UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Multifilter spectrophotometry of the environment of the quasar 3C281 Craven, Sally Eve 1993

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

Item Metadata

Download

Media
831-ubc_1993_spring_craven_sally_eve.pdf [ 5.46MB ]
Metadata
JSON: 831-1.0086092.json
JSON-LD: 831-1.0086092-ld.json
RDF/XML (Pretty): 831-1.0086092-rdf.xml
RDF/JSON: 831-1.0086092-rdf.json
Turtle: 831-1.0086092-turtle.txt
N-Triples: 831-1.0086092-rdf-ntriples.txt
Original Record: 831-1.0086092-source.json
Full Text
831-1.0086092-fulltext.txt
Citation
831-1.0086092.ris

Full Text

MULTIFILTER SPECTROPHOTOMETRYOF THE ENVIRONMENT OF THE QUASAR 3C281BySally Eve CravenB. Sc. (Astronomy) University of VictoriaA THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESGEOPHYSICS AND ASTRONOMYWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAApril 1993© Sally Eve Craven, 1993In presenting this thesis in partial fulfilment of the requirements for an advanced degree atthe University of British Columbia, I agree that the Library shall make it freely availablefor reference and study. I further agree that permission for extensive copying of thisthesis for scholarly purposes may be granted by the head of my department or by hisor her representatives. It is understood that copying or publication of this thesis forfinancial gain shall not be allowed without my written permission.Geophysics and AstronomyThe University of British Columbia2075 Wesbrook PlaceVancouver, CanadaV6T 1Z1Date: AbstractThe technique of multifilter spectrophotometry is applied to identify stars and deducegalaxy redshifts and morphological types in a field centred on the quasar 30281, withthe intention of both testing the method and identifying members of the galaxy clustersurrounding the QS0. Photometry is performed on CCD images taken in twenty-fournarrowband filters, and the resulting spectral energy distributions are compared withredshifted model galaxy spectra from Rocca-Volmerange and Guiderdoni (1988) and withfiducial star spectra from Gunn and Stryker (1983) to determine the physical character-istics of the best-fitting spectral template. For twelve of the 184 objects in the 3.7'x6'field, spectroscopic redshifts have been published (Ellingson, Green, and Yee 1991), andan attempt is made to examine the accuracy of the technique using this small sample: arelationship between object signal-to-noise ratio and error in assigned redshift is sought.The results of a complementary study using simulated spectral energy distributions(Callaghan, Gibson, and Hickson 1992; Callaghan 1992) are briefly described, and com-pared with results from this data set.A broadband R filter is simulated by combining twenty-one narrowband filters, andthe instrumental magnitudes are calibrated with published magnitudes from Yee, Green,and Stockman (1986). The photometric errors are found to increase substantially withdecreasing signal-to-noise ratio, but the results suggest that broadband multicolour pho-tometry could be fairly accurately performed for many of the objects in the field, if theappropriate narrowband filters were present. Of the commonly used broadband filtersUBVRI, this data set has spectral coverage adequate to simulate only the R filter.Multifilter spectrophotometry has the potential to be a versatile and efficient surveytechnique, and will be employed in the sky survey to be made in forty filters by the UBCLiquid Mirror Telescope (Hickson et al. 1993, Gibson & Hickson 1991).ContentsAbstractTable of Contents^ ivList of Tables^ viList of Figures^ viiAcknowledgements^ viii1 Introduction 11.1 History ^ 21.2 Perspective 32 The Data and the Reductions 82.1 The Filter Set ^ 82.2 Observations 152.3 Basic Reductions and Object-Finding ^ 163 Photometry 183.1 Aperture Photometry ^ 183.2 Calibration ^ 20iv3.2.1 Zero Points ^  213.2.2 Spectral Energy Distributions ^  223.3 Uncertainty Estimates ^  223.3.1 Zero Point Uncertainty ^  243.3.2 Calibration Uncertainty  264 Redshifts and Type Classification^ 294.1 Overview^  294.2 The Templates ^  304.3 Details of the Fitting Process ^  314.4 The Test Sample ^  374.5 Systematic Photometry Errors ^  424.6 Results ^  435 Broadband Magnitudes^ 646 Accuracy of the Technique^ 716.1 Results from Modelling  716.2 Accuracy of the Derived Quantities ^  737 Discussion^ 808 Conclusion^ 84Bibliography^ 87List of Tables2.1 Equivalent widths for the filter set ^ 102.2 Transmission curve data for the filter set ^ 113.1 All zero point values derived from the standard stars ^ 234.1 Key to galaxy and star types ^ 324.2 Data for the galaxies of known redshift ^ 414.3 Best fits to all objects in the field 444.4 Positions, magnitudes, and S/N information ^ 504.5 Calibration galaxy data from Yee, Green, and Stockman (1986) ^ 564.6 Objects with derived redshifts within 0.1 of the quasar redshift ^ 60viList of Figures2.1 Transmission curves for the filter set ^ 94.1 Evolution of galaxy spectra ^ 334.2 The 12-Gyr model galaxy spectra used in the template-fitting ^ 344.3 The effects of redshift on spectral energy distributions ^ 364.4 Spectral energy distributions for galaxies with known redshift ^ 384.5 Distribution of best-fit galaxy and star types ^ 574.6 Distribution of the derived redshifts ^ 594.7 The redshift distribution limited by signal-to-noise ^ 635.1 The simulated Bessell R filter ^ 655.2 Relationship between signal-to-noise ratio and uncertainty in r magnitude. 685.3 Calibration of instrumental r magnitudes ^ 695.4 Distribution of the r magnitudes ^ 706.1 Redshift accuracies for the first set of EGY SEDs ^ 766.2 Redshift accuracies for the second set of EGY SEDs 777.1 The filter set of the UBC Liquid Mirror Telescope ^ 83viiAcknowledgementsI would first like to express my gratitude to my supervisor, Dr. Paul Hickson, whogave me the opportunity to work on a fascinating project with many possibilities, and inso doing taught me that nothing in astronomy is ever as straightforward as it looks atfirst. I would also like to acknowledge his financial support over the last two and a halfyears, and offer my thanks to him for making it possible for me to attend graduate school.Thanks are also due to the co-data-owner and collaborator Dr. Howard Yee, professionaloptimist ("You know, the data look cleaner than I thought they would") and author ofPPP, who was always prepared to suggest a new way to approach the problem. I thankDr. Claudia Mendes de Oliveira for helpful discussions and much-needed encouragement atthe right time, and Dr. Greg Fahlman and Dr. Harvey Richer for perceptive and valuablecomments on this thesis.I must offer my thanks to my fellow graduate students, many of whom have helpedme through the rough spots. Brad Gibson was always ready to help on questions aboutgalaxy spectra and evolution, filters, the LMT, computers, and just about anything else.Ted Kennelly exhibited boundless and contagious enthusiasm for astronomy, career de-velopment, artistic impression, and life in general. He was also really good about the Vic-toria Inner Harbour UFO incident. (Vergessen Sie nicht: dem Blinden hilft keine Brine.)Phil Hodder provided invaluable assistance with extended IATEX tutorial sessions, andknew the short answer to every computer question lever asked. Dave Hogg ("astonomer"),Dave Woods, and Andy Walker, among many other things, helped straighten out my un-derstanding of the theory and practice of error analysis, for which I am indebted to allthree. Every astronomy student in this department has been supportive throughout thelast few years; my thanks to all of them.I'm forever grateful to my family, who have always encouraged me and have neveronce remarked on the job prospects in astronomy. Thanks, you guys.Thanks to Pat Durrell (aka Tarpick Di Rella, Hatrack Umbrella, and Josie of theQuasi-Stellar Sheep Ranchers), for friendly competition ("By the way, I'm finishing mythesis this weekend"), years of conversation about astronomy and other things, preprints,many movies, and reams of email.Last but by no means least, my thanks to my husband Scott Tinis — for keeping mehappy, keeping me sane, and keeping me from sleeping in. May it last forever.viiiChapter 1IntroductionThe study of the large-scale structure and evolution of the universe requires extensivesamples of galaxy redshifts to test cosmological models. In particular, redshifts of faintgalaxies are needed to extend the physical limits of the surveys deeper into the universe.The best way to obtain the necessary numbers of faint-galaxy redshifts is to use a bigtelescope dedicated to a survey project, but it is difficult for a single project to claimenough observing time for an extensive faint-object redshift survey.With large liquid mirror telescopes (Hickson, Gibson, and Hogg 1993; Gibson andHickson 1991) it is possible to have an inexpensive, and therefore private, telescope whichcan be dedicated to a redshift survey. In fact, since the reflecting surface of a liquidmirror telescope is composed of mercury forming a parabolic shape due to rotation ofthe mirror, such telescopes are zenith-pointing instruments and are therefore intended assurvey "machines". Because the telescope can only point at the zenith, it cannot trackon individual objects and conventional spectroscopy is difficult (although not impossible;it is conceivable that one could employ a grism, or moving slits or fibres — P. Hickson,pers. comm.). However, low-resolution spectral energy distributions (SEDs) of objects inthe field of view can be obtained by observing through numerous narrowband filters usinga large-format CCD operated in driftscanning mode, read out at the speed with which1the image moves across the chip.The goal of this thesis is to explore the method of multifilter spectrophotometry,or analysis of the spectral energy distributions produced by performing photometry onthe images in the multiple narrowband filters. In parallel with a separate analysis ofsimulated data (Callaghan, Gibson, and Hickson 1992; Callaghan 1992; Hickson, Gibson,and Callaghan 1993; and also reviewed in chapter 6 of this thesis), some problems withthe method are identified and the accuracy of the results investigated.1.1 HistoryIn the days of photographic spectra, the Palomar Observatory responded to the needfor spectroscopy of faint objects by developing a nine-filter photometric system, reachingfrom the ultraviolet (3730 A) to the infrared K (9875 A). Baum (1962) used this equipmentto produce early magnitude-redshift plots in an attempt to fit observations to worldmodels.Shortly thereafter, a prime-focus photoelectric spectrum scanner (Oke 1966) was builtat Palomar. This instrument was designed to obtain spectral energy distributions (SEDs)of sources to about 19th magnitude in "a reasonable amount of observing time". Thatobserving time was determined by both the brightness of the source and the chosenspectral resolution, usually between 50A and 200A. The instrument was tunable to anywavelength and bandwidth in its allowed range, but was only able to measure light in oneband at a time. Such single-channel photoelectric scanners were popular in the late 1960s:the Lick Observatory also had a pulse-counting prime-focus scanning spectrophotometer,which sampled the spectrum every 100A with bandwidth 60A in the red and 45A in theblue (Wampler, 1966, 1967).Spectrum scanners were efficient instruments (Oke 1969) because of their wide band-2passes, small number of reflecting surfaces, and high quantum efficiency photomultipliers.However, observations of objects fainter than 19th magnitude were not practical due tobackground sky limitations. Spectrographs at that time had useful limiting magnitudesof about 20, since their slits admitted less sky than the scanner apertures. The goal of theobservers was to push back the magnitude limits at which usable spectroscopic informa-tion could be obtained, and that inspired the construction of a multichannel photoelectricspectrometer at Palomar.The multichannel spectrometer, first mounted on the 200-inch telescope in 1968, wasdescribed by Oke (1969). Its thirty-three simultaneous channels covered a spectral rangefrom the atmospheric limit, near 3100A to 11000A, where water vapour significantlycontaminates the spectrum. Even before the blue channels of the spectrometer were func-tional, useful results could be obtained: Oke, Neugebauer, and Becklin (1970) presenteda set of QSO observations made in part with the spectrometer over a range of 5600Ato 11000A, in fifteen channels with bandwidths of 360A. The multichannel spectropho-tometer data are similar in nature to the data from the multi-narrowband filter systemdescribed in this thesis.1.2 PerspectiveThe development of linear integrated array detectors, with their increased detectionefficiency, immensely improved faint-end limits for spectrographs. However, the best lim-its of state-of-the-art instruments are never considered good enough, as witnessed bythe unceasing demand for larger numbers of redshifts in a smaller amount of observingtime. Today, the observer can choose from at least three approaches besides conven-tional spectroscopy: multi-object spectrography, multicolour photometry, and multifilterspectrophotometry.3Most large telescopes have among their instrumentation a multi-object spectrographof some type. Whether employing optical fibres or masks with multiple slits, these in-struments allow one to make use of more of the light collected by the telescope thanin ordinary single-slit spectroscopy, by simultaneously obtaining spectra of a number ofobjects in the field. Multi-object spectrographs are particularly suited to galaxy clusterstudies (see for example Ellingson, Green, and Yee 1991). Their main advantage over theother two techniques described below is the resolution of the spectra obtained, typicallyon the order of 15 A. MOSs, however, are not ideal survey instruments: the observer mustknow which are the target objects, such as cluster members, and must have accurate po-sitions for all of them. The number of target objects is also limited, by the field size, chipsize, and desired spectral resolution; but this is a minor limitation in view of the abilityof a MOS to obtain many spectra at reasonably high resolution in a short time.Multicolour photometry has also been used to find redshifts of faint galaxies. Baum(1962) showed that photometry of a galaxy in nine colours, using the Palomar pulse-counting photoelectric photometer, permitted its redshift to be determined by identifyingthe position of the 4000A break. In 1980, H. Spinrad, in a discussion of methods oflocating high-redshift elliptical galaxies to use as standard candles, suggested "a newmethod" of using the colour signature of high-redshift E/SO galaxies to identify themin the vicinity of quasars, and showed that it was possible to detect evolution in thesegalaxies. Broadband photometry of galaxies in clusters has often been used to describethe morphological makeup of the cluster population (where the cluster redshift is known);for example, see papers referred to by Couch et al.(1983): Butcher & Oemler (1978a),Couch & Newa11 (1980), Mathieu & Spinrad (1981), Koo (1981).D. Koo (1981, 1985) made use of Baum's ideas and an extensive literature of previouswork on two-colour plots of galaxies at low and high redshifts, to prove that colour-colour4plots, calibrated with Bruzuars (1983) model galaxy spectra, can be an effective toolfor estimating galaxy redshifts. His agreement with spectroscopic redshifts for a sampleof one hundred galaxies of magnitude 17 to 23 is quite impressive: for redshifts below0.35, two-thirds of the eighty redshifts agree with spectroscopic values within ±0.04.Complications of an approach of this type to redshift determination include the effects ofgalaxy evolution, dust, and reddening (internal and external to the galaxy in question).The advantages over multi-slit or multiple narrow-band filter approaches are the shorterobserving times required in the broadband filters and the ability to compensate for poordetector sensitivity in the blue by taking longer exposures.The third of these options is a compromise between multi-object spectroscopy andmulticolour photometry, including the flexibility and blue sensitivity of colour-colour ana-lyses while providing higher resolution and making the effects of spectral evolution moreevident. Multifilter spectrophotometry provides data similar to that from the the multi-channel spectrum scanner. It employs photometry of CCD images made in a single fieldthrough a number of narrow-band filters, and has the potential to produce hundreds ofspectral energy distributions in a field, depending on the nature and size of the field andon the length of the exposures. It is suited to an exploratory approach: the observer neednot select the particular objects for which spectra are to be obtained as with multi-objectspectroscopy, but may rather investigate the character of a field in any location, whetherit be set on a galaxy cluster or elsewhere. The spectral range and resolution depend on thechoice of number, central wavelength, and bandwidth of the filters, and the depth of thedata set is controlled by the filter characteristics as well as the amount of observing timeavailable. The quality of the spectral energy distributions, however, is highly dependenton the accuracy of the photometry in all the filters.To improve spectral resolution, narrower filter bandpasses are needed, and therefore5the exposure times must be longer. For a limited field, a multi-object spectrograph canprovide a lot of high-resolution spectra in the same time required for a multifilter study.However, surveys over large areas can benefit from the exploratory nature of the multifilterapproach, especially in the case of a zenith-pointing telescope like the UBC Liquid MirrorTelescope discussed later in this section.The key to the technique of multifilter spectrophotometry is to find a spectral templatewhich best fits the spectral energy distribution of a given object, as opposed to seeking acolour difference which crosses the 4000 A break. In principle, this allows one to determinethe redshift and morphological classification of the source. Couch et al. (1983), Ellis et al.