Optical Astronomy from Antarctica

PLEAS£ TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Project Report Sheet Surname or Fam ily name: Dempsey Fi rst name : Jessica Other name/s: Tui Abbreviation for degree as given in the University ca len dar : PhD

School: Physics Faculty: Science

Title : The view from the ice at the bottom of the world: Optical Astronomy from Antarctica

Abstract 350 words maximum: (PLEAS£ TYPE)

The high Antarctic plateau may offer the best site on earth for optical astronomy. This thesis work includes the construction of an infrared cloud-observing instrument, COBBER, which utilises a thermopile detector optimised at lOf.Lm. COBBER was installed at Dome C in January 2003. In 71 observing days, only four days of cloud were measured.

A detailed study of the effect of auroral emission on optical observations is conducted. Analysis of auroral measurements at South Pole show that in an average winter season, the B band sky brightness is below 21.9 mag/arcsec2 for 50% of the observing time. In V band, the median sky brightness contribution is 20.8 mag/arcsec2 in an average winter. Calculations are used to show that at Dome C, the contribution to sky background in Band Vis up to 3.1 magnitudes less than at South Pole.

The first optical stellar spectra observed from the high Antarctic plateau were taken at South Pole station with the Antarctic Fibre Optic Spectrometer (AFOS). The AFOS was installed on a dual telescope alt-az mount in January of 2003. A thorough instrument analysis revealed tower sinkage and telescope flexure problems that were overcome with more frequent pointing runs.

Two years of AFOS observations are described, including selection of sources, design of observing scripts and the creation of a data reduction method for the data. AFOS data was analysed to determine if the H20 atmospheric absorption bands in the spectra could be used to detect daily variations in the precipitable water vapour (PWV) content of the atmosphere. The PWV values obtained by comparing the AFOS data with synthetic spectra created with MODTRAN were compared to similar measurements taken at South Pole with a 350flm radiometer and balloon-borne radiosondes. The PWV values from these instruments showed good agreement with the AFOS results.

Observations of the moon were used to study the earthshine spectrum to detect variations in the earth albedo over a 24-hour period. The earth albedo was successfully detected though poor weather conditions prevented study of any long-term trends in the data.

Declaration relating to disposition of project report/thesis

I am fully aware of the policy of the University relat ing to the retention and use of higher degree project reports and theses, namely that the University retains the copies subm itted for exam ination and is free to all ow the m to be consulted or borrowed . S u ~ect to the provisions of the Copyright Act 1968. the University may issue a project report or thesis in whole or in part, in photostat or microfilm or other copying medium .

I also authorise the publication by University Microfilms of a 350 word abstract in Dissertatian Abstracts International (applicable to doctorates only) .

...... ' "'" .!2/!.1/9!1...... Date

The University recogn ises that there may be exceptional circumsta,,._.,, , "'-tu " ,, ~ ,.,,"''""u'', uo -.ul-'~ ,,~ uo onditions on use . Requests for restriction for a period of up to 2 years must be made in w riting to the Registrar. Requests for a longer period of restr iction may be considered in exceptional circumstances if accompanied by a letter of support fro m the Supervisor or Head of School. Such requests must be submitted with the thesis/project report .

FOR OFFICE US£ ONLY Date of completion of requirements for Award:

Lf( ( /(JS Reaistrar and Deoutv Pr incioal

THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS

The view from the ice at the bottom of the world: Optical Astronomy from Antarctica

Jessica Tui Dempsey

A thesis submitted in satisfaction of the requirements for the degree of Doctor of Philosophy in the Faculty of Science.

THE UNIVERSITY OF NEW SOUTH WALES_

I~ SYDNEY· AUSTRALIA UNSW 1 1 FEB 2005 LIBRARY

For Rodney (1968-2000) Not Without Peril. Statement of Originality

I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, nor material which to a substantial extent has been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis.

I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.

(Signed). £ Abstract

The high Antarctic plateau has shown exceptional potential for infrared and sub millimeter astronomy. Though ground level seeing at South Pole is comparatively poor, new measurements at Dome C have shown the best seeing of any site in the world. As the science case for the optical potential of Dome C strengthens, it is now necessary to quantify the properties of the Antarctic atmosphere for optical astron omy, and to demonstrate that astronomical telescopes can be built, maintained and successfully operated in Antarctic conditions.

This thesis work includes the construction of an infrared cloud-observing instru ment, COBBER (Cloud OBserver), which directs radiation from the sky through a hemispherical ZnSe lens onto a thermopile detector optimised at 10.um. COBBER was installed at Dome C in January of 2003. In 71 observing days, only four days of cloud were recorded.

A detailed study of the effect of auroral emission on optical observations has been conducted. Analysis of auroral measurements at South Pole show that in an aver age winter season, the B band sky brightness is below 21.9 magnitudes per square arcsecond for 50% of the observing time. In V band, the median sky brightness contribution is 20.8 magnitudes per square arcse,eond in an average winter. Calcu lations are used to show that at Dome C, the contribution to sky background in B and Vis up to 3.1 magnitudes less than at South Pole. The use of notch filters to reduce the contribution of the strongest emission lines and bands is also calculated. The science that optical astronomy could potentially achieve with an ELT at Dome C is discussed with reference to the effect auroral emission would have on particular astronomical observations.

January of 2003. A thorough instrument study revealed tower sinkage and telescope flexure problems that were overcome with more frequent pointing runs. An analysis of wavelength-dependent attenuation observed in some of the stellar spectra is also presented.

Two years of AFOS observations are described, including the selection of sources, the design of observing scripts and the creation of a data reduction method for the stellar spectra collected with the AFOS. AFOS data were analysed to determine if the H20 atmospheric absorption bands in the spectra could be used to detect daily variations in the precipitable water vapour (PWV) content of the atmosphere. The AFOS data was compared with spectra created using the MODTRAN atmospheric modelling program. The resulting PWV values from these daily fits were compared with measurements obtained by a 350J.tm radiometer, and daily balloon-borne ra diosondes, at South Pole. The PWV values measured with both instruments showed excellent agreement with the AFOS results, with an accuracy of 10%.

In late 2003, the AFOS collected several 24 hour-long observations of the dark side of the moon. These observations were used to study the earthshine spectrum in an effort to detect variations in the earth albedo over a 24 hour period. These obser vations are of interest in the search for life on extra-solar planets. The earth albedo was successfully detected in the earthshine measurements. Poor weather conditions at South Pole affected a large percentage of the observations and prevented study of any long-term trends in the earth albedo. Contents

Abstract ...... 1 Acknowledgments . . viii

1 Introduction 1 1-1 The Antarctic Continent 2 1-2 Site Characterisation . . 8 1-2.1 Automated Weather Stations 9 1-2.2 AASTO ...... 10 1-2.3 Optical Site Testing . 14 1-3 DomeC ...... 18 1-3.1 Cloud cover 18 1-3.2 AASTINO. 20 1-4 AFOS ..... 24 1-4.1 History. 26 1-5 Thesis Goals . . 27

2 Auroral contribution to optical sky brightness 29 2-1 Introduction ...... 29 2-1.1 Auroral line intensities and spatial extent . 31 2-1.2 Previous Studies 34 2-2 AFOS Measurements . . 36 2-2.1 2000 Lunar Eclipse, South Pole 36 2-3 Auroral Data from South Pole . . . . . 37 2-4 Auroral Intensity at Dome C and Dome A 40 Contents iv

2-5 Sky brightness in standard photometric bands 43 2-5.1 B band. 44 2-5.2 Vband 46 2-5.3 R band. 48 2-6 Discussion . 49 2-7 Conclusions 53

3 Cloud cover at Dome C 60 3-1 COBBER ...... 60 3-1.1 COBBER design 61 3-1.2 Testing. 63 3-1.3 Results. 65

4 Instrumentation 70 4-1 AFOS .... 70 4-1.1 Telescope 71 4-1.2 Fibre Optics . 74 4-1.3 Spectrograph and CCD . 77 4-1.4 Generic mount .. 80 4-2 Testing and Modification . 81 4-2.1 Tower tilt and sinking 82 4-2.2 Telescope flexure ... 83 4-2.3 Pointing improvements . 85 4-2.4 Fibre Mapping ... 86 4-2.5 Spectral attenuation 90 4-2.6 Polarisation Effects 93 4-3 Summary ...... 105

5 Observing and Data Reduction 106 5-1 Observational Strategy .. . 106 5-1.1 Manual Observing . 107 5-1.2 Automating AFOS Observations . . 113 Contents v

5-2 Data Reduction ...... 116 5-2.1 CCD procedures . 116 5-2.2 Spectral Extraction and Calibration . . 120 5-2.3 Flat-fielding . . . 128 5-2.4 The UV Cutoff . 133 5-2.5 Second order leakage . 137 5-2.6 Normalisation . . 141 5-3 Summary ...... 145

6 Atmospheric Modelling 146 6-1 The Antarctic Atmosphere . . 146 6-1.1 Radiosonde Data . . . 147 6-1.2 Millimeter and Submillimeter PWV Measurements . 150 6-2 MODTRAN ...... 152 6-2.1 Input Parameters . 153 6-2.2 Constructing the Model . 154 6-3 Fitting the data ...... 156 6-3.1 Determining the Resolution . 157 6-3.2 Modelling the temperature . . 159 6-3.3 Water Vapour Bands . . . . . 165 6-3.4 Comparison with 350 micron data . . 169 6-4 Conclusions ...... 170

7 Earth Albedo Measurements 173 7-1 Earth's signature . . . . . 173 7-1.1 Previous Studies . 17 4 7-1.2 Method ...... 177 7-2 Observations and Reduction . 178 7-3 Results ...... 181 7-3.1 Scattering . 182 7-3.2 Earth Albedo Features . 185 Contents vi

7-4 Conclusions . 190

8 Conclusions 192 8-1 Future Work . . 195

Appendices 197

A Auroral Intensity Calculations 197 A-1 Rayleighs to Magnitudes .. . 197

B Observation Logs and Scripts 199 B-1 G-mount Command Logs and Scripts . 199 B-1.1 Standard G-mount command log . 199 B-1.2 Gmount Pointing and Co-ordinate files . 202 B-2 AFOS Command logs ...... 203 B-3 Automated Observing Scripts . 206 B-3.1 Instrument control script . . 206 c Reduction Routines 217 C-1 IRAF ...... 217 C-1.1 CCD Reduction . . 217 C-1.2 Aperture extraction and Flat fielding . 220

D MODTRAN input and output files 223 D-1 Input File . 223 D-2 Output file . . 225 D-2.1 Tape 7. . 225 D-2.2 output.plt . 226

E Fitting Programs 227 E-1 chisquare.£90 . . . 227

F Earth Albedo IRAF scripts 232 F-1 divvy.cl ...... 232 Contents vii

F-2 ratio.cl ...... 234

G Published Works 237 G-1 Published Papers . 237

References 276 Acknowledgments

Impossible tasks .. .it is certain that I would not have been been able to complete this without the support, guidance, good humour and unfailingly patient advice of my supervisor, Professor John Storey. Even at my most stubborn, lazy or mis guided, John had time for me and faith in me when I would have said there was little reason! Thank you, John. There are many members of the UNSW Antarctic group who have helped me in this endeavour: thanks to Michael Burton, Michael Ashley, Jon Everett, Andre Phillips, Paolo Calisse and Tony Travouillon for in stalling (and fixing!) my gadgets and telescopes in both Australia and Antarctica, and for support with instruments, programs, papers, and in the writing of this thesis.

Thanks to Mark J arnyk, Gary Hovey and Ralph Sutherland at the RSAA, Mt Stromlo, for all their technical support with the G-mount and ADIMM systems, without which I would have been lost. I extend my sincere thanks to D. Detrick and T.S. Rosenberg from the University of Maryland for providing the South Pole photometer data used in the auroral calculations in Chapter 2. I am also grateful to Stephen Mende and Gary Burns for very informative discussions on auroral physics.

I am very grateful to the UNSW School of Physics for my scholarship to under take this work, and to the Australian Research Council, for financial support of our group, and thus my work. In my Antarctic adventures: I am extremely grateful to the U.S. National Science Foundation, the Australian, United States, Italian and French Antarctic programs for logistical support of my instruments, and myself, while on the Ice. At South Pole, I have had the pleasure of meeting and working with some wonderful people, and in particular I thank Charlie Kaminski, Chris Mar tin and the Carpenters for their help and company, and Dana Hrubes for taking the amazing photo of my telescope, shown on the second page. Acknowledgments ix

Particular thanks must go to Jon Lawrence, who has been subjected to a de manding regime of stupid questions and not a few bouts of hysterics. My thanks for the good dose of perspective, always when I was in need of some! To Melinda: for dealing with EVERYTHING and never without a smile, thank you. You are one-in a-billion, and you must be looking forward to an office without the Soap Opera! To Lou: for the introduction to the Joys of Fortran, and helping me at an absolutely critical moment, thanks darlin. For every laksa and hug and for understanding be cause you have been there, thanks. For proof reading of parts of this work, I also thank Matthew and Wilfred: your advice was much appreciated.

To all my fellow students, particularly the lunch gang, thanks for fun and silli ness, which never goes astray. I thank my long-suffering fl.atmates for their support. To Chris, for Conversations, cheers mate. To Bub for his incredible, and unending support and love: Fuzzy, you are amazing, thank you. To mum, again and always, thank you for being the rock that never doubts. To my friends and family, for their love and support for all of my crazy adventures, thank you and much love. Chapter 1

Introduction

One of the most well-known quotes in history was uttered by the astronaut, Neil Armstrong as he stepped effortlessly onto the lunar surface: "One small step for man, one giant leap for mankind." Arguably, the most well-known quote about Antarctica was written by the frostbitten hand of the English adventurer, Robert Falcon Scott when he was narrowly beaten by Roald Amundsen to the South Pole, and only weeks before he perished slowly in the Antarctic wilderness: "Great God, this is an awful place." It is therefore little wonder that space is often depicted as a glorious and challenging frontier, while Antarctica is imprinted in so many minds as an unforgiving wasteland of winds and deadly blizzards.

To the true adventurers, scientists and explorers, it is quite often the case that "impossible" tasks hold a far greater attraction than merely "difficult" ones. This is perhaps a small part of the reason that the budget for space-based astronomical projects is orders of magnitude greater than for those in Antarctica. An astronomer may require little encouragement to agree that the scientific benefits of a space-based telescope outweigh the risks, cost and short lifetime of most of these projects. Sug gest to the same astronomer that a telescope in Antarctica offers excellent scientific opportunity at a fraction of this cost and risk, and the response is often much less enthusiastic. 1-1. The Antarctic Continent 2

Atmospheric and astronomical investigations on the high Antarctic plateau have been underway for the past twenty-five years. To dispel the widespread belief that Antarctica is the "land of the eternal blizzard", as coined by Mawson, potential sites for astronomy, South Pole and Dome C, are now two of the most extensively characterised astronomical sites in the world.

This chapter will outline the key features of climate and atmosphere above the high Antarctic plateau that make it uniquely suitable for a range of astronomical projects. A brief description of the chief site-testing programs on the Antarctic plateau will be given, as well as a discussion of the key results.

The remainder of the chapter will address two questions: What makes a good optical observing site? and: How do we test the optical astronomical potential of the Antarctic plateau? Some of the key results from previous experiments will be discussed. This thesis investigates some previously unstudied properties of the Antarctic atmosphere and the resulting implications for optical astronomy. The ori gins and aims of the work presented in this dissertation will be outlined.

1-1 The Antarctic Continent

The Antarctic continent is vast. Including both the permanently ice-covered land and oceans, the continent is nearly twice the area of Australia. The continent is a giant both in breadth and towering height. The average elevation of the Antarctic surface is 2300 m (the continent with the next highest average elevation is Asia at 800 m (Schwerdtfeger 1984)). A single span of ice the size of Australia has an elevation of 2300 m at its lowest edges and rises, at the deceptive slope of less than a degree, to heights of 3250 m (Dome C) and 4100 m (Dome A) at its highest 1-1. The Antarctic Continent 3 points. Thi region is the high Antarctic plateau, a vast desert atop the polar icecap that extend from the Transantarctic mountains in west Antarctica to the eastern Antarctic coastline.

A contour map of the Antarctic topography can be seen in Figure 1.1, indicating the po itions of the geographic South Pole (site of the United States Admundsen Scott station) Dome C, the third-highest local maximum on the Antarctic plateau and the site of the joint French-Italian Concordia Station, the Russian base Vo tok at 3400m, and Dome A the highest point on the plateau at approximately 4100m. The elevation of Dome A has been derived from satellite measurements and is ap proximate as no human being has yet traversed the centre of the Antarctic plateau to reach the top of this dome.

Figure 1.1. Contour map of the Antarctic continent with the positions of the geographic South Pole [2835m] (Admundsen-Scott base), Dome C [3250m] (French-Italian Concordia. station), Vo tok [3400m] (Ru ian station) and Dome A [ 4100m] (currently unexplored). Map courte y of U.S. Geophysical Survey webpa.ge http:/ /TerraWeb.wr.usgs.gov (USGS 2004). 1-1. The Antarctic Continent 4

The massive, towering icesheet that covers such a large portion of the Antarctic landmass plays an important role in the climate and atmospheric conditions above the Antarctic surface. A selection of key meteorological statistics for South Pole and Dome C are compared to well-defined astronomical sites such as Mauna Kea (Hawaii) and Atacama (Chile) in Table 1.1. The data are derived primarily from the work of Aristidi et al. (2004), though some comparative statistics from Hawaii and Chile are from site testing results of Roddier et al. (1990) and Giovanelli et al. (2001). Table 1.1. Mean annual meteorological statistics of South Pole and Dome C taken from Automated Weather Station data (Aristidi et al. 2004), compared with those measured at two temperate astronomical sites: Mauna Kea, Hawaii (Roddier et al. 1990)*, and Atacama, Chile (Giovanelli et al. 2001)**.

Quantity South Pole Dome C Mauna Kea Atacama

Elevation(m) 2839 3250 4200 5000 Temperature(degC) -53 -53 6.5* -2.6** Average pressure(hPa) 682 644 650 550 Median wind speed(m/s) 5.5 2.9 4.5 6.0

At a mid-latitude site, seasonal variations in temperature are gradual, decreas ing over six months to a minimum at mid-winter and then slowly increasing at the same rate to the summer peak. The high Antarctic plateau instead experiences a dramatic temperature drop at the onset of winter, whereupon the temperature then remains relatively constant until a rapid increase in ground temperature as the sun rises again (Turner 2003). It is worth noting that South Pole and Dome c· show the same mean annual temperatures though in other meteorological aspects, such as wind speed, they show marked differences (Aristidi et al. 2004). The dramatic variation in temperature occurs as a result of rapid heat loss from the snow-covered plateau as the sun sets, and the isolation of the high plateau from warming, mar itime airmasses over the ocean. This protection from maritime winds, and a lack of surface topography results in a lower tropopause and a troposphere which is strongly stratified and very stable (Valenziano & dall'Oglio 1999; Turner 2003). 1-1. The Antarctic Continent 5

The stability of the atmosphere and the intense infrared emission from the cold surface produce a strong, low-altitude temperature inversion layer. This inversion or 'boundary' layer is a few hundred metres in height above South Pole (Marks et al. 1996), though at the highest areas of the plateau, this inversion layer is much closer to the ground, approximately fifty to a hundred metres above the surface (Aristidi et al. 2004). This inversion layer acts like a 'lid', trapping circulating air beneath it. The strong temperature and wind speed gradient in this layer results in the formation of optical turbulence (Aristidi et al. 2004), which degrades the quality of astronomical observations. This will be discussed in detail in the next section.

Table 1.1 shows that Dome C has extremely low wind speeds, not only when compared with other astronomical sites, but also when compared to the Antarctic coast. Mawson Station, on the eastern coastline, is famous for winds of up to 300km/hr (Schwerdtfeger 1984). Winds are generated by air moving from a region of high pressure to one of lower pressure. In the northern hemisphere, cold air moves down to lower latitudes from the North Pole where it meets warmer equatorial air masses. Coriolis forces produce a stream of high-velocity air at an altitude of approximately 10km with speeds of up to 300km/hr. This is known as the polar jet stream (Schwerdtfeger 1984).

A sub-tropical jet stream of somewhat lower intensity dominates the upper atmo sphere at latitudes closer to the equator. A snapshot of the jet stream wind speeds in the southern hemisphere is shown in Figure 1.2. These jet streams play a dominant role in the climate of most mid-latitude sites. The Antarctic plateau, as can be seen from Figure 1.2, is almost entirely unaffected by these high-altitude jet stream winds.

Air in contact with the ground at the highest domes of the Antarctic plateau (the high ridge seen in Figure 1.1) is extremely cold and dense. If the ground is sloping, the air close to the ground at the highest elevation on the plateau is colder 1-1. The Antarctic Continent 6

Figure 1.2. Map of high- peed upper atmospheric winds above Antarctica. Note how the jetstream does not extend down to the latitudes of the Antarctic high plateau. Map courtesy of the California Regional Weather Service Web ite (2004), http://squall.sfsu.edu/crws/jetstream.html. than air at the same level but at orne horizontal di tance. The result is downslope gravitational flow of the colder denser air beneath the warmer, lighter air. The airflow generated at ground level by this process is known as a 'katabatic wind'.

Thi colder, den er air is driven radially down from the highest dome of the vast central Antarctic icesheet towards the coast, traveling downslope for up to 2000 km without any obstruction or impediment. As a result the katabatic winds can reach extremely high speeds, channeled by steep sloping ground and deep valleys near the coast. Station such as Mawson and Dumont D urville have recorded katabatic wind speeds of over 300km/hr.

A diagram howing how the elevation and contours of the Antarctic icecap affect the directions of the katabatic winds i plotted in Figure 1.3. These winds are char- 1-1. The Antarctic Continent 7

Figure 1.3. Diagram showing how the elevation and contours of the Antarc tic icecap a££ -t the uirectious of the katabatic wiuds. Tll origiu of the wiuds can clearly be seen as the highest arc of the plateau extending from Dome A across to Dome C. Map courte y of the British Antarctic Survey webpage (2004), http://vvv.antarctica.ac.uk/About_Antarctica/Weather/. acterised by extremely constant directionality a trong dependence on the trength of the inversion layer (high urface wind when there is a large temperature increase from the ground to the top of the boundary layer) and large increase in wind speed when the slope of the terrain is pronounced ( chwerdtfeger 1984). These defining properties mean that at the genesis of these inver ion layer winds on the high, fiat plain of the Antarctic plateau, the very low lope of the ice and a weaker inversion layer results in con istently low average wind speeds with a mean of 2.9 m/s at Dome C.

The cold till air above the high Antarctic plateau cannot contain a large amount of precipitable water. The polar ice sheet is the driest desert on earth, with an ex tremely de iccated atmo phere and almo t no precipitation. Snow accumulates on the plateau by gradual fall out of 'diamond du t' or 'clear sky precipitation' at the highest domes. At lower elevations, now accumulation is higher as a re ult of blowing snow driven by the katabatic winds de cending from the higher region of 1-2. Site Characterisation 8 the plateau. At South Pole, this results in a snow accumulation rate of 84.5 'water equivalent' mm per year (Mosley-Thompson et al. 1999; van der Veen et al. 1999). At Dome C the snow accumulation rate is much lower, with an average of 36 mm per year (Petit et al. 1982).

The high Antarctic plateau, sheltered from the weather patterns that affect all more temperate sites, is extremely high, dry, cold and still. The highest domes of the plateau do not experience the violent blizzards generated by katabatic winds that are endured on the coastal regions of the continent. Thus, these conditions prompt the proposal that the high Antarctic plateau would offer a clear, dark, still window to the universe during the long, cold austral winter.

1-2 Site Characterisation

Some of the key characteristics of an excellent astronomical observing site are:

• Clear skies

• Good seeing

• High elevation

• Dark skies (low light pollution)

• Low precipitation

• Low wind speeds

• Stable weather conditions

• Clean air

To fully characterise a site, measurements of these properties must be taken year round, preferably over a period longer than a year, and address all of the aspects of a site that could prove detrimental to astronomical observations. In addition to the 1-2. Site Characterisation 9 atmospheric and climatological properties of the site, logistical requirements must also be taken into account. The site must be accessible, able to support personnel and provide power. High altitude, very low temperatures and extremely dry air can be physically demanding, and so conditions at a potential site cannot be so extreme as to prevent people from working.

