i n order to 2  take the error bars into account before computing the correlation coefficient. We obtain 1  We choose the number of bins and the bin occupancy based on the best compromise between noise  reduction and a sensible number of points.  CHAPTER  4. SDF  SOURCES  2 4  74  11 1111111  111111 i  o E  »  111 1 1 1 1 1 1 1  J  h  2 2  &  1111111111  2 0  1 6 I n 0  111 1  1111  I 2  i i i 111  3  I 4  I 5  6  Redshift z  Figure 4.5: A plot of the 20 brightest i f ' - b a n d objects per redshift b i n . T h e vertical error bars show the 1 a standard deviation of the K' magnitudes i n each redshift bin. T h e asterisks mark the mean redshift of each b i n and the horizontal error bars represent the coverage of each redshift b i n . T h e 7th b i n only contains 14 objects due to the low number of objects i n this redshift b i n . We remark that the the faintest i f ' - b a n d objects also have the highest photometric redshifts.  CHAPTER  4.  SDF  SOURCES  75  E o LD  oo  Redshift  z  Figure 4.6: A plot of observed average 850 //m flux per redshift b i n (Az  = 1). Starting  w i t h the lowest redshift bin, the number of galaxies contained i n each b i n is 183, 150, 100, 36, 46, 6, 4, 1. It is evident that the highest redshift bins contain the smallest number of objects and so w i l l have the largest error bars. T h e lowest redshift bins contain a large number of objects and so the average flux w i l l tend to be closer to 0 m J y , since the measurement is beginning to be constrained by the map average.  CHAPTER  4.  SDF  SOURCES  76  E  a.  o  in X  - 1 0 h  Redshift z  Figure 4.7: A plot of observed average 450fim flux per redshift bin (A.2 = 1). Starting w i t h the lowest redshift bin, the number of galaxies contained in each bin is 183, 150, 100, 36, 46, 6, 4, 1. It is evident that the highest redshift bins contain the smallest number of objects and so w i l l have the largest error bars. The lowest redshift bins contain a large number of objects and so the average flux w i l l tend to be closer to 0 m J y , since the measurement is beginning to be constrained by the average of the map.  CHAPTER  4.  SDF  77  SOURCES  correlation coefficients of  0.49 and 0.38 at 850 and 450 jura respectively.  In order to see how likely it is to achieve these outcomes and correlation coefficients simply by chance, we perform 1000 statistical Monte C a r l o simulations of this procedure (see Figs. 4.8, 4.9, 4.10, 4.11). T h e Monte C a r l o simulation selects positions (the same number of times as there are objects in the bin) from the S D F map at random and measures the flux there. For each b i n , we calculate the mean and error, again weighting the measurements by the corresponding value on the noise map. We calculate a correlation coefficient between the error-weighted  (1/tr ) mean flux and redshift b i n for a complete 2  set of redshift bins and repeat the whole procedure 1000 times i n order to get a welldefined distribution of correlation coefficients. It is now possible to estimate how likely it is to obtain a certain correlation coefficient at random. L o o k i n g at these plots, getting a correlation coefficient of 0.49 or higher by chance occurs about 80 per cent of the time for the 850 pm map and getting a coefficient of 0.38 or higher by chance occurs about 40 per cent of the time for. the 450 fxm map.  From  these results, there is no strong case for a correlation between the stacked submillimetre fluxes and redshift. A l t h o u g h there does seem to be a hint of positive flux spread over the redshift range 3 < z < 6 in the 450 pm map (see F i g . 