(1983), and MacLaren et al. (1988) used six intermediate-band filters and three Hubble-type templates at a range of redshifts to study the populations of galaxy clusters. Part oftheir goal was to investigate the effects of galaxy evolution upon the method. Loh & Spillar(1986), with their six intermediate-band filters and three fiducial galaxy templates (whichwere linearly interpolated to construct templates of intermediate type) estimated redshiftsfor approximately a thousand galaxies. The position and size of the 4000A break werethe primary criteria in making the fit, but the rest of the shape of the spectral energydistribution was also considered. Successful comparisons of some of their results withpublished spectroscopic redshifts led them to conclude that the technique was successfulin determining accurate redshifts for both early and late-type galaxies.Despite the promise of multi-filter "redshift machines", cluster studies can only becarried out to a certain level with these low-resolution spectral energy distributions. Ulti-mately, higher resolution is needed for fine dynamical structure studies. A complementaryapproach incorporating both multifilter spectrophotometry and multi-object spectroscopycould provide very useful results, in particular if the former is used to find targets of in-terest for the latter.6The real power of multifilter spectrophotometry is its application to surveys. In par-ticular, it is an essential part of the survey to be carried out by the University of BritishColumbia Liquid Mirror Telescope (or LMT) (Hickson et al. 1993; Gibson & Hickson1991). The present UBC LMT has a 2.7-metre-diameter reflecting surface, and fortynarrow- and intermediate-band filters will be used in a driftscan-survey of a strip of skytwenty arcminutes wide and of total area thirteen square degrees. The survey is expectedto take two years, and to observe 10000 galaxies and 1000 QS0s during that time. Therelatively low cost of building such telescopes makes them ideal survey instruments, andthe intention of the UBC group is to follow this experiment with a larger LMT.The primary reason for undertaking the project described in this thesis was to testthe narrowband filter technique of obtaining redshifts and morphological classificationsto be used for the LMT project, although cluster fields of scientific interest, containing aquasar apparently at the centre of a cluster of galaxies, were chosen as targets. Previouswork in the area of multifilter spectrophotometry has been expanded upon: the SEDs areconstructed from twenty-four filters of much narrower bandwidth than previous groupshave used, and the fiducial spectra include seven galaxy types and eighty-one stellarspectra. The galaxy templates are redshifted from zero to one in increments of about 0.01for comparison with the spectral energy distributions.The redshifts obtained from this technique are expected to be accurate enough toidentify members of the cluster surrounding the quasar, and to describe the populationand distribution of galaxies in the cluster down to a certain magnitude limit. Surveysof fields such as this can also provide information on the morphology of the field galaxypopulation.7Chapter 2The Data and the Reductions2.1 The Filter SetA multifilter approach to studying galaxy clusters, as compared to higher-resolutionmulti-slit observations, has the advantage of efficiency: all objects in the cluster areobserved simultaneously, with no need for an observer to identify the few cluster membersor objects of interest beforehand to create the mask; and less light is lost in a filter systemthan in a spectrograph. It is also less expensive than a multi-object spectrograph, as thecost depends on the quality of the filters. This is a prime consideration for a dedicatedsurvey telescope which may be operated by a small institution or by an individual.The filter set used for these data was substantially less expensive and therefore oflower quality (in terms of peak transmission value and transmission curve shape) thanthe set of filters now in place on the UBC Liquid Mirror Telescope. Figure 2.1 shows thetransmission curves 1 for the twenty-four filters used to observe the field surrounding theQS0 3C281. The filter bandpasses are approximately 100A wide at their half-maximumpoint, with equivalent widths on the order of 50 to 70A. The equivalent widths of thefilters are listed in table 2.1, and table 2.2 contains the data for the filter transmissioncurves.1These data were obtained from the Corion Corporation, manufacturers of the filters.8I^II^I^I^I^I^I^I^I^I'At■.•^1I^II^I,5000v A tII^I 6000^7000Wavelength■ \I^/ \A\8000^9000Filter Characteristics — 3C281 Field0.800 6II-00.4cnt) 0.2Figure 2.1: The transmission curves used for the twenty-four filters in the 3C281 field.(Data from Corion Corp.)9Table 2.1: The equivalent widths of the filtersFilter A Equivalent width (A)4511.0 57.764993.0 66.315502.0 52.365692.0 56.725792.0 65.635905.0 64.305988.5 51.196104.0 55.036190.0 51.606289.8 56.566380.0 58.366512.0 53.836585.0 73.646714.0 78.126809.0 74.296998.0 68.857312.0 70.007492.0 84.407791.0 54.328019.0 95.328206.0 68.638283.0 90.508490.0 50.718991.0 63.8910Table 2.2: The transmission curve data for the filter setA (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff4420 0.000 4900 0.000 5400 0.000 5580 0.000 5710 0.000 5810 0.0004430 0.013 4910 0.004 5410 0.004 5590 0.000 5720 0.013 5820 0.0084440 0.040 4920 0.030 5420 0.018 5600 0.003 5730 0.047 5830 0.0404450 0.088 4930 0.090 5430 0.058 5610 0.010 5740 0.194 5840 0.1084460 0.200 4940 0.270 5440 0.130 5620 0.029 5750 0.480 5850 0.2904470 0.365 4950 0.600 5450 0.230 5630 0.080 5760 0.563 5860 0.5254480 0.450 4960 0.640 5460 0.357 5640 0.200 5770 0.539 5870 0.5694490 0.473 4970 0.600 5470 0.450 5650 0.470 5780 0.550 5880 0.5644500 0.475 4980 0.600 5480 0.500 5660 0.500 5790 0.574 5890 0.5644510 0.475 4990 0.614 5490 0.515 5670 0.512 5800 0.570 5900 0.5704520 0.472 5000 0.609 5500 0.525 5680 0.510 5810 0.520 5910 0.5404530 0.472 5010 0.598 5510 0.500 5690 0.503 5820 0.508 5920 0.5074540 0.472 5020 0.610 5520 0.480 5700 0.503 5830 0.530 5930 0.5074550 0.461 5030 0.628 5530 0.450 5710 0.500 5840 0.540 5940 0.5504560 0.420 5040 0.430 5540 0.400 5720 0.503 5850 0.445 5950 0.5504570 0.370 5050 0.180 5550 0.266 5730 0.500 5860 0.250 5960 0.3104580 0.250 5060 0.075 5560 0.165 5740 0.400 5870 0.115 5970 0.1304590 0.135 5070 0.030 5570 0.093 5750 0.230 5880 0.056 5980 0.0564600 0.070 5080 0.018 5580 0.047 5760 0.127 5890 0.030 5990 0.0254610 0.040 5090 0.005 5590 0.024 5770 0.050 5900 0.020 6000 0.0124620 0.020 5100 0.000 5600 0.010 5780 0.023 5910 0.012 6010 0.0054630 0.009 5610 0.008 5790 0.011 5920 0.007 6020 0.0004640 0.005 5620 0.004 5800 0.005 5930 0.0004650 0.001 5630 0.002 5810 0.0034660 0.000 5640 0.000 5820 0.00011A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff5880 0.000 6020 0.000 6100 0.000 6150 0.000 6290 0.000 6420 0.0005890 0.003 6030 0.010 6110 0.017 6160 0.000 6300 0.010 6430 0.0045900 0.008 6040 0.030 6120 0.060 6170 0.004 6310 0.044 6440 0.0165910 0.011 6050 0.100 6130 0.230 6180 0.006 6320 0.135 6450 0.0955920 0.030 6060 0.250 6140 0.385 6190 0.010 6330 0.375 6460 0.3005930 0.075 6070 0.405 6150 0.410 6200 0.020 6340 0.557 6470 0.5255940 0.228 6080 0.496 6160 0.433 6210 0.034 6350 0.578 6480 0.5635950 0.450 6090 0.528 6170 0.490 6220 0.065 6360 0.567 6490 0.5345960 0.515 6100 0.530 6180 0.519 6230 0.130 6370 0.553 6500 0.5225970 0.510 6110 0.506 6190 0.493 6240 0.252 6380 0.502 6510 0.4905980 0.495 6120 0.494 6200 0.474 6250 0.418 6390 0.452 6520 0.4405990 0.505 6130 0.518 6210 0.500 6260 0.505 6400 0.448 6530 0.4236000 0.520 6140 0.543 6220 0.571 6270 0.541 6410 0.500 6540 0.4506010 0.525 6150 0.460 6230 0.375 6280 0.543 6420 0.551 6550 0.4816020 0.490 6160 0.300 6240 0.130 6290 0.541 6430 0.340 6560 0.3206030 0.400 6170 0.157 6250 0.043 6300 0.549 6440 0.140 6570 0.1136040 0.235 6180 0.080 6260 0.019 6310 0.542 6450 0.050 6580 0.0586050 0.075 6190 0.048 6270 0.009 6320 0.512 6460 0.020 6590 0.0256060 0.026 6200 0.030 6280 0.002 6330 0.414 6470 0.010 6600 0.0126070 0.010 6210 0.012 6290 0.000 6340 0.270 6480 0.004 6610 0.0066080 0.005 6220 0.006 6350 0.154 6490 0.000 6620 0.0036090 0.003 6230 0.000 6360 0.075 6630 0.0036100 0.000 6370 0.032 6640 0.0006380 0.0186390 0.0116400 0.0056410 0.0046420 0.0016430 0.00012A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff6460 0.000 6590 0.000 6700 0.000 6870 0.000 7160 0.000 7370 0.0006470 0.005 6600 0.002 6710 0.004 6880 0.005 7170 0.002 7380 0.0046480 0.020 6610 0.009 6720 0.012 6890 0.014 7180 0.010 7390 0.0146490 0.060 6620 0.018 6730 0.042 6900 0.026 7190 0.017 7400 0.0386500 0.160 6630 0.045 6740 0.100 6910 0.052 7200 0.028 7410 0.1156510 0.365 6640 0.090 6750 0.220 6920 0.111 7210 0.042 7420 0.2806520 0.490 6650 0.183 6760 0.358 6930 0.218 7220 0.065 7430 0.4606530 0.544 6660 0.375 6770 0.470 6940 0.370 7230 0.108 7440 0.6046540 0.564 6670 0.585 6780 0.530 6950 0.470 7240 0.160 7450 0.6106550 0.592 6680 0.629 6790 0.585 6960 0.538 7250 0.265 7460 0.5956560 0.600 6690 0.650 6800 0.628 6970 0.540 7260 0.395 7470 0.6006570 0.604 6700 0.650 6810 0.635 6980 0.530 7270 0.485 7480 0.6126580 0.591 6710 0.647 6820 0.620 6990 0.521 7280 0.505 7490 0.6126590 0.586 6720 0.650 6830 0.594 7000 0.510 7290 0.505 7500 0.5956600 0.602 6730 0.650 6840 0.594 7010 0.490 7300 0.500 7510 0.5816610 0.584 6740 0.650 6850 0.602 7020 0.470 7310 0.508 7520 0.5856620 0.456 6750 0.615 6860 0.550 7030 0.455 7320 0.525 7530 0.6016630 0.250 6760 0.550 6870 0.390 7040 0.455 7330 0.534 7540 0.5606640 0.140 6770 0.410 6880 0.220 7050 0.423 7340 0.520 7550 0.4606650 0.070 6780 0.215 6890 0.113 7060 0.330 7350 0.490 7560 0.2466660 0.036 6790 0.100 6900 0.070 7070 0.190 7360 0.451 7570 0.1126670 0.020 6800 0.048 6910 0.039 7080 0.095 7370 0.382 7580 0.0706680 0.011 6810 0.022 6920 0.023 7090 0.040 7380 0.250 7590 0.0396690 0.007 6820 0.011 6930 0.015 7100 0.018 7390 0.133 7600 0.0236700 0.004 6830 0.005 6940 0.010 7110 0.010 7400 0.066 7610 0.0136710 0.003 6840 0.002 6950 0.005 7120 0.004 7410 0.029 7620 0.0096720 0.000 6850 0.001 6960 0.000 7130 0.000 7420 0.014 7630 0.0026860 0.000 7430 0.006 7640 0.0007440 0.0047450 0.0017460 0.00013A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff A (A) Coeff7670 0.000 7810 0.000 8080 0.000 8120 0.000 8340 0.000 8840 0.0007680 0.004 7820 0.000 8090 0.005 8130 0.002 8350 0.005 8850 0.0037690 0.009 7830 0.004 8100 0.012 8140 0.010 8360 0.005 8860 0.0057700 0.018 7840 0.004 8110 0.018 8150 0.012 8370 0.007 8870 0.0107710 0.037 7850 0.010 8120 0.035 8160 0.021 8380 0.010 8880 0.0157720 0.070 7860 0.015 8130 0.070 8170 0.030 8390 0.018 8890 0.0257730 0.140 7870 0.020 8140 0.130 8180 0.050 8400 0.021 8900 0.0357740 0.252 7880 0.030 8150 0.240 8190 0.078 8410 0.040 8910 0.0607750 0.410 7890 0.045 8160 0.383 8200 0.115 8420 0.070 8920 0.0957760 0.500 7900 0.063 8170 0.511 8210 0.177 8430 0.130 8930 0.1507770 0.520 7910 0.098 8180 0.583 8220 0.240 8440 0.225 8940 0.2257780 0.507 7920 0.140 8190 0.620 8230 0.330 8450 0.350 8950 0.3157790 0.497 7930 0.190 8200 0.623 8240 0.439 8460 0.430 8960 0.4197800 0.499 7940 0.267 8210 0.623 8250 0.520 8470 0.500 8970 0.4957810 0.502 7950 0.370 8220 0.626 8260 0.565 8480 0.513 8980 0.5357820 0.450 7960 0.480 8230 0.626 8270 0.611 8490 0.474 8990 0.5557830 0.330 7970 0.535 8240 0.555 8280 0.635 8500 0.430 9000 0.5647840 0.235 7980 0.580 8250 0.416 8290 0.645 8510 0.400 9010 0.5727850 0.160 7990 0.610 8260 0.310 8300 0.650 8520 0.350 9020 0.5687860 0.100 8000 0.618 8270 0.200 8310 0.650 8530 0.310 9030 0.5367870 0.060 8010 0.618 8280 0.116 8320 0.635 8540 0.248 9040 0.4107880 0.043 8020 0.610 8290 0.065 8330 0.590 8550 0.190 9050 0.2907890 0.030 8030 0.600 8300 0.040 8340 0.520 8560 0.125 9060 0.1967900 0.022 8040 0.585 8310 0.022 8350 0.430 8570 0.085 9070 0.1217910 0.016 8050 0.560 8320 0.016 8360 0.314 8580 0.050 9080 0.0717920 0.012 8060 0.515 8330 0.009 8370 0.220 8590 0.033 9090 0.0427930 0.006 8070 0.440 8340 0.006 8380 0.164 8600 0.022 9100 0.0277940 0.003 8080 0.370 8350 0.003 8390 0.117 8610 0.014 9110 0.0187950 0.000 8090 0.291 8360 0.000 8400 0.080 8620 0.010 9120 0.0128100 0.222 8410 0.056 8630 0.006 9130 0.0088110 0.164 8420 0.045 8640 0.000 9140 0.0068120 0.117 8430 0.031 9150 0.0038130 0.088 8440 0.024 9160 0.0028140 0.070 8450 0.018 9170 0.0018150 0.049 8460 0.011 9180 0.0008160 0.040 8470 0.0108170 0.028 8480 0.0058180 0.024 8490 0.0008190 0.0188200 0.0148210 0.0108220 0.0088230 0.0058240 0.0048250 0.0038260 0.0008270 0.000142.2 ObservationsCCD images were obtained at the Canada-France-Hawaii Telescope by P. Hicksonand H. Yee on the nights of February 14/15 through 17/18, 1988. The UBC focal reducer(Hickson, Richardson, and Grundmann 1992) was used at the f/8 Cassegrain focus, fora focal ratio of f/2.5. The detector was the RCA4 chip, and the total field size wasapproximately 3.7 x 6 arcminutes, with a pixel size of 0.344 arcseconds. Seeing over thefour nights ranged from 0.9 to 1.3 arcsec. The observers recorded cirrus at the horizonin the morning following the second night. The two hours prior to their noticing thepresence of cirrus had been spent taking dome flat field observations, so an unknowndegree of extinction may or may not have affected the night's observations in the filtersat central wavelengths 4500A, 5700A, 5900A, 6000A, 6200A, 8000A, 8200A, and 8300A.The possibility of calibration error is dealt with as discussed in section 4.5.Two standard stars were observed in order to calibrate the photometry (section 3.2).The stars were HD19445 and HD84937, from Oke and Gunn's (1983) list of subdwarfstandards. Exposure times were one or two seconds. 11D19445 was observed at least oncein all the filters and twice in four of them, while HD84937 was observed in only four filters.