Meteorological measurements at South Pole and from automated weather stations and satellites have been logged since 1956 (Schwerdtfeger 1984; Turner 2003), and provide the foundation for experiments which characterise more specific astronom ical properties of the sites. With this aim, numerous site testing experiments have been conducted at South Pole and Dome C by several groups (Storey et al. 1996; Burton 1996; Valenziano & dall'Oglio 1999; Chamberlin 2001; Stark et al. 2001), us ing a large suite of instruments, probing the astronomical potential of the Antarctic plateau for observing in the infrared, submillimetre and optical wavebands.

1-2.1 Automated Weather Stations

Automated weather stations (AWS) like the one in Figure 1.4 are installed at nearly all of the stations on the Antarctic coastline, and at a number of strategic and sig nificant sites in the interior. They provide year-round statistics on the temperature, pressure, wind speed and direction of the chosen site, powered by a battery pack buried in the snow and taking a logged measurement every 3 or 10 minutes. The stations require no maintenance, making them a valuable tool in characterising areas of the Antarctic plateau which are unattended and impossible to access during the austral winter.

Automated weather station programs are operated by the Australian Antarctic Division, and the Antarctic Meteorological Research Centre at the University of Wis consin, who provide a database of their AWS data (Australian Antarctic Division 1-2. Site Characteri ation 11

tructed at South Pole station in 19 4. This was followed in the early 1990's by VIPER, a co mic microwave background experiment and AST /RO the Antarctic ubmillimetre Tele cope and Remote Observatory (Stark et al. 2001). A detailed history of astronomy from Antarctica is written by Indermuehle et al. (2003).

AASTO with NISM and MISM F igure 1.5. The AASTO installed at South Pole in January, 2000. Photo courte y of Michael Ashley UNSW.

A the scale of astronomical projects on the Antarctic plateau increased the need for a thorough evaluation of the astronomical properties of South Pole, and other potential sites on the Antarctic plateau was obvious. To fully characterise the astronomical potential of sites on the Antarctic plateau, a dedicated site testing project was instigated at South Pole in a joint venture by the University of New South Wales and the Au tralian ational Univer ity.

The AASTO - Automated Astrophysical Site Testing Observatory - ( een in Figure 1.5) is a modified Automated Geophysical Ob ervatory {AGO). A set of six AGO s small insulated portable laboratories were constructed by the S National Science Foundation (Doolittle & Mende 1995) to measure geomagnetic properties of the high plateau (AGO magnetic impulse results are presented, for example, in 1-2. Site Characterisation 12

Sato et al. (1999)). The AGO dimensions were calculated to fit exactly in a Hercules LC130 plane. The LC130 was used to transport the AGOsto several remote sites on the Antarctic plateau which, after a week of set up in the austral summer, would then be left unattended to run autonomously without intervention for a period of 12 months.

The AASTO was powered by a propane-fuelled thermoelectric generator with an output of 2.5kW of heat and 50 watts of electrical energy. The instruments therefore had a power budget of 7 watts each, as well as needing to run autonomously without maintenance, and require no calibration or cryogen refills. The aim of the AASTO project was to characterise the astronomical properties of the Antarctic atmosphere across a range of wavebands from the optical to submillimetre. Three instruments installed and operated on the AASTO between 1997 and 2003 were designed to characterise the excellent potential for infrared and submillimetre astronomy:

• The Near Infrared Sky Monitor (NISM): The NISM (operated in the win ters of 1998/1999) measured the near-infrared sky brightness with a Stirling cycle cooled InSb detector. The cooled NISM filter has a central wavelength of 2.379J.tm and a bandwidth of 0.226J.tm. The instrument looked out at the sky through a sapphire window. A reflecting chopper wheel alternated the NISM observations between two beams that were angularly separated by 45°. The de tector measured the difference in flux between the two beams. Results showed an average wintertime sky brightness of 210±80j.tJy arcsec-2 (Lawrence et al. 2002), which is comparable to other instruments at South Pole with similar bandwidth filters. These results are between 10 and 40 times lower than the

2 Kdark spectral brightness at Mauna Kea of 1000-2000j.tJy arcsec- •

• The Mid Infrared Sky Monitor (MISM): The MISM (operated for the full winter of 1998) is similar in design to the NISM, but with a Stirling-cycle cooled 1-2. Site Characterisation 13

HgCdTe detector. An ambient temperature filter wheel carries a set of filters that provide a spectral coverage from 4 to l4J.Lm. A chopper frequency of 1kHz is used to avoid 1/f noise. Results from the MISM showed a typical flux of 20Jy in the 8.78 to 9.09 J.lm window, an order of magnitude lower than typical values at Mauna Kea and other temperate sites (Chamberlain et al. 2000).

• Subrnillirnetre Tipper (SUMMIT): This instrument (operated at South Pole during winters of 2001/2002) monitors the atmospheric transmission and effective temperature of the atmosphere at 350J.Lm. One of three tippers con structed by the US National Radio Astronomy Observatory (NRAO), the tipper utilises a fixed bandpass metal-mesh filter and room temperature pyroelectric detector. The instrument observes a 6 degree field of view on the sky, observing a range of zenith angles from zenith to horizon by means of an off-axis parabolic mirror driven by a stepper motor. A second tipper of the same design was in stalled in 1997 next to the AST /RO telescope at South Pole, allowing direct calibration between the instruments (and two others at Mauna Kea and Chaj nantor, Chile). Measurements taken by SUMMIT at South Pole (Calisse et al. 2003) confirm those taken by the existing South Pole tipper (Peterson et al. 2003). The results show an average sky opacity (r) at 350J.Lm of 1.20, (Mauna

Kea r = 1.88 and Chajnantor r = 1.39) and excellent stability in comparison to mid-latitude sites.

These instruments confirmed predictions (Townes & Melnick 1990) that the cold, dry, stable Antarctic atmosphere offered good and in some cases exceptional observ ing conditions in the infrared and submillimetre wavebands, and offered improve ments on both existing and other potential ground-based sites at lower latitudes. 1-2. Site Characterisation 14

1-2.3 Optical Site Testing

The primary measure of the quality of an optical astronomical site is quantified by the "seeing". As light from a star travels through the earth's atmosphere, the wavefront passes through turbulent layers of differing temperatures, and therefore different refractive index. These turbulent cells distort the incoming wavefront, causing the image of the star to wobble and spread out. This process also results in the "twinkling" of a star observed from the ground, which is called scintillation.

The seeing is quantified as the angular size of the object's point spread function, E, in arcseconds as given by (Fried 1966):

3 5 5 EFWHM = 5.25..\ -l/ (laoo C't(h)dh) / (1.1) where ..\ is the wavelength of the light and C'fv is the refractive index structure con stant as a function of height: a derivation of the refractive index structure constant can be seen in Fried (1966). Convection in the atmosphere causes strong, continuous temperature gradients, and where layers of different wind speed and direction meet, shear between the layers causes additional turbulence.

At mid-latitude sites, the strongest contribution to seeing comes from the ground layer and tropopause layer (Turner 2003; Fried 1966). Sites such as Mauna Kea, noted for excellent optical seeing, quote average seeing values of 0.74 arcseconds, and a portion of nights during the year with 0.25 arcsecond seeing (Roddier et al. 1990).

As discussed in the previous section, this high altitude turbulence is virtually absent from the upper atmosphere above the Antarctic plateau. This stable up per atmosphere and low surface wind speeds instead confine the turbulence in the atmosphere to the region between the plateau surface and top of the temperature in version layer which is between 100 and 250m in altitude at South Pole (Marks et al. 1996). 1-2. Site Characterisation 15

At sites further from the summits of the plateau, the height of this boundary layer increases to an average of between 300 and 500m (Schwerdtfeger 1984). Above this boundary layer, high-speed turbulent layers are virtually absent and thus the "free atmosphere" seeing was predicted to be exceptional (Gillingham 1993).

A number of experiments have sought to quantify the ground level seeing at South Pole, beginning with the work of Marks et al. between 1994 and 1996. The first experiment in 1994 quantified the seeing contribution of the lowest 27 metres of the atmosphere. Between April and August of 1994, measurements of the temperature structure constant Cf,, from which the refractive index structure constant can be derived, were recorded with microthermal sensors positioned on a 27 metre high mast constructed at South Pole station. The measurements over the austral winter showed strong optical turbulence in the lowest 27 metres with an average seeing contribution of 0.67 arcseconds from this layer (Marks et al. 1996).

In the season of 1995, the group used balloon-borne microthermal sensors to ob tain integrated values of the seeing over the entire atmosphere, and to observe the altitude profile of the optical turbulence. The results confirmed the prediction that the primary contributor to ground level seeing was confined below the temperature inversion layer, which was approximately 220m in height on average. The mean integrated seeing was 1.86±0.02 arcseconds. The free atmosphere seeing (the con tribution of the atmosphere above the 220m altitude boundary layer) was very low at 0.37 arcseconds, and relatively constant throughout the season (u = 0.07") while the range of the integrated seeing was very large (u = 0.75").

This result indicated that the majority of the turbulence was produced by varia tions in the lowest 200m of the atmosphere. This confirmed that the boundary layer turbulence resulting from katabatic wind flow and temperature inversion was the dominant contributor to optical seeing at South Pole. Higher on the plateau aver age wind speeds are much lower and the inversion layer is closer to the ground, and 1-2. Site Characteri ation 10

Figure 1.4. An Australian Antarctic Division Automatic Weather Station measuring temperature, pre ure, wind speed and direction. Photo courtesy of Australian Antarctic Division (2004).

2004; Antarctic Meteorological Research Centre 2004). Aristidi et al. (2004) pro vide an analysis and comparison between the data from an AWS situated near Dome C, and South Pole, with measurement taken at existing and planned astronomical ites such as Mauna Kea and the Atacama de ert in Chile.

1-2.2 AASTO

A number of pioneering astronomical project howed that the unique properties of the high Antarctic plateau were ideal for ground-based astronomy. Muon, Cosmic Ray and Neutrino experiments were conducted at the South Pole and the coastal bases Maw on and McMurdo from the 1960s. At the South Pole the extremely dry. cold atmosphere offered excellent conditions for observations in the millimetre and submillimetre wavebands. The first submillimetre telescope EMILE was con- 1-2. Site Characterisation 16 at time possibly nonexistent (Gillingham 1993). In such conditions the integrated eeing could be con iderably better at best approaching the free atmo phere eeing measured above the boundary layer at South Pole (~arks et al. 1999).

Figure 1.6. AFOS (left) and ADIMM (right) installed on the G-mount at South Pole January 2000. Photo court y of Paolo Calise, UNSW.

Two instruments in the AASTO program, the ADIMM and AFOS continued the quantification of the optical properties of the atmo phere above South Pole.

The ADIMM (Antarctic Differential Image ~1otion Monitor) is n to the right of the mount in Figure 1.6. The ADIMM was built by Dopita et al. (1996) and was designed around the optics of a 35cm diameter Celestron C14. It measure the integrated seeing at ground level by observing an object through two separate apertures. The aberration in the wavefront of the object, which is observed in the differential motion of the two images. can be used to calculate the integrated seeing produced in the atmo phere at the time of the measurement. In the ADIMM this method was optimised with the u e of 24 ub-aperture to obtain multiple imulta neous measurement of the eeing (Travouillon et al. 2003).

Mea: urements carried out with the ADIMM were completed throughout the win- 1-2. Site Characterisation 17 ter of 2001. The analysis showed an average seeing of 1.9" with a standard deviation of 0.6" (Travouillon et al. 2003), confirming the results of Marks et al. (1996) and showing that the ground level seeing at South Pole was on average quite poor in comparison with the best existing optical sites such as Mauna Kea.

In the winters of 2000 and 2001, a SO DAR (Sound Detection and Ranging) device was installed on the roof of the AASTO to characterise the turbulence profile of the atmosphere above South Pole. The SODAR antenna emits a series of 0.2 second long pulses at five different frequencies. The antenna then switches to a receiving mode and records the echo pattern reflected back to it from the turbulent layers in the lower atmosphere. In this way it can derive the three dimensional wind profile of the lowest 890m of the atmosphere, and the echo strength can be used to calcu 2 late the temperature fluctuation constant CT • These results can then be used to determine the turbulence profile (Travouillon et al. 2003).

Results from the SO DAR experiment confirmed that the winter turbulence above the South Pole is confined to a single, low altitude component up to 200-300m in height. This agrees with the results by Marks et al., recording a similar average seeing at ground level of 1. 73", peaking in the zone of vertical wind inversion and improving dramatically above the boundary layer, with an average 'free atmosphere' seeing of 0.61".

Somewhat 'downhill' from the summit domes of the plateau, South Pole is not the optimal site on the plateau for optical observing. The results of the seeing measure ments at the site were disappointing, but revealed that the dominant contribution to the poor seeing was from the turbulent boundary layer. This result promised that higher on the plateau at a site such as Dome C, where this inversion layer was predicted to be closer to the ground, and of less intensity (Schwerdtfeger 1984; Gillingham 1993) the seeing, and atmospheric stability could be vastly improved. 1-3. Dome C 18

Site testing measurements from the suite of AASTO instruments, in addition to observations by Marks et al. (1996, 1999); Marks (2002), Loewenstein et al. (1998) and Chamberlin (2001) have characterised the atmospheric and meteorological prop erties of South Pole as a potential site for infrared, submillimetre and optical as tronomy. The analysis of this data showed that South Pole offered excellent con ditions for submillimetre and infrared observing, though inconsistent weather and high ground layer turbulence reduced the quality of the site for optical astronomy.

The AASTO project allowed the testing of the automated observing, instrumen tation and power sources for the furthering of automated site testing on the plateau. Upon the completion of site testing measurements at South Pole in 2003, attention was turned to the high domes of the Antarctic plateau, which possibly offered a significant improvement in observing conditions across all wavebands.

1-3 Dome C

1-3.1 Cloud cover

Concordia Station, in the final stages of construction at Dome C, is a joint collabo ration between the Italian and French Antarctic programs (Fossat & Candidi 2003). For the past ten years, the base has supported a small population in the austral summer, and will host the first wintering contingent in the austral winter of 2005. Prior to the construction of a full autonomous laboratory, the AASTINO, the first wintertime measurements of optical properties at Dome C were undertaken by two instruments constructed at the University of New South Wales.

The first instrument, ICECAM, uses a small CCD camera and frame grabber to take images of the sky every two hours. Four sample images from ICECAM, collected in the austral winter of 2002 at Dome C, are shown in Figure 1.7. This 1-3. Dome C 19

Figure 1. 7. Four sample images of the 2095 recorded by I CECAM during 2001 at Dome C. (a) cirrus cloud, (b) clear twilight sky, (c) patches of frost on the CCD camera window, and (d) tars down to magnitude 6 ob erved during midwinter. The ICECAM CCD cover a 30x30 degree field-of-view centred at a declination of -53° (Ashley et al. 2003).

Figure 1.8. COBBER, lower left installed below ICECAM. January 2002. Photo courtesy of John Storey UNSW. 1-3. Dome C 20 figure illustrates the variety of conditions at Dome C during the austral winter. A calibration LED in the field of view was used to distinguish between cloud and any ice build-up on the window of the instrument. Further details of ICECAM and results from the experiment can be found in Ashley et al. (2003).

The second instrument, COBBER (Cloud OBservER), is seen installed below ICECAM in Figure 3.1. COBBER is an infrared cloud-observing device, containing a small thermopile detector to measure the integrated infrared flux from the sky at 8-12 p,m. A clear sky is much colder in the infrared than an overcast sky, the change in detector signal is used to monitor the cloud cover at Dome C throughout the winter season. The design, testing and analysis of results were a part of this thesis work, and the description and results are presented in Dempsey et al. (2003). The details of the instrument and the results will be discussed in Chapter 3.

1-3.2 AASTINO

The AASTINO (Automated Astrophysical Site Testing International Observatory), designed and built at the University of New South Wales (Lawrence et al. 2003), rep resents the next generation of fully automated site testing laboratories. Figure 1.9 shows the AASTINO, with the nearly completed Concordia Station in the back ground. As the station is not yet manned during the winter months, the AASTINO was designed to run independently from the beginning of February through to Oc tober of 2003.

The primary source of heat and power in the AASTINO is a Whispergen Stir ling engine manufactured by Whispertech, New Zealand. Two such engines were installed for redundancy, each capable of producing 6kW of heat and 900W of elec trical power. Figure 1.10 shows the internal layout of the AASTINO. The engines are located at the end of the building in this photo. The AASTINO is controlled 1-3. Dome C 21

Figure 1.9. AASTINO, with Concordia station behind, at Dome C, austral summer 2003. Photo courtesy of Tony Ti:avouillon. via a central 'supervisor" computer which communicates with the AASTINO sys tems and can be acces ed via an Iridium atellite link. This data link is also used to monitor data from the engine and control system and to download data from the instruments. In its first year of operation in 2003, the SUMMIT and SODAR instruments were installed on the roof of the AASTI 0 and both collected data from the end of summer in February until July of that year. Both in truments had previously been deployed at South Pole over several seasons, enabling the direct comparison between results from both ites.

Initial data from the SUMMIT showed an increase in the stability of the Dome C atmo phere in comparison to South Pole (Calisse et al. 2003). The SODAR howed a marked decrease in turbulence in comparison with the South Pole and appeared to confirm the predictions that the boundary layer was much lower at Dome C perhap less than fifty metres in altitude and decreasing further during the winter months (Travouillon et al. 2003). In addition the turbulence was o low at times that the antenna did not detect any echoes returning from the atmosphere. 1-3. Dome C 22

Figure 1.10. Internal layout of the AASTINO foreground shows the heat exchangers which warm the room and instruments. The far end of the room are the Whispergen Stirling engines and the control panel (top left). Photo courtesy of Tony Travouillon.

In the austral ummer of 2003/2004 a Multi-Aperture Scintillation Sen or (MASS), was in tailed in the AASTINO to measure turbulence distribution throughout the atmosphere above Dome C (Lawrence et al. 2004). This instrument can determine the turbulent profile of the atmosphere by measuring the scintillation of a single star. The scintillation indexes measured by the MASS are used to compute the turbulence inten ity within six atmospheric layers centred at heights of 0.5, 1 2, 4, 8 and 16km above the ground. From this turbulence profile it is possible to calculate the free atmosphere seeing above 500m.

The in trument collected turbulence profile from February until mid-May of 2004, when the AASTINO experienced a power fault. The results from the first months of the winter season confirmed the vast improvement in seeing from South Pole to Dome C as a result of the reduction in height and strength of the inversion layer above the ite. A full description of the instrument is provided in Lawrence et al. (2004). Results, to be published in Nature later this year (Lawrence et al. 2004), record the lowest seeing measured from any site on earth, 1-3. Dome C 23

Table 1.2. Comparison of seeing c:o, in arcseconds, measured at Dome C with the MASS instrument in 2004, and other premier optical observing sites, Mauna Kea (Hawaii) and Parana! (Chile). As shown in Lawrence et al. (2004).

Site I Seeing (co) I Reference South Pole 1.8 Marks et al. (1999) Dome C 0.27 Lawrence et al. (2004) Mauna Kea (Hawaii) 0.5-0.7 Racine & Ellerbroek (1995) Cerro Paranal (Chile) 0.80 Sarazin (1999)

with an average seeing of Eo = 0.27 arcseconds, and seeing below 0.15 for 25% of the time. The average seeing measured at Dome C with the MASS is compared to the seeing results from other premier optical observing sites in Table 1.2, taken from Lawrence et al. (2004).

The initial poor seeing results at South Pole by Marks et al. (1999) deterred most astronomers from considering the Antarctic plateau as an optical astronomical site. The measurements at Dome C confirm that the atmosphere is quite different from that at South Pole. The mild, extremely stable weather, low fraction of winter cloud cover (Ashley et al. 2003; Dempsey et al. 2003) and extremely low winds (Aristidi et al. 2004) already differentiate it from South Pole which can ex perience blowing snow, ice and 'diamond dust' for a significant fraction of the winter.

The SODAR and MASS results now confirm the predictions that the turbulent boundary layer that produces poor seeing at South Pole is lower and weaker at Dome C. The measurements show seeing lower than any other site on the planet. The Antarctic plateau is now a viable and exciting new option for ground-based optical astronomy. There is therefore renewed relevance in the assessment of other important optical properties of the Antarctic atmosphere for astronomy. 1-4. AFOS 24

1-4 AFOS

What else i required in a ite for good ob erving at optical wavelengths? At mo pheric components including H20 and ozone absorb incoming radiation in the UV-visible band. Other gases in the upper atmosphere, including atomic and molecular oxygen and nitrogen emit radiation when bombarded by energetic particles and radiation from the sun. These collisions produce auroral and airglow emi sion lin in the visible region of the spectrum. The AntarcLic Fibre Optic pectrometer (AFOS) was designed to measure the ky transmission at wavelengths from the ultraviolet (300nm) through to the red end of the "visible pectrum (900nm). AFOS was the first telescope to collect spectra of stellar object from the high Antarctic plateau.

Figure 1.11. AFOS telescope on the G-mount, austral summer 2002/2003. The silver foil covers a mall heater that warms the front window plate of the telescope to keep it free from ice. Photo courte y of Paolo Calis e, l:.JN W.

A small 30cm tele cope with a V-visible pectrometer mounted at the focu was constructed by Ro coe et al. (1994) and operated at the British Halley base on the Antarctic coast. This telescope used tars to measure the quantity of atmospheric constituents by a quantitative comparison of the depth of atmospheric 1-4. AFOS 25

absorption lines (03 , N02 , N03 and OClO) in the stellar spectra when observed at two different zenith angles. Any difference in the depth of the absorption can be attributed to the quantity of the molecule present in the atmosphere (Fish et al. 1994). This novel technique offered possibilities for a similar instrument to quantify the atmospheric constituents above the South Pole.

The particular questions asked iii the design and optimisation of the AFOS were: does the reduction in the quantity of ozone over Antarctica result in more ultraviolet light reaching the ground? At wavelengths below 295nm, all radiation incident upon the atmosphere is absorbed by atmospheric gases, primarily ozone. Between 295-400nm, the radiation is partially absorbed, and there was interest in discovering if this "UV cutoff wavelength" was at a shorter wavelength at South Pole than at a mid-latitude site like Mauna Kea.

Secondly: what was the depth of various molecular absorption bands in the visible region of the spectrum? In particular, could the spectra be used to model these absorption bands and determine the quantities of certain atmospheric constituents (such as H2 0 and ozone) in the atmosphere and how these quantities vary over an austral winter?

A significant question was also: how strongly do auroral emissions, and airglow lines, affect optical observations? No existing large optical observatories are positioned near the auroral regions in the Northern hemisphere, and although auroral phenomena have been extensively studied in Antarctica, there has been no quantification of the effect of auroral emissions on optical astronomical observations. 1-4. AFOS 26

1-4.1 History

The AFOS was designed and built at the University of New South Wales during 1997 by Boccas et al. (1998). It was tested at Siding Spring Observatory during September of that year, where the first spectra were successfully observed. In late 1997, the AFOS was installed at South Pole station on a temporary, single-axis T-mount, on a 7.5 metre tower to reduce snow drifting and minimise local air turbulence problems.

Difficulties with the mount prevented measurements from being collected during that austral winter. The AFOS was returned to Australia, pending the completion of the dual-telescope, altitude-azimuth G-mount (Hovey et al. 1998) which would support and control both the ADIMM (described in the previous section) and the AFOS on two adjacent supports, as seen in Figure 1.6.

Both telescopes and mount were installed on the tower in the summer of 1999/2000. A fault in the azimuth axis of the mount prevented a full set of stellar spectra from being observed during this winter but the telescope was able to observe the full lunar eclipse of June 16th 2000. Michael Ashley at UNSW wrote a script to collect AFOS images of the moon for four hours bracketing the eclipse. The resulting spectra were encouraging: a strong auroral storm occurred across the face of the moon during the observations, and these emission lines were observed in the lunar spectra at the height of the eclipse. An example of the lunar eclipse spectrum is shown and discussed in Chapter 2.

The mount was repaired in 2001 and the two telescopes were refitted to the mount and reinstalled on the tower in January of 2002. All of the data analysis conducted before February 2002 was completed by Boccas et al. (1998) at the University of New South Wales. 1-5. Thesis Goals 27

The AFOS instrument analysis, observations and results presented in this thesis were undertaken at the start of 2002. The AFOS collected data successfully for part of the winter of 2002 and for a large part of the winter of 2003. The AASTO project was completed at South Pole at the end of the winter of 2003, and the AFOS electronics, CCD and spectrometer were returned to Sydney in February of 2003/2004. The AFOS currently remains on the G-mount at South Pole, while the ADIMM has been replaced by a new optical telescope, Vulcan, funded by the SETI institute. Vulcan will be briefly discussed in Chapter 2.