4.7), these results suggest that the flux we detect i n the submillimetre is not well-described by the AT'-band galaxy photometric redshifts. Because the first few bins each contain a large number of objects, we are inherently just taking the average of the map, and we therefore expect a stacked flux close to 0 m J y . T h e bins w i t h fewer objects w i l l inherently posess very large error bars, as they suffer from small number statistics. It is therefore more useful to rebin the objects into bins w i t h an equal number of objects. We redo the analysis for Figs. 4.6 and 4.7 but instead divide the objects into bins w i t h equal numbers of objects to see if this makes a trend visible (see Figs. 4.12 and 4.13). We get correlation coefficients of 0.40 and 0.43 for the 850 and 450 fj,m data-sets, respectively. We perform 1000 Monte C a r l o simulations again (see Figs. 4.14 and 4.16), but this time w i t h an equal number of objects per b i n , 2  2  All bins contain 40 objects except for the last bin which contains 46 objects, mimicking how the real  CHAPTER  4.  SDF  SOURCES  78  250  n  1 1  _i  I  1  1  1  1  r  200  150 CL  I  100  50  •1.0  -0.5  i_  0.0  0.5  1.0  Correlation Coefficient Bin  Figure 4.8: T h i s plot shows the distribution of correlation coefficients obtained from 850 / i m stacked fluxes versus redshift for a set of Monte C a r l o simulations of stacked 850/xm flux in redshift bins of Az = 1.  CHAPTER  4.  SDF  SOURCES  79  Figure 4.9: T h i s plot shows how many times a correlation coefficient is obtained for the 850 / i m stacked fluxes versus redshift from a set of Monte C a r l o simulations of stacked 850 / i m flux i n redshift bins of Az = 1.  CHAPTER  4. SDF  100  SOURCES  -i  1  80  r  i  1  1  r  80  c  CD  60  <D CL  c  Q  40h  20 h  0l_C -1.0  —i  -0.5  i  i_  _L _J  I  1_  _l  0.0  I 0.5  L_  I  I  I  1.0  Correlation' Coefficient Bin  Figure 4.10: T h i s plot shows the distribution of correlation coefficients obtained from 450 fim stacked fluxes versus redshift for a set of Monte C a r l o simulations of stacked 450 pm flux i n redshift bins of Az = 1.  CHAPTER  4.  SDF  SOURCES  81  Correlation  Coefficient  R  Figure 4.11: T h i s plot shows how many times a correlation coefficient is obtained for the 450 / i m stacked fluxes versus redshift from a set of Monte C a r l o simulations of stacked 450 / i m flux in redshift bins of Az — 1.  CHAPTER  4.  SDF  SOURCES  82  in order to evaluate the likelihood of getting these correlation coefficients at random for b o t h maps. U p o n examination of the output of the simulations (see Figs. 4.15 and 4.17), obtaining a correlation coefficient of 0.40 or higher by chance occurs about 15 per cent of the time for the 850 / m i map, and obtaining a coefficient of 0.43 or higher by chance occurs about 15 per cent of the time for the 450 / m i map. These results seem more suggestive of a flux-redshift correlation, but we are not confident enough to reject the possibility that these correlation coefficients arose randomly. Therefore, we ascertain that the submillimetre flux we detect is not very well characterized by the photometric redshifts of the i f ' - b a n d selected galaxies. Using the table of I R positions, multi-colour magnitudes, and photometric redshift estimates, we investigate any other possible correlations of I R position w i t h the flux at the corresponding position in our S C U B A map. We investigate possible correlations of stacked submillimetre flux w i t h K' magnitude of an equal number of objects per b i n (see Figs. 4.18 and 4.19). We get correlation coefficients of —0.08 (a null result) and —0.76 for 850 and 450 /tm respectively. F i g . 4.18 shows no trend of increasing 850 /tm flux w i t h A ' - b a n d faintness. F i g . 