(See the table of zero points in section 3.2.1.)On the cluster fields, exposure times were 300 to 1800 seconds, depending on the filterwavelength. Night sky emission-line interference fringes were visible in the cluster imagestaken in the 8300A filter. Normally this signal would be removed by dividing the databy a "fringe frame", a median of all the data frames showing the fringes. However, therewere only four images, two of the 3C281 field and two of the field around the quasarPK50812-1-020, which showed the fringing signal, and the median image of these four stillretained artifacts of the objects in the fields. Since this would have seriously affected the15photometry, no correction was made to the image. The effect of the fringing is discussedfurther in the section dealing with uncertainties, 3.3.The sources in the cluster frames show significant coma near the edges of the images.The effect of the coma is to spread the light from the object over a larger area than itwould cover at the centre of the frame. The peak brightness and the signal-to-noise ratioare therefore reduced, as more light is lost to the faint edges of the image. As a resultless light is measured in any given aperture than would be detected in the absence of thedistortion, and the object appears fainter than it should.2.3 Basic Reductions and Object-FindingThe IRAF package 2 was used for the initial reduction of the data. Median biases wereconstructed for each night, and the images were bias-subtracted, flat-fielded using domeflats, and trimmed of the overscan region.The images were converted from real to short integer pixels and read into H. Yee'saperture-photometry program, PPP (described in chapter 3 and more fully in Yee 1990).The positions of reference objects in the field were used to align and trim the images, sothat a given object would lie at the same coordinates in every frame.Object detection was performed on median images of the field. These were constructedby scaling the background levels of an image in each of the twenty-four filters to the samevalue, and then taking the median value for each pixel. Because two observations of thefield were made in every filter, two independent median images were made. In such animage, the sky background is smoothed out, improving the signal-to-noise and visibility2IRAF is distributed by the National Optical Astronomy Observatories, which is operated by theAssociation of Universities for Research in Astronomy, Inc., under contract to the National ScienceFoundation.16of real objects in the field.The object-finding routine in PPP is a modified version of that used by Kron (1980).When a local maximum is found, a flux is measured from the central nine pixels aroundthe maximum, and a local sky, computed as the mode of the pixel values in an annulus ofinner and outer radii five and eleven pixels, is subtracted. If the "object flux" lies abovea given threshold, then the position of the maximum is recorded as a detection. If thethreshold is set low enough to detect faint and diffuse objects, a number of noise detectionswill also be made. Cross-checking the detection lists from the two independent images,and eliminating detections not seen on both frames (within two pixels of the position),removes most, although not all, of the noise-spike detections.The threshold level for object detection was selected to be 1.2 times the standarddeviation of the sky background pixels across the frame. Object-finding was done atvarious threshold levels, from 0.8o- to 2cr, and it was found that after cross-checking thetwo object lists, the number of detections did not increase substantially below a thresholdlevel of 1.2cr. Such a threshold is low for putting confidence in real detections, but helpsavoid missing objects which might be bright only in some filters. (The median image ofsuch objects could be quite low in intensity.) The signal-to-noise value for every object ineach filter was later computed, and the median of these values taken as a representativesignal-to-noise for the detection (section 4.6).The cross-checked list was visually compared with both median images. Detectionswere removed from the list if they were obviously due to chance alignment of noise spikes,or if they lay in the haloes of relatively large galaxy images. The object-finding routineoccasionally missed extended diffuse objects. If evidence could be seen for an objectin both sum frames, the object was added to the list. After this, the preliminary listcontained 184 detections.17Chapter 3PhotometryGalaxy photometry can be approached in a variety of ways. Some methods involvethe fitting of elliptical profiles, with a range of eccentricities and angles of rotation, togalaxy images. Others measure the light within an aperture of fixed size, or attempt toselect an aperture of appropriate size for each object. Because the object images in thisstudy were not large or well-resolved enough to allow profile-fitting, the applicability ofthe latter two techniques was studied.3.1 Aperture PhotometryPhotometry of the objects in this field went through several incarnations. The firstmethod tested was growth-curve aperture photometry. Light from an object is measuredin a series of concentric circular apertures. The resulting "growth curve" is analyzed todetermine the size of the object and hence the appropriate aperture. For example, thegrowth curve of a relatively compact, isolated object will stop rising when the apertureis larger than the object; the aperture at which this occurs is selected as the appropriateone for photometry of the object. 1 Any error in the calculated sky level will cause the1This is not necessarily the aperture in which the signal-to-noise ratio is maximized; an alternativetechnique employs that method, but it was not tested on this data set.18growth curve to turn up or down in the apertures where little light from the object isdetected. The aperture prior to the drop (if the sky is overestimated) or to the upturn (ifunderestimated) is identified as the best one to use.A disadvantage of the growth-curve technique for aperture selection is that it maytend to underestimate the brightness of an extended isolated object. This is due tothe "termination condition" of setting the aperture just inside a perceived drop in flux,because the possibility is not considered that noise superimposed on the object may causethe growth curve to fluctuate both up and down. In this way, a significant amount ofsignal from the outer regions of the object can be lost, especially for faint or diffuse objectswhose outer regions will be badly affected by noise. This is also a problem for objects thatmight lie close to low-valued columns or pixels. The growth-curve technique, as describedhere, can therefore introduce a bias in the flux measurements.The potential bias from the growth-curve method could have a serious effect on spectralenergy distributions composed of multi-filter data, since the difference between fluxes inadjacent filters can be crucial in later determining object type and redshift. For thatreason, a fixed-aperture approach was chosen for the photometry.H. Yee's photometry program PPP was used to compute the total flux in a set of aper-tures with diameters up to thirty-five pixels (where the pixel size is 0.344 arcseconds). Thebackground sky level is calculated with the modal sky estimator, 3 xmedian — 2 x mean,where the mean and the median are computed, with an iterative outlier rejection algo-rithm, in an annulus surrounding the object and the largest object aperture.The flux for each object in each filter was taken from an aperture of diameter ap-proximately twice the full width at half maximum, seven pixels or 2.4 arcseconds across.The light within this aperture represents the light from the whole object, without thecomplications introduced by estimating a size for the object in each filter image.19The spectral energy distributions produced later are compared with model galaxyspectra. Because the model spectra of spiral galaxies are composed of separate bulgeand disk components, it is important to be sure that spiral arms as well as centralbulges fall within the aperture so that spirals will not be misclassified as ellipticals. Theangular size-redshift relation (e.g. as in Sandage 1988, page 610) allows the calculationof the linear size d at a given redshift corresponding to the angular size of the aperture.For example, for a universe with qc, = d = 1.6/h100 kpc at z = 0.05, d 3.0/h100 kpcat z = 0.1, and at z = 0.2, d 5.0/li100 kpc. A galaxy for which identification might be aproblem, then, would be a face-on spiral at low redshift, especially if Ho= 100 km/s/Mpc.However, at redshifts above 0.1 or 0.2, there should be no such confusion, so use of anaperture this size seems reasonable.3.2 CalibrationThe conversion from instrumental (PPP) magnitudes to real intensity units was madeusing the following calibration equationm = MPPP + 2.5 log T A0 + EXwherem is -2.5 log f, where f is the total flux in ergs s-1 cm-2 through the filter,Mppp + 2.5 log r is the magnitude based on flux from a one-second exposure,Mppp is the magnitude given by PPP,T is the exposure time,Ao is the zero point for the calibration,X is the airmass, andE is the extinction at the filter wavelength, which is from the "Mauna Kea Extinction20Curve", figure [5]-3a in the 1990 CFHT Users' Manual.No colour term was included in the equation because the colours of the standard starswere not sufficiently different for a solution to be found. If the colour term is significant,it could introduce errors in the calibration.3.2.1 Zero PointsTwo standard stars from Oke and Gunn (1983) were used to determine the zero pointfor the calibration of each filter. Oke and Gunn give spectral energy distributions overthe range 3080A to 12000A for a set of subdwarf standards in A1379 magnitudes. Thesewere converted into flux units of ergs s- cm-2 Hz-1 by the equation 2AB79 = —2.5log f, — 48.60.The filter transmission curves provided by the manufacturer were sampled every tenangstroms. The flux through each filter was computed by summing the product of the filtertransmission coefficient T times f, multiplied by the sampling width in Hertz, effectivelyevaluating the integral f T My over the filter. The resulting flux, f, was substitutedinto the calibration equation, along with the PPP magnitude for the star in that filter,exposure time, extinction coefficient and airmass, and the zero point was calculated.For those four filters in which both standard stars were observed, the difference inzero points derived from the two stars was on the order of 0.04 magnitudes. This is ameasure of the uncertainty in the derivation of the zero point; it does not represent anight-to-night variation in the value because all observations in a given filter were madeon the same night. Table 3.1 lists all the zero points derived in these filters. The tablealso lists the derived uncertainties in the zero point, which include uncertainty in Oke2The sign of the constant in this equation is different from that given by Oke and Gunn. The change isjustified by noting that absolute flux from a star of apparent magnitude V=0.000 at 5480 A is 3.65 x 10-20ergs s' cm-2 Hz" (Oke & Schild 1970), and that AB79 -a V at 5480 A.21and Gunn's spectral energy distribution for the standard as well as uncertainty in theflux measured from the image. Calculation of these values is described in section 3.3.1.3.2.2 Spectral Energy DistributionsThe measured magnitude of each object was converted to a true magnitude by thecalibration equation, using the airmass from the middle of the exposure for the bestapproximation. Estimates of the uncertainties in the contributing quantities are describedin section 3.3.In the case that PPP was unable to calculate an object magnitude in a given filterbecause background subtraction left a negative flux value, the flux was set to zero. Suchdata points were excluded from the later SED-fitting procedure.A list was then compiled with one object spectral energy distribution per line. Thisconsisted of the object number and position in the field, followed by the intensity and itsuncertainty for each filter, ordered by central wavelength. Fluxes in this list were recordedin units ergs s-1 cm-2 so as to be compatible with the model-fitting program. However,the data displayed later (section 4.4) are shown in the form log Mergs cm' A-1 vs. A.3.3 Uncertainty EstimatesThe uncertainty estimate for the points is of great importance, since these uncertaintiesappear in the x2 function which is minimized to determine the template which best fits theobject spectral energy distribution. Uncertainties arise in every term of the calibrationequation, in the zero point itself, and in external influences on the data such as galacticextinction and interference fringes.22Table 3.1: Table of the zero point values derived from all the observations of standardstars in these filters. There were multiple standard observations in only four filters.Filter A HD19445 Obsn 1 11D19445 Obsn 2 HD84937 Obsn 14511 37.632 + 0.0404993 38.101 ± 0.0295502 37.474 ± 0.0445692 37.310 ± 0.0395792 37.512 ± 0.0335905 37.662 ± 0.0315989 37.520 ± 0.0416104 37.605 ± 0.0376190 37.547 ± 0.0356290 37.616 ± 0.046 37.609 37.6586380 37.661 ± 0.033 37.631 37.6736512 37.594 + 0.039 37.573 37.6296585 37.574 ± 0.0356714 37.501 ± 0.0336809 37.574 ± 0.0346998 38.448 ± 0.0377312 37.556 ± 0.0427492 38.037 + 0.0317791 37.224 + 0.0508019 37.691 ± 0.0458206 37.064 ± 0.0408283 36.987 ± 0.0408490 37.912 ± 0.057 37.917 37.9118991 37.386 ± 0.052233.3.1 Zero Point UncertaintyAn estimate of the zero-point uncertainty is made by considering each term of thecalibration equation in turnAo mAB — mppp — 2.5 log r — EX.To estimate the uncertainty in the "true magnitude" of the standard, the 0.01-magnitudeerror bar in Oke and Gunn's SED measurements is considered. The filter transmissioncurve is sampled every 10 A, so at each of these sample points, a Afy is computed from1^(.1-48.60).-Am^(m+48.60)+A. Af, = [10^-2.5^- 10^-2•5This quantity is multiplied by the sample width (i.e. 10 A) in Hertz, and summed overthe width of the filter.The uncertainty in mppp is primarily ascribed to the level of sky noise in the aperture;this will dominate the counts for the faint objects, at least. Yee (1991) calculates this un-certainty AF for an object with F counts in the aperture as the product AF = o-skyN/Npix,where crsky is the rms value per pixel in the local sky, and Npix is the total number ofpixels in the aperture. The local sky is the mode of the pixel values in an annulus withinner diameter 35 pixels and outer diameter 73 pixels. The area of the annulus is largeenough to represent a reasonable statistical sampling of the sky around the object. The"signal-to-noise" calculated by PPP's photometry routine is the value , and hence AFcan be computed in flux units (ergs s-1 cm') by taking F as the true flux calculatedfrom the SED.The square root of the number of photons detected from the object is a further sourceof error for the bright standard stars, although it is completely dominated by the skynoise in faint objects.24The exposure time T is very short (either one or two seconds) for the standard stars.This time may be in error by as much as five per cent (H. Richer, pers. comm.), due tothe finite time taken for the shutter to close. The error due to this factor could not beestimated because no longer exposure of a standard was taken.Because the change in zenith distance over the exposure time is almost infinitesimal,no error is assumed in the airmass X. The differential atmospheric extinction was obtainedfrom the graph of mean extinction for the site published in the CFHT Users' Manual.Unfortunately the manual does not quote the typical night-to-night variation in the ex-tinction, nor can it be estimated from the data, since repeated observations of standardson different nights in the same filters were not made. The uncertainty in reading thevalues from the curve is taken as ±0.003 magnitudes per airmass.The total uncertainty in Ao is computed by the method of partial derivatives, as de-scribed by Bevington (1969, Chapter 4). The uncertainties in the terms of the calibrationequation are assumed to be uncorrelated, and therefore no cross terms are included in theerror calculation.The calibration equation is rewritten in the formAo = —2.5 log f + 2.5 log C — 2.5log T - EXwhere f is the calibrated flux and C is the number of counts representing the objectbrightness. Then the zero-point uncertainty 0rA0 is given by :2^2(8A0'\22 09A0)2_,_ 2(8A0\2 j_ 2 (8A0)2^2 (5A0)2uAo= Crf a f^ac ac ) '^) CrE DE ) crx ax )where cr.r and crx are zero for the standard star observations.Substituting for the partial derivatives, and including the appropriate values for theuncertainties a-, yields the zero point uncertainty. Since the number of photons counted25for a star, after sky subtraction, is the sum of the contribution from the star itself andthe contribution from sky noise, the partial derivative analysis givesCC =^0.02^+ 472obi^— sky—noise^C elcy—stibtrand the zero point error is2 „ 2CT 2A0 = (2.5log e)2 [()^) +^X2.