1-5 Thesis Goals

The South Pole and Dome C have been shown to have great potential as sites for infrared and submillimetre observing. Though ground level seeing at South Pole is comparatively poor, new measurements at Dome C have shown the best seeing of any site in the world. The work of this thesis endeavours to further the understanding of the optical properties of the atmosphere above the high Antarctic plateau. This work began with the design, testing and cloud-cover measurements of COBBER, which is discussed in Chapter 3.

With the increasing interest in Dome C as a site for optical astronomy, the effect of auroral emission on observations is a question of high importance. A detailed study of existing auroral measurements at South Pole, and an extrapolation of these intensities to those which may be observed at Dome Cis presented in Chapter 2.

The AFOS measurements in 2002 and 2003 were a primary part of the work, and Chapters 4 and 5 discuss the instrumentation analysis, observing method and data reduction techniques that were a large part of this thesis. 1-5. Thesis Goals 28

Analysis of the AFOS spectra and modelling using the MODTRAN radiative transfer code allowed a measure of the precipitable water vapour in the South Pole atmosphere to be derived from the data. The analysis and results of these model comparisons are presented in Chapter 6.

Finally, a unique set of measurements of the dark side of the moon were made with the AFOS in August and September of 2003 to observe the earth shine spectrum. The geographic location of the telescope allowed continuous 24 hour observations of the moon to be taken, one of the particular advantages of the South Pole site for a range of optical astronomical projects including the detection of extra-solar planets. The details of the observation, data analysis and results are presented in Chapter 7. Such measurements are of interest to as tronomers seeking to use the technique in the detection of life on extra-solar planets.

It is thought-provoking that during its two winters of successful remote observa tions AFOS was ten times further from, and less accessible for its observers, than the Hubble Space Telescope. Throughout this work, the final important goal of this thesis is continually addressed: to show that astronomy is possible on the Antarctic plateau.

As the science case for the optical potential of Dome C strengthens, there is a need to demonstrate that astronomical telescopes can be built, maintained and successfully operated in Antarctic conditions. It is hoped that this work is a step, a small step, towards showing that Antarctica is not, perhaps, such an awful place. Chapter 2

Auror_al contribution to optical sky brightness

2-1 Introduction

Auroral emission over Antarctica has been studied intensively for the last fifty years by space physicists. In the last twenty years, satellite data from NOAA and other sources provide a nearly continuous stream of images of the auroral activity in the auroral circle, centred on the geomagnetic south pole. Ions and electrons are ejected from the sun's corona and are transported towards the earth as the solar wind. Aurorae are a result of collisions between atmospheric gases and precipitating charged particles (mostly electrons) guided by the geomagnetic field from the magnetotail (Omholt 1971). Each gas (oxygen and nitrogen molecules and atoms) emits in its characteristic spectral lines when bombarded. Atmospheric composition varies with altitude with predominantly molecular gas at lower altitude giving way to atomic gas at higher altitude.

Since the faster precipitating particles penetrate deeper, the colour of the aurora will depend on particle energy. The auroral altitude range is 80 to 1000 km, with typical aurorae 100 to 250 km above the ground. The colour of the typical aurora 2-1. Introduction 30

Table 2.1. Potential optical astronomy sites on the high Antarctic plateau, and their GMT (Geomagnetic) Latitude and Longitude, calculated for 2004. Skitbotn is a Norwe gian observatory situated below the northern auroral oval.

Potential Site Latitude Longitude GMT Latitude GMT Longitude South Pole -76° 13' 11° 53' Dome C -75° 06' S 123° 23' E -89° 10' 264° 56' Dome A -81° 58' 51° 10' Skibotn 20° 19'E 105° 24'

is yellow-green, from a transition of atomic oxygen at 557.7nm (Vallance Jones 1974). Auroral light from lower levels in the atmosphere is dominated by blue and red bands from molecular nitrogen and molecular oxygen. Above 250 km, auroral light is characterized by a red spectral line of atomic oxygen.

Auroral emissions are limited to an elliptical band around each geomagnetic pole. This band ranges from about 75 degrees geomagnetic latitude at local noon to about 67 degrees geomagnetic latitude at midnight under average conditions. An image of the southern auroral oval, with the positions of the South Pole, Dome C and Dome A, can be seen in Figure 2.1. The geomagnetic pole locations are arrived at by calculation using magnetic reference field models (MacMillan & Quinn 2000). This provides a constant reference point for geophysical study of the aurora throughout a given year of observation. At the beginning of 2004, the position of the geomagnetic south pole was 79.3S, 108.5E, a shift of 0.24S and 0.1E from 2003.

The aim of this chapter is to evaluate the contribution by auroral emissions to the overall photometric and spectroscopic sky brightness above the polar plateau, and the consequences for optical astronomy. The results presented in this chapter have been submitted for publication in Dempsey & Storey (2004). The three sites on the Antarctic plateau that will be considered are South Pole, Dome C and Dome A. South Pole is located at a geomagnetic (GMT) latitude of -76° 13', 2-1. Introduction 31

Figure 2.1. South Pole, Dome C and Dome A locations in respect to a map of the southern auroral oval extrapolated from measurements taken during a recent polar pass of the NOAA POES atellite Feb 26, 2004. Activity index, Kp = 6 (high). Courtesy of National Oceanic and Atmospheric Administration (2004) webpage.

directly under the dayside auroral oval. Dome A is closer to the geomagnetic pole at a geomagnetic latitude of -81o 58' and Dome C is situated very near to the geomagnetic south pole itself at the geomagnetic latitude of -89° 10'. The geographic and 2004 geomagnetic coordinates for the e sites are given in Table 2.1.

2-1.1 Auroral line intensities and spatial extent

Auroral intensity is described by the International Brightness Coefficient (IBC) and is scaled to the intensity of the 557.7nm [OI] line on a logarithmic scale from IBC1 (Iss7.7nm = 1kR) to IBC4 (Iss7.7nm = 1000kR) (Vallance Jones 1974). Most of the calculations in the following sections follow model calculations from Vallance Jones (1974) assuming a bright aurora of intensity IBC3, corre ponding to an intensity of 100kR for the 557. 7nm [OI]line.

A medium resolution spectrum of some of the visible wavelength auroral emission 2-1. Introduction 32

0.2

0 4000 6000 8000 10000 12000

Figure 2.2. The position of the visible auroral lines in respect to the UBVRI pass bands. Auroral spectrum is for a bright auroral event and covers from 3000A to 8900A (Paxton et al. 1999; Swenson et al. 1998). lines is shown in Figure 2.2 (Swenson et al. 1998). Between 400nm and 600nm, the 427.8nm Nt band emission and the 557.7nm [OI] forbidden oxygen line make the most significant contributions. At 630.0nm and 636.4nm are forbidden oxygen transitions [OI] and beyond that the red end of the spectrum is dominated by molecular oxygen and N2 band emission. The time evolution of the intensities of the 427.8nm (blue), 557.7nm (green), and 630.0nm (red) lines for this particular measurement are shown in Figure 2.3.

The intensities are given in Rayleighs, the standard unit used by atmospheric physicists when describing emission intensities (Chamberlain 1961). By definition: 1 Rayleigh= 7.96x108 photons s-1 m-2stec1 (Baker & Romick 1976).

If we consider a 1kR emission at 557.7nm, then we can calculate the intensity in V magnitudes that it would produce in a V band filter centred at 556nm and with a bandwidth of 85nm. If the total flux from the 557.7nm line were distributed as a continuum uniformly across the V band, it would, at each wavelength, be equal to 11.8 Rayleighs/nm. This intensity, I, in R/nm, can be converted to a flux in 2-1. Introduction 33

10000 03/23<96, 9000 Sandre, Up B i 8000 -~... 7000 :;. 6000 & '""J 5000 1!· 4000 ·=~.. 3000 ..:: 2000 -· 1000 0 ~ 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 Time (UT)

Figure 2.3. Time history intensity for the 427.8nm (blue), 557.7nm (green), and 630.0nm (red) auroral emission event shown in Figure 2.2 (Swenson et al. 1998). p,Jy j arcsec2 at the central wavelength of the band, using the equation (derivation of this equation is provided in Appendix A):

6 2 F =I x (1.24 x 10 ) x .:\ J.tJanskys/arcsec (2.1)

where.:\ is the central wavelength in metres and the resulting flux, F is in J.tJanskys per square arcsecond. At a central wavelength of 556nm, an emission intensity of

2 2 11.8 R/nm equals a flux of 8.14 p,Jyjarcsec . This flux in J.tJyjarcsec is then converted to V magnitudes per square arcsecond by the equation:

2 Magnitudesjarcsec = 20- 2.5log10 (~) (2.2)

where F is the auroral flux in the band in j.tJanskys per square arcsecond, and K is 42.6, 36.4 and 30.8 for the B, V and R bands respectively (Bessell 1979; Benn & Ellison 1998). A lkR auroral emission at 557.7nm is observed in V band at a V magnitude of 21.6 per square arcsecond. For a B filter centred at 436nm and a bandwidth of 94nm, a lkR emission in the Nt band at 427.8nm observed in 2 the B band is equivalent to 5.75 j.tJyjarcsec , and thus a B magnitude of 22.2 per square arcsecond.

For the particular aurora in Figure 2.2, we observe an intensity of 7000 R at 557.7nm, 2000R at 427.8nm and 800R at 630.0nm. A list of auroral optical line and 2-1. Introduction 34 molecular band intensities is provided in Tables 2.5 to 2.11 (Vallance Jones 1974; Omholt 1971) for a IBC3 aurora. These intensities will be used in calculations in Section 5. Vallance Jones (1974) states that the ratio of the 427.8nm band intensity to the 557.7nm line intensity is always close to 0.33, while the ratio of the 557.7nm intensity to the 630.4nm line can vary from 0.02 up to 1. The 427.8nm Nt band emission and 557.7nm [OI] line emissions are dependent only on the incident electron flux, whereas the 630.4nm and 636.4nm [OI] line intensities are also determined by the energies of the incident electrons (Vallance Jones 1974).

For an IBC3 aurora producing the intensities in Tables 2.5 to 2.11 the results will be a sky brightness contribution of 16.6 V magnitudes per square arcsecond from the 557.7nm line, 18.5 B magnitudes per square arcsecond for the 427.8nm Nt band emission, and 17.0 R magnitudes per square arcsecond from the 630.0nm [OI] line (assuming the maximum emission of 100kR). The 630.0nm emission line in particular is rarer and less strong on average than the 557.7nm [OI] line and 427.8nm band emission intensities (Vallance Jones 1974).

2-1.2 Previous Studies

There are no large optical observatories at high northern latitudes where auroral light is a frequent and dominant contributor to sky brightness at visible wavelengths. The only observatory where auroral light is of significant impact is the small optical observatory close to Skibotn, Norway. Skibotn is situated within the northern auroral oval (see Figure 2.1). The northern and southern auroral ovals are identical in shape. Studies during 1979-1984 by Myrabo (1978, 1979, 1980, 1982) inves tigated the viability of photometry in a region of strong and variable auroral activity.

Myrabo used a double-aperture chopping photometer (at 2-20Hz), with fields of view of 3' each, and care taken to ensure the throughputs of both fields were 2-1. Introduction 35

identical. An auroral photometer, with a 1o field of view and a 2.0nm bandpass centred on 427.8nm, observed the same stellar field to allow subtraction of the measured auroral brightness during the observation. Over three years, the mean sky brightness measured in the UBV colour bands was found to be U 17.2, B 18.8, and V 18.1, all in star magnitudes per square arcsecond (Myrabo 1980).

These papers also show that the 427.8nm band emission of the aurora is a very good indicator of the total sky brightness in the B band, however the rapid variation in its brightness means that this has to be constantly monitored. Rapid changes in auroral intensity and spatial extent cause high-frequency variations in irradiance seen from the ground. Fast moving aurorae and rapid brightening can cause variations from a few kR up to several hundreds of kR within a fraction of a minute (Myrabo 1982). High speed photometric measurements (McHarg et al. 1998) observe the rapid variation in flickering aurora, and though most intensity fluctuations observed are below 80Hz, frequencies as high as 180Hz are detected. Deep field, long exposure observations and sky background subtraction will obviously be compromised by such rapidly varying features.

Auroral arcs can also possess fine spatial structure on extremely small scales (Trondsen & Cogger 1998). Arcs are narrow linear auroral features that can range from the order of lOOm across down to structures as fine as 10m in the ionosphere. A 10m structure in the ionosphere, approximately 100km from the ground, corresponds to an angular size of 20 arcseconds. Bright, dynamic arcs of this scale can also be observed embedded in more diffuse airglow. Small scale structure of this order will affect even narrow field-of-view observations and prevent uniform sky background subtraction.

Auroral structures do not often dominate the entire sky, but even when observing unilluminated regions, a bright aurora outside of the telescope beam will contribute to the sky brightness in the field of view (Gattinger et al. 1991). The high ground 2-2. AFOS Measurements 36 albedo of the Antarctic Plateau will increase this effect (Carroll & Fitch 1981).

2-2 AFOS Measurements

Stellar measurements were made with the AFOS throughout 2002 and 2003, and for the greater part were short exposures ( < 60 second) of very bright and well-known objects (mag < 2.0), as the primary goal of the instrument was to investigate the atmospheric extinction. None of these measurements has sufficient sensitivity to detect auroral lines, though comparison with auroral data from South Pole at the times of our observations show that there were occasions when there was auroral activity in the sky. The instrument, observations and data reduction will be described in detail in the following chapters. Only one set of data, taken of the moon dur ing the full lunar eclipse of 16th of July, 2000, has auroral lines evident in the spectra.

2-2.1 2000 Lunar Eclipse, South Pole.

600

A r b i 400 t r a r !,j 200

3000 4000 5000 6000 7000 8000 Wavelength (angstroms)

Figure 2.4. Spectrum of the moon observed during the total lunar eclipse of July 16th, 2000, with the AFOS. Strong auroral lines can be seen.

These data taken with the AFOS were of the lunar eclipse of July 2000. During the peak of the eclipse a strong auroral storm occurred directly over the face of 2-3. Auroral Data from South Pole 37 the moon and was evident in the spectra collected from the AFOS for a period of approximately forty minutes, until the increasing moonlight exceeded the faint signal. Only in the longest exposures during the full eclipse, of 100 and 600 seconds respectively, were the auroral lines detectable. An example of the spectrum is shown in Figure 2.4. The reduced intensity of the blue emission lines is a result of a lower transmission of the fibres towards the UV. This observation is during the complete eclipse and so the reflected solar spectrum of the moon is only observed in the far-red wavelengths.

2-3 Auroral Data from South Pole

Figures 2.5, 2.6 and 2.7 present data collected by the University of Maryland (De trick, private communication 2004) binned as a function of intensity. These data are from two photometers, each with a field of view of approximately 60°, used in conjunction with a narrowband filter of width 3.0nm. One photometer is centred on the auroral emission 427.8nm of the Nt band, and the other on the forbidden oxygen [OI] transition at 630.0nm. The intensities presented are ten minute averages, for the entire winter periods (approximately 125 days) of 1985 (solar minimum) and 1990 (solar maximum). The 427.8nm photometer has a limiting sensitivity of 23.7 B magnitudes per square arcsecond. The 630.0nm photometer sensitivity is 24.5 R magnitudes per square arcsecond.

The intensity ratio of I427.s/I557.7 = 0.33 (Vallance Jones 1974) is used to arrive at the intensities of the 557.7nm emission. In the plots of the 427.8nm and 557.7nm emission, the intensities have been converted to magnitudes per square arcsecond using the method outlined in Section 1.1. The measurements were averaged over 10 minute intervals; the histogram values are the number of hours per winter that a particular intensity is observed.

From the percentile distribution curves in Figures 2.5 and 2.6, we can conclude 2-3. Auroral Data from South Pole 38

60 100 1985 Solar Minimum 50 80

40 (l) 60 = ~ 30 c: :::1 - 0 40 :I: 20 e 20 0... 10

0 0 17 18 19 20 21 22 23 6016 100 I 1990 Solar Maximum 50 80 40 60 :E ~ 30 c: ::I 0 40 e :I: 20 0... 10 20

o+-~--~--~-r~~~~~LL~~_u,u~J_~~~~~~ 0 16 17 18 19 20 21 22 23 V magnitudes I arcsec2 Figure 2.5. Intensity of the 557.7nm [OI] line at the South Pole in V magnitudes per square arcsecond. The histogram (left axis) is the number of hours per winter that that intensity is observed, while the cumulative probability is referenced to the right hand axis.

60 100 1985 Solar Minimum I I i :_ ! 50 ~ 80 -~ - ,. 40 (l) 60 = ~ 30 / t: ::I r- - 0 / 40 ~ :I: 20 r ~ (1) ,. 20 D.. 10 I-"' 0 ~rmJk 0 17 18 19 20 21 22 23 6016 100 I.--1-9_9_0_S_o_la_r_M-ax_l_m_u_m_...., 50 80 40 (1) 60 :E ~ 30 c: ::I (1) 0 40e :I: 20 (l) D.. 10 20 0 0 16 17 18 19 20 21 22 23 B magnitudes I arcsec2 Figure 2.6. Intensity of the 427.8nm nitrogen band at the South Pole in B magnitudes per square arcsecond. The histogram (left axis) is the number of hours per winter that that intensity is observed, while the cumulative probability is referenced to the right hand axis. 2-3. Auroral Data from South Pole 39

80 1985 Solar Minimum 70 60 50 ~:::s 40 0 30 ::1: 20 10

O+-----r-~~--~~--~~~--T-~~~~-4~~~~ 8016 17 18 19

70 'I_1_9_9_0_S_o_la_r_M_a_x_lm_u_m--, 60 -./80 --

17 18 19 20 21 22 23 24 R magnitudes I arcsec2 Figure 2.7. Intensity of the 630.0nm [OI]line at the South Pole in R magnitudes per square arcsecond. The histogram (left axis) is the number of hours per winter that that intensity is observed, while the cumulative probability is referenced to the right hand axis.

that for 50% of winter observing time in a solar minimum year, the emission intensity of the 557. 7nm line is less than 2000 Rayleighs. For a V band centred at 556nm and bandwidth of 85nm, an emission at 557. 7nm of 2kR is equivalent to 23.6 R/nm at each wavelength when averaged across the V band. This results in a sky background in V of 20.9 V magnitudes per square arcsecond. In a solar maximum year the median value is slightly lower at 21.7 V magnitudes per square arcsecond. In a solar minimum year, if we consider a B band centred at 436nm and with a bandwidth of 94nm, the 427.8nm Nt(O,l) band emission produces an intensity of 22.8 B magnitudes per square arcsecond or less for 50% of the winter period and 23.5 B magnitudes per square arcsecond or less for 50% of observing time during solar maximum.

The 427.8nm and 557. 7nm line intensities are dependent on the electron flux incident on the atmosphere above the site (Vallance Jones 1974). South Pole is 2-4. Auroral Intensity at Dome C and Dome A 40 positioned close to the inner edge of the auroral oval. During solar maximum the auroral oval extends down to lower latitudes and away from the geomagnetic pole. At such times South Pole experiences lower electron flux and thus less intense aurora, while lower latitude sites experience increased auroral activity.

The red 630.0nm [OI] line photometer data is plotted in Figure 2.7, and from this we conclude that for 50% of winter observing time in a solar minimum year, the emission intensity that the 630.0nm line would produce in an R filter is less than 23.6R magnitudes per square arcsecond. In a solar maximum year, the median value is 23.4 R magnitudes per square arcsecond. Unlike the 427.8nm and 557.7nm emissions, the 630.0nm emission intensity is dependent on the energy of the incident electrons (Vallance Jones 1974). There are more high-energy electrons incident upon the atmosphere in a solar maximum year, and so we see an increase in the intensity of this auroral transition.

2-4 Auroral Intensity at Dome C and Dome A

Unlike South Pole, there are as yet no ground-based measurements of auroral in tensity at Dome C and Dome A. However there are satellite measurements of the average flux of electrons that strike the atmosphere above all three sites (Hardy et al. 1985). Models by Vallance Jones (1974) allow us to use these electron fluxes to esti mate the auroral intensities that would be observed at the ground. These data can then be used to calculate the contribution that aurora would make to the B and V sky background at the given site.

Hardy et al. (1985) presented a statistical study of the Defense Meteorological Satellite Auroral Program over Antarctica. This study aimed to determine the average characteristics of auroral electron precipitation as a function of magnetic local time, magnetic latitude, and geomagnetic activity as measured by the Kp index. The K-index is measurement of the maximum fluctuation in the geomagnetic 2-4. Auroral Intensity at Dome C and Dome A 41

Table 2.2. Average electron flux in 107keV/cm2 s sr, for the listed sites for Magnetic Local Times: 06:00hrs, 12:00hrs, 18:00hrs and 24:00hrs. Difference between MLT and UT for SP is 3:43hrs, Dome C is 2:34hrs, Dome A is 2~05hrs.

MLT I 6hrs l12hrs 118hrs 124hrs I SP 12.14 27.14 8.93 6.07 DomeC 2.0 1.64 1.64 1.64 Dome A 6.07 10.71 2.93 1.71

field during a three-hourly interval. The Kp index is deriv!=ld from an algorithm that averages the K-indices measured at several stations around the world. Kp is expressed as a logarithmic scale between 0 (low magnetic disturbance) and 9 (high magnetic disturbance). See, for example, Michel (1964). This scale can be used as an estimate of the level of auroral activity. For an average level of auroral activity (Kp = 4) and knowing the geomagnetic positions of South Pole, Dome C and Dome A we can arrive at an estimate of the typical electron flux at these sites as a function of Magnetic Local Time (MLT). These values are shown in Table 2.2.

Several theoretical models have been derived to relate the incident electron flux to the intensity of aurora that it produces. Using the model described by Vallance Jones (1974), we can use the average electron fluxes from Hardy et al. (1985) to arrive at an estimate of the auroral intensity (kR) in a particular emission line or band. Vallance Jones (1974) calculates the electron flux required to pro 2 duce an 1BC3 aurora (1557.7 = 100kR, 1427.8 =30kR) is 154 ergsjcm .s. The rate of emission varies as a function of height, and the intensity observed at the ground is the emission rate integrated along the line of sight. Because auroral structures are aligned along the magnetic field, observations away from the magnetic zenith in tersect aurora of different intensities, thicknesses and height distributions ( Omholt 1971; Vallance Jones 1974). All line and band intensities calculated here are there fore for the magnetic zenith. An observer's magnetic zenith is located southward (for an observer in the southern hemisphere) of their true zenith, by a value correspond- 2-4. Auroral Intensity at Dome C and Dome A 42

Table 2.3. Sky background in B and V magnitudes produced by aurora at South Pole with incident electron flux levels in Table 2.2.

I MLT I 6hrs l12hrs 118hrs 124hrs I B 24.93 24.05 25.26 25.68 v 23.58 22.71 23.92 24.34

Table 2.4. Difference (in magnitudes) between the auroral contribution to the sky brightness at Dome C and Dome A, relative to the contribution at South Pole. Values are calculated as a function of magnetic local time, for an average level of auroral activity.

MLT I 6hrs 112hrs 118hrs I 24hrs I DomeC 1.96 3.05 1.84 1.42 Dome A 0.75 1.01 1.21 1.38

ing to their geomagnetic latitude. For example, magnetic zenith for an observer at South Pole will be 14 degrees from their true zenith, as their geomagnetic latitude is 76 degrees.

From the relative line and band strengths in Table 2.5 to 2.11 by Vallance Jones (1974), we can compare the total contribution that auroral emission makes to the B band and the V band at the three sites. These tables are also provided at the end of this chapter, for reference. Table 2.3 shows the sky background at South Pole in B and V magnitudes per square arcsecond calculated from the incident electron flux levels in Table 2.2. Table 2.4 shows the relative auroral background magnitude at Dome C and Dome A in comparison to South Pole. In regions of high auroral activity, such as South Pole, emission intensities are dependent on Magnetic Local Time (Vallance Jones 1974) and so the comparisons are shown for four intervals over a 24hr period.

South Pole, located directly under the auroral oval, experiences auroral activity at a greater intensity than at Dome A and Dome C. The location of Dome C close 2-5. Sky brightness in standard photometric bands 43 to the geomagnetic pole results in a sky background that is lower by up to 3.1 magnitudes. Dome A experiences less auroral intensity than South Pole, but as it is closer to the edge of the auroral oval than Dome C the decrease in auroral activity is less.