4.19 does not show conclusive evidence for a connection between 450 / m i flux and AT'-band brightness but there is a slight hint that brighter A ' - b a n d objects have slightly higher (and more positive) 450 / m i fluxes. F i n a l l y , we investigate if there is any correlation of stacked submillimetre flux w i t h absolute V magnitude (My)  for different redshift cuts. We look to the F - b a n d this time,  as it is a measure of the absolute optical luminosity and galaxy mass if the galaxy is not too heavily enshrouded in dust, to see if there is a stronger submillimetre signal correlation w i t h optically luminous objects.  We use absolute magnitude because it removes the  d i m m i n g effect of distance, resulting in a better representation of the power output of the galaxy. W e calculate the absolute magnitude by using the well-known equation: m  v  - M  v  = 5 l o g D - 5,  (4.1)  where D is i n parsecs. We note that since we do not know the shape of each galaxy's data was sampled and binned (see Figs. 4.12 and 4.13).  CHAPTER 4. SDF SOURCES  0.5 I  83  —1  1  r  0.0  E 3. o m oo X 3  -0.5  -1.01 Redshift z  Figure 4.12: A plot of observed average 850 / i m flux per redshift b i n w i t h an equal number of objects per b i n . A l l bins contain 40 objects except for the last which contains 46 objects. T h e asterisks mark the mean redshift of each b i n and the horizontal bars represent the extent of bin redshift coverage. T h e vertical error bars represent the flux errors i n each redshift bin. A horizontal line through (0,0) is drawn to guide the eye to easily distinguish the number of redshift bins above and below a flux level of 0 m J y .  CHAPTER 4. SDF SOURCES  84  6  E a. o m  3  0  Mr*  _J  I  L_ _J  Redshift  1  L_  z  Figure 4.13: A plot of observed.average 450 fim flux per redshift b i n w i t h an equal number of objects per bin. A l l bins contain 40 objects except fot the last b i n which contains 46 objects. T h e asterisks mark the mean redshift of each b i n and the horizontal bars represent the extent of b i n redshift coverage. T h e vertical error bars represent the flux errors in each redshift b i n . A horizontal line through (0,0) is drawn to guide the eye to compare the number of redshift bins above and below a flux level of 0.  CHAPTER  4.  SDF  SOURCES  1001  80  1  1  1  85  i  1  1  1  1  1  —i— — —'— —r 1  1  1  ~i  1  I  !  r~  h  60  2  40  20  OLLJ -1.0  i  i  i  L -0.5  _1  I  I  I  0.0  I 0.5  L_J  L_  1.0  Correlation Coefficient Bin  Figure 4.14: T h i s plot shows the distribution of correlation coefficients obtained from 850 u.m stacked fluxes versus redshift for a set of Monte C a r l o simulations of stacked 850 fxm flux w i t h an equal number of objects per redshift b i n .  CHAPTER  4.  SDF  SOURCES  86  Correlation  Coefficient  R  Figure 4.15: T h i s plot shows how many times a correlation coefficient is obtained for the 850 / i m stacked fluxes versus redshift from a set of Monte C a r l o simulations of stacked 850 / i m flux w i t h an equal number of objects per redshift bin.  CHAPTER  4.  SDF  SOURCES  87  100  n  1  r  ~i  1  r  _j  i  i_  ~i  1  r  _j  i  i_  -  80  c CO  60 h  CL c Q  40  h  20  0 -1.0  J_ -0.5  0.0  0.5  1.0  Correlation Coefficient Bin  Figure 4.16: T h i s plot shows the distribution of correlation coefficients obtained from 450 / i m stacked fluxes versus redshift for a set of Monte C a r l o simulations of stacked 450 / i m flux w i t h an equal number of objects per redshift b i n .  CHAPTER  4.  SDF  SOURCES  88  Correlation  Coefficient  R  Figure 4.17: T h i s plot shows how many times a correlation coefficient is obtained for the 450 / i m stacked fluxes versus redshift from a set of Monte C a r l o simulations of stacked 450 /tm flux in bins w i t h an equal number of objects per redshift bin.  