^/c C3.3.2 Calibration UncertaintySources of uncertainty in the calibration other than those from the zero point includethe contribution from sky noise in the aperture (as discussed in section 3.3.1), the con-tribution from atmospheric extinction, and that from the airmass. The other sources ofuncertainty in the derived magnitudes originate in galactic extinction and the presence ofsky emission interference fringes in some of the data frames.A mid-exposure airmass was used in the calibration equation. The uncertainty inthis quantity is the difference between the mid-exposure airmass and the average airmassover the exposure. For twenty-minute exposure times and fields near the meridian, thecontribution of the airmass to the uncertainty was taken to be so small that it was notincluded in the error calculation.The colours of the galaxies in the field, as compared to model galaxy spectra, maybe affected somewhat by galactic absorption, although the quasar is at a high galacticlatitude : = 314.500, b = 69.20° (epoch 1950). An analytical calculation was madeto estimate the effect of galactic absorption in the the filter bandpasses using a typicalhydrogen density for the galaxy 3 and an extinction formula from Lang (1986, p. 565). The3 Mihalas and Binney (1981) describe the most common constituents of the ISM as (1) hot neutralgas surrounding cool clouds, occupying 20% of the volume and having number density 0.3 cm-3, and (2)26amount of absorption through 100 pc was found to be on the order of 0.036 magnitudesfor the reddest filter and 0.02 for the bluest. Since the true nature of the interstellarmedium in the direction of the cluster is not included in this calculation, these valuescannot be used as a correction, but they are taken as representative of the uncertainty inthe object magnitude due to galactic absorption. The appropriate value was included inthe total uncertainty for each object in each filter.Although the estimated uncertainties due to extinction quoted above were applied tothe flux uncertainty in the data used for the project, a more appropriate course of actionwould be to make an extinction correction in each filter based on the column densityalong the line of sight to the field. Burstein and Heiles (1978) published column densitiesand reddenings for three objects within about 20 degrees of the 30281 field. The E(B-V)values they listed for the objects (0.03, 0.00, and —0.02, with an estimated uncertainty of0.03) yield maximum extinction values by the method of Cardelli, Clayton, and Mathis(1989) of about 0.003 at the red end of the filter set and 0.057 at the blue end. Giventhe errors in the measured reddening, the uncertainties applied to the flux data are notunreasonable, although a correction to the flux would have been better.Interference fringe patterns were observed in the 8300A filter images. They wereprobably due to an atmospheric emission line near this wavelength entering the filter. Anattempt was made to create a fringe pattern image which could then be removed from thedata. Because there were only four images showing the fringe pattern, however, it provedimpossible to successfully remove the fringes without seriously affecting the quality of theobject images. It was not thought necessary to add an extra error term to describe theeffect of the fringing on the photometry, because the fringing occurs on a small enoughhot, low-density gas, occupying 70% of the volume and with number density 10-3 cm-3. The hydrogennumber density of the former was used in the extinction estimate, although areas of the galaxy containingclouds will have higher densities.27scale (about twenty pixels) that it is perceived as a further source of sky noise in the localsky aperture. Its effect is therefore included in the signal-to-sky-noise ratio calculated byPPP.As was done for the zero-point error, the uncertainty in the photometry is propagatedusing the method of partial derivatives. This time, however, the uncertainty is needed inthe linear domain (intensity units), so the calibration equation is rewritten asCf = 10 -0.4(EX-1-A0)— Twhere the variables are as defined above in section 3.3.1.The total uncertainty cri in the calibrated intensity derived for an object is given by2^2^2^(crf) ^( 1 )^(  ln 10  ) 2 E2 „2^(  ln 10 2 ,2 2^ln 10^f ) — SIN) + — 2.5^-.1. + —2.5 - E—y + —2.5 a-A°.The term representing the square root of the number of photons received from the objecthas been left out of the equation above, since it is generally overwhelmed by the sky noisefor the faint objects.The value f lux I uncertainty, or -L, is a measure of signal-to-noise ratio in the aperture.of28Chapter 4Redshifts and Type Classification4.1 OverviewThe classification of the field objects based on the nature of their spectral energydistributions (SEDs) is accomplished by the brute-force method of making a quantitat-ive comparison between the observed SED and each one of an array of artificial SEDs,produced by multiplying the filter transmission curves by a set of galaxy and star fiducialspectra.The galaxy model SEDs, or templates, are calculated over a range of redshifts, fromzero to one in increments of 0.005 in log(l+z). This corresponds to an effective resolutionin redshift of from 0.01 to 0.02. Each model SED is multiplied by a scale factor whichminimizes x2, based on the ratio of the raw model intensity to the data intensity. Thenthe goodness of the model fit is quantified by calculating a reduced ,c2 value:I, no. /iltera —model^data 12x2=^ )—v s(f.where v is the number of filters defining the SED, and a is the uncertainty in the intensityof the data point (as discussed in section 3.3). The model yielding the lowest x2 value isunderstood to best represent the physical nature and redshift (for galaxies) of the object.It should be noted that this x2 does not have the usual statistical meaning. The29quantity v represents the number of degrees of freedom and is taken as the number offilters, but the filters overlap and are therefore not completely independent. The x2calculated here is useful for comparing the fits of models to a single data set (i.e. theSED for a single object), but cannot be used to estimate the statistical significance of afit, and cannot even be compared between fits to different data sets.A further problem with the x2 value is related to the uncertainties cri. For a usefulstatistical x2 value, the errors must be uncorrelated and have a gaussian distribution.However, the errors in each filter are primarily dependent on the sky noise in the image,which varies significantly for observations in different filters. The uncertainties in a givenSED are therefore unlikely to be gaussian in distribution.The fiducial spectra available represent a sequence of normal galaxies and stars. Anyobjects with spectra not included in this sequence, such as emission-line galaxies and QS0sfor example, cannot be recognized as such without visual inspection. The best-fittingmodel is assigned; if the uncertainties were well understood and the filters independent,the value of the x2 would illustrate how close the best-fit model is to the real SED. Suchan estimate of goodness-of-fit would be valuable in determining confidence in the derivedobject type and redshift.The following sections describe the sets of fiducial spectra, then some of the detailsand complications of the fitting procedure, and finally the results of applying the fittingprogram to the real data from the 30281 field.4.2 The TemplatesP. Hickson's fitting program compares spectral energy distributions of the objects inthe fields with SEDs constructed from fiducial star and galaxy spectra in the literature.The eighty-one star spectra, from Gunn and Stryker's (1983) stellar atlas, range from30type 05 to M8III and M8V. The galaxy templates are taken from Rocca-Volmerangeand Guiderdoni's (1988) atlas of synthetic spectra, which consists of eight morphologicaltypes of galaxies, evolved to twenty ages from 0.08 Gyr to 19.08 Gyr using the galaxyspectrophotometric evolution package 1 of Guiderdoni and Rocca-Volmerange (1987). Theseven types used for template-fitting in this study range from E/SO through the spiralsto an Jr galaxy. Intermediate templates are produced by linearly interpolating betweenspectra of successive types. Table 4.1 lists the galaxy and star types in the set of templates,together with their reference numbers.While the fitting program is able to fit galaxy templates of varying ages, the data arenot of high enough quality to justify introducing the "spectral age" of the galaxy as aseparate parameter. The appropriate age for the galaxy templates should correspond tothe model spectra which are likely to match the largest number of galaxies in the observedfield. The most significant change with time in the spectrum of a galaxy occurs early inits life; after about twelve to fourteen gigayears of evolution, the spectrum does not altermuch more, as demonstrated by figure 4.1. This data set is unlikely to reach deep enoughto see galaxies early in their formation, so young galaxy templates were not considered.The 12.08—Gyr templates were adopted for fitting to the entire sample. Figure 4.2 showsthis set of templates at zero redshift.4.3 Details of the Fitting ProcessThe points in the model galaxy templates represent the flux that would be measuredin each filter for a galaxy of the model type and redshift. Each point is the sum over the1This package assumes solar metallicity and does not take chemical evolution into account. However,no models currently available do otherwise. At any rate, errors due to this assumption will occur primarilyin the blue end of the spectrum, and will not be important in the spectral range observed for most of thegalaxies in this study.31Stars8 = B3III 9 = B3IV 10 = B3V 11 = B4V12 = B5Ib 13 = B6V 14 = B7III 15 = B7IV16 = B7V 17 = B8Ia 18 = B9IV 19 = B9V20 = AOIV 21 = AOV 22 = A1V 23 = A2V24 = A3III 25 = A3IV 26 = A3V 27 = A4IV28 = A5III 29 = A5IV 30 = A5V 31 = A7V32 = A9IV 33 = FOIV 34 = F2IV 35 = F4V36 = F5IV 37 = F6V 38 = F7IV 39 = F8V40 = F9V 41 = GOV 42 = G2IV 43 = G2V44 = 03IV 45 = G4IV 46 = G5IV 47 = G5V48 = G6IV 49 = G7IV 50 = G8III 51 = 08IV52 = G8V 53 = KOIII 54 = KOIV 55 = KOV56 = K1III 57 = K1IV 58 = K2III 59 = K2IV60 = K3III 61 = K3V 62 = K4III 63 = K4V64 = K5III 65 = K7V 66 = K8V 67 = MOIII68 = MOV 69 = M1III 70 = M2III 71 = M2V72 = M3III 73 = M4V 74 = M5III 75 = M5V76 = M6III 77 = M6V 78 = M7III 79 = M8III80 = M8V0 = 05^1 = 06^2 = 08V 3 = BOIbB2III^7 B2V4 = BlIV 5 = B1 6 = =Galaxies0 = cold EO 1 = hot EO2 = Sa^3 = Sb4 = Sc^5 = Sd6 = ImTable 4.1: Key to the numerical classifications used to refer to galaxy and star types.3214.6^14.8^15^15.2log frequency15.4Figure 4.1: The change in a zero-redshift galaxy spectrum with age. The upper windowshows evolution of a model UV-cold E/SO spectrum (Rocca-Volmerange and Guiderdoni1988) from age 2.08 Gyr to 18.08 Gyr, while the lower figure shows the same for a spiralgalaxy. The vertical axis is in units 5.32 x 1029ergs s-1 A-1 M-01- Notice the differencein the vertical scales of the two plots. The shape of a galaxy spectrum changes less as thegalaxy ages.3314.6^14.8 15^15.2log frequencyFigure 4.2: The model galaxy spectra (from Rocca-Volmerange and Guiderdoni 1988) atage 12.08 Cyr and zero redshift. These spectra are used as the fiducials in the template-fitting program and have been scaled to the same level at 5500 A. In this figure the spectrahave been slightly offset from one another so that they can be more clearly seen at thelower-frequency end.34filter transmission curve of the quantityfv,(z)A logwhere Ti is the ith transmission coefficient of the filter, f„(z) is the redshifted modelgalaxy flux (in W MV) at the filter sample frequency, and Llog v = 0.001 is theseparation between samples in log frequency space.Model SEDs are computed for galaxy redshifts from zero to one, in increments of0.005 in log(l+z). For an elliptical galaxy, the change in spectral shape with redshift isquite dramatic, a fact made use of by Koo (1985) and others who used optical colours todeduce large numbers of redshifts. However, later-type galaxies have flatter spectra andsmaller 4000-A breaks, and as figure 4.3 demonstrates, it is much more difficult to assignan accurate redshift to a given SED. Erratic SED data points due to noise and errors inthe photometry can make it still harder to discern the redshift.The model SEDs must be multiplied by some scale factor before they can be quanti-tatively compared with the real SED. This scale factor is computed by equating to zerothe first derivative, with respect to the scale S, of the reduced x2 formulano.filters f tobj^S X fnadel22 i Ex = -v^ i^)cri21=1The resulting scale, representing the minimum x2, isv■no.filters ;obi pnodel cri7S = L.4=1^Ji Ji v■no.filters 10612 /0.?2Li=1The effect of the a in the denominators of the two quantities is to weight the scale valuetowards the value of f°bj I frn°del where fc'bj has the smallest uncertainty.351 's^ -n^• • .1,^\ ... ............. ,23 \ • ■■,)Cold E/SOI^I^II^II22z24I^I^I^Iz = 0.0z = 0.2z = 0.4z = 0.6z = 0.8z = 1.0Sc^NAN .V:>'----e—'e4"-`^.._...,^. .-0— • ___,0.___ ^ ,...,-.- -,............._0„4,_ --..--. ......... ,...... ..._ , ._...^.. - -...•*-- - tr-o^-n,._•-::2 \ \' -^---4^..Neo---e-:-Z------_,8------ -"----- .---...- .-----^".0An....- - --14.5^14.6^14.7^14.8log frequencyFigure 4.3: The effects of redshift on the spectral energy distributions of an E/SO and anSc galaxy. These SEDs are generated by multiplying the redshifted model spectra by thetwenty-four filters used to produce the data. Note that the flux units in this figure areergs s - 1 cm- 2 rather than ergs s-lcm-2:4-1. This figure depicts the templates as they arecompared with the data in the fitting program, and the "flux" is a measure of the totallight through each filter.2423221364.4 The Test SampleIn order to test the accuracy of the fitting program, a test sample of spectral energydistributions was identified, consisting of galaxies in the field for which Ellingson, Green,and Yee (1991; hereafter EGY) have measured and published redshifts from multislitspectroscopic data. EGY give redshifts for galaxies in the fields of sixteen quasars, andYee (pers. comm.) believes about 85 to 90% of the published redshifts are correct. Twelveof the galaxies EGY looked at in the vicinity of 3C281 correspond to the positions ofobjects identified in the data for this study.Because two observations of the field were made in each filter, there exist two separateSEDs for each object. A fit was made to each of these, so the consistency of the fit couldbe studied in the presence of uncertainty in the photometry.The best-fit redshifts and types assigned to the EGY objects are listed in table 4.2,together with the published redshifts. Both SEDs for object 95 are best fit by a startemplate (of type 73, M4V). Figure 4.4 shows both spectral energy distributions for thesetwelve objects, with their best-fit galaxy models overplotted.There are seven objects in the first list of SEDs for which the difference between thederived and the published redshifts is less than about 0.1, and there are five such objectsin the second list. A discussion of the accuracy of the technique based on these results isgiven in section 6.2.375000 6000 7000 8000 9000Objects 36, 44, 75, 86- 16.5 —16.5—17 —17—17.5 —17.5—18 —18—18.5 —18.55000 6000 7000 8000 9000__ta00,——16.5—17.5-- 18-—17111■111i^i-^ ^—16.5—17.5—17 _—18_A--_-—--18.5 —_—18.5, .l,,,,I,,,,,,,,,I,,-5000 6000 7000 8000 9000- -__5000 6000 7000 8000 9000-Central Filter WavelengthFigure 4.4: Spectral energy distributions for the galaxies from Ellingson, Green, andYee (1991). The flux units are ergs s- 1 crn' A-1. Filled triangles with dotted-line errorbars represent the SED produced from the first observation in each filter, hollow circlesand solid-line error bars the second SED. The dashed line represents the best-fit galaxytemplate for the first SED, and the solid line is the best-fit galaxy for the second SED.381^1^1 1^1^1 _5000 6000 7000 8000 90005000 0 6000 7000 8000 9000Objects 87, 92, 95, 961-'1^1^1^1^1^I^1^1^1^1^I^1^1^1^I^I^1^I^1 5000 6000 7000 8000 9000-18.5- 17.5-18- 18.5-5000 6000 7000 8000 9000Central Filter Wavelength-16.5- 17-17.5- 18-18.5-16.5-17-17.5-18- 18.5- 16.5- 17- 17.5- 18-16.5039- 16.5- 17- 17.5- 16- 18.5Objects 106, 120, 128, 1515000 6000 7000 8000 90005000 6000 7000 8000 9000-18.5- 17.5- 16.5-18-175000 6000 7000 8000 90005000 6000 7000 8000 9000Central Filter Wavelength40Table 4.2: EGY's published redshifts for objects in the field, and the derived redshiftsand galaxy types for both SEDs are listed here. Notice that object 95 is best fit by astellar model, and one SED for object 92 is also identified as a star. AZ1 and AZ2 arethe differences between published and derived redshifts. The R magnitude and mediansignal-to-noise ratio is shown for each galaxy.ID ZEGy ZSED1 Gall ZSED2 Gall AZi AZ2 MR S/N36 0.6090 0.718 3.1 G 0.718 3.9 G 0.109 0.109 21.65 3.444 0.46 0.462 4.5 G 0.641 6.0 G 0.002 0.181 22.02 2.775 0.6067 0.641 1.4 G 0.679 1.5 G 0.034 0.072 21.65 2.686 0.5668 0.245 3.5 G 0.318 4.5 G -0.322 -0.249 20.90 5.887 0.5037 0.567 0.2 G 0.549 0.3 G 0.063 0.045 21.10 4.492 0.6053 0.862 4.2 G 0.995 3.4 S 0.257 0.390 22.19 3.095 0.4349 0.950 1.3 S 0.841 1.9 S 0.514 0.406 21.67 3.196 0.5024 0.603 1.8 G 0.567 1.2 G 0.101 0.065 21.44 3.6106 0.3266 0.318 6.0 G 0.318 6.0 G -0.009 -0.009 21.72 3.7120 0.6025 0.567 2.2 G 0.567 1.3 G -0.036 -0.036 22.35 2.5128 0.5513 0.603 1.2 G 0.799 2.7 G 0.052 0.248 21.64 3.3151 0.5951 0.035 3.2 G 0.972 6.0 G -0.560 0.370 20.99 5.3414.5 Systematic Photometry ErrorsWhen the first set of spectral energy distributions and best fits for the EGY test samplewas plotted, it appeared that the data points corresponding to certain filters were consis-tently lower or higher than the best-fit model values. This was confirmed when, for eachfilter, the mean value of fl"—b"t r"del was calculated.modelIt was clear that the 6200-A data were significantly lower than the best-fit model inalmost every data set. Multiplying the flux level in this filter by a factor of 1.4 tended tobring the point closer to the best-fit model and improve the fit.It is possible to justify such an alteration to the data because the 6200A observationswere made on the second night, meaning there may have been some extinction from cirrusclouds. No other filters from that night show such a dramatic difference from the models;in fact, some data from that night even produce mean residuals higher than the modelflux level.A high mean residual may indicate a problem with the calibration. In particular, thezero point may have an error ascribable to the nature of the standard