2-5 Sky brightness in standard photometric bands

In Figure 2.2, high-resolution spectra of auroral emission lines from Swenson et al. (1998); Paxton et al. (1999) are superimposed upon the transmission curves for the UBVRI wavebands. We will use these to investigate the contributions of auroral emission to sky background flux in the B, V and R wavebands, first when standard filters are used, and secondly when additional 'notch' filters are used to remove the strongest auroral lines and bands.

Individual auroral lines are very narrow, with line widths dependent on the tem perature of the atmosphere where the excitations occur. The Doppler-broadened FWHM of a single auroral line is of the order <0.1nm (Vallance Jones 1974; Omholt 1971). Molecular band emissions are much broader, up to the order of 5nm (Bashkin et al. 1991). The data presented in Section 3 were measured by photometers with filters of width 3nm.

The line and band lists in Table 2.5 to 2.11 indicate that there are a signif icant number of contributors to the auroral emission spectrum in the optical wavebands. In addition to this, observations show that in between these pre dicted auroral emission bands and lines, a quasi-continuum emission is also seen (Gattinger & Jones 1974; Sharp 1978). This emission is likely contributed by weaker molecular bands of nitrogen and oxygen that lie below the resolution of the spectrometer (Bashkin et al. 1991). For a strong IBC3 aurora, Gattinger & Jones 2-5. Sky brightness in standard photometric bands 44

(1974) calculate a quasi-continuum level of 270 Rayleighs per nanometer (R/nm) between 450-650nm, dropping to 100 R/nm at 900nm. In the following calculations for an IBC3 aurora, we have included a quasi-continuum of 270 R/nm.

In the following sections we will consider the effect of using a commercial holo graphic notch filter such as one from Kaiser Optical Systems Inc (2004). An other technique for producing extremely sharp, deep notch filters is the Rugate method (Johnson & Crane 1993).

2-5.1 B band

In the B band, the Nt(0,1) vibrational band emission with band head at 427.8nm is the significant auroral contributor. All calculations use the intensities in Table 2.5 to 2.11. Fortunately the stronger Nt(O,O) band emission at 391.4nm lies beyond the edge of a typical B band filter. The Nt(1,3) vibrational band at 465.2nm, the Nt(0,1) at 423.6nm and Nt(0,2) at 470.9nm are the other strong emissions in the B band. A typical B band filter has a central wavelength of 436.0nm and an effective width of 94.0nm.

For an IBC3 aurora, the quasi-continuum intensity in the B band is 270 R/nm, equivalent to 85% of the emission intensity at 427.8nm. The intensity ratio,

Iquasd I 427.s = 0.85, is therefore used to estimate the quasi-continuum contribution in the B band from the 427.8nm photometer data.

Tables 2.5 to 2.11 show the intensities of the dominant auroral emission lines for an IBC3 aurora. The 427.8nm band emission intensity is 30kR. The intensities of the other strong emission in this band produce a combined intensity of 11.3kR. The total flux from these auroral emissions, if distributed as a continuum uniformly across the band, would, at each wavelength, be equal to 429 R/nm. If we add to this a quasi-continuum contribution of 270 R/nm, the emission from the auroral 2-5. Sky brightness in standard photometric bands 45

60 100 1985 Solar Minimum 50 so 40 .9:! 60 ::;::::; ~ 30 ~ :::s C2> 40 0 . •' 20 i: :I: , .. ··. C2> D.. 10 20

0 0 6016 17 18 19 20 21 22 1990 Solar Maximum 50 I

40 .·. ·... . . •' . ~ ,• .. :::s~ 30 . . . , . , ~ 0 ·,.,.. · . :I: 20 . . ..,.

10 20 rJ.rf : : .: ::· .. : ··> .. ·:-: . : ..:~~-~-- ' 0 ~ ·.-·.·· ..... ·.· .. ··.' 0 16 17 18 19 20 21 22 23 24 25 B magnitudes I arcsec2 Figure 2.8. Sky brightness (in B magnitudes per square arcsecond) at the South Pole for a standard B filter (solid grey line) and for a notch-filter at 427.8nm (dashed line). The cumulative probability for the filter histogram is referenced to the right hand axis. lines, bands and continuum for an IBC3 aurora is equivalent to 709 R/nrn when averaged across the B band. This results in a B band auroral intensity of 17.6 B magnitudes per square arcsecond.

Using this method the measurements from the 427.8nrn South Pole photometer have been used to calculate the probability distribution of B magnitudes per square arcsecond for the winters of 1985 and 1990, years which correspond to the solar minimum and maximum respectively. This is represented by the solid grey histogram in Figure 2.8.

We now consider a notch-filter that reduces the flux at 427.8nm to negligible levels. The remaining emission bands and quasi-continuum are equivalent to an intensity of 390 R/nm at each wavelength, when averaged across the B band. This results in a B magnitude of 18.3 per square arcsecond. We can also ask what depth of notch is required to reduce the contribution from the 427.8nm line to below 10% 2-5. Sky brightness in standard photometric bands 46 of the B band total. We find that an attenuation of 5 would suffice. Thus, a single notch filter centred at 427.8nm can decrease the background sky emission by 0.7 magnitudes, while removing only 11 percent of the broad-band B flux. This is shown as the dashed line histogram in Figure 2.8.

2-5.2 V band

A standard V band filter, from 500nm to 650nm, includes the strongest auroral emission feature, the forbidden oxygen [OI] transition at 557.7nm. Other lines included in the following calculation are the Nt vibrational bands at 522.8nm, 612.3nm and 628.5nm, with relative intensities as shown in Table 2.5 to Table 2.11.

For an IBC3 aurora, the quasi-continuum intensity in the Vband is 270 R/nm, equivalent to 23% of the emission intensity at 557.7nm. The intensity ratio

Iquasd ! 557.7 = 0.23 is therefore used to estimate the quasi-continuum contribution in the Vband from the 557.7nm photometer data.

A typical Vband filter is centred at 556.0nm, and has an effective width of 85.0nm. For an IBC3 aurora, the intensity of the 557.7nm line is, by definition, 100kR. The combined emission from the other strong emission bands in the V band produces an intensity of 5.5kR. The total flux from these auroral emissions, if distributed as a continuum uniformly across the band, would, at each wavelength, be equal to 1242 R/nm. If we add to this a quasi-continuum contribution of 270 R/nm, the emission from the auroral lines, bands and continuum for an IBC3 aurora is equivalent to 1512 R/nm when averaged across the V band. This results in a sky background intensity of 16.4 V magnitudes per square arcsecond.

Now consider a notch-filter with a FWHM of 10.0nm that reduces the emission from the 557.7nm [OI] line to negligible levels. The remaining bands and quasi continuum contribute emission equivalent to an intensity of 338 R/nm at each 2-5. Sky brightness in standard photometric bands 47

60 100 I 1985 Solar Minimum 50

40 Q) :.·., . 60 !E ~ 30 ::I .', .. ' ... ··. ,, ·'. ~ 0 ·.:., : ::· :: ~- : : 40 e ::I: 20 . . . ~ a,)

10 £L .. ~ . <:' ...... <~ :· 0 16 17 18 19 20 21 22 23 24 25 100 1990 Solar Maximum 50

40 . :. -~:>.> / so ...... · ..... •.. 60 ..!E ..... :.. ·.· .. ·. >.. ~ 30 :e C1) :::s . ' . ~ 0 40 . .. ~ .. '• ~ ·: : ' ...... ' . e ::I: 20 C1) 0.. 10

17 18 19 20 21 22 23 24 25 V magnitudes I arcsec2 Figure 2.9. Sky brightness (in Vmagnitudes per square arcsecond) at the South Pole for a standard V filter (solid grey line) and for a notch-filter that reduces the photon flux from the 557.7nm oxygen line to negligible levels (dashed line). The cumulative probability for the filter histogram is referenced to the right hand axis.

wavelength when averaged across the band. This results in a V magnitude of 18.0 per square arcsecond. In the case of this band the filter is able to decrease the background by 1.6 magnitudes. A filter with a notch attenuation of only 10 would suffice to reduce the 557.7nm to below 10% of the band total.

The South Pole 427.8nm photometer measurements have been scaled up using the relative intensities in Table 2.5 to Table 2.11 and used to calculate the probability distribution of the total V band emission for the winters of 1985 and 1990. This is shown by the solid grey histogram in Figure 2.9. The calculations were then repeated for the case with a notch filter at 557.7nm and this probability distribution is shown by the dashed line histogram in Figure 2.9. 2-5. Sky brightness in standard photometric bands 48

2-5.3 R band

It is clear from Figure 2.2 that in the R band removing the contribution from auroral lines will not be as simple as in the B or V bands. The forbidden transitions of oxy gen that produce the 630.0nm and 636.4nm auroral lines are less easy to categorise because their intensities vary quite widely depending on the energy of t~e incident electrons and ions striking the upper atmosphere. It is certainly the case that they are less frequent than the 557.7nm emission (Vallance Jones 1974; Omholt 1971), and of a lesser intensity than the lines in the blue and visible wavebands. However in addition to these lines the R band contains a number of strong emission lines of N2 as well as lines of atomic and molecular oxygen ( Omholt 1971). However, observa tions in the R band are always affected by numerous atmospheric lines, since many of them are produced in airglow processes above any ground-based astronomical site.

The best compromise in R band will depend on the observation to be made. As an example we consider the Vulcan South 'Transit Search project, currently at South Pole station. Vulcan South 'Transit Search (Caldwell et al. 2003) has a wide-field imaging telescope monitoring six square degree fields of sky, searching for extra-solar planets. The waveband they have selected to observe in has a FWHM of 60nm with a central wavelength of 675.0nm. This waveband constitutes a good compromise between avoiding the 630.0nm red oxygen auroral emission line and the increasingly dense distribution of night sky lines which begin at around 720nm.

From 640nm to 720nm the primary band contributions are from first order N2 excitations, with a small contribution from atmospheric molecular oxygen (which increases towards the 760nm atmospheric oxygen band).

Unlike the case of the Band V bands discussed previously there is little to be gained by removing a few of these lines when such a narrow passband is already in use. Sky background in this narrow band is almost completely dominated by auroral emission. This means that if an observation's signal to noise ratio is limited 2-6. Discussion 49 by the sky background, as opposed to other noise sources, the aurora will be the limiting factor in observations during peak auroral activity.

2-6 Discussion

To evaluate the effect on optical astronomy at Dome C, we now estimate the con tribution auroral emissions would make to Dome C sky brightness. Figures 2.10 and 2.11 plot the winter auroral intensity in the B and V bands at Dome C using the South Pole B and V band intensities in Section 5 and the relative auoral inten sities in Section 4. In B band, for 50% of the wintertime in a solar minimum year, Figure 2.10 shows that the auroral sky background is less than 23.7 B magnitudes per square arcsecond. In a solar maximum year, the auroral sky background is below 24.5 B magnitudes per square arcsecond for 50% of the winter. In the v; Figure 2.11 shows the sky background contribution in a solar maximum year by aurora is less than 24.7 magnitudes per square arcsecond for 50% of the winter. In a solar mim imum, the background in Vis 23.7 magnitudes per square arcsecond. The use of notch-filters, as discussed in Section 5, would reduce the auroral sky background contribution further still if required.

We now examine the science case for one of the proposed ELTs, the Giant Magellan Telescope (GMT), and ask what impact the aurorae would have on key projects. A second GMT has already been proposed for Dome C (Angel et al. 2003). The GMT design consists of a segmented primary mirror with an effective collecting area of 21.5m, and a 3.5 m Gregorian secondary. The operating modes for the GMT have limiting resolutions (from seeing measurements for Las Campanas) ranging from 0.4 arcseconds (natural seeing) down to 0.007 arcseconds for a full multi-conjugate adaptive optics system that uses natural and laser guide stars. Further details on the telescope and science priorities of the GMT can be seen at Giant Magellan Telescope homepage (2004). 2-6. Discussion 50

60 100 1985 Solar Minimum 50 80

40 Q) 60 :s e 30 c ~ 0 40 s...~ ::I: 20 Q) 20 10 c..

0 0 18 19 20 21 22 23 24 25 26 100 50 1990 Solar Maximum 80 40 Q) 60 :s e 30 c ::s Q) 0 20 40 ~ ::I: c..Q) 10 20

o+-~-r~~--~~~~~~liilgu~~~~~~ 0 18 19 20 21 22 23 24 25 26 B magnitudes I arcsec2 Figure 2.10. Calculated auroral contribution to sky brightness at Dome C (in B mag nitudes per square arcsecond). The cumulative probability is referenced to the right hand axis

60 100 1985 Solar Minimum 50 80

40 Q) 60 = e::s 30 c 0 - 40 ~ ::I: 20 Q) 20 10 c..

0 0 18 19 20 21 22 23 24 25 26 100 50 1990 Solar Maximum 80 40 Q) 60 :s e 30 c ::s Q) 0 20 40 ~ ::I: c..Q) 10 20

o+-~-r~~~~~~~~~~~4ll~~il4llll~ 0 18 19 20 21 22 23 24 25 26 B magnitudes I arcsec2 Figure 2.11. Calculated auroral contribution to sky brightness at Dome C (in V mag nitudes per square arcsecond). The cumulative probability is referenced to the right hand axis. 2-6. Discussion 51

Most of the science goals of the GMT focus on near-IR and IR wavebands, as these are wavelengths where high-order AO allows diffraction-limited observations. Beyond 800nm, where OH airglow lines begin to dominate a mid-latitude site, the sky background emission at Dome C is not yet characterised, though the atmospheric emissions that produce airglow are also observed in aurorae. The same methods that are used to remove airglow from observations can also be used to remove aurorae. A majority of the science projects that are guiding the specifications for the GMT place emphasis on high-resolution spectroscopy. Narrow auroral emission lines can be removed from these spectra, while the reduced slit-width of the instrument made possible by the better seeing condi tions at Dome C will further decrease the sky background. The science goals include:

a) Intergalactic medium: Distant quasar observations using high-resolution spectroscopy allow comparisons of the inter-galactic medium along different lines-of-sight. When observing quasars that are reasonably bright, at 15 to 17 V magnitudes, high-resolution spectroscopy should allow the easy identification and removal of the narrow auroral lines in the spectra.

b) Multi-aperture spectrographs: Redshift galaxy surveys are looking to collect continuum spectra of galaxies down to 26 I magnitudes, in 15 degree fields at intermediate resolutions. Slit-spectroscopy would allow better sky-subtraction than fibre-fed spectrographs, but auroral light in the continuum spectra could prove a limiting factor.

c) Integral field (IF) spectroscopy: IF spectroscopy of nearby galaxies can determine stellar populations, orbital structure and the kinematics around nuclear black holes. Spectroscopic mapping allows discrete auroral lines in the galactic spectra to be removed prior to integrating the field and is already carried out with bright sky lines. These observations should not be affected by aurora at Dome C. 2-6. Discussion 52

d) Faint-end AGN: Broad emission lines in AGN are black hole signatures which, even at modest redshifts, are lost in the light of their host galaxies. Using an ELT in a high resolution mode and masking strong telluric OH emission features, the spectra can be rebinned at lower resolution to improve the sensitivity of these weak broad lines. Removing the light from the host galaxy will be the primary limiting factor in these observations, not aurorae.

e) Resolved Stellar populations: Imaging in the near-IR with large aperture telescopes will extend the study of stellar environments that are currently confusion limited. These observations will be limited to the seeing that can be achieved at the site and on the AO capabilities of the telescope. Auroral sky background should not be a factor in such measurements.

f) Single object, deep field studies: Continuum measurements and imaging of faint objects may be affected by a fast-varying background and a large number of contributing sky lines in the band. Such projects include measuring abundances of the supergiants and studies of the abundance of red-giants in the local group. Long exposures would be affected by short time-scale variations in the background.

g) Extra-solar planets: Circumstellar debris disks could be probed down to scales less than lOAU with the GMT. Better seeing at Dome C reduces the sky background per pixel (as the pixel size can be scaled with the resolution) and means that the flux from the sky will be lower in seeing-limited observations.

h) Direct detection of young planets: An ELT allows direct detection of planets close to the parent star, using coronography and apodization, or nulling interferometry. Both methods require AO to reach the spatial resolutions required (60-120 milli-arcseconds in J band) at a temperate site, while at Dome C such 2-7. Conclusions 53 resolutions can be achieved in natural seeing. The light from the parent star is the fundamental limit to such observations and so additional sky background from aurorae will make little difference to these measurements.

2-7 Conclusions

At sites such as Skibotn and South Pole, intense auroral emissions produce an increase in sky background photon flux over all three visible wavebands (V,B,R). For photometric observations, the variability of the background over short time periods (Myrabo 1980) requires careful and frequent sky background measurements and subtraction. In spectroscopy, the positions of the emission lines may prevent observations of specific lines in sources, but otherwise should pose little problem.

For a bright IBC3 aurora in the B band a notch filter that eliminates the 427.8nm Nt vibrational band emission will reduce the auroral contribution to the sky background by 0.7 B magnitudes. In the V band, a notch filter at 557.7nm decreases the auroral contribution to the sky background by 1.6 V magnitudes. Specially designed passbands could give significantly darker backgrounds than the standard B and V bands. However, a fully optimised Rugate filter could yield great improvements, and this could form the basis for very fruitful future research.

Projects such as the Vulcan South 'Iransit Search have avoided the wavelengths of strong and variable auroral emission lines by using carefully selected filter pass bands.

A site such as Dome C, higher on the plateau and closer to the geomagnetic pole, will be more favourable for optical astronomy as the contribution to sky brightness by aurora is reduced by up to three magnitudes when compared to sites of high auroral activity such as South Pole. 2-7. Conclusions 54 2-7. Conclusions 55

Table 2.5. Line intensities for the observed forbidden atomic multiplets (Vallance Jones 1974). These intensities are calculated relative to a 557.7nm emission line of strength 100kR.

Emitter A Intensity(kR) [OI] 297.2 6 [OI] 557.7 100 [OI] 630.0-636.4 2-100

[OII] 373.7-372.9 1 [OII] 731.9-733.0 0.4-100

[NI] 346.6 1 (NI] 519.9-520.1 0.1-2 [NI] 1039.5-1040.4 6

Table 2.6. Line intensities for the observed allowed atomic multiplets (Vallance Jones 1974). These line intensities are calculated relative to a 557.7nm emission line of strength lOOkR.

Emitter A Intensity(kR) Emitter A Intensity (kR) or 374.9 1 NI 493.5 0.19 OI 397.3 0.3 NI 532.9 0.22 OI 436.8 0.5 NI 644.2 0.11 OI 441.5 0.5 NI 746.8 0.55 OI 459.1 0.13 NI 821.6 3.3 OI 464.9 0.6 NI 862.9 3.84 OI 595.9 0.13 NI 868.0 10.5 OI 615.8 0.14 NII 399.5 1 OI 645.4 0.14 NII 404.4 0.5 OI 700.2 0.3 NII 424.4 0.5 OI 777.4 9.6 NII 500.1 0.6 OI 799.5 1.2 NII 568.0 0.65 OI 844.6 11.5 NII 648.2 0.4 2-7. Conclusions 56

Table 2. 7. Band intensities for the M Nt Band (Vallance Jones 1974). These band intensities are calculated relative to a 557.7nm emission line of strength lOOkR.

A Intensity(kR) A Intensity(kR) A Intensity(kR) 504.8 0.05 785.3 65.49 1277 2.28 517.4 0.08 792.4 0.11 1326 0.8 530.5 0.08 808.1 38.91 1379 0.20 544.2 0.06 823.3 10.99 1460 36.67 552.9 0.33 857.8 2.50 1521 32.00 567.1 0.46 884.5 0.85 1581 10.06 581.9 0.37 912.7 0.02 1654 1.40 597.4 0.22 918.1 158.0 1728 0.05 612.3 2.11 947.0 64.43 2126 3.21 628.5 2.43 977.4 9.63 2240 5.77 645.5 1.55 1010 0.36 2363 3.84 663.3 0.74 1044 0.05 2498 1.40 681.9 0.29 1080 0.15 2648 1.40 687.3 14.22 1110 122.69 2813 0.08 701.6 0.10 1118 0.11 3856 0.08 706.4 10.50 1147 10.06 4186 0.19 726.3 5.23 1158 0.05 4578 0.17 747.2 1.84 1188 1.84 5016 0.08 769.3 0.51 1231 4.67 5552 0.03 2-7. Conclusions 57

Table 2.8. Band intensities for the lP N2 Band (Vallance Jones 1974). These band intensities are calculated relative to a 557.7nm emission line of strength lOOkR.

A Intensity(kR) A Intensity(kR) A Intensity(kR) 575.5 0.21 738.7 23.74 1051 96.36 580.4 0.43 750.4 48.70 1078 4.57 585.5 0.75 761.2 0.96 1117 1.03 590.6 1.28 762.7 63.56 1152 3.24 595.9 1.80 775.2 1.94 1159 0.97 601.4 2.36 775.4 53.61 1193 26.41 606.9 2.27 789.7 3.17 1237 54.88 612.7 1.76 804.8 3.01 1306 12.68 618.7 0.75 820.6 1.04 1363 12.68 625.3 0.29 837.0 0.90 1427 21.13 632.3 0.81 854.2 17.84 1489 0.89 639.5 2.14 872.3 74.50 1498 17.27 646.8 4.52 891.2 124.5 1571 5.20 654.5 8.93 898.3 1.35 1664 8.80 662.4 13.5 920.3 6.61 3 1706 0.97 670.5 18.2 943.6 17.130 1769 8.12 678.9 17.5 968.0 23.439 1887 3.81 687.5 8.80 984.2 1.58 1962 3.03 706.0 0.27 993.9 15.09 2121 3.27 716.5 2.24 1013 3.36 2306 2.00 727.4 8.36 1045 5.26 2-7. Conclusions 58

Table 2.9. Band intensities for the IN Ni Band (Vallance Jones 1974). These band intensities are calculated relative to a 557.7nm emission line of strength 100kR.

A Intensity(kR) A Intensity (kR) A Intensity(kR) 330.8 0.02 419.9 0.05 470.9 6.04 356.3 0.15 423.6 3.87 514.9 0.32 358.2 6.80 427.8 30.0 522.8 0.99 388.4 3.93 459.9 0.03 575.4 0.06 391.4 98.51 465.2 1.34 586.5 0.14

Table 2.10. Band intensities for the 02 Atmospheric Band (Vallance Jones 1974). These band intensities are calculated relative to a 557.7nm emission line of strength lOOkR. The 761.9nm(O,O) transition is not observable from the ground because of self-absorption by ground state 0 2 (Vallance Jones 1974).

A Intensity (kR) A Intensity(kR) A Intensity(kR) 596.0 0.02 761.9 1176.1 906.2 1.63 628.9 0.05 770.8 13.96 918.0 1.93 637.1 0.38 780.2 6.79 996.8 1.44 645.7 0.29 790.1 11.53 1007 0.07 654.8 0.56 800.3 3.04 1018 0.09 688.2 1.66 811.1 2.25 1030 0.33 697.0 1.93 864.5 58.00 1042 0.19 706.0 6.09 874.2 1.61 1055 0.30 715.4 2.75 884.4 1.40 1173 0.02 725.3 3.44 895.1 3.89 1235 0.02 2-7. Conclusions 59

Table 2.11. Band intensities for the lN ot Band (Vallance Jones 1974). These band intensities are calculated relative to a 557.7nm emission line of strength lOOkR.

A Intensity (kR) A Intensity(kR) A Intensity(kR) 498.7 0.2 632.1 0.38 785.4 0.33 525.2 0.90 638.9 3.47 799.1 0.04 527.4 1.884 665.7 0.08 813.7 0.12 554.2 0.06 673.5 0.86 829.8 0.21 557.3 1.08 682.2 1.90 847.5 0.12 560.8 5.90 700.5 0.02 857.2 0.02 585.7 0.23 709.7 0.19 875.6 0.07 590.0 0.65 719.7 0.74 896.0 0.09 594.7 0.23 730.7 0.85 918.3 0.04 599.9 3.98 747.1 0.04 945.9 0.03 620.4 0.07 758.7 0.18 971.6 0.04 625.9 0.02 771.4 0.43 1000 0.02 Chapter 3

Cloud cover at Dome C

3-1 COBBER

In this chapter, the design, testing and results obtained from COBBER (Cloud OBserver) are presented. COBBER was one of two instruments designed for automated observations of the winter cloud cover over the Antarctic plateau and was first deployed to Dome C in the summer of 2002. The design and results of this instrument are published in Dempsey et al. (2003).