CHAPTER  4. SDF  SOURCES  89  E a. o m CO  Figure 4.18: T h i s is a plot of stacked 850 /xm flux measured i n K magnitude bins containing 48 objects each (except for the last bin which contains 49 objects). T h e asterisks mark the mean K magnitude value of each b i n and the horizontal bars represent the extent of b i n K magnitude coverage. T h e vertical error bars represent the flux errors i n each K magnitude bin. T h e correlation coefficient of —0.08 suggests a null result.  CHAPTER  4. SDF  SOURCES  -4 I  16  ,  ,  i  90  I  ,  ,  L_J  ,  18  20  i  ,  I  22  ,  ,  ,  I  24  K mag  Figure 4.19: T h i s is a plot of stacked 450 pm flux measured i n K magnitude bins containing 48 objects each (except for the last bin which contains 49 objects). T h e asterisks mark the mean K magnitude value of each b i n and the horizontal bars represent the extent of bin K magnitude coverage. T h e vertical error bars represent the flux errors i n each K magnitude b i n . T h e correlation coefficient of —0.76 is slightly suggestive of a relation of increasing 450 fim flux w i t h brighter K magnitudes.  CHAPTER  4.  SDF  SOURCES  91  S E D , we have not included a K-correction term that corrects for the effect of light i n a wavelength band of interest being redshifted to longer wavelengths.  T h i s effect is  obviously more pronounced at higher redshifts. We take D to be the luminosity distance, which can be easily found by using the redshift information. F i g . 4.20, shows a series of scatterplots of 850 /xm flux versus V - b a n d absolute magnitude w i t h different redshift cuts of z > 0 (all objects), z > 2, and z > 4. We notice that the mean 850 /xm flux does not vary w i t h My,  but this can be more clearly seen if we stack the submillimetre flux.  T h e leftmost plots i n Figs. 4.21 and 4.22 represent stacked 8 5 0 / a n flux for equal-size bins of My.  These plots generally give null results, except for a hint of a correlation  for objects w i t h z > 4 of 850/xm flux increasing w i t h the faintness in My.  T h i s would  be plausible considering that strong submillimetre detections are associated w i t h a great deal of dust extincting the optical light.  T h e rightmost plots i n Figs. 4.21 and 4.22  represent stacked 850 /xm flux w i t h an equal number of objects i n each b i n and all show no trends. We conclude that the lack of a stong trend of submillimetre flux w i t h  My  indicates that the submillimetre emission i n the S D F map is not well-described by the absolute V magnitudes of A ' - b a n d selected objects i n the S D F .  4.1.1  Detections  We use C o l i n B o r y s '  BS_FINDSOUR.CE  program to pick out sources by fitting the point  spread function ( P S F ) of the beam to each pixel on the map.  Bright peaks i n the  convolved (i.e. smoothed) map are selected as "detected" sources when the P S F matches a source i n the m a p (see B o r y s [6] for further details).  W e note detections of three  S N R > 3.0 a S C U B A sources at 450 and 850 /xm combined, but unfortunately they all lie outside of the region covered by this deep A^'-band image (see Figs. 2.8 and 2.9 for the maps). Table 4.1 lists the detections in the 450 and 850 /xm maps. Using the source counts of 850/xm submillimetre sources (cf. F i g . 1.6), we expect to see « 1 source detected at the 3.5 o~ level, given a map rms of 2 m J y . T h i s estimate is consistent w i t h the single 3.5 a detection we make. T h e number counts at 450 /xm are not as well known, however,  CHAPTER  4.  SDF  SOURCES  92  E a.  -20 -15 Absolute V magnitude  -20 -15 Absolute V mognitude  -20 -15 Absolute V magnitude  Figure 4.20: F r o m left to right: these scatter-plots represent 850 am flux for every object detected in the V - b a n d (488 objects total), for objects detected i n the Vband w i t h a photometric redshift higher than 2 (158 objects total), and for objects detected i n the V - b a n d w i t h a photometric redshift higher than 4 (31 objects total).  CHAPTER 4. SDF SOURCES  93  0.6 0.4 E  0.2  £  0.0  LO CO  X  r  -0.2 r  * *  -0.4 r -0.6 -25 -20 -15 -10 Absolute V magnitude  -25 -20 -15 -10 Absolute V magnitude  Figure 4.21: T h e leftmost plot represents stacked 850 fxm flux measured i n M y bins, two magnitudes i n w i d t h , for all objects detected i n the F - b a n d (488 objects total).  T h e rightmost plot represents 850/zm flux measured in M y bins,  w i t h an equal number of objects per bin (49), for all objects detected i n the V - b a n d . T h e horizontal bars represent the actual range of magnitudes i n a bin and the vertical error bars show the standard deviation of the mean stacked flux i n each bin.  CHAPTER  4. SDF SOURCES  94  E 3. o in oo X 3  -25 -20 -15 -10 Absolute V magnitude  -25 -20 -15 -10 Absolute V magnitude  -20 -15 -10 Absolute V magnitude  -25 -20 -15 -10 Absolute V magnitude  E =1 o tn co X 13  Figure 4.22: T h e leftmost plots represent stacked 850 / i m flux measured i n M  v  bins, two  magnitudes in w i d t h , for all objects detected in the 7 - b a n d w i t h z > 2 (158 objects total) (top plot) and for all objects detected i n the F - b a n d w i t h z > 4 (31 objects total) (bottom plot). T h e rightmost plots represent 850 /xm flux measured in M y bins, w i t h an equal number of objects per b i n , for all objects detected in the F - b a n d w i t h z > 2 (40 objects per bin) (top plot) and for all objects detected in the V - b a n d w i t h z > 4 (10 objects per bin) (bottom plot). T h e horizontal bars represent the actual range of magnitudes in a b i n and the vertical error bars show the standard deviation of the mean stacked flux in each b i n .  CHAPTER  4.  SDF  95  SOURCES  and we make no attempt to estimate how many sources we expect to see at 450 / i m , given the depth of the map and the area surveyed. ID  Position (2000.0)  F l u x (mJy)  SNR  S D F 850.1  13 24 16?91  +27°29'46'.'01  S  8 5 0  = 4.40 ( ± 1.25)  3.5 cr  S D F 450.1  13 24 25?25  +27°30'3r.'01  S  4 5 0  = 75.45 ( ± 18.52)  4.1 cr  S D F 450.2  13 24 19?39  +27°30'3r.'01  S  4 5 0  = 50.29 ( ± 16.50)  3.0 cr  h  h  h  m  m  m  Table 4.1: T h e positions, fluxes and signal-to-noise ratios ( S N R ) of significant (> 3cr) S C U B A detections in the S D F .  CHAPTER  5.  CONCLUSIONS  96  CHAPTER 5 CONCLUSIONS We have mapped the 850 and 450 /zm continuum emission of the Subaru Deep F i e l d using S C U B A on the J C M T . We found that the observed high J - K' (K 3 - 4) sources produce an amount of S C U B A flux which is near to or below the ~ 1 m J y confusion level, m a k i n g their true flux difficult to estimate. The four H E R O sources have an average 850 /zm flux of 1.15 ± 0.46 m J y or a combined 95 per cent upper confidence l i m i t of 1.76 m J y , a hint that they are dusty galaxies emitting a small amount of submillimetre  flux.  