star spectrum. TheOke and Gunn spectral energy distribution for the standard star is smoothed, averagingover emission or narrow absorption lines. However, the filters used for these observationsare so narrow that, say, an emission line falling in the middle of the filter can affect thecalibration perceptibly.The success of the 6200A correction in improving the fit was such that similar cor-rections were made in every filter. Using the set of EGY galaxies with derived redshiftswithin 0.1 of the published values, an average residual between the observed and modelflux was calculated at each filter. The values of these mean residuals were used to applya multiplicative correction to each of the flux values in the data sets (effectively bringing42the average residual between the "corrected" data and the original best-fit models to zerofor the subsample of EGY galaxies).Best-fit models were assigned to all the "corrected" SEDs. The mean residuals in eachfilter were then calculated using the entire list of detected objects. If the "correction"of the data made up for real calibration errors or real extinction in the observations,one would expect the mean residuals between this data set and its best-fit models to besignificantly smaller than the residuals from the unaltered data set. In fact, the averageof the twenty-four mean residuals was very slightly higher for the "corrected" SEDs thanfor those in which only the value of the 6200A point was changed.The attempt to take into account an unknown amount of extinction in the flux valuesfrom some filters was therefore unsuccessful. The SEDs from which the redshifts and typeswere finally derived were chosen to be those in which only the 6200A flux was adjusted.4.6 ResultsThe final list of best fits to the SEDs is shown in table 4.3. The table contains theaverage redshift, galaxy type, and star type of the fits made to the two SEDs, as well asthe difference between the two derived values (indicated by A), and a key to whether thebest-fit model for each SED is a galaxy or a star. It also lists the x2 values for the beststar and galaxy fits for both SEDs. As previously discussed, the errors are different inthe two data sets and therefore the x2 values are not really comparable even between thetwo SEDs for the same object.Table 4.4 contains further data which could not be fit into table 4.3 : the position interms of a distance from the QSO position in arcseconds of right ascension and declination,the simulated R magnitude (the derivation of which is discussed in chapter 5, and someinformation on the signal-to-noise: the median value of 1-, the maximum, and the number43Table 4.3: Identifications for the objects in the field. The identifications were made usingtwo separate SEDs for each object. The second column summarizes the classificationmade for each SED: a "G" indicates that the best fit was to a galaxy template, while "S"stands for a star. The average best-fit redshift, galaxy type, and star type for each objectare listed here, together with the difference (A) between the two derived values. The X2values for both best-fit galaxy and star templates are in the last four columns. Furtherinformation on the objects is in the table following this one.ID Z AZ Gal AGal Star AStar X2.7an. X2q012 X!tarl XLar21 S G 0.023 0.023 3.2 5.7 33.0 58.0 4.455 0.452 3.485 0.5212 G G 0.274 0.000 0.9 0.5 64.0 0.0 7.367 5.353 13.723 12.1073 S S 0.718 0.000 0.0 0.0 74.5 5.0 0.752 1.341 0.620 0.9954 S S 0.873 0.022 0.0 0.0 80.0 0.0 14.068 16.317 2.940 3.2705 G G 0.445 0.000 3.6 3.5 62.0 4.0 0.356 0.203 0.399 0.2686 G G 0.427 0.279 1.7 3.4 65.5 3.0 0.470 0.281 0.589 0.3277 G G 0.173 0.202 1.5 0.7 62.0 0.0 0.910 1.038 1.491 1.7298 S G 0.167 0.334 4.8 2.3 46.0 34.0 0.640 0.290 0.633 0.3129 G G 0.282 0.104 2.2 0.3 62.0 0.0 0.277 0.779 0.333 1.01410 G G 0.238 0.015 3.7 0.0 53.0 0.0 0.880 1.039 1.270 1.24111 S G 0.504 0.891 5.8 0.5 61.0 12.0 0.525 0.285 0.520 0.43312 S S 0.678 0.327 0.0 0.0 75.5 1.0 1.375 0.622 0.939 0.29413 G G 0.755 0.481 0.7 1.4 70.5 13.0 1.746 0.496 2.674 0.73114 G G 0.134 0.222 0.6 1.2 63.0 2.0 0.547 0.558 0.612 0.66415 S S 0.841 0.000 1.1 0.2 71.5 3.0 4.366 4.441 2.363 2.07816 G G 0.245 0.086 1.1 1.9 63.0 2.0 0.429 0.318 0.537 0.37417 G S 0.632 0.057 2.5 5.1 66.0 18.0 0.507 1.770 0.570 1.43618 G G 0.883 0.043 3.7 1.3 68.0 2.0 0.435 0.395 0.511 0.49219 S G 0.479 0.069 0.0 0.0 73.5 9.0 3.067 0.360 2.556 0.42720 G G 0.259 0.029 1.7 0.0 62.0 0.0 0.379 1.020 0.699 1.55821 G G 0.523 0.017 0.6 1.1 68.0 2.0 0.367 1.049 0.697 1.32522 G G 0.738 0.000 1.9 0.0 69.0 0.0 3.249 1.338 4.415 3.15023 G G 0.513 0.035 1.0 0.5 68.0 2.0 0.677 0.566 0.824 1.14124 G G 0.357 0.046 3.8 1.0 60.0 6.0 0.275 0.215 0.306 0.24225 G G 0.799 0.042 3.8 0.2 67.0 0.0 0.610 0.500 0.688 0.57926 G G 0.763 0.464 1.1 2.3 70.5 3.0 0.383 0.360 0.685 0.49327 G G 0.471 0.085 2.8 5.3 66.5 1.0 1.013 0.407 1.060 0.55528 G G 0.343 0.139 1.9 1.7 64.0 0.0 0.539 0.555 0.653 0.70529 G S 0.867 0.257 3.0 6.0 69.0 14.0 0.330 2.481 0.412 2.12330 G S 0.417 0.563 0.6 1.3 66.0 18.0 0.263 1.178 0.296 1.08631 G G 0.274 0.000 1.3 0.2 63.0 2.0 2.283 2.264 3.721 4.56932 G G 0.274 0.000 2.1 0.7 62.0 0.0 0.248 0.754 0.321 0.99833 G G 0.448 0.703 3.4 4.6 63.0 2.0 0.230 0.394 0.274 0.50934 S S 0.841 0.000 0.9 0.1 75.0 4.0 2.430 1.464 1.325 0.91744ID Z AZ Gal AGal Star AStar X!mall X2ga12 XLarl Xtar2!35 G G 0.304 0.490 0.0 0.0 66.0 8.0 0.620 0.502 0.800 0.56736 G G 0.718 0.000 3.5 0.8 67.0 0.0 0.352 0.172 0.463 0.45937 G G 0.906 0.131 3.0 6.0 53.5 39.0 0.334 12.644 0.408 12.84938 G G 0.135 0.000 5.8 0.4 41.0 8.0 8.658 12.281 9.613 13.97539 S G 0.067 0.110 0.4 0.8 62.0 0.0 0.465 0.627 0.448 0.92940 G G 0.950 0.090 5.4 1.1 51.0 34.0 0.471 0.306 0.560 0.35041 S G 0.412 0.065 3.7 4.6 58.5 19.0 0.447 0.327 0.415 0.37142 G G 0.623 0.149 3.7 3.7 67.5 1.0 0.676 0.832 0.708 1.09143 G G 0.298 0.135 2.0 1.1 62.0 0.0 1.180 1.894 1.366 2.39044 G G 0.551 0.179 5.2 1.5 60.0 0.0 0.347 0.228 0.379 0.24745 G G 0.908 0.175 0.1 0.2 70.0 0.0 0.424 0.518 0.720 0.70746 G G 0.650 0.095 1.8 1.1 69.0 4.0 0.252 0.363 0.391 0.54947 G G 0.711 0.217 6.0 0.0 65.0 2.0 1.385 0.625 1.611 0.86448 G G 0.641 0.000 5.2 0.4 60.5 7.0 0.714 4.480 1.075 5.03049 G G 0.424 0.213 3.1 2.0 59.5 5.0 0.380 0.532 0.382 0.67750 G S 0.824 0.252 0.0 0.0 73.0 4.0 5.461 4.283 5.572 4.23251 G G 0.708 0.020 2.5 2.1 69.0 4.0 0.570 0.342 0.598 0.39652 G G 0.516 0.174 0.5 1.0 69.0 0.0 0.377 0.203 0.509 0.21453 G G 0.287 0.383 3.0 5.9 52.5 25.0 3.603 0.234 3.795 0.26254 S G 0.506 0.844 5.8 0.3 48.0 32.0 0.815 0.413 0.785 0.48355 S G 0.336 0.648 4.1 3.9 37.5 67.0 3.900 0.248 3.712 0.32056 G G 0.304 0.060 3.5 0.8 61.0 10.0 0.312 0.509 0.352 0.58557 S S 0.104 0.089 5.9 0.1 43.5 3.0 2.957 2.561 2.783 2.53358 G G 0.604 0.074 2.2 0.5 67.5 1.0 0.712 0.786 1.538 1.37259 S G 0.132 0.169 3.8 4.3 32.0 60.0 1.205 0.261 1.026 0.29660 G G 0.669 0.057 2.7 1.5 67.0 0.0 0.367 0.475 0.442 0.60961 G G 0.634 0.632 6.0 0.0 31.5 9.0 0.571 2.080 0.661 2.09862 G G 0.709 0.098 5.3 0.7 60.5 7.0 0.425 0.453 0.576 0.59163 G G 0.622 0.038 2.7 1.4 67.5 7.0 0.411 0.792 0.488 0.86964 G G 0.589 0.219 0.3 0.7 71.5 1.0 0.615 0.398 0.683 0.81165 G G 0.536 0.247 2.2 1.9 66.5 3.0 0.291 0.521 0.332 0.58266 G G 0.505 0.018 3.8 4.1 67.0 2.0 1.116 0.123 1.189 0.13267 G S 0.972 0.045 0.9 1.8 75.0 8.0 0.746 1.003 0.776 0.84768 G G 0.282 0.294 4.2 3.4 57.5 7.0 0.244 0.726 0.331 0.99869 S S 0.098 0.126 6.0 0.0 36.5 9.0 0.578 0.361 0.554 0.36070 G G 0.773 0.264 5.2 1.6 68.0 0.0 0.317 0.567 0.322 0.62971 S G 0.299 0.573 6.0 0.0 23.0 38.0 5.379 0.858 4.492 0.87072 S G 0.388 0.286 3.2 2.9 59.5 5.0 0.437 2.396 0.436 2.78273 G S 0.126 0.207 6.0 0.0 28.0 20.0 0.276 0.776 0.288 0.65274 G G 0.558 0.054 5.5 0.6 61.5 9.0 0.191 2.059 0.284 2.49545ID Z AZ Gal AGal Star AStar Xg2 all X2g012 4arl XLar275 G G 0.660 0.038 1.5 0.1 71.0 0.0 0.450 0.398 0.696 0.46876 S G 0.775 0.306 1.9 3.8 73.5 5.0 2.280 0.345 1.941 0.35577 G G 0.524 0.158 0.8 1.6 70.0 2.0 1.235 0.372 1.296 0.45378 G G 0.711 0.217 1.2 2.5 72.0 2.0 0.746 0.702 1.132 0.86279 S S 0.283 0.567 5.9 0.1 55.0 0.0 2.185 2.926 1.876 2.75180 S S 0.018 0.011 6.0 0.0 33.5 3.0 1.739 3.135 1.638 2.93481 G G 0.728 0.060 2.8 1.9 69.0 4.0 0.513 0.622 0.602 0.76382 G G 0.650 0.019 3.8 4.3 64.0 14.0 0.489 0.382 0.529 0.43083 G S 0.576 0.325 0.2 0.4 72.0 8.0 0.215 2.312 0.235 2.27184 G S 0.492 0.960 6.0 0.0 52.5 31.0 0.250 0.491 0.327 0.41385 G S 0.275 0.147 6.0 0.0 39.0 6.0 0.310 0.513 0.314 0.51086 G G 0.281 0.073 4.0 1.0 57.0 2.0 0.320 0.704 0.614 0.83187 G G 0.558 0.018 0.2 0.1 69.0 0.0 0.593 0.517 1.817 1.07188 G G 0.497 0.069 2.7 1.2 63.5 3.0 0.425 0.789 0.576 1.11789 G G 0.961 0.067 0.1 0.2 74.5 9.0 0.579 1.068 0.866 2.74490 S G 0.799 0.392 0.0 0.0 69.5 13.0 3.210 0.787 2.944 0.98591 G G 0.672 0.646 1.9 2.6 67.5 1.0 0.415 0.423 0.447 0.51292 G S 0.928 0.133 3.8 0.8 69.0 4.0 0.452 0.737 0.538 0.72793 G S 0.030 0.059 6.0 0.0 38.5 7.0 0.742 1.015 0.756 1.00194 S S 0.047 0.000 6.0 0.0 19.0 4.0 44.314 36.643 16.509 19.33195 S S 0.896 0.109 1.6 0.6 73.0 0.0 0.817 0.769 0.655 0.56996 G G 0.585 0.036 1.5 0.6 67.0 0.0 0.716 0.366 1.222 1.09897 G G 0.578 0.834 4.2 3.6 54.0 0.0 0.381 0.170 0.421 0.22698 G G 0.928 0.133 3.9 4.0 68.5 7.0 0.857 2.874 1.183 3.72699 G S 0.012 0.023 4.1 3.8 30.0 56.0 0.253 1.596 0.265 1.382100 G G 0.514 0.000 0.3 0.6 71.5 1.0 0.724 0.155 0.945 0.248101 S S 0.054 0.061 6.0 0.0 24.5 15.0 0.457 0.762 0.401 0.641102 S S 0.067 0.110 6.0 0.0 25.5 17.0 0.523 0.878 0.487 0.760103 G G 0.527 0.228 3.5 1.7 64.5 1.0 0.903 0.640 0.957 0.681104 S S 0.650 0.019 0.0 0.0 71.0 2.0 0.554 0.571 0.534 0.452105 G G 0.737 0.516 4.8 2.5 59.5 11.0 0.424 0.410 0.449 0.502106 G G 0.318 0.000 6.0 0.0 53.5 3.0 0.796 0.438 0.835 0.467107 G G 0.245 0.029 4.6 1.1 55.5 7.0 0.925 0.844 1.128 1.019108 G G 0.350 0.582 3.9 0.9 58.0 12.0 0.898 0.424 0.960 0.459109 G G 0.780 0.164 3.7 4.5 67.5 7.0 0.073 0.415 0.106 0.475110 S G 0.252 0.044 4.2 0.8 53.5 1.0 4.651 5.488 3.792 6.682111 G G 0.691 0.175 0.9 1.9 71.5 1.0 0.242 0.208 0.429 0.338112 G S 0.385 0.474 0.0 0.0 72.0 14.0 1.862 1.768 1.969 1.527113 S S 0.875 0.151 3.5 0.1 69.5 3.0 0.656 0.695 0.650 0.604114 G G 0.540 0.018 3.6 0.2 57.0 0.0 0.884 0.906 0.997 1.26846ID Z AZ Gal AGal Star AStar XLii X29a:2 XLarl XLar2115 G G 0.531 0.035 3.0 6.0 60.5 37.0 1.776 3.788 1.951 3.903116 G G 0.429 0.032 5.4 1.1 57.5 7.0 0.445 1.049 0.498 1.171117 G S 0.758 0.040 0.6 1.3 73.5 3.0 0.289 2.810 0.321 2.705118 G G 0.972 0.045 1.6 3.2 34.5 69.0 0.639 1.642 0.825 1.912119 S G 0.189 0.082 6.0 0.0 46.5 3.0 4.031 5.937 2.920 6.185120 G G 0.567 0.000 1.8 0.9 67.0 0.0 0.316 0.600 0.389 0.863121 G G 0.809 0.021 0.2 0.1 70.0 0.0 0.174 0.609 0.434 0.979122 G G 0.591 0.293 6.0 0.0 60.0 8.0 0.355 3.436 0.454 3.544123 S S 0.841 0.000 1.9 0.2 71.0 0.0 18.359 18.142 5.966 9.188124 G G 0.405 0.049 4.7 2.7 56.5 11.0 0.471 1.756 0.635 2.578125 S G 0.153 0.212 5.4 1.1 37.5 7.0 0.653 0.245 0.645 0.262126 S S 0.660 0.000 1.8 0.3 71.0 0.0 1.740 1.598 1.220 0.799127 G G 0.790 0.410 3.0 6.0 71.5 11.0 0.546 0.331 0.737 0.337128 G G 0.701 0.196 2.0 1.5 70.0 2.0 0.763 0.758 1.155 1.152129 G S 0.203 0.055 5.6 0.9 48.5 7.0 0.236 0.662 0.308 0.640130 G S 0.800 0.124 3.0 6.0 72.5 7.0 1.154 5.520 1.351 3.716131 G G 0.702 0.235 4.8 2.3 66.5 1.0 0.414 1.139 0.460 1.197132 G S 0.785 0.287 2.5 5.0 70.5 11.0 0.589 3.174 0.617 2.438133 G G 0.679 0.077 1.0 0.6 70.0 2.0 0.375 0.565 0.494 0.655134 G G 0.906 0.088 4.6 2.5 68.5 1.0 0.361 0.420 0.568 0.612135 G G 0.603 0.000 2.8 0.7 65.5 3.0 0.731 0.363 1.001 0.468136 G G 0.672 0.646 5.8 0.4 49.5 15.0 1.084 0.366 1.109 0.400137 G G 0.728 0.060 6.0 0.0 62.0 4.0 0.256 0.253 0.355 0.326138 G G 0.303 0.000 4.0 2.4 56.0 12.0 0.149 0.439 0.201 0.491139 G G 0.732 0.259 6.0 0.0 50.5 17.0 0.492 11.389 0.574 11.537140 G G 0.820 0.084 5.5 1.0 53.0 32.0 0.830 3.916 0.896 4.280141 G G 0.484 0.508 1.0 2.1 64.5 15.0 0.289 0.137 0.369 0.160142 G G 0.153 0.212 5.2 1.6 41.5 5.0 1.315 1.520 1.434 1.701143 5 S 0.906 0.044 0.9 1.8 75.5 9.0 3.001 0.113 2.958 0.107144 G G 0.837 0.316 3.0 6.0 59.0 20.0 0.983 0.691 1.302 0.735145 G G 0.672 0.646 3.0 6.0 52.5 33.0 0.563 0.936 0.622 1.030146 G S 0.291 0.178 5.3 1.3 54.5 1.0 0.515 0.596 0.588 0.583147 S S 0.503 0.276 0.2 0.5 71.0 0.0 0.433 0.403 0.402 0.352148 G S 0.605 0.111 0.1 0.1 71.0 8.0 0.376 0.726 0.529 0.620149 G S 0.157 0.291 5.1 1.8 33.5 53.0 0.284 1.190 0.311 0.920150 G G 0.650 0.057 3.8 4.3 66.5 1.0 0.565 0.424 0.725 0.446151 G G 0.503 0.937 4.6 2.8 51.0 4.0 0.442 0.632 0.662 1.183152 S G 0.867 0.257 2.3 4.7 66.5 19.0 2.234 2.028 2.223 2.145153 G G 0.667 0.306 5.2 1.7 64.5 5.0 0.312 0.242 0.367 0.262154 G G 0.641 0.000 6.0 0.0 57.0 0.0 2.490 0.455 2.941 0.78847ID Z AZ Gal AGal Star AStar X29all X2qa12 XLarl X82 i ar2155 G S 0.457 0.184 5.9 0.1 49.5 1.0 0.476 0.403 0.495 0.398156 G G 0.467 0.664 1.8 0.2 65.5 7.0 0.627 0.277 0.692 0.320157 G G 0.748 0.060 4.6 2.9 64.5 1.0 0.305 0.316 0.310 0.371158 G G 0.632 0.131 0.0 0.0 72.0 0.0 1.599 0.726 1.698 0.875159 G G 0.535 0.212 5.2 1.7 65.0 0.0 0.337 0.256 0.345 0.274160 G G 0.065 0.061 5.8 0.1 41.0 0.0 0.582 0.330 0.632 0.349161 S S 0.529 0.540 1.3 2.6 67.0 0.0 18.138 16.562 11.747 12.411162 G G 0.698 0.000 0.3 0.6 70.5 3.0 0.520 2.331 0.841 3.036163 S G 0.739 0.120 0.9 1.7 73.0 4.0 0.912 0.265 0.868 0.307164 G G 0.831 0.147 3.0 6.0 26.0 0.0 0.782 0.832 0.875 0.869165 S G 0.585 0.036 0.0 0.0 75.5 7.0 1.385 0.372 1.246 0.408166 G G 0.140 0.210 4.7 0.8 48.5 7.0 0.608 0.638 0.678 0.654167 G G 0.176 0.108 4.0 1.0 46.0 0.0 1.668 1.007 1.744 1.178168 G G 0.467 0.583 6.0 0.0 46.0 18.0 0.300 0.600 0.309 0.637169 G G 0.326 0.016 5.1 1.7 52.5 7.0 0.523 0.279 0.572 0.303170 G G 0.516 0.174 5.0 2.0 53.0 22.0 0.487 0.418 0.503 0.486171 G G 0.811 0.187 3.7 4.7 56.5 31.0 0.503 0.212 0.609 0.317172 G G 0.625 0.187 6.0 0.0 53.0 8.0 0.502 0.421 0.529 0.490173 G G 0.533 0.744 5.2 1.6 56.5 23.0 0.275 0.370 0.285 0.462174 S G 0.972 0.045 4.0 2.2 68.0 2.0 0.570 0.405 0.541 0.513175 G G 0.365 0.095 0.6 1.3 64.0 0.0 1.175 3.902 2.848 5.400176 G G 0.084 0.050 4.4 0.7 46.5 1.0 0.322 0.633 0.400 0.654177 G G 0.281 0.014 1.1 0.0 64.0 0.0 6.781 3.752 9.500 6.536178 G G 0.887 0.217 0.4 0.6 72.0 0.0 0.634 0.526 0.806 0.546179 G G 0.567 0.675 0.9 1.9 65.5 7.0 0.341 4.530 0.360 4.565180 G G 0.738 0.000 0.2 0.5 72.0 0.0 0.160 0.520 0.286 0.919181 G G 0.238 0.015 5.8 0.5 53.5 3.0 4.424 4.499 4.609 5.973182 G S 0.730 0.179 0.7 1.4 71.5 5.0 0.216 1.783 0.226 1.668183 G G 0.577 0.127 3.0 6.0 65.5 17.0 0.744 0.830 0.866 1.005184 G G 0.877 0.237 4.6 0.8 65.5 3.0 0.575 0.357 0.709 0.44848of filters (out of a possible twenty-four) in which -L > 3.0.of —Forty-two of the objects in the list were classified as galaxies according to one SED andas stars according to the other. Most of these objects were very faint, so the noise in theSEDs probably obscured the spectral shape and did not permit a definite identification. Afew, however, are quite bright, notably those numbered 39, 110, and 119, for which all thedata points have signal-to-noise above 3.0. Objects 72, 92, and 129 have 12, 12, and 19points with SIN > 3 respectively. It is likely that these belong to classes of galaxies notincluded in the range of models considered in the fitting program; they may be emission-line objects. Objects such as 50, which has a median signal-to-noise in its data set of 1.6,and a maximum of 9.4, with only two points having SIN > 3, may represent emission-lineobjects.Photometric data for the 30281 field were taken from Yee, Green, and Stockman(1986), and Yee (pers. comm.) provided classifications for these objects (i.e. galaxy orstar) based on the shape of each image. Table 4.5 summarizes the type classifications andmagnitudes of the thirty-three objects which appear in this data set. One of the objects,number 94, is the QS0 itself. Of the remaining thirty-two objects, Yee's analysis indicatesthat eight are stars and twenty-four are galaxies. The template-fitting program identifiedfive of the objects as a star from one SED and a galaxy from the other. Excluding thosefive objects, star types have been assigned to 2/6 of Yee's remaining stars, and galaxytypes to 20/21 of his galaxies.The field size for the data presented by Yee, Green, and Stockman was 3.64 squarearcminutes, and they listed the limiting R magnitude for the 30281 field as 23.4. Thecorresponding limit in V is roughly 24 magnitudes, and the Bahcall-Soneira model (Ratna-tunga and Bahcall 1985) predicts between four (for Vun, = 23) and seven (for Vun, = 25)stars in the field. This suggests the possibility that one or more of Yee's stars is actually49Table 4.4: The position of each object is given as a distance in arcseconds of right ascensionand declination from the position of the QSO. The R magnitude and uncertainty arederived in chapter 5 but listed here for completeness. The three signal-to-noise valuesrepresent the median S/N value from all the observations, the maximum value of theS/N, and the number of filters in which the S/N was 3.0 or greater.ID ARA (") ADec mR cr„„ Median S/N Max S/N No. S/N>3.01 -159.78 20.66 22.54 