The work completed as part of this thesis included the mechanical and optical design of the instrument, all electronic component and circuit testing and data analysis. John Storey designed the COBBER electronic circuits and raytracing of the lens and detector optics was completed by Jon Everett. Paolo Calisse, John Storey, Jon Lawrence and Tony 'Iravouillon installed COBBER and replaced electronic and optical parts at Dome C in the summers of 2002-2004. 3-1. COBBER 61

Figure 3.1. COBBER in tailed at Dome C, January 2002. Photo courte y of Paolo Calisse, 2002.

3-1.1 COBBER design

COBBER was fir t designed in late 2001, to add to the complement of instruments on an Au tralian Antarctic Division automated weather station {AWS). At the time an expedition to Dome A was planned for the approaching summer, and this first journey to the highest point on the Plateau was seen as ideal for the installation of an AWS to log the conditions on the dome, 4100m above sea level. A pivotal piece of information for astronomy, the percentage of cloud cover during the austral winter, could not be provided with the existing AWS system o COBBER was designed to fill this role.

COBBER's basic design was modelled on an instrument built by Clay et al. (199 ). The most ignificant alterations to this design were to adapt the instrument for the extreme temperatures of a Dome C winter (temperatures down to - 0°C) and the long period without any handling or maintenance. All circuit components were carefully elected and rigorou ly tested in a laboratory freezer down to temperatures below - 0°C. The TLC2201 op-arnp s excellent tability in freezer tests and an offset drift of only 0.5J-LV ;oc ensured satisfactory low-temperature operation. Likewise the Perkin-Elmer TP 534 detector was submitted to extensive 3-1. COBBER 62

P4 Ll 100 33ul! .rtm +SV "' I 1_ cz Io.lur C4 O.lul i ' .. 111~~17

I I Ul I I !LC2201 uz TLC2201 ~ 16 RS -4 .f$ 100 !Ok Cl ~ 17 Signal lo, Dl O:f;':' ll! . ~ !'1115817 lcs II lOUJ iheaopil dotector lU 40k C3 ,Signal P3 return ll! 0~1~ x~SS34 !

RZ iS !Ok 180k '

I 20 metre <------COBE!R BOWS ------> <-- 7-> Paver supply -:- i cable 1 and ~ignal board

Figure 3.2. Circuit configuration of COBBER, as installed at Dome C, January 2003. cold-testing to determine the drop in responsivity, if any, and performance as a function of detector, and source, temperature.

The ZnSe lens served a dual function in the design: firstly to limit the in strument's field of view to 30° to eliminate the sun and secondly, its smooth anti-reflection coated hemispherical surface acts as a deterrent to snow build-up on the instrument. The thermopile detector has an integrated IR filter with a passband of 8-12 f-Lm. Twenty-metre long cables connect the device to the power supply and DC/DC converter, buried in a crypt beneath the ice. This power supply, which also powered the ICECAM experiment described in the introduction, consisted of 5 kilograms of lithium thionyl chloride batteries. Powering up once every two hours for only a few minutes, the electronics were designed to have a settling time of less than a minute, and very low power consumption.

A sketch of the original electronic layout of COBBER can be seen in the Figure 3.2. The overall DC gain is 600. While the operational amplifiers and other electronic components were tested in the freezer, the final set-up including the lens 3-1. COBBER 63 was not tested under laboratory conditions as a result of lack of time. However, the maximum offset drift of the op-amps can be calculated, with the view to assessing its contribution to the offset voltage seen in the COBBER data. The max imum offset drift of the TLC2201 op-amps is 0.5J.tV /degC, which for a !::,.T of SOC (from 20C to -60C) results in only 12mV of offset at the output of the x600 amplifier.

3-1.2 Testing

The UNSW Antarctic group owns a Forma Scientific chest freezer (Model 8558, Forma Scientific Inc. (1994)), which can be cooled to -86°C, and has internal dimensions of 168.9cm x 71.7cm x 47cm. The freezer is vital for testing instru mentation prior to deployment to Antarctica to ascertain the material properties or device response at temperatures below -40°C.

To test the low temperature response of the TPS534 thermopile device, the de tector was mounted on a small circuitboard in a shielded tube 10cm in length. The detector observed the infrared emission from a 333n surface-mount resistor mounted on a board at the other end of the tube. A 15V DC voltage was applied across the resistor to heat it, resulting in a power output of approximately 1 Watt. This test apparatus was placed in the freezer, and then the temperature of the freezer was set to -80°C. Outside the freezer, the detector was connected to a small amplifier circuit and was run from a ±5V supply.

Once the freezer had been allowed to stabilise at its set temperature (a process that can take up to two hours) the resistor was turned on and the response of the detector recorded in millivolts. The resistance of the internal thermistor in the detector was also measured. A thermocouple close to the resistor was used to measure the air temperature directly in front of the heat source. Once the detector voltage was recorded, the freezer would be allowed to warm by approximately 5 3-1. COBBER 64 + +

0.95

0.85

0.8 + -80 -60 -40 -20 0 20 T(detector) degC

Figure 3.3. Laboratory results; TPS534 signal responsivity as a function of the detector temperature, normalised to the responsivity at 20°C. degrees, and left to stabilise again before another reading would be taken.

The results of the laboratory cold testing of the TPS534 thermopile detector can be seen in Figure 3.3. This plot shows the responsivity of the detector as a function of the detector temperature, normalised to the responsivity of the detector at room temperature (T = 20°0). When the thermopile detector was cooled to -80°0, there sponsivity of the device decreased by 20% in comparison to its performance at 20°0.

As the thermopile detector was cooled, the performance of the internal thermistor of the detector was also monitored. Its resistance as a function of temperature can be seen in Figure 3.4. The results at room temperature agree well with the manufac turer's measurements and, as can be seen, the thermistor has very high resistances at low temperatures. To simplify the electronic design of the device the thermistor was not used to log the ambient temperature of the detector. A Dallas tempera ture sensor in the ICECAM instrument (which was located close to COBBER as can be. seen in Figure 1.8) was instead used to calibrate the COBBER measurements. 3-1. COBBER 65

TPS534 thermistor resistance vs Temperature 100 ;------

iii' • Freezer Measurement E 10 ... J: .... 6 Manufacturer ~ • B .. c .m .. .!!! 1 • ...e ~ ~ ~ 0.1

0.01 '------·80 ·60 -20 0 20 40 Temperature (deg C)

Figure 3.4. Laboratory results; TPS534 internal thermistor resistance as a function of temperature (blue squares), as compared to similar measurements provided by Heinmann (red triangles).

3-1.3 Results

COBBER was first deployed to Dome C in the summer of 2002. Unfortu nately, the first measurements recorded by COBBER after station close showed no sensitivity to cloud cover and it became clear that an electronic fault had occurred. As a result, no cloud cover data was collected by COBBER in this first austral winter at Dome C. When the station reopened in late 2002, inspec tion of the wiring revealed that this fault was caused by a disconnected ground wire.

A second set of COBBER electronics was manufactured in 2002 and was deployed in the austral summer of 2003. After installation on January 17th, 2003, COBBER 3-1. COBBER 66 took a measurement once every two hours, and returned the data via an ARGOS satellite link 12 times a day.

Figure 3.5 shows the full set of data up obtained from January until the middle of April, 2003. The large scale daily variation is due to the diurnal cycle of the sun heating the instrument and its detector, while the daily clear sky temperature remains constant. To show this, the ambient temperature has also been plotted in Figure 3.6. If the detector temperature is expressed as Tdetector and the sky temperature is given by Tsky then COBBER's signal, Vsignal can be described simply as:

"Vsignal CX: (Tdetector - Tsky) (3.1)

As described by Equation 3.1, a thermopile detector produces a voltage propor tional to the difference between the temperature of the detector and the temperature of the sky in its field of view. The clear sky has an effective brightness temperature

(at ~190K) that is lower than the ambient ground temperature at Dome C (average

Tno = 230 K). Clouds absorb and emit infrared radiation. A cloud-covered sky has therefore a higher brightness temperature than a clear sky. When the sky is over cast, the sky temperature T sky is equal to or warmer than the ambient temperature

T detector. A zero level voltage thus indicates a completely clouded day (T Detector

:::; Tsky), whilst a large positive voltage indicates clear conditions (Tnetector ~ Tsky)·

In the full data set to the middle of April 2003, only four days of complete cloud were recorded by COBBER. To confirm this encouraging result, the days in question were compared to images from the AASTINO webcamera, which showed that these days were indeed overcast. The two top plots in Figure 3.6 show an example of a clear day, in comparison to a cloudy one, with their corresponding ambient temperatures.

As winter approached at Dome C the ambient ground temperature, and therefore 3-1. COBBER 67

0 0>

0 CXJ

....--... (/) >- 0<( r--..0 '---"' w 2 I-

0 tO

0 11")

0 I") tO I") 0 0 c:i 0 (/\) 3~\fllO/\ Figure 3.5. Full COBBER data set from January to April 2003. Even without removing the diurnal variations caused by the sun heating the instrument, the four recorded days of cloud (Vsignal = 0) are clearly seen. 3-1. COBBER 68

-20 .. -20 uc:n s ...... , w ...... "0 -4tr" ~0.5 a.. ~ ::::;; w §;> 1- -60

0 '-'-~...... L~~...l..-o..~...... J.-'-'~~ 0 38 39 40 41 42 48 TIME (days)

0.4

&o.3 (!) !:§ o0.2 >

0.1

49 50 51 52 84 86 TIME(doys) TIME(days) Figure 3.6. Top Left: COBBER signal voltage (measurements every two hours) for four clear days (solid line) and the corresponding T ambient in degrees Celsius (triangles). Top Right: COBBER signal voltage for two clouded days (solid) and the corresponding Tambient (triangles). Bottom Left: COBBER signal (as shown in top right plot) measured in February (T ambient = 235K) and Bottom Right: COBBER signal in early April (Tambient = 220K). Overcast conditions are still detectable.

the detector temperature, decreased. One detrimental effect of the declining Tambient was that the ~ T between a clear, cold sky and the detector was reduced. This is shown by the drop in Vsignal in Figure 3.5. The bottom two graphs in Figur'e 3.6 compare four days of data taken in early February (bottom left) with four days in early April (bottom right). Though the signal strength in the April data is weaker, overcast conditions can still be detected by a sharp drop to Vsignal = 0.

Beyond mid-June, the gain of the instrument proved insufficient for measurement of sky/cloud variations. In the summer of 2003/2004, the gain of the COBBER circuit was increased to 1000, allowing an increase of the average voltage on a clear sky to approximately 2.8 Volts in mid-summer. The results of the instrument as logged in March and April of 2004 are presented in Figure 3.7. A fault in the AASTINO power supply prevented any further measurements in the winter of 2004. 3-1. COBBER 69

Cobber AASTI!ID - Dcae C, .lntarctica - ~inte: 2004 2.0·.-,-..,--.--,--,--,---,--,---,._:_...,,..--,-.,--.---.--,--,--,--,---,-_:_...,-1...---n 2. 0 I I

I •• •• •• • • • t nl ~ • • ,., t,\t t t t ;;. 1,, • t• ••t + \ •• + f 1..1 tt tt I /• • t ,• •'• , , • ', \ ! ,\ / r-1 1.0 • \tt • ... • •• t t t t t t ~,: : + \t. t.:\ f +\. ,.t 1. 0 ~ t l \ •• t ... t t t: tit f t /t+ t t,• \ \ ·~ ' •• t •• \.\\' ,\t ''\"> • •• .. t\¥.1! • 5 + t t~t... + t •1"1 I t • Cll t

\ t

o. 0L--.L...-..I..-..l.-...I--'---1--J....--L---l.--li--l-..l..-..l..-..l.-...l--'---1---'----l.--l--L...-....llI I I o. 0 24 Mar 30 liar 6 Apr 13 Apr urc fGC, u Apr ~~~ ll:ll:ll m Figure 3.7. COBBER data set from late March to April 2004.

COBBER has produced some useful and informative results during its current installation at Dome C. In the summer months its results can be confirmed by the webcamera currently working at the site; however once the sun sets, only COBBER and ICECAM are able to detect clouds. An expedition to Dome A will be undertaken in the austral summer 2004/2005. Future versions of the detector are planned for an Automated Weather Station at Dome A and other potential astronomical sites in Antarctica, and development and testing continues to improve the device for this purpose. Chapter 4

Instrumentation

4-1 AFOS

Antarctic instrumentation poses two sets of challenges to the designers and observers. It requires specialised design for the extreme cold temperatures that are otherwise only experienced by space-based instruments. In addition to working at temperatures averaging -60°0 during the winter, Antarctic instruments must either prevent or withstand the formation of ice on apertures, windows and moving parts. Finally, these instruments must be well tested so that they require minimum maintenance during the austral winter.

The AFOS was a pioneering telescope in this area. One of the goals of this project was to complete the evolution of AFOS into an entirely automated, low-power Antarctic instrument. A number of sites on the high Antarctic plateau promise excellent observing conditions but do not have the necessary infrastructure to support habitation for any extended length of time. It is therefore essential to design low-maintenance, low-power instrumentation which requires little or no manual intervention for months at a time. Achieving successful operation of such a telescope demonstrates the feasibility of Antarctic sites for large-scale telescope projects. 4-1. AFOS 71

At the beginning of 2002, when the work that is detailed in this thesis was begun, the AFOS was already fully constructed and deployed at South Pole station on a dual-telescope alt-az mount. A small data set of spectra had been collected of the June 2000 total lunar eclipse. No observations had been successfully completed with the telescope in a fully operational state, and there had been no assessment or study of the operation of the telescope since its deployment on the new mount.

The observing techniques, data reduction and results will be presented in later chapters. This chapter firstly describes the telescope, instrumentation and software configurations as they existed during the observations in 2002 and 2003. The work undertaken in this thesis and detailed in this chapter includes:

• The investigation of the fibre-optic alignment and spectral response

• Analysis of the mechanical properties of the tower and mount including sinking and flexure

• Trouble-shooting the pointing of the telescope

• Investigation of the cause of wavelength-dependent losses seen in the spectra

4-1.1 Telescope

The design and construction of the AFOS are detailed in Boccas et al. (1998). A Newtonian optics system was chosen to minimize complexity where Cassegrain optics would have been expensive and complicated, and a prime focus system too large and unwieldy. The most stringent design limitation was the need for the telescope to survive the extreme temperature range that it would experience (20°C to -80°0), without compromising the precisely aligned optics. Without athermalisation of the design there is risk of both defocussing and mechanical strain on the optics. 4-1. AFOS 72

Invar 36, with an almost negligible expansion coefficient of a = 0.9 x 10-6 K-1, was used in the construction of all the mechanical parts, while the primary parabolic mirror was formed from Astrosittal, a low thermal expansion glass (a = 3.2 x

10-6 K-1). These selected materials provide sufficient athermalisation for the AFOS to successfully operate at both potential test sites in Australia and also in the sub -sooc Antarctic winter.

70.0

e::6o.~ o

Av s

30.0

20.0

10.0

Incident angle= 20 degrees 0 · 0 J-__,.....:.:...... ,_..~o.;::lo ....z;;..;__so,_o---:-6.,..oo_..;;;..~7:"!"0o:----a:"!o~o--~goo Wavelength {nanometres) Figure 4.1. Beamsplitter transmission response as a function of wavelength for the P state polarisation (top), Average polarisation (centre) and S-state polarisation (bottom). Manufacturer's data.

The primary mirror has a diameter of 318mm while the secondary is an elliptical flat mirror with a major axis of 108mm. Both were given UV-enhanced coatings of AlMgF 2 to yield >85% reflectivity from 300 to 850nm. The primary has a focal ratio of f/3.35 and is focused into an Invar injection module. A dichroic beamsplitter (R2:: 50% from 300-550nm, and T2:: 50% from 550-850nm), directs the light into two fibre bundles, one optimised in the red and the other in the blue region of the spectrum to maximise the performance of the fibres across the 4-1. AFOS 73

llFOS "Ins-trument. Sigunt.urc 1

.E...... 6 ~- ~... .~t ~ ~

0 3000 .•1000 5000 6000 7000 3000 0000 'l'fnvclcnglh (ang,.troms)

Figure 4.2. Quartz-halogen lamp spectrum, in the blue and red AFOS fibres, divided by a blackbody function at the same temperature as the Quartz-halogen lamp. The quotient thus represents the instrument signature of the AFOS, describing the transmission of all the optical components in the light path from the window to the CCD camera. The Quartz halogen lamp is at a temperature of approximately 3100K. The transmission percentages are normalised to the peak transmission value. range of the visible wavelength band. The manufacturer's plot of the spectral transmittance of the beamsplitter is shown in Figure 4.1. A quartz-halogen lamp positioned in the AFOS telescope barrel was used as the standard flat-fielding source. Figure 4.2 shows the red and blue fibre spectral response of AFOS, obtained by dividing a spectrum of the quartz-halogen lamp by a blackbody function at the same temperature as the lamp filament (TQH ~ 3100K).

The beamsplitter also acts as the order-sorter for the spectrograph. When a grating spectrometer is adjusted to transmit a given wavelength A in the first order, a certain fraction of radiation is also transmitted from wavelength .\/2 in the second order, and .\/3 in the third and so on. To prevent this radiation from adversely affecting the measurements, the unwanted orders must be filtered out. Passband and interference filters are the most common method of order-sorting in the visible range (James & Sternberg 1969). In the case of the AFOS, the beamsplitter prevents blue radiation in the second and third order wavelengths from reaching the red fibres and the spectrometer. 4-1. AFOS 74

4-1.2 Fibre Optics

Fibre optics are a favoured, and now common design feature in astronomical spec troscopy (Barden 1998). Particularly for multi-object and wide-field applications, fibre optic transmission provides versatility and stability (Vaughnn 1994). In the particular case of the AFOS, fibre optic transmission provided us with the ability to house the delicate and expensive detection system consisting of a spectrograph and a 1024 x 256 pixel CCD camera within the protected environment of the AASTO, while the telescope could be positioned up to 40 metres away, on a 7 metre tower, to avoid local thermal emission and ground turbulence.

Fibre optics add a delicacy to the instrumentation. Damage and stress on the 45 metre length of fibres, or inefficient coupling to the instruments at either end, could decrease sensitivity by several orders of magnitude. Careful design and installation reduces these risks and removes the need to heat any part of the telescope or mount.

Wave~gth fnm) Figure 4.3. Transmission response of Optran UV fibre as a function of wavelength. The dotted lines indicate the attenuation, A, (in dB/km) that would allow 99.9% (A=4) and 99% (A=40) transmission through a 1 metre length of Optran UV optical fibre(CeramOptec Industries Inc. 2003).

The AFOS was required for observing deep into the blue and UV bands, but observations of atmospheric absorption required reasonable transmission in the red wavelengths as well. Thus two Polymicro fibre types were chosen, one 'blue', (FV series, Polymicro Technologies, 1994), and one 'red', (FI series, Polymicro 4-1. AFOS 75

OptranWF

WiNel~flaih lnml Figure 4.4. Transmission response of Optran WF fibre as a function of wavelength. The dotted lines indicate the attenuation, A, (in dB/km) that would allow 99.9% (A=4) and 99% (A=40) transmission through a 1 metre length of Optran WF optical fibre (CeramOptec Industries Inc. 2003).

Technologies, 1994, ultra-low OH, and thus poor UV transmission). During the summer of 2002, these sets were replaced with more durable fibres from Ceramoptec Industries. Again one set of three fibres was selected for their UV transmission (UV 100/140P CeramOptec Industries Inc. (2003)), with operating wavelength from 160 to llOOnm whereas the second set optimised for visible and red wavelengths (WF 100/140P CeramOptec Industries Inc. (2003)) with an operating wavelength between 350 to 2400 nm. The spectral transmissions of these fibres are displayed in Figure 4.3 and Figure 4.4. Figure 4.3 shows that the attenuation of the UV fibre is 1000 dB/km at 160nm, which would transmit only 0.003% of the input signal through at 45 metre length of fibre. The effective wavelength range of the fibres was therefore reduced as a result of the length of fibres used in the experiment.

The dichroic beam-splitter injects light into the two sets of fibres, the red and blue optimised bundles, each consisting of three fibres with 100,um diameter fibre cores. To increase the ease of subtracting sky brightness, separate fibres were provided for the observation of the star and the sky background. Both the blue and red fibre bundles consist of three fibres, one to collect the light from the star and two bracketing the 'star' fibre, to measure the sky background. The bundles are aligned and set at 90° to each other in an X-shaped configuration as depicted ------

4-1. AFOS 76

...... ,.. . Figure 4.5. X- haped projection on the sky of the two fibre bundles. The central ' tar fibre of the two bundle overlap while the four remaining 'sky fibres aid in aligning the star and for sky subtraction. in Figure 4.5. The blue and red 'star fibres are positioned so that they overlap the same position on the sky to collect the tellar flux simultaneously.

100~-tm core diameter fibres were selected as a compromise between a small fibre, for high spectro copic resolution, and a large fibre, to maximi e the field of view on the sky to make target acquisition easier. The plate scale of a telescope, in arcseconds per millimetre is given by:

S= 206265 (4.1) f where f is the telescope focal length in millimetres (Chapman & Nations 1996). For the AFOS, with a focal ratio of 3.38 and primary mirror diameter of 319mm, the focal length f equals 1078mm. The resulting plate scale S is 191.3 arcseconds of sky per millimetre. So each 100~-tm diameter fibre subtends a 19.13 arcsecond field of view on the sky.

The e bundles were placed in Teflon tubing for protection, but this casing is loose to allow for thermal expansion of the fibres. Teflon cable has reasonable flexibility 4-1. AFOS 77 at Antarctic temperatures and was selected to avoid any stress on the fibres which would cause an increase in focal ratio degradation, and flux loss through the fibres. In addition, the Teflon assembly was then threaded through steel-braided cable to insure against any rough treatment during removal or assembly.

4-1.3 Spectrograph and CCD

The six fibres are aligned single-file in a grooved holder which creates a pseudo-slit for the spectrograph. The axes of the fibres were made parallel with a precision better than 3' by careful machining of the grooves. A commercial imaging spectro graph from Jobin Yvon (1989, model CP200, Jobin Yvon Inc. (1993)) was selected for image collection from the fibres. This model has a fixed concave holographic grating of 200 g mm-1, and was chosen for its stability and excellent stray light

4 rejection ratio of 1 part in 10 . One problem that can be caused by optical fibres is a 'speeding up' of the beam, and this phenomenon is called focal ratio degradation (FRD) (Ramsey 1988). The entrance pupil was matched to the calculated focal ratio that would exit the fibres given the measured focal ratio degradation, which is f /2.9.

A second problem results from the fact that the fibre subtends a large angle on the sky, and so a stellar image will not fill the whole input face of the fibre. The radial modes of the fibre are not completely scrambled for input beams faster than f /5, even over a fibre length as long as 45 metres. This means that at the output face of the fibre there will still be some memory of the image position on the input fibre face (Watson & Terry 1995). If the image is centred on the input face, there will be a centred spot on the output face. If, however, the image is progressively decentred on the input face this will result in a ring of increasing diameter on the output face.

As the output face of the fibres is used as a pseudo-slit for the spectrograph, decentering of the image on the input fibre face will therefore change the effective 4-1. AFOS 7

0 u u

0 f-< Cl) · ~ ~ I · ;; · .S ,__ ------'

~ ~ I> ... ! .b ~ 1 h c u e c :: u c;: c u .. 0 0 _g ...... u ] l ·" 01 ::s Q !.. s ~ ... c.. L u c .... ~ t.::

Figure 4.6. The light path through the telescope and fibres to the CCD camera. 4-1. AFOS 79 slit width of the spectrograph. Thus a decentred image will increase the effective slit width and decrease the resolution of the resulting spectrum. The following calculations assume that the output image is 100,um which is the width of the fibre core. For a well-centred image, the spot size (and therefore effective slit width) will be smaller than this value, and the corresponding resolution will be higher. The following calculation therefore indicated the lower limit on the AFOS spectral resolution.