Naively, the H E R O s are most likely to be starburst galaxies based on their apparent irregular morphologies rather than compact regularly-shaped elliptical galaxies. Based on the assumption that H E R O s are well-represented by an S E D that fits A r p 2 2 0 , then they are at z < 1.6, consistent w i t h known redshifts for S C U B A sources, but at the low redshift e n d . If we assume instead that the H E R O s lie at average S C U B A redshifts 1  (2 « 2 — 5), then the typical elliptical galaxy model from Totani et al. [63] would also fit the data for at a redshift of z = 2.3. Using an Arp220-like S E D , we need to increase the brightness of the average H E R O by a reasonable amount (about a factor of 1 — 1.5) and add a modest amount of dust (about 1 magnitude i n J ) i n order to obtain an adequate fit. Thus, there is no reason to suspect that these objects are J-dropout galaxies at extraordinary redshifts. T h e H E R O s are not well-represented by an N G C 3 9 3 8 - l i k e normal face-on spiral galaxy, given the flux constraints, since they would need to be brightened by about a factor of about 20 — 30, J  At optical and near-IR wavelengths, deep images are crowded with galaxies with no more than a  few percent of them lying at z > 3 (see Guhathakurta, Tyson & Majewski [34]), as the galaxies do not benefit from the negative K-correction at near-IR wavelengths as they do in the submillimetre. We note that we may be only be picking up the low end of the redshift distribution by selecting objects in the iY'-band.  CHAPTER 5.  CONCLUSIONS  97  a larger than acceptable amount for a galaxy of this nature. Because the H E R O s are faint i n optical-IR-submillimetre wavebands w i t h current instrument sensitivities, these objects may still play an important role i n tracing the dustenshrouded part of the early star formation picture. If the H E R O s are dusty starbursts, as this work suggests, we see that the optically derived star formation rate ( S F R ) would be underestimated due to the presence of dust. O u r strongest detection (S 5o = 2.15 ± 8  0.92 m J y , is the most convincing case for the H E R O s being dusty starbursts. However, they could well be a mixed class of objects. Even for this object, its A ' - b a n d morphology looks superficially to be a merging pair of galaxies with' z < 1.6, as determined from the Arp220 starburst template, consistent w i t h a hierarchical formation scenario, where galaxies are made over a long period of time through successive mergers or through accretion of clumps of matter.  Barger et a l . [2] find that most of the submillimetre  extragalactic background is emitted by 1 < z < 3 sources, so the peak of the starburst activity lies at moderate redshifts. If there are many H E R O s of this type at these redshifts then this has serious implications for the S F R as a function of redshift, since we m a y be missing more than was previously thought. Our measurements of the H E R O s do not support the prediction that H E R O s should be S C U B A - b r i g h t . S C U B A galaxies are very often identified w i t h E R O s and we have tested if the inverse statement also holds true. O u r results clearly demonstrate that measuring the submillimetre flux of H E R O s w i l l not necessarily select very bright S C U B A sources. S C U B A galaxies may just comprise one sub-set of a more diverse population of E R O s . F r o m Figs. 4.1 and 4.3, we infer that we are detecting a population of sources just below the confusion l i m i t of the maps. We correlate the I R A ' - b a n d selected galaxy positions w i t h 450 and 850 / i m flux and find that there is no correlation at the sensitivity l i m i t of our measurements of the S D F . We conclude that the submillimetre flux we detect i n our data is not characterized by the A ' - b a n d selected I R galaxies or by their photometric redshifts, and is presenting a view of the Universe different from that of the I R S D F . A l t h o u g h there is no apparent correlation w i t h the I R S D F galaxies, the S C U B A map could still potentially be correlated w i t h U V , optical, X - r a y or radio data.  CHAPTER  5.1  5.  