0.84 2.2 3.7 32 -159.09 0.34 17.55 0.04 23.9 31.0 243 -158.06 18.25 21.35 0.18 4.5 8.3 174 -157.37 -27.55 20.19 0.08 7.6 20.4 235 -156.68 8.26 22.79 0.73 1.5 3.9 16 -154.27 82.64 21.82 0.26 2.6 4.8 97 -153.92 54.06 20.08 0.10 8.4 11.6 248 -152.20 -39.26 23.09 1.40 1.1 3.4 29 -151.86 -24.45 21.29 0.16 4.1 6.6 1710 -146.35 19.97 20.38 0.09 6.4 14.3 2311 -141.53 -27.89 23.14 0.91 1.4 3.1 112 -141.53 26.51 23.34 0.00 1.1 2.6 013 -139.46 88.15 22.25 0.63 2.3 3.6 314 -135.67 -14.12 21.00 0.11 5.2 7.9 2215 -135.67 -64.05 19.96 0.06 10.7 20.8 2416 -134.99 2.07 21.93 0.21 2.9 6.9 1017 -131.20 61.64 23.23 0.74 1.4 2.5 018 -130.85 13.77 22.04 0.33 2.5 6.6 819 -130.16 26.17 22.50 0.38 1.9 2.8 020 -125.69 -11.36 20.21 0.08 8.7 13.2 2421 -125.69 -51.31 21.93 0.29 3.1 6.1 1422 -125.00 105.72 20.59 0.13 6.3 11.2 2323 -121.56 -48.21 21.30 0.18 4.4 7.5 1824 -118.46 54.41 22.32 0.37 2.2 3.3 625 -117.42 80.58 22.16 0.32 3.0 6.0 1226 -112.95 18.25 22.83 0.64 1.8 3.9 127 -109.50 -26.86 22.76 0.71 1.8 5.3 228 -108.81 -9.30 20.36 0.07 8.9 14.7 2429 -107.78 -1.72 23.32 0.60 1.4 3.3 130 -105.37 46.14 23.70 0.85 1.2 2.0 031 -104.68 -79.89 18.48 0.04 18.6 25.1 2432 -103.99 40.29 20.80 0.09 5.7 11.0 1933 -103.99 25.14 22.57 0.48 1.7 2.9 034 -102.62 1.03 20.73 0.09 7.7 17.9 2050ID ARA (") ADec mg cr„,„ Median S/N Max S/N No. S/N>3.035 -101.58 14.12 23.49 0.00 1.3 2.3 036 -96.07 23.76 21.65 0.22 3.4 6.9 1337 -95.73 49.93 23.43 0.91 1.1 2.1 038 -88.50 -91.94 18.86 0.04 19.3 27.9 2439 -87.46 4.13 20.66 0.08 7.8 12.5 2440 -83.33 -9.30 23.37 2.50 1.1 2.8 041 -81.61 -41.32 22.93 0.00 1.8 4.1 142 -77.48 95.04 22.90 0.85 1.6 3.3 143 -77.48 -60.61 19.84 0.06 10.7 15.7 2444 -76.79 -2.07 22.02 0.25 2.7 4.9 945 -70.59 15.15 22.78 0.49 2.1 5.7 546 -69.90 100.89 22.75 1.48 1.5 4.8 647 -64.39 -73.00 22.14 0.38 2.1 6.5 348 -62.67 97.80 21.36 0.16 3.9 7.6 1949 -62.33 6.89 20.57 0.10 7.1 11.3 2350 -60.26 68.53 23.32 0.00 1.6 9.4 251 -59.57 30.99 21.88 0.26 3.2 8.6 1352 -57.16 37.19 22.84 0.75 1.2 2.3 053 -54.06 10.67 22.32 0.32 2.2 12.4 554 -48.55 -49.24 22.81 0.57 1.3 2.8 055 -48.21 -33.06 22.83 0.61 1.8 3.2 456 -47.86 48.21 22.29 0.39 2.2 3.6 357 -46.49 29.61 21.12 0.11 4.7 12.2 1758 -45.11 -14.81 20.92 0.12 5.4 12.3 2159 -44.77 102.62 23.31 0.77 1.1 1.9 060 -44.08 22.04 22.57 0.44 2.1 4.0 361 -44.08 -107.09 23.23 2.23 1.2 2.2 062 -43.04 88.15 22.28 0.36 2.1 4.1 663 -40.63 14.12 22.84 0.51 1.5 3.2 164 -39.60 -66.12 21.56 0.22 3.6 8.1 1765 -35.12 -28.24 22.21 0.36 2.2 4.0 566 -33.75 52.00 22.85 0.00 1.3 6.2 167 -33.40 12.05 23.51 0.93 1.5 4.0 168 -32.71 -49.93 21.03 0.14 4.7 9.0 1969 -27.89 -95.04 22.50 0.33 1.8 3.7 470 -26.17 -21.69 23.09 0.91 1.3 2.2 071 -25.48 73.69 22.56 0.53 1.8 7.7 272 -25.48 -90.56 21.73 0.23 2.9 5.7 1273 -24.45 -81.27 23.37 0.89 1.4 2.4 074 -23.76 -42.70 21.78 0.24 3.1 5.6 1151ID ARA (") ADec mn a„,„ Median S/N Max S/N No. S/N>3.075 -19.63 48.55 21.65 0.26 2.6 6.5 1076 -19.28 16.87 23.10 0.46 1.5 2.6 077 -18.59 19.97 22.28 0.34 2.4 7.1 978 -17.56 -34.09 22.92 0.72 1.7 3.3 279 -17.56 -85.74 21.17 0.12 4.4 11.6 2080 -14.46 50.28 22.83 1.28 1.6 3.3 181 -14.12 25.14 22.65 0.55 2.0 3.2 182 -13.09 -71.62 23.02 0.61 1.5 3.1 183 -12.74 -6.20 22.66 0.44 1.8 3.0 084 -11.71 -76.45 22.66 0.52 1.4 3.0 085 -9.64 -94.01 23.18 0.71 1.2 3.4 186 -8.26 68.87 20.90 0.10 5.8 9.8 2287 -7.23 14.12 21.10 0.17 4.4 9.0 1988 -6.89 5.51 21.58 0.20 3.3 5.1 1489 -6.89 -37.53 22.33 0.34 2.4 7.0 990 -5.17 70.94 23.06 0.98 1.5 3.2 191 -3.79 -19.63 22.84 0.70 1.7 2.8 092 -1.03 -5.17 22.19 0.34 3.0 4.5 1293 -0.34 3.44 22.45 0.38 2.2 4.2 694 0.00 0.00 17.37 0.04 26.9 36.4 2495 0.69 -98.83 21.67 0.21 3.1 6.6 1296 2.41 20.32 21.44 0.19 3.6 6.5 1597 2.75 68.18 23.24 1.06 1.2 2.3 098 3.10 -0.34 22.50 0.35 2.7 6.2 1099 4.48 -56.82 23.24 0.56 1.3 3.0 0100 5.85 48.21 23.13 0.00 1.3 3.7 1101 7.58 -17.91 22.64 0.34 2.0 5.6 4102 7.58 97.45 22.63 0.49 2.2 4.2 1103 10.33 -52.00 20.83 0.09 6.7 11.2 22104 12.05 -8.26 22.63 0.55 1.9 5.6 3105 12.40 -3.44 23.07 0.89 1.3 2.4 0106 13.43 -22.38 21.72 0.23 3.7 5.5 14107 13.77 -108.47 21.10 0.14 5.4 8.3 20108 15.50 -32.02 22.45 0.38 2.1 3.7 4109 15.84 82.64 23.40 0.61 1.3 2.2 0110 20.32 -105.37 16.89 0.04 26.7 35.3 24111 21.35 15.84 22.32 0.33 2.6 4.4 7112 23.07 72.31 23.00 0.57 1.4 8.2 1113 23.76 -42.70 22.92 0.64 1.6 4.3 2114 24.79 86.78 21.44 0.21 3.3 5.8 1752ID ARA (") ADec mR mi, Median S/N Max S/N No. S/N>3.0115 24.79 -17.56 22.94 0.00 1.6 3.1 1116 25.14 35.47 21.25 0.14 5.0 8.7 22117 25.48 -61.29 22.58 0.71 1.6 3.4 2118 26.51 59.23 22.77 0.72 2.0 4.5 1119 27.55 33.40 18.22 0.04 24.0 28.9 24120 27.89 -71.28 22.35 0.34 2.5 4.3 5121 32.02 -60.61 21.72 0.25 3.3 8.4 13122 32.71 25.14 23.20 0.00 0.8 2.4 0123 35.12 -76.45 16.55 0.04 27.2 34.5 24124 36.16 51.65 20.42 0.11 7.2 10.9 24125 38.22 17.22 23.30 1.82 1.4 2.9 0126 39.94 -6.54 20.85 0.11 6.2 11.1 23127 39.94 -63.70 23.29 1.58 1.0 3.7 1128 40.29 -11.71 21.64 0.25 3.3 6.0 13129 42.36 -75.76 21.39 0.15 3.7 5.8 19130 45.45 66.12 22.91 1.11 1.8 3.4 1131 47.52 -11.71 22.92 0.65 1.4 3.0 0132 47.86 -59.23 23.25 0.00 1.2 2.5 0133 47.86 -61.98 22.52 0.48 1.5 4.2 5134 50.96 61.64 21.95 0.33 2.8 5.3 10135 53.03 -44.42 22.34 0.36 2.5 4.0 11136 53.03 22.38 22.97 0.47 1.2 4.2 1137 53.37 54.41 23.02 0.60 1.6 2.9 0138 54.06 99.17 23.00 0.53 1.1 2.6 0139 54.41 -86.78 23.00 0.80 1.5 3.1 1140 57.51 87.46 23.25 1.33 1.0 3.6 1141 77.82 3.44 22.99 1.10 1.1 2.8 0142 78.51 58.88 21.05 0.13 4.1 8.4 19143 79.89 26.86 23.95 0.00 0.8 6.6 1144 81.27 -50.62 23.48 0.00 1.4 5.0 1145 82.99 22.38 23.28 0.00 1.1 2.2 0146 83.68 -80.23 22.52 0.39 2.0 4.1 4147 88.84 -24.10 22.28 0.40 2.5 4.4 6148 88.84 12.05 23.64 0.00 0.8 2.1 0149 89.19 -101.58 23.10 0.69 1.0 2.3 0150 90.91 21.35 22.84 0.66 1.3 3.1 1151 92.29 -84.71 20.99 0.12 5.3 7.7 23152 93.32 2.41 22.97 0.47 1.5 2.5 0153 94.01 -44.42 23.62 0.00 0.9 2.0 0154 98.48 61.98 20.73 0.11 4.6 13.5 2053ID ARA (") ADec mR an,„ Median S/N Max S/N No. S/N>3.0155 98.48 43.39 22.41 0.33 2.1 3.5 2156 101.58 -32.71 22.61 0.60 1.8 3.4 1157 103.31 66.12 22.94 0.70 1.2 3.0 0158 103.31 -52.34 22.55 0.51 1.8 8.1 3159 103.65 4.13 22.85 0.52 1.3 2.8 0160 113.64 67.15 21.66 0.27 3.2 6.1 13161 116.39 78.17 15.68 0.04 27.5 36.0 24162 116.39 35.12 22.24 0.65 2.4 6.0 7163 117.42 -39.26 23.24 0.00 1.3 1.9 0164 121.56 -70.59 23.35 1.04 1.2 2.6 0165 123.28 -38.57 22.66 1.58 1.2 3.2 1166 125.34 -14.12 21.82 0.22 3.3 5.4 14167 126.03 10.33 20.85 0.12 5.8 8.6 22168 127.07 -7.92 22.53 0.44 1.6 3.5 1169 127.75 79.89 22.28 0.42 2.2 3.1 2170 127.75 -90.56 23.05 0.00 1.3 2.4 0171 129.13 -44.42 22.61 0.56 1.8 5.1 3172 133.26 88.84 21.73 0.23 3.2 5.8 13173 133.26 16.18 22.92 0.53 1.6 2.6 0174 136.02 -97.45 22.24 0.42 2.1 4.3 7175 136.02 -102.62 19.44 0.06 12.3 17.5 24176 138.77 69.21 22.15 0.35 2.5 6.1 7177 141.18 32.71 18.93 0.05 16.2 22.8 24178 141.87 -45.80 22.65 0.66 1.7 3.8 3179 142.91 -101.24 23.01 0.92 1.0 2.7 0180 143.25 -40.29 21.77 0.33 2.5 5.1 8181 143.94 -28.24 18.34 0.04 21.5 27.5 24182 143.94 -92.29 22.72 0.60 1.5 2.7 0183 146.00 -73.69 22.72 1.48 1.4 2.6 0184 147.04 45.45 22.32 0.41 2.1 4.2 354a compact or unresolved galaxy, or bright galactic nucleus, with the appearance of a star.In all, the list contains 184 objects, of which 122 are classified as galaxies according toboth SEDs, and twenty as stars. The QSO 3C281, object 94, is classified as a type B9Vstar; its best-fit galaxy type is 6.0 at z= 0.047. Its real redshift, from EGY, is 0.602. Ofcourse, the fitting program includes no model spectrum for an unusual object like a QS0,so neither class nor derived redshift was expected to be accurate.The distributions of best-fit galaxy types and best-fit star types are plotted in the twohistograms shown in figure 4.5. The solid-line histogram shows the distribution for thewhole sample, while the shaded histogram shows the distribution only for those objectsclassified by the fitting program as galaxies in the first plot and as stars in the second.The distribution of identified galaxy types seems approximately uniform when onlythose objects identified as galaxies in both SEDs are considered. Other objects seem tobe preferentially assigned extreme types (0.0 and 6.0) but, as discussed earlier, many ofthese objects either are too faint to be uniquely fit by a model or are unusual types notrepresented in the set of galaxy models used by the fitting program.Recalling that types 0 and 1 represent UV-cold and -hot E/SO galaxies and 2 through 5are spirals, the distribution of galaxy types suggests that there are very roughly three timesas many spirals as early-type galaxies in the field. The Revised Shapley-Ames Catalog ofBright Galaxies (Sandage and Tammann 1987) finds a slightly lower ratio of spirals toE/S0s for relatively local galaxies (the completeness limit of the catalog is mpg = 13.2).It is possible that a Butcher-Oemler effect is observed in the cluster at z 0.6, increasingthe number of blue late-type galaxies in the field.There are only twenty objects classified as stars according to both SEDs, of whichthirteen are of type 64 (K5) or later. This is not unexpected given that the line of sightto the cluster passes through the disk of the Galaxy. The QS0's position in 1950 galactic55Table 4.5: Magnitudes and type classifications for the set of calibration galaxies. R magni-tudes, shown in the second column, were taken from Yee, Green, and Stockman (1986) tocalibrate the instrumental magnitudes, and the resulting calibrated magnitudes appear inthe third column, followed by their uncertainties. Yee's classifications (s=star, g=galaxy)based on the image shape are given in the fifth column, next to the classifications derivedfrom template-fitting. Where the two SEDs classed the object as a star and a galaxy,both are noted. A "y" following indicates that both image shape and SED-fitting methodsyield the same classification. The last column contains the median signal-to-noise.Object Ry0s86 Retail, 0R Class (Yee) Class (SED) Match? SiN49 20.74 20.568 0.100 g g Y 7.151 22.15 21.878 0.264 s g n 3.252 22.51 22.845 0.749 g g Y 1.253 22.63 22.315 0.318 g g Y 2.255 22.63 22.826 0.607 g sg 1.856 22.02 22.293 0.386 s g n 2.257 20.94 21.116 0.110 g s n 4.758 20.84 20.917 0.121 g g Y 5.460 22.61 22.569 0.444 s g n 2.163 22.27 22.839 0.506 g g Y 1.565 22.28 22.208 0.356 g g Y 2.267 23.35 23.511 0.927 s gs - 1.575 21.72 21.645 0.258 g g Y 2.677 22.04 22.283 0.342 g g Y 2.478 22.49 22.924 0.717 s g n 1.781 22.05 22.647 0.545 g g Y 2.083 22.56 22.662 0.435 g gs - 1.887 20.96 21.105 0.166 g g Y 4.488 21.11 21.577 0.205 g g Y 3.389 21.76 22.330 0.340 g g Y 2.491 22.48 22.845 0.698 g g Y 1.792 22.25 22.191 0.339 g gs 3.094 17.27 17.374 0.037 qso s - 26.996 21.58 21.444 0.188 g g Y 3.6101 22.38 22.641 0.343 s s y 2.0104 22.21 22.631 0.546 s s y 1.9105 23.01 23.066 0.891 g g Y 1.3106 21.60 21.715 0.230 g g Y 3.7108 23.16 22.455 0.375 g g Y 2.1111 22.06 22.319 0.334 g g Y 2.6116 20.98 21.245 0.142 g g Y 5.0119 18.38 18.223 0.038 s sg - 24.0128 22.00 21.639 0.248 g g Y 3.35630-z10I^I^I0 " 'Galaxy type0-^----_---,^1 -,^I^,60 8040— I30 —.,a)-E 20 —0z10 —0 ^O ,^I^,20 ,^I^,40Star typeFigure 4.5: The upper histogram is the distribution of average best-fit galaxy types forthe entire list of detections. The shaded histogram shows the distribution only for objectsclassified as galaxies according to both SEDs. The lower histogram shows the distributionof best-fit star types, and the twenty objects in the shaded histogram are those identifiedas stars from both SEDS.57coordinates is 1 = 314.50°, b = 69.20°. Star counts in the direction of the field may be es-timated using the Bahcall-Soneira model (e.g. Ratnatunga and Bahcall 1985). Assuminga limiting R magnitude for these data of about 22.5 (chapter 5), which corresponds overall star types to an approximate limit of V=23, the model predicts that about twenty-onestars will be seen in the field, of which fourteen will be of types K and M. These numbersare close to those observed in the field.The distribution of assigned redshifts, shown in figure 4.6, is quite interesting. In thisfigure, objects classified as galaxies according to both SEDs are shown by the shadedhistogram. All but two of the objects with best-fit galaxy redshifts of 0.1 or below areclassed as stars. One would expect the number of galaxies to increase with redshift,since at greater distances a larger volume is being sampled per redshift bin. At highredshifts, however, the faintest objects will not be seen, so only the most luminous will besampled, and the volume density of detected galaxies will decrease with redshift (ignoringevolution).These biasing effects are almost impossible to correct with these data, since there arelarge uncertainties in both redshift and magnitude. This distribution therefore does notrepresent a luminosity function for the field. It is interesting to notice, however, thatthere seems to be an enhancement in the number of galaxies near the quasar redshift,where one might expect to find cluster galaxies. Table 4.6 lists the fifty-eight objectswith redshifts within 0.1 of the quasar redshift 0.602. However, the "enhancement" in theredshift distribution may only represent the natural peak created by the combination ofthe two biases, one causing a dropoff in counts at high redshift and the other causing adropoff at low redshift.The type identifications and redshifts should be more accurate for objects of highersignal-to-noise ratio. Figure 4.7 shows the redshift distribution from figure 4.6, but in-58-_---_-i^1^,0.2I^1^I^I^I0.4 0.6Redshift.^1^.0.80Figure 4.6: Distribution of the redshifts derived by template-fitting for all the objects inthe field, shown by the solid histogram. The shaded histogram shows the distribution forobjects classified as galaxies for both SEDs.59Table 4.6: Possible cluster members by redshift. The best-fit galaxy templates for theseobjects are at redshifts within 0.1 of the quasar redshift.ID Z AZ Gal. Star ABA ADec mR cr„,,, S/N147 S S 0.503 0.276 0.2 71.0 88.84 -24.10 22.28 0.40 2.5151 G G 0.503 0.937 4.6 51.0 92.29 -84.71 20.99 0.12 5.311 5 G 0.504 0.891 5.8 61.0 -141.53 -27.89 23.14 0.91 1.466 G G 0.505 0.018 3.8 67.0 -33.75 52.00 22.85 0.00 1.354 S G 0.506 0.844 5.8 48.0 -48.55 -49.24 22.81 0.57 1.323 G G 0.513 0.035 1.0 68.0 -121.56 -48.21 21.30 0.18 4.4100 G G 0.514 0.000 0.3 71.5 5.85 48.21 23.13 0.00 1.352 G G 0.516 0.174 0.5 69.0 -57.16 37.19 22.84 0.75 1.2170 G G 0.516 0.174 5.0 53.0 127.75 -90.56 23.05 0.00 1.321 G G 0.523 0.017 0.6 68.0 -125.69 -51.31 21.93 0.29 3.177 G G 0.524 0.158 0.8 70.0 -18.59 19.97 22.28 0.34 2.4103 G G 0.527 0.228 3.5 64.5 10.33 -52.00 20.83 0.09 6.7161 5 S 0.529 0.540 1.3 67.0 116.39 78.17 15.68 0.04 27.5115 G G 0.531 0.035 3.0 60.5 24.79 -17.56 22.94 0.00 1.6173 G G 0.533 0.744 5.2 56.5 133.26 16.18 22.92 0.53 1.6159 G G 0.535 0.212 5.2 65.0 103.65 4.13 22.85 0.52 1.365 G G 0.536 0.247 2.2 66.5 -35.12 -28.24 22.21 0.36 2.2114 G G 0.540 0.018 3.6 57.0 24.79 86.78 21.44 0.21 3.344 G G 0.551 0.179 5.2 60.0 -76.79 -2.07 22.02 0.25 2.774 G G 0.558 0.054 5.5 61.5 -23.76 -42.70 21.78 0.24 3.187 G G 0.558 0.018 0.2 69.0 -7.23 14.12 21.10 0.17 4.4120 G G 0.567 0.000 1.8 67.0 27.89 -71.28 22.35 0.34 2.5179 G G 0.567 0.675 0.9 65.5 142.91 -101.24 23.01 0.92 1.083 G S 0.576 0.325 0.2 72.0 -12.74 -6.20 22.66 0.44 1.8183 G G 0.577 0.127 3.0 65.5 146.00 -73.69 22.72 1.48 1.497 G G 0.578 0.834 4.2 54.0 2.75 68.18 23.24 1.06 1.296 G G 0.585 0.036 1.5 67.0 2.41 20.32 21.44 0.19 3.6165 S G 0.585 0.036 0.0 75.5 123.28 -38.57 22.66 1.58 1.260ID Z AZ Gal. Star ARA ADec mR cr„,,, S/N64 G G 0.589 0.219 0.3 71.5 -39.60 -66.12 21.56 0.22 3.6122 G G 0.591 0.293 6.0 60.0 32.71 25.14 23.20 0.00 0.8135 G G 0.603 0.000 2.8 65.5 53.03 -44.42 22.34 0.36 2.558 G G 0.604 0.074 2.2 67.5 -45.11 -14.81 20.92 0.12 5.4148 G S 0.605 0.111 0.1 71.0 88.84 12.05 23.64 0.00 0.863 G G 0.622 0.038 2.7 67.5 -40.63 14.12 22.84 0.51 1.542 G G 0.623 0.149 3.7 67.5 -77.48 95.04 22.90 0.85 1.6172 G G 0.625 0.187 6.0 53.0 133.26 88.84 21.73 0.23 3.217 G S 0.632 0.057 2.5 66.0 -131.20 61.64 23.23 0.74 1.4158 G G 0.632 0.131 0.0 72.0 103.31 -52.34 22.55 0.51 1.861 G G 0.634 0.632 6.0 31.5 -44.08 -107.09 23.23 2.23 1.248 G G 0.641 0.000 5.2 60.5 -62.67 97.80 21.36 0.16 3.9154 G G 0.641 0.000 6.0 57.0 98.48 61.98 20.73 0.11 4.646 G G 0.650 0.095 1.8 69.0 -69.90 100.89 22.75 1.48 1.582 G G 0.650 0.019 3.8 64.0 -13.09 -71.62 23.02 0.61 1.5104 S S 0.650 0.019 0.0 71.0 12.05 -8.26 22.63 0.55 1.9150 G G 0.650 0.057 3.8 66.5 90.91 21.35 22.84 0.66 1.375 G G 0.660 0.038 1.5 71.0 -19.63 48.55 21.65 0.26 2.6126 S S 0.660 0.000 1.8 71.0 39.94 -6.54 20.85 0.11 6.2153 G G 0.667 0.306 5.2 64.5 94.01 -44.42 23.62 0.00 0.960 G G 0.669 0.057 2.7 67.0 -44.08 22.04 22.57 0.44 2.191 G G 0.672 0.646 1.9 67.5 -3.79 -19.63 22.84 0.70 1.7136 G G 0.672 0.646 5.8 49.5 53.03 22.38 22.97 0.47 1.2145 G G 0.672 0.646 3.0 52.5 82.99 22.38 23.28 0.00 1.112 S S 0.678 0.327 0.0 75.5 -141.53 26.51 23.34 0.00 1.1133 G G 0.679 0.077 1.0 70.0 47.86 -61.98 22.52 0.48 1.5111 G G 0.691 0.175 0.9 71.5 21.35 15.84 22.32 0.33 2.6162 G G 0.698 0.000 0.3 70.5 116.39 35.12 22.24 0.65 2.4128 G G 0.701 0.196 2.0 70.0 40.29 -11.71 21.64 0.25 3.3131 G G 0.702 0.235 4.8 66.5 47.52 -11.71 22.92 0.65 1.461cludes only the objects with median signal-to-noise ratio of 2.0 or more (99 objects) andof 3.0 or more (61 objects). Again, those objects classed twice as galaxies are shown bythe shaded histograms.Some apparent enhancement in the number of galaxies near redshift 0.6 is still seen ineach of these distributions, again suggesting that some cluster galaxies are being detected.A more thorough statistical discussion of the significance of this apparent cluster cannotbe made, due to the size of the redshift and magnitude errors and to the presence of thetwo biases described above. However, there is another enhancement in the number ofgalaxies with redshifts around 0.25 to 0.3, which is clearly seen in all three figures. Thenearest redshift published by EGY was at z=0.3266, but they did not find evidence for acluster at that redshift. Without obtaining spectra of these objects, we cannot be certainwhether this enhancement represents a second cluster or just an artifact in the fittingprocess.62I0.4^0.6Redshift0.4^0.6Redshift1^1^0 . 