The spectrograph focal length is 190mm and creates a flat image of 25x8 mm. The spectrograph was required to have best possible transmission in the UV, and resolution good enough to be limited only by the pixel size of the CCD array. The reciprocal linear dispersion, D, of the spectrograph is given by the following equation (Jobin Yvon Inc. 1993):

D = 1 x 106 cos(3 [nmjmm] (4.2) f g Where (3 is the angle of diffraction as measured from the normal to the grating, f is the focal length of the spectrograph camera (in this case effectively equal to that of the collimator), and g is the reciprocal grating line spacing in grooves/mm. The resulting reciprocal linear dispersion for the Jobin Yvon spectrograph was calculated to be 24.0 nm/mm. Now the resolution of the system can be described as the product of the reciprocal linear dispersion, D, and the slit width, d, which in this case is the diameter of the optical fibres, 0.1mm. The resultant resolution, R, is given by:

R - Dxd (4.3)

- 24 X 0.1 (4.4)

- 2.4 nm (4.5)

The CCD camera, and controlling software, is an Instaspec IV (Oriel Instru ments, 1995 Oriel Corporation (1990)). The format of the CCD is 1024 x 256 4-1. AFOS 0 pixels each pixel is 27 x 27 J-Lm in area, resulting in a total active area of 27.6 x 6.9 mm. The spectrograph re olution element is therefore oversampled with approximately four spectral pixel per resolution element. However,The CCD output is digiti ed to 16 bits. The diagram in Figure 4.6 de cribes the light path from the telescope to the CCD.

4-1.4 Generic mount

Figure 4. 7. The AFOS telescope (on the right) and the ADIMM(left) installed on the G-mount, summer 2002. Photo courtesy of Michael Ashley, 2002 .

The AFOS was deployed in 1997 on a temporary mount for initial te ting, while awaiting completion of the Generic mount (Hovey et al. 1998) as een in Figure 4. 7. The G-mount was designed and built by Gary Hovey, Ralph Sutherland, Mark 4-2. Testing and Modification 81

Jarnyk and their team at the Research School of Astronomy and Astrophysics (RSAA) at Mt Stromlo Observatory. Designed for extremely low-power and high precision the G-mount is a dual-telescope altitude-azimuth mount that simulta neously cradles the AFOS and another Mt Stromlo instrument, the Antarctic Differential Image Motion Monitor (ADIMM). First deployed during the summer of 1999/2000, the mount and its two telescopes were assembled at South Pole and placed on a 7m tower approximately 10 metres from the AASTO building.

In the first year of operation, the mount's altitude axis seized, and prevented data collection for nearly all of the winter season of 2000. Early in April 2000 the mount was manually moved to a fixed altitude, and as the azimuth axis was still functional, the AFOS collected its first data set capturing the lunar eclipse in June 2000. In the summer of 2000/2001 repairs were completed and the mount has operated successfully, though with some intermittent electronics problems, in the winters of 2002 and 2003.

4-2 Testing and Modification

The AFOS has no pointing camera to improve the ease of centering the telescope on a star. The companion instrument ADIMM can be seen to the left of the AFOS in Figure 4.7. The ADIMM has a wide-field camera and an automated pointing routine that had been well-tested with the G-mount prior to deployment. Further details on the ADIMM instrument and results can be seen in Travouillon et al. (2003). Once a pointing sequence was undertaken by the ADIMM and a new pointing model entered into the mount's software, a single fixed offset in altitude and azimuth should centre the star in the AFOS aperture. However, the first spectra taken with the AFOS revealed that when an exposure was taken with the star more than 5 arcseconds from the centre of the fibre, strong wavelength-dependent atten uation was seen in the spectrum. This shall be discussed further later in the chapter. 4-2. Testing and Modification 82

It was decided that a thorough instrument and telescope analysis should be con ducted at the very beginning of the 2003 winter season, and the results of these tests are presented here. The observing and software modifications adopted as a result of this analysis allowed a higher quality and quantity of spectra to be taken through the 2003 winter. In addition a limited number of fully-automated observations near the end of the season were successfully collected.

4-2.1 Tower tilt and sinking

The ADIMM telescope was programmed with an extensive, 300 star pointing program. During the previous years of operation (2001-2002), Mt Stromlo Obser vatory ran the mount control, data collection and pointing programs. The pointing software was run only three times during 2002, being time-consuming both in telescope time and for the person analysing the run. The program T-POINT was used to calculate the telescope pointing model (Wallace 1998).

Once this pointing model was derived, we would be provided with the name of the file to enter into the G-mount. This model would then control the telescope pointing for the AFOS, after entering a single azimuth and altitude offset to allow for the physical separation of the ADIMM and AFOS focal planes. While drift was seen in the pointing, it was not until we collected the first pointing data of 2003, nearly six months beyond the previous model, that a strong trend was seen in the pointing logs.

The G-tower is a hexapod configuration, originally constructed during 1997. By 2002, nearly half of the tower was embedded in packed snow. This was presumed to actually have a stabilising effect on the tower and mount, and this is partially true. However the first pointing model collected in June 2003 showed that the tower was sinking in a tilted fashion towards one particular leg. The plot of this trend is displayed in Figure 4.8 (the 'west' and 'north' indicated in the plot are constructs 4-2. Testing and Modification 83

lsqfit y=-1 .487x -215.079

-600

~ '-' ~550 ~ ~ -=t: ~500

-450 September 2002

-400 ~~--~~~~~--~~~~~--~~--~~~~~~~--~~ 150 200 West (orcsec)250 300 Figure 4.8. Drift of centre of pointing towards the telescopes north and west axis as a function of time. of the pointing program).

It was reasonably assumed in 2002 that a pointing run every month or so should be sufficient and for the large field of view camera on the ADIMM, it was. Unfortunately this sinking, amounting to approximately 5 arcseconds of offset per week, was significant for the AFOS. As initial analysis showed that even an offset of 5 arcseconds from the centre of the AFOS fibre caused attenuation the stellar spectra, it was concluded that a pointing run would have to be conducted at least once a week to take into account the tilt.

4-2.2 Telescope flexure

Even with weekly pointing corrections, daily adjustments were still required when acquiring targets with the AFOS. The standard AFOS aperture offset was a preset value of D..AZ = 602.4", D..ELT = 355.6", determined by calculating the average value of the azimuth and elevation offsets required for each object in the first observing runs of 2003. The nine sources were selected for their large range of zenith angles and while objects near the zenith centred quite easily in the fibre, objects closer to the horizon required large additional offsets even when the pointing model 4-2. Testing and Modification 84 had been updated on the day of observation.

640

~620 c:: c:J 0 u ... f.:-~8--± .. 1.Q .•9.!".

"'~ ~600 AVERAGE AFOS APERTURE ~ 0 t:;:; ········------·--·--·---·------·------·------·------' ,-'------······ • AVE - 1 0 orcsec:: X t£0 580

560 -

0 5 10 1 20 AZIMUTH ~DEG) Figure 4.9. Daily offsets (arcsec) in azimuth required to centre each object, with respect to the average AFOS aperture offset b.AZ = 602.4". AVE+10" and AVE-10 "define the edge of the fibre aperture from this average.

400

= c:J 38o r- c:J c:J ~ c:J 1360"' ;-~~!'::~f.'.~~+_!_!?~:;>-~_c ______., ...... ~---··'$·····------.q}················

~ ~A~V~ER~A~GIE~A=FO~S~A~P~E~RT~U~R~E------~~xr----~~~Q------· ~ ~ 1- ··"AVERAGE"-::-;-a··orcsec------·------·················

~340 1- l:l C>

c:J 320

300 30 40 0 70 - - ~~TITUDE (DEG~ - Figure 4.10. Daily offsets in elevation required to centre each object, with respect to the average AFOS aperture offset b.ELT = 355.6". AVE+10" and AVE-10 "define the edge of the fibre aperture from this average.

The daily azimuth and elevation corrections for each object were recorded during the observing season and used to conduct an analysis of any trends in the data. Figure 4.9 and Figure 4.10 show the daily focal plane offset required to centre the object once the pointing model had been entered. The symbols indicate six separate days of observations over a three month period. For reference, the average AFOS aperture offsets are plotted, as well as the width of the fibre from this centre, 4-2. Testing and Modification 85 as projected on the sky. It can be seen from Figure 4.9 that the average AFOS aperture offset in azimuth was usually accurate enough to acquire each object in the star fibre, though small adjustments might be required to optimise the signal.

In the elevation axis, however, the offset had a strong dependence on the elevation of the star above the horizon. Figure 4.10 shows that this offset varied by nearly an arcminute between a star at a declination of -30 and one at -70. This variation was clearly a result of flexure in the telescope and mount while pointing low in the sky, and little could be done during the winter observing period to improve the telescope stiffness.

4-2.3 Pointing improvements

The instrument analysis revealed the sources of difficulty when trying to program remote observing runs. While the tower sinking was a slow and linear deviation which could be removed by frequent pointing runs, another solution was required to combat the flexure of the telescope as a function of elevation. It was fortunate in this case that a routine existed in the G-mount software to set artificial 'apertures'.

A focal plane offset could be typed at the command line of the G-mount, in Azimuth and Altitude, for example Az: 320" Alt: 45" and called 'Aperture 4'. Initially a single "Aperture AFOS" was programmed (as mentioned above this offset was D.AZ = 602.4", D.ELT = 355.6") and each star's position would be optimised with incremental offsets from this central position.

Now it seemed necessary to derive an entire set of apertures, for each stellar object, so as to remove the need for manual offsets to centre each star from a common aperture. It required several observing runs to find acceptable offsets for each object, and after a new pointing run it was often necessary to 'tweak' each 4-2. Testing and Modification 86 aperture position by a slight amount.

The first measurements also showed that the central blue and red 'star' fibres were not completely overlapping on the sky, but were offset from each other by over half a fibre width (this is illustrated in Figure 4.11 in the next section). To optimise the signal in each fibre respectively, two sets of apertures were required for each object, one 'blue-centered' and one 'red-centred'. As a result, a total of twenty aperture commands were set for each observing run. This improved the pointing enough to allow a reduction in the amount of manual observing, increasing the potential for automated observing schedules.

4-2.4 Fibre Mapping

Red fibre core Figure 4.11. Graphic of the mapping across the face of the fibres, showing the offset between the blue 'star' fibre and red 'stae fibre. The azimuth and altitude axis are oriented as shown with respect to the face of the fibres. The graphic details the mapping of the blue fibre. The same process was repeated centred on the red fibre.

The first observing runs of the 2003 winter season were used to determine the response of the telescope and fibres by mapping across the face of a suitably strong 4-2. Testing and Modification 87

> = 6 ~ 1-

= 4 =

0 ....,.,. -15 5 10 15 OFFSET AZIMUTH (arcsec) Figure 4.12. Integrated stellar flux for the blue (circles) and red (stars) fibres as a function of azimuth from the centre of the blue fibre .

'- 6 I --~~ 1-= = 4 \\

Once an image was taken at this edge-most position, the telescope was shifted back towards the centre of the fibre in increments of 2 arcseconds. A spectrum was collected at each position until the telescope had moved through the centre and 4-2. Testing and Modification 88

' 6 ' ' ' ' \ .. ,'',,, 4 ...... 2

OFFSET AZIMUTH (arcsec) Figure 4.14. Integrated stellar flux in the blue (solid line) from the centre of the blue fibre compared with the integrated stellar flux in the red (dashed line) from the centre of the red fibre, as a function of azimuth . onto a position 14 arcseconds to the other side of the fibre. This process was then repeated in the elevation axis. This allowed a comprehensive cross-section of the fibre response to be collected.

To assess the overall fibre transmission as a function of position from the centre of the fibre each spectrum was integrated over its entire wavelength range. In the blue this was from 300-550nm, and in the red from 550-850nm. The integrated stellar flux was plotted as a function of offset position from the cen tre of the fibre. This was to check the alignment of the fibres with respect to the beamsplitter and to ensure the optical components were focused sufficiently well.

Figure 4.12 plots the beamshape centred on the blue fibre as the telescope shifts the star position in the azimuth axis. The integrated flux is plotted in arbitrary units, and the red fibre flux collected in the same exposure is also plotted (in the same units). The shift in thered fibre peak and the reduced flux clearly show the offset between the two central fibres. Figure 4.13 shows the blue fibre beamshape again, but this time as the star position is shifted in elevation. It is likely that the broadening of the beam in the elevation axis when compared to the azimuth axis is a result of the flexure of the telescope's elevation axis, discussed in the previous 4-2. Testing and Modification 89 section.

Figure 4.14 compares the normalised beamshape of the blue fibre map (centred on the blue fibre) with that of the red fibre (centred on the red fibre). The red fibre has a much broader beamshape. The minimum angular separation between two point sources that can be resolved by a telescope is a function of the diameter of the primary mirror and the wavelength of the incident light. This angle is called the diffraction limit of a telescope, 0 (in radians), and can be expressed (Zombeck 1990): = 1.22 x A 0 (4.6) D where A is the wavelength of light, and D is the telescope diameter. With a 0.3 meter diameter primary mirror, the AFOS has a diffraction limit of 0.25 arcseconds at 300nm. At 800nm, the diffraction limit is 0.67 arcseconds. The fibre core is

100~tm in diameter, corresponding to 20 arcseconds on the sky. The increase in the diffraction spot of the stellar image as a function of wavelength is not likely be observed in such a large fibre.

At visible wavelengths, the diffraction limit of the telescope is rarely the limiting factor at any ground-based observing site. The seeing, or blurring of a point source as a result of the atmospheric turbulence above the telescope, is usually the dominant cause of image degradation. As discussed in the introduction, an excellent optical observing site such as Mauna Kea experiences average seeing of 0.7 arcseconds, with best conditions of 0.25 arcseconds (Roddier et al. 1990).

Average seeing at South Pole has been measured at approximately 1.8 arcseconds (Marks et al. 1996). Importantly, the angular size of an observed object for a given level of atmospheric turbulence can be given as (Fried 1966):

(4.7)

where A is the wavelength of light from the object. This shows that the angular 4-2. Testing and Modification 90 size of an observed object will be larger at shorter, bluer wavelengths than in the red, which is opposite to the effect shown in Figure 4.14.

When a stellar object was offset from the centre of the optical fibre, spectrally dependent attenuation was observed in both the blue and the red AFOS spectra. An analysis of these effects is presented in the following section. The intensity, and spectral shape of the attenuation is different in the blue and red optical fibres. This may therefore be the explanation for the different beamshapes in Figure 4.14, and will be discussed in the following analysis.

4-2.5 Spectral attenuation

The first stellar spectra collected with the AFOS in 2002 were observed prior to the improvements in pointing just discussed. Initially it was thought that the pointing would not need to be particularly accurate as the benefit of large-diameter, long optical fibres is that the image information from the star is scrambled during its path through the fibres. This should mean that if the star is only partially centred in the fibre the signal will be attenuated but the shape of the stellar spectrum will not be altered (Chapman & Nations 1996).

Analysis of the spectra, however, showed a substantial wavelength-dependent attenuation in the spectra when the star was poorly centred in the fibre. The effect is clearly seen in the blue fibre spectra collected in the Figures 4.15 and 4.16. The solid line spectrum in these two figures is from the blue star fibre when the star is centred in the aperture. The dashed-line spectrum was taken with the star positioned on the edge of the fibre.

When the ratio of these two is taken, as in Figure 4.17, the curve shows a wave length dependent attenuation that increases sharply at the the wavelength decreases. 4-2. Testing and Modification 91

1,00E5 Centred spectrum

!3 soooo ..;:::~ 60000 t ..s $3 § 0 40000 u ~ "'.-.__ ,_ E: ) ,.. •.J v '--...,..-..·--·- -.. Misaligned spectrum ! ..,.·~-~ ---·-·-·-·-· ) ~- ...... / ~·-·- 0 3000 4000 5000 6000 w~ve~en~th (an~stroms) Figure 4.15. Comparative flux level from star on the edge (solid) compared to when it is centred (dashed) .

1.25E5

_:t.,OOE5 i::' j ~ 75000

J::::; 50000 ~ tZ

3000 4000 6000 6000 Wavelength {angstroms} Figure 4.16. A normal stellar spectrum, centred correctly in the blue fibre (solid) and the same stellar object scaled to the same size, this time positioned at the edge of the blue fibre (dashed) .

• 35 !3 _g .:E ~ ,g s .25 ~ E:

.:1.5 3000 3500 4000 4600 5000 5500

Figure 4.17. The ratio of the spectrum of the star at fibre edge to that when it is centred. Noise below 3000A is a result of lack of signal . 4-2. Testing and Modification 92

Blue fibre ratio (edge/centre)

0.35

0.3 0 ~ 0.25

0.2

0.15 .J...------,,..------.------=-----' 3200 3700 4200 4700 5200 5700 6200 Wavelength (angstroms) Figure 4.18. The ratio from Figure 4.17, fitted with a .>.-4 curve modeling the attenu ation that would be seen as a result of Rayleigh scattering.

Rayleigh scattering in the core of the optical fibre is the most likely cause of this attenuation. Rayleigh scattering is the primary cause of attenuation in optical fi bres (Barden 1998). The attenuation coefficient, aR, is related to the wavelength of light, A, being propagated in the fibre by:

(4.8)

This .A - 4 dependence means that red wavelengths are not as strongly attenuated in an optical fibre as the blue. To determine if the attenuation seen in the blue spectral ratio is caused by Rayleigh scattering, a .A - 4 curve was least-squares fit to the ratio in Figure 4.17. The resulting fit, Figure 4.18 shows good agreement with the data. The attenuation seen in the blue fibre was the limiting factor in the observations. When the star was more than 5 arcseconds from the centre of the fibre, attenuation at 400nm increased to 10%. At 10 arcseconds, as seen in Figure 4.18, the value at 400nm was 25%. To counter this, a pointing accuracy of better than 5" was required before each exposure was taken. This greatly improved 4-2. Testing and Modification 93 the consistency and quality of the data for the 2003 observing season.

4-2.6 Polarisation Effects

The Rayleigh scattering attenuation was also observed in the red fibre spectrum, though due to the attenuation's .A - 4 dependence, the loss due to this effect was not as detrimental as in the blue fibre. Another trend was observed in the red fibres, however, when the fibre mapping described in Section 3-2.4 was analysed. Figures 4.19 and 4.20 plot a selection of the spectral ratios calculated as the star was shifted diagonally across the red fibre in elevation and azimuth respectively. The ratios are scaled to a common mean. The ratio of the edge-aligned spectra to the centred spectra shows a periodic variation as a function of wavelength.

..R t J. :1. C>

( "".,. b J. .,.t .,...... )

6500 7000 7500 8000 8500 Wave~en~th c~nsstrcms>

Figure 4.19. Ratio of spectra to a 'perfectly centred' spectrum as the star is shifted from the top left edge of the fibre (red), towards the centre (yellow), past the centre (light blue) to the bottom right edge (dark blue), as the telescope is move past the star in altitude.

Let us consider the offset of the object from the centre of the fibre as a function

of the angle1 !:J.e, from the optical axis (where i::J.8=0 when the object is centred in the fibre). If we take the case when the object is at the edge of the fibre, I::J.Omax, then we define the spectral ratio, EL, at this point as: 4-2. Testing and Modification 94

R~t~o oP O~Fset Spectr~ to Centr~d Spectra

I +:1.0" of'"f'":ec PI I - R ..."'t 0 F ••• J !::.. • -~ iii..!"'~· ~ ." . ',·, ·\.·~ ,...,..,~, ;~~~'' Hlo " ~,.;_·,., ~~ < ~( . f _.. ·,,.,/~t .. '~\.~J. ,. "''r b i. t 'r ·"'rl"'V-·,.·.,'. ·'"·\ . r~/~.;1\\:h· '·~:/'~1;1-~~~"- .. .'\j· -9", o.P-Fsot ~!.i.,jo,f, ~~w ~ r "')

1 1.0 ' oF'.Poot.

6500 7000 7500 8000 8500

Figure 4.20. Ratio of spectra to a 'perfectly centred' spectrum as the star is shifted from the bottom left edge of the fibre (red), towards the centre (yellow), past the centre (light blue) to the top right edge (dark blue), as the telescope is moved across the star in azimuth.

1? __ ~(~Br.nax) (4.9) L-- ~(0)

where ~(~Br.nax) is the spectrum collected at the maximum offset from the central position, and ~(0) is the reference spectrum at the centre of the fibre. The ratios were then calculated for each object position in Figure 4.11 with respect to the central reference spectrum, ~(0). Following from Equation 4.9, the ratio at the centre is expressed as E0 = ~(0)/~(0) = 1.

If the angular offset from the centre is considered ~ -~0 on the right-hand side of the fibre face, then the spectral ratio at the right edge of the fibre can be written as ER = ~( -~Or.nax)/ ~(0). Plotted together in Figure 4.19, the periodic variation shows a 180° phase-shift as the star is shifted diagonally (see Figure 4.11) from one side of the fibre to the other, such that EL = -ER.

When the ratios in the elevation axis were compared to the same ratios calculated in the azimuth axis it was observed that the ratio of the top left-most spectrum (elevation axis) EL was identical in phase to that of the bottom left-most spectrum

(azimuth axis) A£. When expressed in terms of the angle of offset: ~0, this is given by: 4-2. Testing and Modification 95

...... ,..---..._ ; .... " z ; " w £2' I :!: -w I z (!) ::i <( ~ :i w Ol I " I u:: ::J --' ' ; . w -<( ' " N - ' ...... "

U) w ..Ju zoo <(0:: I :i::J: I <(~ we Ol::J: tnO -;:)

~0)0:: ' ' LLJ: ' ' , LLI- ' ' , , 0 """'...... _____ .. -*' ,' Figure 4.21. Graphic of injection module and two possible causes for the position dependent wavelength response of the red fibre spectra. LEFT: wavelength-dependent variation in the transmission of the dichroic beamsplitter as a function of incident angle of the beam on the dichroic as the star position was moved across the fibre. RIGHT: Misalignment of the red fibre causing preferential transmission of one polarisation state as the star is moved across the fibre. This diagram is not to scale: the angle .6.8, the change in incident angle on the dichroic, is very small(<< 1°). 4-2. Testing and Modification 96

100.0

90.0

eo.o

70.0

60.0 ,o ;:-. e ~so.o $ E

30.0

20.0

~0.0

lnclden1: angle = 45 degrees o.o 300 400 500 600 700 BOO 900 Wavelength (nm) Figure 4.22. Transmission of the dichroic beamsplitter in P, S and average polarisation as a function of wavelength, incident angle 45°. Manufacturer's data.

(4.10)

The plot of the ratios in the azimuth axis displayed the same 180° phase shift as the star's position was shifted diagonally from the left to the right hand side of the red fibre (AL = -AR)· The data from the azimuth and elevation axis are plotted against each other in Figure 4.24 when the star was on the left-hand side (EL and AL) of the central fibre position (6.() > 0). The result is the same, though opposite in phase, when this comparison is repeated between ER and AR. This symmetry suggested that the effect was a result of a change in the transmission properties of the dichroic beamsplitter as a function of the incident beam angle 6,.(), or a misalignment of the red-fibre bundle with respect to the optical axis of the injection module. The two possibilities are shown in Figure 4.21.

The S and P polarisations of incident light are transmitted differently through 4-2. Testing and Modification 97

Increasing Incident Angle

100

300 400 500 600 700 800 Wavelength (nm)

Figure 4.23. Shift in the transmission of the beamsplitter as a function of the beam incident angle, measured by a laboratory optical spectrometer. the beamsplitter as a function of wavelength. The manufacturer's plot of the dichroic transmission as a function of wavelength (and polarisation) is seen in Figure 4.1. For comparison, Figure 4.22 shows a second transmission plot of the same beamsplitter at an incident angle of 45°. The P polarisation transmission curve has a higher transmission but a lower peak-to-peak amplitude than the S state. Secondly, as the angle of incidence on the beamsplitter is increased, the transmission peaks are shifted slightly in wavelength, and the peak-to-peak amplitude of the transmission curves increases. In addition, the transmission of the S polarisation state increases as the incident angle is reduced, and the difference between the S and P state transmission curves decreases.

The angle of the incident beam from the telescope onto the dichroic changes by only a small fraction of a degree ( < < 1°) as it is moved from one edge of the fibre (.6.Bmaa:) to the other (-.6.Bma:z:). To investigate whether such a change 4-2. Testing and Modification 98

E(m.in)

A~min)

A 'I' b E

6500 7000 7500 8000 8500

Figure 4.24. Comparison of the ratios EL and AL as a function of beam angle. in angle could produce the observed effect, a sample of the dichroic beamsplitter was obtained from the manufacturers, Thompson & Co. This sample was placed on a rotating stage in an optical spectrograph and the transmission of the sample was plotted as a function of the incident beam angle. The results are shown in Figure 4.23.