CONCLUSIONS  98  Future Work  Future S C U B A observations could provide even deeper limits (down to an rms of « l m J y ) , but w i l l be constrained by the confusion l i m i t (~ 0.5 m J y ) .  Deep V L A ob-  servations would allow us to obtain further redshift constraints through the well known r a d i o / f a r - I R correlation  2  (see C h a p m a n et al. [13]). A L M A (submm), S I R T F and J W S T  (IR) w i l l have the required sensitivities to detect these faint objects and possibly reveal their nature and redshifts w i t h wide-band spectroscopy instruments. If the H E R O s are at z < 1.6 then they should be detectable by B L A S T (Balloon-borne Large Aperture Submillimetre Telescope) or w i t h S C U B A at 450 / i m in very good dry weather. Those data, i n addition to C S O / S H A R C - I I 350/tm observations, could help to better constrain the assumed shape of the S E D , thereby providing tighter constraints on the inferred redshift. A more complete study might include the use of a wider range of S E D s , rather than just the two templates that were employed here. In addition, without spectroscopic redshifts, substantial progress may also require model S E D s w i t h a proper treatment of the dust at the same time as the stellar population evolution.  2  The underlying power causing the emission originates from the same short-lived massive stars  99  BIBLIOGRAPHY I] A r c h i b a l d E . N . , Wagg J . W . , Jenness T., 2000, C a l c u l a t i n g Sky Opacities:  a re-  analysis for S C U B A data, S C D system Note 2.2 2] Barger A . J . , Cowie L . L . , S m a i l I., Ivison R . J . , B l a i n A . W . , K n e i b J . - P . , 1999, A J , 117, 2656 3] B l a i n A . W . , S m a i l I., Ivison R . J . , K n e i b J . - P , Frayer D . T . , 2002, preprint (astroph/0202228) 4] B l a i n A . W . , B a r n a r d V . E . , C h a p m a n S . C . , 2003, M N R A S , 338, 733 5] Bolzonella, M . J . , Miralles, M . , Pello, R., 2000, A & A , 363, 476 6] Borys C , 2002, P h D . thesis, U n i v . of B r i t i s h C o l u m b i a 7] Borys C , C h a p m a n S . C . , Halpern M . , Scott D., 2000, preprint (astro-ph/0009143) 8] Borys C , C h a p m a n S . C . , Halpern M . , Scott D., 2002, M N R A S , 330, L 6 3 9] C a l z e t t i D., K i n n e y A . L . , Storchi-Bergmann T . , 1994, A p J , 429, 582 10] C a r d e l l i J . A . , C l a y t o n G . C . , M a t h i s J . S . , 1989, A p J , 345, 245 II]  C a r r o l l S . M . , Press W . H . , Turner E . L . , 1992, A R A & A , 30, 499  12] C h a p m a n S . C . et al., 2000, M N R A S , 319, 318 13] C h a p m a n S . C , Lewis G . E . , Scott D., Borys C , Richards E . , 2002, A p J , 570, 557 14] C h a p m a n S . C , Scott D., Borys C , Fahlman G . G . , 2002, M N R A S , 330, 92 15] Condon J . J . , 1974, A p J , 188, 279  100 [16] D a v i s M . , Efstathiou G . , Frenk C . S . , W h i t e S . D . M . , 1985, A p J , 292, 371 [17] Dey A . , G r a h a m J . R . , Ivison R . J . , 1999, A p J , 519, 610 [18] Dickinson M . E . et al., 2000, A p J , 531, 624 [19] Dole H . , 2000, P h D . thesis, U n i v . of Paris-Sud X I , Orsay [20] Dole H . et al., 2001, A & A , 372, 364 [21] Downes D. et a l , 1999, A & A , 347, 809 [22] Downes A . J . B . , Peacock J . A . , Savage A . , Carrie D . R . , 1986, M N R A S , 218, 31 [23] D u n l o p J . S . , 2001, N e w A R , 45, 609 [24] D u n l o p J . S . et a l , 2002, preprint  (astro-ph/0205480)  [25] Dunne L., Eales S., Ivison R., Morgan H . , Edmunds M . , 2003, N a t , 424, 285 [26] Eales S., L i l l y S., Webb T . , Dunne L., Gear W . , Clements D., Y u n M . , 2000, A J , 120, 2244 [27] Eggen O . J . , L y n d e n - B e l l D., Sandage A . R . , 1962, A p J , 136, 748 [28] E l b a z D., Flores H . , C h a n i a l P., M i r a b e l I.F., Sanders