8^1^I^ICVArn 105a.)coA0) 100z1^I^10.2I 0.2 0.8Figure 4.7: The distribution of the derived redshifts, for objects with median signal-to-noise ratio 2.0 or greater (upper plot) and 3.0 or greater (lower plot). The shadedhistograms represent those objects classified as galaxies according to both SEDs.63Chapter 5Broadband MagnitudesThe narrowband filters can be combined in such a way as to simulate the trans-mission curves of some standard broadband filters. Besides allowing presentation of themagnitudes of the objects in a recognizable form, this should make it possible, with theappropriate spectral coverage, to construct the filters used by Loh and Spillar or Koo,and compare results from SED-fitting with redshifts derived from optical multicolours.Unfortunately, due to the spectral range and distribution of the filters used in thisstudy, it was only possible to make a reasonable simulation of one filter, R. The R trans-mission curve modelled is from Besse11 (1983). The transmission curves of twenty-onefilters were each multiplied by an appropriate coefficient and combined to produce amodel R filter transmission curve with the same area, in transmission-wavelength space,as the Bessell R. Both the real and the model R filters are shown in figure 5.1. Thecoefficients for the filters were chosen so that the difference between the areas is as smallas possible regardless of the width of the spectral subregion considered.The simulated filter will provide a good approximation of an R filter for flat-spectrumobjects, and a less-good approximation for objects with steep spectra or other spectralfeatures. Because of its comb-like shape, it would not be a satisfactory filter for studyingobjects with significant energy in emission lines, since the observed brightness would be641111.^I^.6000 7000Wavelength (1)___--1\\--1-1il,^I^.8000--__-----1.50Figure 5.1: The true Bessell R filter is the dashed line, and the solid line represents a sumof narrowband filter transmission curves which have been multiplied by coefficients tosimulate the broadband filter. The area below the two curves in wavelength-transmissionspace is the same.65a function of redshift (depending on whether or not the emission lines fall in the regionsbetween peaks in the transmission curve).To compute an "instrumental" r magnitude for an object, the total energy throughthe model R filter was computed as the sum of the coefficient (Ci) times the flux (fi) ineach filter:#filter sMinstr = 2.5 logio EC2 x fi.i=1Flux from the small aperture used to construct the SEDs was not used here, because ameasure of all the light from the objects was sought. The possibility of using a singlerelatively large aperture was investigated, but the high sky contribution for objects ofsmall angular size caused extra uncertainty in the fluxes for the faintest objects. For thisreason the fluxes in the apertures selected by PPP's growth-curve algorithm were used.Although the growth-curve approach can underestimate the true flux value (as discussedearlier in section 3.1), the scatter in the R-magnitude calibration plot was smaller forgrowth-curve instrumental magnitudes than for those in a fixed large aperture.Some objects are so faint in some filters that the subtraction of the background levelleaves a negative signal. In these cases, PPP was not able to compute a magnitude, andthe intensity for that filter was set to zero in the SED of the object. When such a pointwas encountered in the calculation of the instrumental magnitude, an interpolation wasmade from the fluxes in the two adjacent filters; that is, the intensity in each of theadjacent filters was divided by the equivalent width of the filter, then an interpolation (orextrapolation, in the case of a zero value in the first or last filters) was made, and finallythe interpolated value was multiplied by the equivalent width of that filter to produce anestimated intensity.The uncertainty in the instrumental magnitude is calculated by keeping a running sumof the total flux uncertainty over the contributions from the filters. This quantity is simply66the product of the flux in the filter, the fractional error of the flux (7), and the modelcoefficient for the simulated R transmission curve. The resulting uncertainties range from0.04 magnitudes for the QSO and a few of the brightest objects, to 2.5 magnitudes forthe faintest galaxy, at magnitude R = 23.37. As expected, in general the estimateduncertainty in the r magnitudes of objects increases as the median signal-to-noise ratiodecreases. This is illustrated by figure 5.2, which suggests that for objects with signal-to-noise ratio below about three or four, this technique does not supply useful R magnitudes.R magnitudes for thirty-three objects in the field were obtained from Yee, Green,and Stockman (1986). These objects were used to establish the calibration betweeninstrumental and true R magnitudes. The data for these objects is listed in table 4.5,and figure 5.3 shows the least-squares fit defining the calibration, which has the formmR = (0.992 + 0.013) minatr — (13.267 ± 0.253). The least-squares fit was weighted by theerrors in the instrumental r magnitudes.For interest, the distribution of the R magnitudes of the 184 objects in the sampleis shown in the histogram in figure 5.4. This figure in no way represents a luminosityfunction for the field, because of the large uncertainties in the magnitudes of the faintestobjects, and because no completeness correction has been made.67 -20100,0 ipo^(43^0 CO 0^O0^8 O^00 — I I^I^I^I^I^I^I^I^I^I^I^I^I^I^ I^i^I^7"0.5 1 1.5 2 2.5Error in r magnitudeFigure 5.2: The estimated uncertainty in the simulated r magnitude shows a trend withmedian signal-to-noise ratio.683836za)343218^20^22R (YGS)Figure 5.3: The instrumental r magnitudes were calibrated using R values from Yee,Green, and Stockman (1986). The first figure shows the calibration relation between theinstrumental and the published magnitudes, and the second shows the deviation of thecalibrated magnitudes from the fit.6916^18^20r magnitude22 2430100Figure 5.4: The distribution of the R magnitudes for the whole sample is shown by thesolid-line histogram. The shaded histogram represents the magnitude distribution forthose objects classified by the program as galaxies according to both SEDs.70Chapter 6Accuracy of the TechniqueIt is crucial to understand the accuracy of redshifts and morphological classificationsderived from multifilter spectrophotometry, if the technique is to be useful for otherstudies. A modelling analysis has helped evaluate the dependence of the results on thequality of the data and the type of object, but associating the modelling results with thequantities derived from real data in this study is difficult because there are so few objectsin the sample for which the redshifts and physical identities are known. It has not beenpossible to investigate the accuracy of the morphological classification (type of galaxyor star) due to the lack of data of this sort for objects in the sample. Obtaining suchcomparison data spectroscopically would be quite demanding of telescope time becausemost of the field objects are very faint.6.1 Results from ModellingK. Callaghan, B. Gibson, and P. Hickson investigated the accuracy of the fittingtechnique using simulated observations in forty gaussian filters. The results of their studyare currently in press (Hickson, Gibson, and Callaghan 1993). An interim report can befound in Callaghan, Gibson, and Hickson (1992), and a summary of the study to datein Callaghan (1992). Their goal was to investigate the effects of signal-to-noise ratio,71morphological type, and evolutionary state on the classifications and redshifts derivedthrough template fitting.Their approach was fairly straightforward. To construct an "observed" galaxy spectralenergy distribution, they chose fiducial spectra from the atlas of Rocca-Volmerange andGuiderdoni (1988), defined an object-to-sky intensity ratio and a desired signal-to-noiselevel, then added to the model spectrum, random noise fluctuations of the appropriatesize and a sky spectrum based on data from Turnrose (1974). The corresponding spectralenergy distribution was made by multiplying this model spectrum by the forty filtertransmission curves. Fitting of model templates was done in the same manner as for thereal data presented in this thesis: x2 was calculated for each scaled template SED overa range of redshifts, and over a range of template ages as well, and the minimum x2identified the best fit.For each galaxy type investigated, ten different simulated observations were con-structed at a given redshift and signal-to-noise level (from three to fifty), and the derivedquantities from the fitting were averaged. The outcome of the study was summarized inthe form of the error as a function of signal-to-noise ratio, for varying redshift at constanttype and for varying type at constant redshift. For all types and redshifts, the error inboth redshift and type decreases very clearly with increased signal-to-noise ratio. It isalso observed that errors in type are larger for later-type galaxies than for E/S0s, andthat errors in both type and redshift are reduced at higher redshifts. Callaghan (1992)concludes that redshifts derived from forty-filter spectrophotometry with signal-to-noiseratios of 3.0 or greater can be expected to have an error less than 0.12 for z < 0.5, andan error below 0.04 for objects at higher redshift.The modelling study found that stars were rarely confused with galaxies, despite thesimilarities in their spectra. The best-fit stars to galaxy SEDs were types K and M for72ellipticals, G and K for early spirals, and A to G for late-type galaxies.An important difference between the simulated observations used in the modelling andthe real data in this study is the filter sets used. The set of forty gaussian filters, apartfrom having more points per SED, extends well into the blue, with the bluest filter ata central wavelength of 4000A (log v = 14.875). The model galaxy spectra in figure 4.2demonstrate the importance of the blue end of the spectrum in distinguishing betweenshapes of the SEDs for different galaxy types. Redshifting causes the spectral differencesto become more apparent, but because this study has only two widely separated filtersblueward of 5500A, galaxies must be at redshifts of 0.38 or more for the 4000A break tobe seen with any resolution. The filters used were chosen to make the break obvious atthe cluster redshift (at z=0.602, the break is at 6400 A), in expectation that this wouldallow galaxy velocities within the cluster to be accurately measured; however, it meansthat types and redshifts of low-redshift galaxies are very difficult to determine.A secondary modelling study based on the filter set used to acquire the data in the3C281 field was carried out, and the results are summarized in Callaghan, Gibson, andHickson (1992). Using a signal-to-noise ratio of 5.0 for the simulated SEDs, they find thatredshifts of early-type galaxies can be determined with an accuracy of 2%; for later-typegalaxies, the redshift uncertainty is on order of 20%. The results also show, however, that"at low signal-to-noise ratios, it is possible to assign an anomalously high redshift to anylocal z 0.0 galaxies". Morphological classification is observed to be accurate to withinone type, and again they find no instance where a galaxy is misclassified as a star.6.2 Accuracy of the Derived QuantitiesIt is of interest to examine the accuracy of the technique in distinguishing betweenstar and galaxy SEDs, as well as in deriving the correct redshifts and types. The first73goal may be approached by comparing the classifications assigned by template-fitting withthose derived by Yee according to the shape of the object for the objects in table 4.5.If Yee's shape-dependent classifications are correct, then five out of thirty of the objectsof "known" class (galaxy or star) in the 3C281 field data may have been erroneouslyclassified by the fitting program. (Yee's list contains thirty-three objects, of which oneis the QS0, and two others are galaxies that were too faint for the fitting program toclassify.) Four of Yee's eight stars were classified as galaxies, but only one of twenty-twogalaxies was best-fit by a star template.The five objects have median signal-to-noise ratios of 1.7, 2.1, 2.2, 3.2, and 4.7 (the lastis for the object identified as a star by SED-fitting). Objects with lower signal-to-noiseratios than these are assigned classifications that are the same as Yee's, so these are not"misclassified" because of their faintness. It is likely that some of the objects classifiedas stars by their shape and physical appearance are actually compact galaxies or brightgalactic nuclei which appear pointlike because of their distance. The object classed as agalaxy by its shape but a star by its SED could be a galaxy of a type not represented bythe models.On the whole, classifications assigned by template-fitting to the SED correspond wellwith the classifications independently derived from the shape of each image.It is also desirable to have a means of estimating the accuracy of the redshifts andmorphological classifications produced by template-fitting. The only galaxies in the fieldfor which redshifts are available are the twelve from EGY, listed in table 4.2.The modelling results described in the previous section suggested a relationship be-tween the signal-to-noise ratio of an object and the error in its assigned galaxy redshift.If such a relationship could be observed for this test sample, it could be extrapolated todescribe the accuracy of the redshifts derived for objects at a given signal-to-noise ratio.74It should be kept in mind that EGY selected their galaxies based on magnitude (r < 22.0)and proximity to the quasar, and therefore the test sample may not represent the generalpopulation of objects seen in the field.Using the twelve EGY galaxies, the redshift error is tested to see if it is related toproperties of the object. One might expect, based on the modelling results, that theaccuracy of the derived redshift would be related to the following quantities: the signal-to-noise ratio, since if the spectrum is better defined the model-fitting should be moreaccurate; the galaxy type, since early types have steeper spectra and their SEDs vary morein shape with redshift (e.g. see figure 4.3); and the redshift, since at higher redshifts the4000A break is shifted into the spectral region well sampled by the filter set. Figures 6.1and 6.2 show the variation in redshift error with the three parameters listed above forthe first and second sets of SEDs independently. The dashed lines in these figures are atredshift error of ±0.1, a useful working limit for "acceptable" redshift errors.The first set of SEDs yields seven redshifts within 0.1 of the true value, and an eighththat is close to being within the limit. Two of these objects, 44 and 128, have much worseassigned redshifts for their second SEDs. Both SEDs in each case appear equally noisy(figure 4.4), so it is unclear why one fit is good and the other bad. Object 44 has a fairlyflat spectrum, which could contribute to confusion in the model-fitting, if the noise spikesare being treated as real spectral features in the fitting procedure.Of the four objects with large redshift errors, two are classed as stars according toat least one SED. Object 92, classed once as a star, is assigned suspiciously high best-fitredshifts for the best galaxy fits to both SEDs. The data look no noisier for this objectthan for others which are assigned reasonable redshifts, but the spectrum is fairly flat,which means the noise is more likely to be fit by minor shape features in the models thanit would be in a steeper spectrum. Object 95 