Figure 4.23 shows that the change in the peak-to-peak amplitude of the transmission curve as a function of incident angle was very small (less than 1% per degree), and is barely discernable on. this plot even over a range of incident angles between 5° and 45o. In comparison, the shift in the wavelength of the transission pcalcs is quite significant.. If the effect was a result of the D.O in the beam angle through the beamsplitter, the spectra would also exhibit a shift in the peak wavelength of the oscillations as the star was moved from one edge of the fibre to the other. This shift in wavelength was not observed, and so this possibility was excluded as a potential source of the oscillations in the spectra.

The second possible cause of the observed effect was a preferential transmission 4-2. Testing and Modification 99

Model dichroic transmission (P, S and Avg polarisations)

~~-~~~

50.5 I

50 r!:' ._ca49.5 I ! -e- 49 <( i 48.5 . ! 48 . - P polarisation -s polarisation 47.5 ' -Average polarisation • 47 ~------~ 0 2 4 6 8 10 12 Arbitrary Figure 4.25. Empirically derived model of the S, P and Average polarisation state transmission curves of the dichroic beamsplitter. of the S or P states of polarisation when the beam was incident on the face of the red fibre, as seen in Figure 4.21. If the red fibre had been jolted out of alignment as shown, the angle of the incident beam on the fibre face would change as the star was shifted from one side of the fibre to the other. Given the 'X-shaped' configuration of the axes on the fibre face, a misalignment in this axis would also explain the symmetry observed in the oscillations.

To determine if such au effect could be produced by a preferential transmission of one polarisation over the other, the dichroic transmission was modeled. This model does not attempt to describe the precise transmission properties of the beamsplitter but instead was simplified to a function that oscillated uniformly as a function of wavelength. The manufacturer's transmission curves were used to record the peak to-peak amplitudes, and peak wavelengths, of the S, P and average polarisation states at an angle of 20° (Figure 4.1). The transmission, T(.A) of each state was described by a simple function of the form:

T(.A) = Asin(.X + .b..A) (4.11) 4-2. Testing and Modification 100

where >. is the wavelength, .6.). is the shift in the peak wavelength of transmission between each polarisation and A is the relative amplitude of each curve. The resulting models for the three states can be seen in Figure 4.25. Two functions were derived, Tp(>.), to describe the P state, and Ts(>.), to describe the S state. The average polarisation state is given by (Tp(>.) + Ts(>.))/2. For brevity, the >. denotation will be omitted for the rest of the analysis.

The empirical equations derived for the two states are given by:

Tp - sin(>.) (4.12) Ts - 0.791sin(>. + 0.0061) (4.13)

A transmission spectrum which was a linear combination of these two states was then computed for six positions across the fibre as a function of .6.8. The six spectra, from .6.8(max) to .6.8(min) and then to -.6.8(max) to -.6.8(min), plus a central spectrum at .6.8(0) are given by:

SMAX 0.1Tp + 0.3Ts (4.14)

SMED - 0.2Tp + 0.35T8 (4.15)

SMIN - 0.4Tp + 0.45Ts (4.16)

So - 0.5Tp + 0.5Ts (4.17)

B-MIN - 0.45Tp + OATs (4.18)

B-MED - 0.35Tp + 0.2Ts (4.19)

B-MAX - 0.3Tp + 0.1T8 (4.20)

These transmission spectral models can be seen in Figure 4.26. As with the data from the fibre map, the ratios of these spectra were then calculated with respect to the central spectrum T0 • Figure 4.27 and Figure 4.28 compares the ratios calculated from the data mapped in the azimuth axis with the derived ratios from the model just described. There is no offset to a common mean in these data. The model 4-2. Testing and Modification 101

Model spectra as a function of beam angle from centre of fibre 55 -S(max) I -·-S(med) 50 S(mln) i -S(·mln) I -S(·med) • ~=----~ -S(-max) i

30~------~ 0 2 4 6 8 10 12 Arbitrary

Figure 4.26. Six model spectra. created from linear combinations of the two transmission curves Tp and Ts as a function of the beam angle flO from the maximum offset from the centre of the fibre (MAX) through the centre (MIN) and then to the other side (-MAX). effectively describes the phase change in the resulting ratios as the beam is shifted diagonally across the fibre.

To investigate the reason for this preferential transmission as a function of the incident angle on the fibre, the Fresnel coefficients of reflection and transmission at the entrance face of the silica optical fibres was plotted. Fresnel's equations describe the reflection and transmission of electromagnetic waves at an interface by providing the transmission coefficients for radiation parallel (P state) and perpendicular (S state) to the interface. These coefficients are given by:

2sin0tCOS0i tp - (4.21) sin(ei + 8t)cos(ei- 8t) (4.22) 2sin0tCOS0i ts - (4.23) sin(ei + Ot) 4-2. Testing and Modification 102

I A<-min) ~~ ,__ .-.~,..,._-...- =~-·------~~ .....__ ...... -~ -~--·£~·-·- --- 1 A

A r b A<-med) i A{med) t r a r !:J

A <-mC1x) 8 (max>

6500 7000 7500 8000 8500 Wavelength (angstroms>

Figure 4.27. Ratios of the spectra to a central reference spectrum as the star is shifted in azimuth from the one edge .6.0(MAX), diagonally through the centre to the other edge -L\O(MAX). The relative flux of the spectra has not been scaled.

Azimuth axis: Ratio of spectra as a function of star position on fibre

·-·~- --·--·-· -·· ··~.-:___~_-:__.·~_:--~~·~-=------·- __-_·-·~-~--~-__:_::===:-! 0.98 I 0.93 I :c Ill .!!! 0.88 ==~======! iii """"'A{max) I ...... ,A{med) e0 -A(mln) ~ 0.83 1 .9 ~A{-mln) I -A{-med) i ~ 0.78 -A{·IJ!liX) : I ' 0.73

0.68 '------' 0.8 1.8 2.8 3.8 4.8 5.8 6.8 7.8 8.8 9.8 Arbitrary

Figure 4.28. Ratios of the six model spectra from .6.0(MAX) to -.6.0(MAX). 4-2. Testing and Modification 103

External Reflection and Transmission Fresnel coefficients

1 - Reflection(ll) -Reflection(+) 0.8 -~-~~ Transmission(ll) 0.6 ~ -Transmission(+)

0.4 ~~-,, C1) ::s n; 0.2 > 1: 0 . .! 0 20 100 iS -0.2 ------.;;,.._---~ ~ .{).4 (J -0.6

-1

-1.2 Angle of Incidence

Figure 4.29. External Fresnel coefficients of transmission and reflection for an air/silica medium.

(4.24) tan(ei- Ot) Tp - (4.25) tan(ei + Ot) (4.26) sin( ei - f1t) rs (4.27) + sin(ei + et)

where ei is the angle of incidence, Ot is the angle of transmission, tp and ts are the transmission coefficients for the P and S states respectively and rp and rs are the reflection coefficients for the P and S states. For the silica fibres in the AFOS, the core refractive index, n2 is 1.436. Knowing that the refractive index of the incident medium is 1, Snell's law can be used to calculate the angle of transmission as a function of the angle of incidence. These in turn can be used to calculate the Fresnel coefficients for the silica fibres. The resulting coefficient curves are plotted in Figure 4.29. 4-2. Testing and Modification 104

The first thing to note is that the face of the fibre would have to be misaligned by a large angle, greater than 30°, before a difference in transmission of more than 5% is seen between the two states of polarisation. This would be unlikely, as the fibres are afixed to the module with shim pins designed to prevent such a misalignment. However, it is worthy of consideration as a 5% difference in the transmission of the two states would be enough to produce an effect such as the one observed. It is also worth noting the angle at which the reflection coefficient, rp, intersects the x-axis. This angle, ()B, is known at Brewster's angle, at which point all of the light in a plane parallel to the surface is transmitted, and the reflected light is linearly polarised in the plane perpendicular to the surface (Hecht 1987). For the air/silica interface however, ()B is 55.7°, and this would be an extreme misalignment in the red fibre bundle.

The optical fibres are approximately 40 metres in length. It could also be the case that the transmission within the fibres of the S and P polarisations states varies with the point, and angle, of transmission. A small preferential transmission of one state over the other could result in a noticeable loss after the beam has traveled through the full 40 metre length of fibre. Finally, there is one other notable possiblity that was not considered in this analysis which may also produce the effect observed in the AFOS spectra. In some multi-object fibre systems, chromatic fringing effects are observed in the spectra when a thin-film (such as an air gap appearing in the cement-layer covering the fibre input-face) is in the beam path (Watson et al. 2000). It is possible that such a film was present in the beam, most likely on the dichroic, as the effect is observed in both the red and blue fibres.

During the winter observing season of 2003 there was no further way of testing these hypotheses without risking further, and potentially more debilitating, mis alignments to the optical system. Further analysis with the entire system at a temperate site would be the best way to thoroughly investigate the cause of the attenuation and oscillation in the spectra as the star is shifted across the fibre face. 4-3. Summary 105

These effects, as mentioned previously, were minimised during the 2003 winter season by careful and frequent pointing runs. The improved pointing reduced the time required to make certain the star was within 5 arcseconds of the fibre centre before an exposure was taken. The periodic variation in the red fibre spectra may have a small affect on the spectral modeling detailed in Chapter 6 and that is discussed briefly in the chapter.

4-3 Summary

Instrument and data analysis from the 2002 and 2003 AFOS observing seasons has been presented. The characterisation of the fibre positions on the sky, telescope flexure and spectral response were crucial to improving the quality and quantity of the data collected. The improvement in pointing accuracy as a result of this study allowed a streamlining of the observation techniques to the point where automated observing sessions (up to twenty-four hours in length) were undertaken during August and September of 2003.

A wavelength-dependent spectral attenuation was observed in both the blue and red fibres, and a detailed analysis of this effect was necessary to understand the cause of these variations. This revealed an effect in the blue fibre when the object is offset from the centre of the fibre which we ascribe to Rayleigh scattering. In the red fibre, a position-dependent periodic variation was observed in the ratio of a misaligned spectrum to a centred spectrum. This attenuation, which changed in phase as it move from one edge of the fibre to the other, was shown to be polarisation-dependent, though further investigation is required to ascertain the exact source of the effect. Understanding of this effect led to a tightening of the pointing accuracy constraints for the 2003 observing season and the data quality and consistency was greatly improved as a result. Chapter 5

Observing and Data Reduction

5-l Observational Strategy

AFOS observations required a selection of stars that fitted the following criteria:

• Bright, well-characterised stellar objects

• Relatively featureless and smooth continuum spectra

• Hot, with strong emission in the ultra-violet

• A range of declinations

Bright sources were needed as the lack of a finding camera made the acquisition of less luminous sources time-consuming and difficult. In addition, a high signal to-noise assured well-defined atmospheric absorption features in the spectra. The pointing file created for the ADIMM instrument already had a list of three hundred stars including the thirty brightest southern stars, and so this object file was used as a starting point to select suitable AFOS objects.

Of this list, a number of the sources (such as Canopus) were too red, and con tained too many stellar absorption features. Only the hottest 0 and B stars had the smooth, featureless continuum spectra necessary to avoid confusion between stellar 5-l. Observational Strategy 107 and atmospheric absorption features. The hot stars also emitted strongly in the ultra-violet, allowing a better study of the UV cut-on wavelength of the atmosphere.

The final requirement was to observe a number of stars at a range of declinations. From the geographic south pole, an object outside the solar system follows an orbit that does not vary in zenith angle. At a mid-latitude site a star travels through a large change in zenith angle, allowing that source to be observed at different airmasses .. The depth of atmospheric absorption lines increases as a function of airmass. A common technique for determining the amount of a particular molecular species in the atmosphere is to compare observations of the same absorption feature at two or more different airmass values (Roscoe et al. 1994; Fish et al. 1994). At South Pole comparisons needed to be made between different objects (of similar or identical spectral type) at a range of zenith angles to achieve the same sampling.

The list of sources observed during the winters of 2002/2003 is shown in Table 5-l. It can be noted that a Centauri does not exactly fit the above criteria, as it is a cooler, redder star with lower emission in the ultra-violet. It was primarily used as an initial source for pointing, and to check conditions at South Pole at the time. This star was observed first in all observing runs by the AFOS and, once centred in the telescope, the flux count would instantly indicate the presence of cloud, or icing on the telescope window.

5-1.1 Manual Observing

Manual observations were required for most of the spectra collected with the AFOS in 2002 and 2003. The standard observing session was about five hours long, increasing to six or seven hours if the satellite connection was stable and the weather was good. Weather delays and poor satellite connectivity were the most common causes of a halt or delay in observations, though a technical fault in the 5-1. Observational Strategy 108

Table 5.1. Stars selected for AFOS observations in winter 2002/2003.

Name Right Ascension Declination App. V Mag. Spec. Type a Centauri 14 39 35.885 -60 50 07.447 0.00 GO+k5 a Eridani 01 37 42.852 -57 14 12.18 0.46 B5 f3 Centauri 14 03 49.408 -60 22 22.79 0.61 B1 a Piscis Austrini 22 57 39.055 -29 37 20.10 1.16 A3 f3 Crucis 12 47 43.237 -59 4119.46 1.25 B1 a Crucis 12 26 35.871 -63 05 56.58 1.33 B1 f3 Carinae 09 13 11.957 -69 43 01.95 1.68 AO

€ Sagittarii 18 24 10.327 -34 23 04.73 1.85 AO

€ Centauri 14 35 30.429 -42 09 28.39 2.30 B1

telescope mount limited the number of successful observing runs conducted late in the winter season of 2003.

Connectivity to the mount and telescope control computers first required connection via telnet to a UNIX server at South Pole, pharlap.spole.gov. A telnet connection from the control computer in Sydney to pharlap was established each day as soon as stable satellite connectivity could be obtained. Satellite connectivity at South Pole is notoriously variable. Four satellites, TDRS1, GOES3, MARISAT2 and LES9 are visible to South Pole for up to six hours each, though the quality and stability of the connections is frequently unreliable.

In general there was a five to six hour window each day when satellite connectiv ity with sufficient bandwidth was available for observations. This six hour window precessed forward by about five minutes each day and the schedule of rise and set times required constant monitoring. A useful website run by Chris Martin at the Harvard Smithsonian Center for Astrophysics (Martin 2004) logs and plots the satellite rise and set times and is updated daily. 5-l. Observational Strategy 109

------... : J..~ _ -f =Satellite dependent 1 connections I

Gmount control AFOS PC/104 computer stack -----I ERIC control I software 1 file.gz I (276k8) I tel net y Data review and connectivity Data retrieval monitoring

pharlap terminal pharlap terminal pharlap terminal pharlap terminal ~ J..~ "1 -"1 J..~!..-f I tel net tel net UNIX !ftp I server ·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·· I \ UNSW, Sydney Figure 5.1. Flow chart of manual observing, mount and telescope control and data retrieval between Sydney and South Pole during satellite up-time.

A flow chart depicting the software connections required to run the mount, telescope and data retrieval can be seen in Figure 5.1. Three telnet sessions to pharlap were established, two for connection to the G-mount and AFOS control computers and one for monitoring data retrieval and satellite connectivity. A fourth ftp connection to pharlap would be established to retrieve the data files as they were downloaded to the pharlap server.

The G-mount control computer located in the AASTO, 15 metres from the telescopes, is a PC/104 CPU and runs on a QNX R/T operating system. The details of the software can be seen in Hovey et al. (1998). The control computer 5-l. Observational Strategy 110

Table 5.2. Common commands used to run the G-mount during observations.

Command Parameters Description cgmount AFOS/ADIMM Identify instrument port axes(l) ON/OFF Controls power to axes servos axes(2) PARK/STOW/DEICE Slews mount to preset coords aperture AP NAME Modifies pointing to predef aperture aoffset /j.x,/j.y Moves object in focal plane gstatus 1-25 Requests status info fromo control system spiral START /STOP /RESUME Spiral scan in focal-plane track OBJECT INDEX Tracks object in selected Coord file cfile *. CFILE/*. COORD Pointing file or object list

has two instrument interface ports, one for the ADIMM and another for the AFOS. From a terminal on the pharlap server, a telnet link was established to the AFOS interface port of the G-mount.

A typical log of the start-up and control commands entered at the G-mount command-line prompt is given in Appendix B. The command "cgmount AFOS" was first entered to establish the identity of the instrument to be used. The mount would then be powered up with the entry "axes on" at which point any faults in the mount or connection were usually apparent. A short list of the most commonly used commands is provided in Table 5.2.

Once contact had been successfully initiated with the mount, a second telnet connection to the AFOS PC/104 stack would be set up. The AFOS control software is called ERIC (Extensible Remote Instrument Control), a C++ based command line language written by Michael Ashley at the University of New South Wales. Full details of the control software are available in Ashley et al. (1996). A standard 5-l. Observational Strategy 111

AFOS log is provided in Appendix B. Once connection was established with the AFOS, the commands "powerup" and "initccd" turned on the power to the CCD camera and established a connection to the CCD interface. The CCD temperature was set to -40°C and monitored with the "gettemp" command until it had reached the setting requested.

A useful feature of the software allowed the user to set the readout area of the CCD. Using this command the CCD readout area was halved from from 256xl024 pixels to 128x1024 pixels - the area in which the six fibres were positioned on the chip. The significance of this was the reduction of the file size from 535kB to 267kB. The hard disk memory on the AFOS computer is only 1MB, which means that only three images can be stored on the disk at one time.

Once the CCD had been peltier cooled to -40°C and an exposure was taken and read out, the file created would be immediately uploaded to pharlap and then erased from the AFOS hard disk to free space for the next exposure. Failure to free the space before reading out the next image would usually hang the AFOS connec tion and require the connection to be reinitiated and all start up commands repeated.

The current G-mount pointing correction file (eg. the file for the 8th of July 2003, c030708.cfile, is shown in Appendix B) would then be entered into the mount computer, as well as the coordinate file containing the list of targets and their co-ordinates. The file used for all AFOS observations, astars.coord, is included in Appendix B.

A run would begin by locating the first object in this file, a Centauri, to check the weather conditions and the pointing of the telescope. Each star had its own 'artificial aperture' (or manual offset in RA and DEC) from the ADIMM pointing file to allow for the flexure and drift of the two telescopes (as discussed in Chapter 4). Once a Centauri was located and centred, the flux level would 5-l. Observational Strategy 112 be evaluated to determine the conditions. It was possible, to a certain extent, to predict the cloud cover from both the ambient temperature and the atmospheric pressure at ground level - two numbers provided hourly by station weather logs. However blowing snow and ice, almost as effective at obscuring observations, were less easy to predict, or even to be seen by an observer on the ground on a moonless night at South Pole. Quite often the conditions would not be known until the first observation was attempted.

Without an acquisition camera, the flux level of the star in each fibre was the only method of determining the position of the star in the aperture, and the observing conditions. Collecting, downloading and examining every single image collected was too time-consuming for this purpose. A quicker method of ascertaining the flux level of each aperture was devised whereby a command in the AFOS software allowed the user to specify six lines along the CCD to be readout at the AFOS prompt. The command "setfibres" could be entered as shown here:

02 Jun 25 03:47:32 AFOS > setfibres 142 168 189 213 234 254 where the values after the command are the lines on the CCD chip where each of the six fibre apertures are centred. Once these fibre positions were set, an exposure-time length was defined and the command "fibres" would expose the CCD and readout the total counts in the specified lines of the CCD directly to the screen. An example is shown here when Alpha Centauri is centred in the red 'star' fibre:

02 Jun 25 03:51:07 AFOS >fibres 10 9 4253 22336 5 8517.2

The fibres, from left to right, are alternatively blue and red, where the third (blue) and fourth (red) fibre are the 'star' fibres. The sixth (red) fibre shows a misleadingly high number of counts as a result of its position over a high bias level on the edge of the CCD, which will be discussed later in the chapter. The "fibres" command was used to optimise the pointing, and to monitor cloud and blowing 5-l. Observational Strategy 113 snow conditions prior to, and during, an observing run.

If conditions were satisfactory then each object would be observed in the follow ing manner. The object's (artificial aperture' was loaded into the mount's pointing file, and the mount was commanded to track that object. Once tracking, a quick offset of a few arcseconds in each direction from the aperture centre determined if the star was truly in the centre of the field. If not, the star flux was optimised using the ((aoffset" commands in the G-mount software while the improvement in the flux level of the (star' apertures was monitored using the ((fibres" command at the AFOS prompt. Once the star was re-centred the (aperture' was updated to the new position.

A series of exposures were then collected, ranging in integration time from 0.1 seconds through to 60 seconds (and 120 seconds for fainter objects or during poor conditions). Eight to twelve exposures would then be taken per object, half of these centred in the blue fibre and half in the red. An offset of 10 arcminutes was then entered, and sky fiats of the same exposure lengths were collected. Bias CCD frames, and two or three long sky fiats (up to 600 seconds) were collected both at the beginning and end of the observing run. The files were downloaded to a local computer at South Pole and then ftp'ed to Sydney for reduction.

5-1.2 Automating AFOS Observations

Part of this thesis work included developing a number of scripts to run the observing routine during the hours of satellite down-time when direct communication with the telescope from Sydney was not possible. A number of instrumentation limitations, including the misalignment problems discussed in Chapter 4, prevented these programs from being utilised for the stellar observations of 2002 and 2003. 5-l. Observational Strategy 114

In late August of 2003, it was decided to use these automated programs in a series of observations of the moon. The science goal of this project was to observe the half-full moon during a complete 24 hour sequence, and see if a measurement of the variation in the earth's albedo could be detected by observing the dark side of the moon. A series of scripts needed to be constructed to input the commands detailed in the previous section, and to respond to prompts and error messages in the two separate control systems for the mount and telescope respectively. The programming language used to achieve this was Expect, a TCL language designed specifically to 'talk' to other interactive systems (Ousterhout 1994). The script adaption and modifications detailed here were undertaken as work in this thesis; however the author is grateful to Michael Ashley who provided the original AFOS command script, which he wrote in 1997.

One comprehensive instrument control script was modified to initiate interaction with a specified instrument (either the G-mount or the AFOS). The program first connected to the instrument via telnet, and upon receiving a response, input the start up and observing commands that were detailed in the last section. If the program received no response for longer than a 10 minute interval, the program timed out and ended the telnet session. Variations of this script were tailored for the AFOS, and four G-mount command sets which centred the mount on different, preset apertures across the face of the moon. These observations will be discussed in further detail in Chapter 7.

The scripts can be seen in Appendix B, and were called runafos, rungmount, agl, ag2 and ag3, where runafos and rungmount were the core start up programs to connect to the instruments and set up the observing parameters. In addition runafos collected a series of six exposures with the AFOS, ranging in exposure time-length from 0.1 to 100 seconds, and uploaded the files to a data folder on pharlap. The programs agl through to ag3 shifted the 'aperture' setting of the G mount, and monitored the tracking state of the mount. Once these programs were 5-l. Observational Strategy 115 completed and tested, a shell script on pharlap was written to call these programs from the command line. The script, moonrun, called the programs in turn as shown here:

#!/bin/bash /home/aasto03/mcba/rungmount gmount /home/aasto03/mcba/runafos afos /home/aasto03/mcba/ag1 gmount /home/aasto03/mcba/runafos afos /home/aasto03/mcba/ag2 gmount /home/aasto03/mcba/runafos afos /home/aasto03/mcba/ag3 gmount /home/aasto03/mcba/runafos afos

The entire program set would be completed in under ten minutes. A cronjob set on pharlap during the satellite up-time at the start of the day, ran moonrun every twenty minutes. The moonrun script collected five partial sets of moon observations during late August and early September, and two full 24 hour sets of data in mid-September before increasing twilight at South Pole prevented further observations. The results of the observations are presented in Chapter 7. 5-2. Data Reduction 116

5-2 Data Reduction

Prior to the work undertaken in this thesis no AFOS observations (not even the small data set of the June 2000 lunar eclipse) had been reduced from the raw COD images. There were no standard procedures for data reduction to follow and so a method to achieve this was carefully developed. The remoteness of the telescope meant that no standard wavelength calibration sources could be used for calibration. The wavelength calibration was therefore determined from the known lines present in the spectra of the stars that were observed. The quartz-halogen lamp positioned inside the telescope barrel was used to remove the instrument signature from the data.

5-2.1 CCD procedures

AFOS produces FITS images. The readout area was set to 129 x 1024 pixels (half the area of the chip) as this was the area over which the six fibres were positioned. This halved the image size and reduced download time from the small AFOS disk considerably. A typical image is 267904 bytes (267kb). A good observing run of a standard nine stars included perhaps fifty to sixty images, four to six quartz-halogen calibration images, and ten to fifteen sky and dark frames. The raw COD image produced when a stellar object is centred in the AFOS can be seen at the top of Figure 5.2. The image at the bottom of Figure 5.2 shows all six fi bres illuminated by the quartz-halogen lamp positioned in the barrel of the telescope.