D., Due P . - A . , Cesarsky C . J . , Aussel FL, 2002, A & A , 381, L I [29] Ferguson H . C . , Dickinson M . , W i l l i a m s R., 2000, A R A & A , 38, 667 [30] Franceschini A . , 2000, preprint  (astro-ph/0009121)  [31] Prayer D.T., S m a i l I., Ivison R . J . , Scoville N.Z., 2000, A J , 120, 166 [32] Gear W . K . , L i l l y S . J . , Stevens J . A . , Clements D . L . , Webb T . M . , Eales S . A . , Dunne L., 2000, M N R A S , 316, L51 [33] Gispert R., Lagache G . , Puget J . - L . , 2000, A & A , 360, 1  101 [34  G u h a t h a k u r t a P., Tyson J . A . , Majewski S.R., 1990, A p J , 357, L 9  [35;  Hildebrand R . H . , 1983, Q J R A S , 24, 267  [36  H o l l a n d W . S . et a l , 1999, M N R A S , 303, 659  [37;  H u E . M . , Ridgway S . E . , 1994, A J , 107, 1303  [38  Hughes D . H . et al., 1998, N a t , 394, 241  [39  Ibata R . A . , Gilmore G . , Irwin M . J . , 1994, N a t , 370, 194  [40  Jenness T . , Lightfoot J . F . , 1998, Starlink User Note 216, Starlink Project, C L R C  [41  Jenness T . , Stevens J . A . , A r c h i b a l d E . N . , Economou F., Jessop N . E . , Robson E.I.,  2002, M N R A S , 336, 14 [42] Johnstone D., W i l s o n C D . , Moriarty-Schieven G . , Giannakopoulou-Creighton J . , Gregersen E . , 2000, A p J S , 131, 505 [43; Kashikawa N . , et al., 2003, A J , 125, 53 [44; K a u f f m a n n G . , 1996, M N R A S , 281, 487 [45; Lagache G . , Dole H , Puget J . - L . , 2002, M N R A S , in press [46  M a i h a r a T . et al., 2001, P A S J , 53, 25  [47  O ' D o n n e l l J . E . , 1994, A p J , 422, 1580  [48  Peacock J . A . et al., 2000, M N R A S , 318, 535  [49  Rowan-Robinson M . , 2001, A p J , 549, 745  [50  Sajina A . , 2002, M S c . thesis, U n i v . of B r i t i s h C o l u m b i a  [51  Sanders D . B . , A p & S S , 266, 331  [52; Sanders D . B . , M i r a b e l F., 1996, A R A & A , 34, 749  102 [53  Schlegel D . J . , Finkbeiner D.P., Davis M . , 1998, A p J , 500, 525  [54  Searle L., Z i n n R., 1978, A p J , 225, 357  [55; Scott D. et al., The Promise of the Herschel Space Observatory. E d s . P h i l b r a t t G . L . , Cernicharo J . , Heras A . M . , P r u s t i T., Harris R., 2001, E S A - S P , 460, 305 [56  Scott S . E . et a l , 2002, M N R A S , 331, 817  [57; Skinner C . J . , 1996, Instrument Science Report N I C M O S - 0 1 4  [ss;  S m a i l I., Ivison R . J . , K n e i b J . - P . , Cowie L . L . , B l a i n A . W . , Barger A . J . , Owen F . N . ,  Morrison G . , 1999, M N R A S , 308, 1061 [59; S m a i l I., Ivison R . J . , B l a i n A . W . , K n e i b J . - P . , 2002, M N R A S , 331, 495 [60; Steidel C . C . , 1999, P r o c . N a t l . A c a d . Sci. U S A , 96, 4232 [61  Takeuchi T . T . , Hirashita H . , Ishii T . T . , Hunt L . K . , Ferrara A . , 2003, M N R A S , 343,  839 [62  Totani T . , Takeuchi T . T . , 2002, A p J , 570, 470  [63  Totani T . , Y o s h i i Y . , Iwamuro F., M a i h a r a T . , M o t o h a r a K., 2001, A p J , 558, L87  [64  Trentham N . , K o r m e n d y J . , Sanders D . B . , 1999, A J , 117, 2152  [65  Webb T . M . A , L i l l y S . J . , Clements D . L . , Eales S., Y u n M . , B r o d w i n M . , Dunne L.,  Gear W . K . , 2003, preprint [66  (astro-ph/0308439)  W h i t e S . D . M . , Frenk C . S . , 1991, A p J , 379, 52  [67; Wright E . L . , Hinshaw G . , Bennett C . L . , 1996, A p J , 458, L 5 3  [68] Y a h a t a N . , Lanzetta K . M . , Chen H.-W., Fernndez-Soto A . , Pascarelle S . M . , Y a h i l A . , Puetter R . C . , 2000, A p J , 538, 493 [69] Zheng Z., W u H , M a o S., X i a X . - Y . , Deng Z . - G , Zou Z . - L . , 1999, A & A , 349, 737  

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.831.1-0091401/manifest

Comment

Related Items