also has very high assigned redshifts for the75—0.51 ZZ".1^0 e-o- 000^2^4^6galaxy type0.5—0.5I * I3^4^5median S/N—0.50.4^ 0.5^0.6redshift (egy)Figure 6.1: The y-axes of these plots are the accuracy of the derived redshifts for the firstset of SEDs of the twelve galaxies from EGY. The accuracy is plotted as a function ofgalaxy type, median signal-to-noise ratio, and redshift. The hollow circles are the early-type galaxies, and galaxies of type 2.0 or later are represented by star-shaped points.76__o^*^ * :_^ * -— o^cs- -o *—o- *--- I^ 1^ I_ * *-0.5ii- I^' _-__ *0 *-* o---o- ■-1 i I-->";ona)0.5*i^ ■^ 1o^ * *- 4^ *^76^*^odr _- *^_-- -_^IF-'rina)0.5N 0ICVN —0.50 2 4^6galaxy type3^4^5median S/N0.4^ 0.5^0.6redshift (egy)Figure 6.2: Redshift accuracies for the second set of SEDs of the EGY galaxies, as afunction of galaxy type, median signal-to-noise ratio, and redshift. The hollow-circlepoints are the early-type galaxies, and the star-shaped points are the later types.77best-fit galaxy models, but both SEDs are classed as stars. The SEDs are steeper thanthose for 92, and again are not noisier than other SEDs with low redshift errors. Whythe star spectra fit better is unknown, but it may be that the galaxy is of a type notrepresented in the set of models.The other two objects with large redshift errors are 151 and 86. The SEDs of object151 are so flat as to suggest that a definite redshift cannot be assigned, and noise in themodels causes two different redshifts and galaxy types to be assigned. Object 86 appearsto be well-fit by the two assigned galaxy models, which are at similar redshifts. It ispossible that EGY's measured redshift for this object is in error, as H. Yee (pers.comm.)believes that 85-90% of the published redshifts in EGY were correct.A survey of the two figures does not immediately reveal any of the anticipated cor-relations between redshift error and other properties. However, if object 95 is excluded(since both SEDs are best fit by a star model), then the early-type galaxies (up to type1.9) all have redshift errors below the limit. This may have to do with the fact that theearly-type galaxies are all at redshifts greater than 0.5, meaning that the 4000A break isredshifted into a spectral region well sampled by the narrow filters. The galaxies assignedlater types cover a wider range of redshifts, and no correlation is visible between accuracyand redshift; however, the 4000A break is smaller for these objects.There does not seem to be a significant relationship between median signal-to-noiseratio and redshift accuracy, as was suggested by the modelling study. However, there areno galaxies in the comparison sample with signal-to-noise ratio higher than six, and thesample itself is too small and limited to allow generalizations about the accuracy of themethod to be made.It is possible that the shape of a curve of X2 vs. redshift for a particular object couldallow an estimate of the redshift accuracy to be made; and similarly, a curve of X2 VS.78model type could give an idea of the type accuracy. However, this would be most effectivefor well-defined SEDs. Most of the SEDs in this sample show signs of substantial noisein the photometry, and are not clear enough to yield a smooth curve of x2 with a singlewell-defined minimum; it is more likely that there exist a number of local x2 minima inredshift-type space.With a larger sample of galaxies with known redshifts, and a wider range in signal-to-noise ratio, it might be possible to extrapolate the relationship between signal-to-noiseratio and expected error found in the modelling study, in order to estimate the redshiftuncertainty for galaxies in the 3C281 field. Unfortunately, no such relationship can bedefined using the sample of known-redshift objects available at present. It is only possibleto agree with the well-known observation, discussed at length by Koo (1985) and upheldby the modelling study using the LMT filters, that it is much easier to obtain accurateredshifts based on the shapes of galaxy spectra for early-type galaxies than for later types.79Chapter 7DiscussionStudies using simulated data suggest that accurate redshifts and object classificationscan be derived from fitting templates to multi-narrowband filter spectral energy distrib-utions. The simulations of Callaghan et al. (1992) using the same twenty-four filters thatwere used to produce the real data predict an error of 20% in the redshifts of late-typegalaxies, with much better accuracy for early-type galaxies, based on trials with signal-to-noise ratio of 5.0. It is not appropriate to compare the results for the small sample ofgalaxies of known redshift with the modelling results, because only two galaxies observedhave median signal-to-noise ratio above five, and neither yields a redshift within 0.1 of thepublished value. However, of the five galaxies assigned early types by template-fitting,four have acceptable redshifts, while the fifth is classed as a star and assigned an inaccurateredshift by its best-fit galaxy model. Three of the seven later-type galaxies have redshiftswithin 0.1 of the published value, and another is close. The overall accuracy of the derivedredshifts is fairly good, especially in view of the reasons, suggested in section 6.2, why thefour galaxies with incorrect redshifts may not be accurately identified.The modelling results anticipated that no confusion would arise between galaxy andstar templates, but one EGY galaxy is classified as a star according to both its independentSEDs, and another has a star template as the best fit to one SED. The misclassification80could occur because of noise in the SED, or because the object is of a type not representedin the set of model spectra, such as an emission-line object.Errors in the photometry which contribute to misclassification can arise within thederived uncertainty term, or from possible extinction in the observations made on onenight. A further source of error may arise in the calibration, particularly if a spectralline in the standard star's spectrum lies within one of the filters. A search for systematicerrors was made by comparing EGY galaxy SEDs with the best-fit models yielding correctredshifts. Total intensities for all the objects were scaled accordingly, and best-fit typesand redshifts derived. However, this procedure did not improve the average goodness-of-fit to models, and produced slightly poorer redshifts for the EGY sample, as comparedto results from spectral energy distributions where only the points from the 6200A filterwere adjusted. The latter set was used for the analysis described in the previous chapters.It is stressed that high-quality photometry will yield the best object identifications.The results presented here are the second set to be derived from these data. The firstversion included several extra sources of error in the photometry, and the accuracy wasvisibly poorer than that seen in these results.Although the comparison sample of known-redshift galaxies did not cover a wide rangeof redshifts, it is expected that better sampling at the blue end of the SED would be ofgreat value in helping discriminate between low-redshift objects and high-redshift later-type galaxies.If good-quality, non-comatic images are used, it might be possible to assign an objecta galaxy or star class based on its shape, before trying to fit models to its SED. Thiswould eliminate some of the confusion due to the similarity of model galaxy and starspectra, although comparisons with shape-derived classifications supplied by Yee suggestthat there are few instances where the fitting program cannot distinguish stars from81galaxies at reasonable signal-to-noise ratio (section 4.6). Since galaxy images, and imagesof other interesting non-stellar objects, can be circular as well as extended, all compactobjects would have to be fitted with both star and galaxy models, but the extendedobjects would not have to be fitted with star models. It was not possible to pre-classifyobjects in this way with the data set used, because the obvious coma near the edges ofthe images suggests some degree of image distortion for many objects in the field.Standard broadband filter photometry can be simulated by combining the fluxes fromthe narrowband filters multiplied by appropriate coefficients. The accuracy of the simu-lated photometry is not yet known, although there is quite a large scatter as shown inthe calibration diagram, figure 5.3. A second data set is available, for the field of thequasar PKS 0812+020, and this will serve to investigate the accuracy of the "multifilterbroadband colours" in the near future. With suitable spectral coverage, it should provepossible to simulate other broadband filters, and perhaps to compare the redshifts derivedfrom template-fitting with redshifts from optical multicolours, as done by Koo (1985) andothers. The UBC LMT data will be extremely versatile in this respect, since the fortyfilters to be used overlap well and cover the same spectral range as the V, R, and I filters,and most of the B filter as well. The transmission curves of the LMT filter set and thebroadband filters are shown in figure 7.1.The second data set will be useful not only in testing the accuracy of the broadbandphotometry, but also for further analysis of the accuracy of the method. EGY havepublished redshifts for thirteen objects in the field of PKS 0812+020, which lies at aredshift of 0.403. If all these objects are detected in the CCD photometry of the fieldimage, they will double the size of the sample of galaxies with known redshifts, and mayhelp define a relationship between signal-to-noise ratio and redshift accuracy, which couldnot be done using the comparison sample from the 3C281 field alone.82Figure 7.1: The transmission curves of the UBC Liquid Mirror Telescope filter set.Standard broadband filters are shown as dotted lines: the filter set covers (from low-est to highest wavelength) most of the B filter, and all of the V, R, and I filters. (Thisfigure is provided courtesy of B.K. Gibson.)83Chapter 8ConclusionThe technique of deriving galaxy redshifts by fitting models based on template spec-tra to the spectral energy distributions produced by multi-narrowband-filter photometryappears to be fairly successful, although the results presented here are not entirely con-clusive. The levels of accuracy predicted by the modelling study cannot be confirmed,but the derived redshift values for objects of known redshift give reason for optimism.An enhancement in the number of galaxies with derived redshifts near that of the QS0,where a cluster is expected, further suggests that the derived redshifts are fairly close tothe true values.Though results from simulated spectral energy distributions show a clear relationshipbetween the signal-to-noise ratio of the observations and accuracy of the derived redshift,the real data for objects in the field of the quasar 3C281 showed no such relationship.However, there were only twelve galaxies in the field with known redshifts, and these allhad median signal-to-noise ratios below six, so whether or not a signal-to-noise ratio—redshift-accuracy relation exists for the data is still undetermined.What is clear from the results for the galaxies of known redshift is that the redshiftsderived from template-fitting for galaxies classified as E/SO are more likely to be accuratethan are the redshifts for objects fit as later-type galaxies. This is to be expected, as84the 4000A break is much more pronounced in the spectra of early-type galaxies, and cantherefore be seen more easily when it is redshifted into the spectral domain of the filter set.It is observed that four of five early-type galaxies of known redshift, with signal-to-noiseratio greater than 2.0, have derived redshifts within 0.1 of their spectroscopic redshift(and the fifth is best-fit by a star model), while three of seven later-type galaxies (andnearly one more) have redshifts errors of this magnitude.Thirty-two objects in the observed field, which were independently classified by H. Yeeas galaxies or stars depending on image shape, were used to test the accuracy of typeclassification. Tests with simulated data, using both the forty-point LMT filter set andthe twenty-four filters used for these data, indicated no confusion between galaxies andstars even at low signal-to-noise ratios. In practice, the template-fitting method yieldedthe same classes as the shape criterion for two out of six of Yee's stars and twenty oftwenty-one galaxies. The remaining five objects (two stars and three galaxies accordingto Yee) were too faint to be consistently fit: the best fit to one SED was a star andto the other, a galaxy. It is entirely possible that some of the objects with the shapesof stars could actually be compact galaxies or nuclei, which would explain the galaxyclassifications assigned to the four objects of stellar appearance.The same list of objects, with the QS0 added, was used to calibrate the instrumental rmagnitudes produced by combining narrowband filters. The estimated uncertainty in theR thus produced ranged from 0.04 magnitudes for the brightest objects to 2.5 magnitudesfor a galaxy of magnitude 23.4. With a suitably high level of signal-to-noise ratio data,it should be possible to do broadband photometry with multi-narrowband-filter data.The results of this study, despite the low number of comparison objects of knownredshift, suggest that useful redshifts and type classifications can be obtained from themethod of multifilter spectrophotometry even at signal-to-noise of three or below. The85most immediate need is for a larger sample of objects with known redshifts, so that a morecomplete analysis may be made of the accuracy of quantities derived by the template-fitting procedure. Improved spectral resolution, higher signal-to-noise data, and bettersampling in the blue end of the spectrum as in the LMT filter set, can only improve uponthe results described here.86BibliographyBaum, W.A. 1962, in Problems of Extragalactic Research, IAU Symposium 15,ed. G.C. McVittie (New York: MacMillan) p. 390Bessell, M. 1983, PASP 95, 480Bevington, P.R. 1969, Data Reduction and Error Analysis for the Physical Sciences,(New York: McGraw-Hill)Bruzual, G. 1983, ApJ 273, 105Burstein, D., and Heiles, C. 1978, ApJ 225, 40Butcher, H., and Oemler, A. 1978, ApJ 219, 18Callaghan, K. 1992, Multi-Filter Spectrophotometry Simulations, unpublished reportCallaghan, K., Gibson, B.K., and Hickson, P. 1992, in The Evolution of Galaxiesand Their Environment, eds D. Hollenbach, H. Thronson, and J.M. Shull,NASA Conference Publication 3190, page 357Canada-France-Hawaii Telescope Corporation 1990, Users' ManualGarde J.A., Clayton, G.C., and Mathis, J.S. 1989, ApJ 345, 245Couch, W.J., Ellis, R.S., Godwin, J., and Carter, D. 1983, MNRAS 205, 1287Couch, W.J., and Newall, E.B. 1980, PASP 92, 746de Vaucouleurs, G., de Vaucouleurs, A., and Corwin, H.G. Jr. 1976, Second ReferenceCatalogue of Bright Galaxies (Austin, Texas: University of Texas Press)Ellingson, E., Green, R.F., and Yee, H.K.C. 1991, ApJ 378, 476Ellis, R.S., Couch, W.J., MacLaren, I., and Koo, D.C. 1985, MNRAS 217, 23987Gibson, B.K, and Hickson, P. 1991, in The Space Distribution of Quasars,ed. D. Crampton (ASP: San Francisco)Guiderdoni, B., and Rocca-Volmerange, B. 1987, ASzA 186, 1Gunn, J.E., and Stryker, L.L. 1983, ApJS 52, 121Hickson, P., Gibson, B.K., and Callaghan, K. 1993, MNRAS, submittedHickson, P., Gibson, B.K., and Hogg, D.W. 1993, PASP, in pressHickson, P., Richardson, E.H., and Grundmann, W.A. 1992Koo, D.C. 1981, ApJ 251, L75Koo, D.C. 1985, AJ 90, 418Kron, R. 1980, ApJS 43, 305Lang, K. R. 1986, Astrophysical Formulae (New York: Springer-Verlag)Loh, E.D., and Spillar, E.J. 1986, ApJ 303, 154MacLaren, I., Ellis, R.S., and Couch, W.J. 1988, MNRAS 230, 249Mathieu, R.D., and Spinrad, H. 1981, ApJ 251, 485Mihalas, D. and Binney, J. 1981, Galactic Astronomy - Structure and Kinematics,(New York: W.H. Freeman 8.6 Co.)Oke, J.B. 1966, ApJ 145, 668Oke, J.B. 1969, P.A.S.P. 81, 11Oke, J.B., and Gunn, J.E. 1983, ApJ 266, 713Oke, J.B., Neugebauer, G., and Becklin, E.E. 1970, ApJ 159, 341-355Oke, J.B., and Schild, R.E. 1970, ApJ 161, 1015Ratnatunga, K., and Bahcall, J. 1985, ApJS 59, 63Rocca-Volmerange, B., and Guiderdoni, B. 1988, A.SLA.Supp. 75, 93Sandage, A. 1988, ARAA 26, 56188Sandage, A., and Tammann, G.A. 1987, A Revised Shapley-Ames Catalogof Bright Galaxies, Carnegie Institute of Washington Publication 635Spinrad, H. 1980, in Objects of High Redshift, IAU Symposium No. 92, ed.G. Abell and P.J.E. Peebles, (Dordrecht: Reidel), p.39Turnrose, B.E. 1974, PASP 86, 545Valdes, F. 1982, Faint Object Classification and Analysis System, Kitt PeakNational Observatory, TucsonWampler, E.J. 1966, ApJ 144, 921Wampler, E.J. 1967, ApJ 147, 1Yee, H.K.C. 1991, P.A.S.P. 103, 396Yee, H.K.C., Green, R.F., and Stockman, H.S. 1986, ApJS 62, 681

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.831.1-0086092/manifest

Comment

Related Items