All images were first corrected for dark current and bias in the COD. The six fibres occupy a region of the COD approximately 5mm in height (128 pixels), with each fibre separated from the other by 640J.Lm (24 pixels) along a column to avoid crosstalk. The Oriel Instaspec IV open electrode COD camera (Oriel Corporation

1 1 1990) was selected for its very low dark current (0.3 electrons s- pixel- , at -40°0), and was peltier cooled to -40°0 for all measurements taken with the AFOS. 5-2. Data Reduction 117

Figure 5.2. Raw CCD image when a star is centred in the two 'star' fibres (TOP) and when the quartz-halogen lamp exposure is taken (BOTTOM). The wavelength scale is from right (300nm) to left (900nm).

F igure 5.3. Top: A raw CCD image of an ob ervation of a Crux centred on the blue fibre. Bottom left: Central row of red star fibre after bias subtraction. Bottom right: Central row of blue star fibre, after bias subtraction. Comparison of the peak flux levels in the two fibres shows the misalignment of the central two fibre , and was the reason two ob ervations of each star were made. 5-2. Data Reduction 118

197 g

§._196. 8 ~ 8

196.6

25 50 75 100 125 Line (pixels)

Figure 5.4. Average bias level for zero length 'exposure' of CCD across the 'short' column axis of the chip.

Figure 5.3 shows a COD image when a stellar object is centred in the blue fibre. The bottom left spectrum is the central row of the red star fibre, after bias subtraction. The bottom right spectrum is the central row of blue star fibre, after bias subtraction. Comparison of the peak flux levels in the two fibres shows the misalignment of the central two fibres, and was the reason two observations of each star were made.

Bias and dark frames were always taken in sets of five to ten to allow a noise-reduced average of the images to be used for all COD reduction procedures. Subtraction of a bias frame is necessary to quantify and remove the electronic zero exposure level for each image frame (Hallet al. 1994). A two-dimensional bias frame is an average of a series of zero-second 'exposures' taken with the COD shutter closed. The bias frames were averaged with ZEROCOMBINE, a COD reduction process used to combine all bias images from a set of observations and average them by a weighted mean (weight is by exposure-time, which was zero for all frames).

The average bias level along the vertical axis (short dimension) is shown in 5-2. Data Reduction 119

198 - I I I I

197,5

196 ,__

I I I I 200 400 600 800 Column {pixels> Figure 5.5. Average bias level for zero length 'exposure' of CCD across the 'long' axis of the chip (spectral dimension).

Figure 5.4. The average bias level across most of the chip is 196.5 ADUs. Along the narrow dimension, shown in Figure 5.4, a sharp increase in the bias level is observed beyond row 125. As can be seen from Figure 5.2, this was unfortunate as the sixth (red) fibre (the topmost fibre in the image) was positioned in this region of the CCD.

The high bias level was not distinguishable from the genuine signal from the star and so this fibre was not used for measurements during the 2003 season. The average bias level in the horizontal axis (long dimension) is shown in Figure 5.5. In the long dimension of the chip, the bias level shows a large fluctuation in the first 50 columns. These 50 columns were cropped from each image using the CCDPROC routine TRIMSEC.

For each set of observations, ZEROCOMBINE produced a single image Zero.fits which records the bias level of the CCD and any bad pixels on the chip. Zero.fits was then subtracted from all star images in that set of observations package CCDPROC. It is necessary to use CCDPROC to carry out the trimming of the data and bias subtraction as all other IRAF routines require the CCDPROC label in the image header to indicate that CCD reduction has been completed. The 5-2. Data Reduction 120 dark frames were then averaged using a weighted mean in ZEROCOMBINE. This weighted dark frame was then subtracted from each image to remove any hot pixels in the active region of the chip. IMPLOT is an interactive task that can be used to inspect images by plotting either lines or columns (or the average of a selected set of lines). IMP LOT was used to visually inspect the images at this point to ensure the bias subtraction was carried out successfully and to check for any remaining bad pixels in the image.

5-2.2 Spectral Extraction and Calibration

DOFIBERS is an IRAF package used to identify, extract and calibrate multi-fibre spectroscopic data. The processes in this package that were carried out on the AFOS data were the identification and extraction of the fibre apertures and the wavelength calibration of the data by identifying known absorption lines in the spectra. At the completion of this step, the data was ready for flat-fielding. The input parameters used in this package are listed in Appendix C.

The first step in the package is an interactive cursor-based process which takes a cross-section of the fibre apertures on the CCD (in the vertical axis) and asks the apertures of interest to be selected and numbered. Figure 5.6 shows an example of the aperture cross-sections for a typical stellar image centred in the blue fibre. The misalignment between the blue and red 'star' fibres is obvious from this plot. For the stellar spectra, where only the central two fibres (one red and one blue) were illuminated, only these two apertures were selected and reduced.

Figure 5.7 shows the same aperture cross-section for a standard quartz-halogen exposure. This is an informative plot as it shows the relative illumination of each aperture by an unfocused, extended source. The six fibres are arranged alternatively blue and red from the left to the right of the image. As mentioned previously, the 5-2. Data Reduction 121

NOAO/IRAF V2.11EXPORT Jtd@phantom Man 19:02:21 14-Jun-2004 Image=2706,21, Sum oF columns 490-499 DeFine and Edit Apertures I I I I 2 I I 1.00E5 f- - ~H ~

80000 f- -

13 .§60000 f- - :e-.e-

f.- - ~40000c..> u::~ 20000- -

o-, ) \. 0 25 50 75 100 125 aperture = 2 beam = 2 center = 79.56 low = -5.00 upper = 5.00 Figure 5.6. Aperture identification using DOFIBERS in IRAF. Cross-section along the narrow dimension of the CCD of the two 'star' fibre apertures illuminated by a star. sixth red fibre was positioned over a high bias level section of the CCD and was not used.

The cross-section of the apertures is taken through the central column of the CCD and so the relative flux levels are slightly misleading. This region of the CCD corresponds to a 'blue' region of the spectrum, just prior to the beamsplitter cut-off at 550nm. Therefore, the red aperture flux levels are lower than the blue at this point. It is worrying, however, to note that the three blue fibres (positions one, three and five from the left of the plot) are not illuminated uniformly, and likewise the flux levels of the two red fibres used (positions two and four) are different. This may suggest that the face of the fibre bundles is damaged, or not at right-angles to the optical axis, confirming concerns about the red fibre misalignment discussed in Chapter 4. There is, however, no guarantee that the fibres are being illuminated by a fiat field, which would also explain the differing intensity levels.

Once the apertures are selected and numbered, DOFIBERS then displays the dispersion of light from each fibre on the CCD chip. Figure 5.8 and 5.10 show the 5-2. Data Reduction 122

NOA~IRAF V2,11EXPORT jtd@phantom Man 19:12:30 14-Jun-2004 Ima~e=~h2706.1, Sum or columns 490-499 DerLne and EdLt Apertures

5000

f4000 :s s. 3000 I 1:3 2000 E:

1000

0 25 50 75 100 125 aperture = 2 beam = 2 center = 79.51 low = -5.00 upper = 5.oo Figure 5. 7. Aperture identification using DOFIBERS in IRAF. Cross-section along the narrow dimension of the COD showing all six fibres illuminated by the quartz-halogen calibration lamp within the AFOS telescope barrel. dispersion axis of the blue and red 'star' fibres respectively. It was important to be sure that the light from a star and the unfocused beam of the quartz-halogen lamp were dispersed along the same axis on the CCD. For comparison, the dispersion axis of a quartz halogen lamp spectrum in the blue (Figure 5.9) and red (Figure 5.11) fibres is shown.

Though there is more scatter in the quartz halogen dispersion (especially in the red), the position of the axes are the same, confirming that the same extraction func tions could be used for the calibration spectra as for the stellar images. The positions of the spectra on the CCD are located using the subroutine CENTER1D, where the user provides an initial guess to the centre of the feature and the approximate width of the feature (in the case of the spread of the light from the fibre in the short axis of the CCD this is approximately four pixels) by solving the equation (Tody 1993):

j (I- Io)f(X- Xc)dX = 0 (5.1) where I is the intensity at position X, ! 0 is the continuum intensity, X is the pixel coordinate, and Xc is the desired feature position. The function f(X- Xc) is a 5-2. Data Reduction 123

NOAO/IRAF V2.11EXPORT jtd@phantom Mon 19:05:23 14-Jun-2004 Func=spline3, order=3, low_rej=3, high_rej=3, niterate=1, grow=O total=99, sample=99, rejected=1, deleted=6, RMS=0.01469 Aperture 1 oF 2706,21 56.5 I I I I >i"" ?<_ 56 - LJ..l. ~--· ~~~~~~~~~~~~~~~ ~ 55,5 ::::r

L 55 - - i n e 54.5 - -

54 - X -

53,5 - -

I I I I 0 200 400 600 800 Column Figure 5.8. Dispersion axis of the blue 'star' fibre across the CCD, when illuminated by a star.

NOAO/IRAF V2.11EXPORT jtd@phantom Mon 19:14:45 14-Jun-2004 Func=spline3, order=1, low_rej=3, high_rej=3, niterate=1, grow=O total=98, sample=98, rejected=3, deleted=8, RMS=0.01684 Aperture 1 oF qh2706.1 I I I I ;x

56 - - ""'x

54 - - X L i n ,__ e 52 - X

50 - -

X X>

48 i- I I I I X" 0 200 400 600 800 Column Figure 5.9. Dispersion axis of the blue 'star' fibre across the CCD, when illuminated by the quartz-halogen lamp. 5-2. Data Reduction 124

NOAO/IRAF V2,11EXPORT Jtd@phantom Mon 19:04:32 14-Jun-2004 Func=spline3, order=3, low_reJ=3, high_reJ=3, niterate=1, grow=O total=98, sample=98, rejected=2, deleted=13, RMS=0,02284 Aperture 2 oF 2706,21 I I I I X X

81 - -

80 - L i n e 79 - -

78 '-- - X

I I I 0 200 400 600 800 Column Figure 5.10. Dispersion axis of the red 'star' fibre across the CCD, when illuminated by a star.

NOAO/IRAF V2.11EXPORT Jtd@phantom Man 19:14:06 14-Jun-2004 Func=spline3, order=1, low_rej=3, high_rej=3, niterate=1, grow=O total=98, sample=98, rejected=2, deleted=33, RMS=0,02024 Aperture 2 oF qh2706,1 I I I I X :XX X 81 1-

80 -

L 79 - X - i X n e X 78 - - X X 77 - X -

76 1- I I I I -; 0 200 400 600 800 Column

Figure 5.11. Dispersion axis of the red 'star' fibre across the CCD, when illuminated by the quartz-halogen lamp. 5-2. Data Reduction 125

0 0 0 0 0 0 0 0 i ••• f[X-XC] i + + :+ + 0 ~-+--+-+-~ .------.t------~ -+-+--+~ • +.+: : ~ 0 . 0 0 0 0 0 0 -width/2 0 +Width/2 fX-XCl Figure 5.12. Sawtooth convolution function f(X-XC) in the IRAF routine, CENTERlD, used to define the width of the AFOS apertures prior to extracting the spectra (Tody 1993). sawtooth convolution function as shown in Figure 5.12 (Tody 1993).

The fitted points from this routine show some point-to-point variation (as seen in Figs 5.8-5.11) due to noise and cosmic rays. A smooth function must be fit to the data to exclude these points. The default fit is a cubic spline (Tody 1993) which was more than adequate for the AFOS apertures, as shown in Figure 5.9 and 5.11. This fit identifies the path along the COD that the spectrograph has dispersed the light from the fibre. Once the dispersion axis is identified for each aperture, the apertures are extracted from the image, using subroutines APEXTRACT in the DOFIBERS package.

APEXTRACT requires the following parameters defined by CENTER1D: the aperture number, centre position, the outer limits of the dispersion relative to the centre and the curve fit describing the shift of the aperture centre across the dispersion axis as a function of wavelength. No weighted variance was used and so the the aperture extraction consisted of summing all the background subtracted pixel values at each wavelength within the aperture limits. At the ends of the aperture, partial pixels are used. The extraction of the two central apertures in each image produced a single file, file. ms.fits, which contained two one-dimensional spectra from the blue and red 'star' fibre apertures respectively. An example of two 5-2. Data Reduction 126

600

I 400 il..._.. j u

~J::t..t 200

200 400 600 800 Column (pixels)

Figure 5.13. Extracted red fibre spectrum of Beta Centauri, prior to wavelength cali bration. Wavelength increases from right to left.

6000

~.}::j ~ ";;;4000

~ 2000

200 400 600 800 Column (pixels)

Figure 5.14. Extracted blue fibre spectrum of Beta Centauri, prior to wavelength calibration. Wavelength increases from right to left. 5-2. Data Reduction 127

Table 5.3. Spectral features identified in AFOS stellar spectra used in wavelength calibration.

Line/feature Wavelength Fibre Ca(triplet) 866.2 Red Ca( triplet) 854.2 Red Ca(triplet) 849.8 Red 02[Atm] 762.2nm Red 02/H20[Atm] 688.6 Red Ha 656.3 Red Fe 544.7nm Blue Mg( doublet) 517.4nm Blue H{J 486.1nm Blue H'Y 434.0nm Blue Ho 410.2nm Blue HE 397.0nm Blue H( 388.9nm Blue

such spectra can be seen in Figure 5.13 and Figure 5.14.

The next step in DOFIBERS carries out the wavelength calibration of the spectra. The spectrum selected as the template for wavelength calibration had to be a particularly good spectrum with strong, clear absorption lines. A list of the absorption lines fitted in each wavelength reference spectrum is given in Table 5.3.

The IRAF routines IDENTIFY and REFSPECTRA were used to inspect a selected stellar spectrum, interactively identify the features shown in Table 5.3 and assign them coordinates. A "coordinate function" mapping pixel coordinates to user coordinates is then determined using the identified features. The CENTER1D routine described previously is again used to identify the central position and width of each absorption feature. 5-2. Data Reduction 128

The beamsplitter complicated the coordinate function assignment as the features present in the red spectra were not present in the blue, and if the IDENTIFY routine was carried out separately on the blue and red apertures, a different dispersion function was obtained for each fibre, as the wavelength dispersion as a function of pixel value was not uniform across the COD. However, if each image was kept in "multispec" format (where the one-dimensional blue and red spectra are in the same image file) the user could toggle between the two spectra in a single IDENTIFY routine and automatically reassign the pixel-to-feature coordinates of one spectrum to the other, prior to a coordinate function being fitted to the identified features.

Once a single coordinate function was applied to this reference spectrum, the dispersion function that was calculated could be reapplied to each spectrum in the observation set. The DOFIBERS package allowed visual inspection of each disper sion corrected spectrum, to check for misidentifications of spectral features. The resulting dispersion corrected spectrum can be seen in Figure 5.15 and 5.16.

5-2.3 Flat-fielding

Because AFOS uses a dichroic to split the starlight into two wavelength regions and uses an internal quartz-halogen lamp for calibration, IRAF flat-fielding programs were not suitable for these spectra. Instead a manual flat-fielding process was designed using the spectrum arithmetic program SARITH. At this point, the data were divided into two sets, a 'red' and a 'blue' set, and all batch processes were carried out on the sets individually.

A quartz-halogen lamp has a 3100K tungsten filament (Osram Sylvania Inc. 2000), which outputs a smooth black-body spectrum. Each optical element, from the telescope mirrors, beamsplitter and fibres to the spectrograph absorbs some of 5-2. Data Reduction 129

2000

'B 1500 .~ ~ ~ "E 1000 ::l 0 (.) ~

3000 4000 5000 6000 7000 8000 9000 Wavelength (angstroms)

Figure 5.15. Extracted red fibre spectrum of Beta Centauri, after wavelength calibra tion.

30000

25000 ,...... j 20000 :e ~ r/.1 15000 § 0 0 ~

3000 4000 5000 6000 7000 8000 9000 Wavelength (angstroms)

Figure 5.16. Extracted blue fibre spectrum of Beta Centauri, after wavelength calibra tion. 5-2. Data Reduction 130 the incoming photons as the light passes through the instrument. Since the shape of the quartz-halogen spectrum is well-known prior to passing through the system, the ratio of the resulting QH spectrum to its initial, smooth blackbody spectrum describes the spectral attenuation in the light path.

Blackbody at 3100!_5.. / / / _.,-·' 60000 '&m Red~ spectrum / £! / ; /' .! -e ./ , s // ~ 40000 .e3 / / ,.r I § _,. I _,.~ ~ ~ 20000 ,."' • e QH spectrum _,.;· / /~·~ .,... ~·...,· / ·-·-·....--· o~-=~~~------T-- 3000 4000 5000 6000 7000 8000 Wavelength (angstroms} Figure 5.17. Blackbody spectrum created with MKlDSPEC in IRAF (dotted line plot), and a standard quartz-halogen spectra in the blue(left) and red(right) fibres.

This is known as the instrument signature of the telescope and, as these features are also present in the star spectra when they reach the CCD, they must be removed to arrive at the original stellar spectrum. A routine in IRAF called MK1DSPEC was used to create a simulated blackbody spectrum at the same temperature as the quartz-halogen lamp. To ensure the wavelength-to-pixel dispersion function for this spectrum was identical to that of the stellar and calibration spectra, a dispersion

corrected quartz-halogen spectrum was converted to a flat spectrum, Stlat, with zero emission, in SARITH by the function:

SQH] (5.2) Sflat = [ SQH -1

where SQH is the quartz halogen spectrum as a function of wavelength. In 5-2. Data Reduction 131

F 1 u X

a r b i t r a r !;!

3000 4000 5000 6000 7000 8000 9000 Wavelength (angstroms)

Figure 5.18. Blue spectrum of Beta Centauri, after fiat-fielding.

MK1DSPEC, S flat was multiplied by a blackbody function of the form: 27rhc2 1 B>.(T) = _AS ehcj(>.kT) - 1 (5.3)

where .A is the wavelength in nanometers, T is the temperature in Kelvin, h is Planck's constant, c is the speed of light and k is Boltzmann's constant (Rybicki & Lightman 1979). The input parameters of MK1DSPEC require the blackbody temperature, T = 3100K, and a continuum emission value. This con tinuum emission was determined empirically by computing a number of blackbody emission spectra of the same temperature but different intensity and selecting the spectrum which most closely matched the output of the quartz-halogen emission spectrum. The fit was qualitative as no absolute flux calibration was necessary for the spectral analysis in the following chapters. An example of the resulting blackbody spectrum plotted against the blue and red quartz-halogen spectra can be seen in Figure 5.17.

Each stellar spectrum, Sob was divided through by the calibration spectrum of the quartz-halogen lamp, SQH, then the resultant spectrum was multiplied by the generated black-body spectrum of the same temperature as the reference quartz- 5-2. Data Reduction 132

F 1 u X

a r b 1 t r a r B

3000 4000 5000 6000 7000 8000 9000 Wavelength {angstroms)

Figure 5.19. Red spectrum of Beta Centauri, after flat-fielding. halogen lamp, SsB· The simple flat field routine can be described by:

(5.4)

where S1 is the resulting, flat-fielded stellar spectrum. The resulting spectrum in the blue and red fibres is shown in Figure 5.18 and 5.19. The spectral shape <600nm in the red spectrum and >600nm in the blue spectrum are remnants of the reduction process. This is a result of both the low stellar flux beyond the beamsplitter cut-off and the shape of the quartz-halogen blackbody emission. This section of each image (shaded sections of Figures 5.18 and 5.19) was cropped from the spectra prior to further reduction or analysis.

It should be noted at this point that a quartz-halogen lamp is less than ideal as a standard lamp for calibrating sources in the blue and UV as there is only a very smalll amount of flux emitted from the lamp at these wavelengths. Ideally, the telescope should observe a hot blue star, and use model atmosphere fluxes for that object (available via VizieR) to remove the instrument signature from a spectrum. A thorough account of this method is provided by Bessell (1999). The wavelength-dependent attenuation observed in a stellar spectrum collected 5-2. Data Reduction 133 at the ground is produced by both attenuation in the atmosphere and losses in the instrument itself. To recover the attenuation curve of the instrument from a stellar observation, the atmospheric extinction curve of the site must be known in advance, or vice versa. Unfortunately, the AFOS was not used to observe standard hot AB stars at a site such as Siding Spring Observatory, where the atmospheric extinction is well-characterised. This would have allowed an evaluation of the absolute transmission characteristics of the instrument in the blue/UV wavelength region. As the instrument response in the blue/UV was not known prior to its deployment to South Pole, to use this method to remove the instrument signature of the AFOS from spectra collected at South Pole would also remove any information about the Antarctic atmospheric transmission that were present in the spectrum.

5-2.4 The UV Cutoff

The AFOS was optimised to transmit and collect light in the blue and near ultraviolet regions of the visible spectrum. By observing bright, hot stellar objects with high flux in the ultraviolet, a determination of the UV cutoff wavelength of the Antarctic atmosphere could be obtained. It was possible that the reduction in ozone in the upper atmosphere above Antarctica may result in an increase in the ultraviolet transmission between 300-350nm.

However, a number of atmospheric processes contribute to the UV extinction at wavelengths longward of 300nm. Rayleigh scattering is the dominant cause of extinction at wavelengths shorter than 400nm. Scattering from aerosols and dust (for example, volcanic dust) in the atmosphere can be comparable in magnitude to Rayleigh scattering at visible wavelengths, though it is more difficult to calcu late than Rayleigh scattering because of the great variability in the nature and concentration of aerosol particles in the atmosphere. It is a combination of these extinction processes, in addition to absorption by terrestrial ozone, that results in 5-2. Data Reduction 134 the UV cutoff in the atmosphere.

Terrestrial ozone (03), exists primarily in the stratosphere and produces prominent absorption bands in near-UV spectra obtained with ground-based telescopes. The Huggins ozone absorption system absorbs ultraviolet radiation between 310-350nm (Schachter 1991). The ozone opacity r(03), for a typical site 2 such as Lick Observatory, records r(03) ::::::: 10- at 335nm, reducing to r(03) ::::::: 1 10- at 320nm and r(03) ::::::: 1 at 310nm (Schachter 1991).

Mauna Kea measurements claim up to a 15-20% improvement in UV trans mission in the 310-330nm band in comparison to lower sites La Palma, Kit Peak, Tololo and La Silla. This is suggested to be a result of the higher altitude, which implies a lower Rayleigh scattering component (Sarazin 2003).

The bluest calibration line used for the standard AFOS dispersion function was the hydrogen Balmer line, HB at 380nm. No standard lamps were observed with the AFOS to determine the dispersion function at wavelengths shorter than this line. It was important, therefore, to understand the pixel-to-wavelength calibration function as accurately as possible. Inaccuracy in the wavelength calibration would make any estimate of the UV cutoff unreliable. Figure 5.20 shows the dispersion function of the spectrograph grating. The dispersion function showed excellent linearity, which suggested that a good estimate of the Antarctic UV cutoff wavelength could be obtained from the AFOS data.

Figure 5.21 shows a calibrated blue spectrum of f3 Centauri plotted against a model of the blackbody continuum produced by the star, which has a surface temperature of 25,500K. The UV cutoff region is easily observed against the blackbody curve. No absolute flux calibration was carried out on the AFOS data, however a measure of the relative transmission in the 300-350nm region can be obtained by taking the ratio of the two spectra in Figure 5.21. 5-2. Data Reduction 135

8000

...... en ~7000 (J) Ol c ~ _c6000 cr.c Q) a; Ei5ooo ~

4000

3000 L---~----~----L---~-----L----~--~----~----~--~ -400 -200 0 200 400 Pixel value Figure 5.20. Pixel versus wavelength dispersion function for the Jobin Yvon spectrograph grating, using the stellar calibration lines from Table 5.3. The grating shows excellent linearity.

\ \ \ '\Blackbody, T=25,500K \ \ \ \ \ ~ i\~-- J UVcuYoff

3000 4000 5000 6000 7000 8000 9000 Wavelength

0~~~--L------L------L------L------L-~ 3000 3100 3200 3300 3400 3600 Wavelength {angstroms) Figure 5.22. Relative transmission of the atmosphere in the region of the Huggins band ozone absorption, calculated as a percentage, by taking the ratio of the calibrated (3 Cen tauri spectrum to the synthetic blackbody spectrum (T=25,500K) shown in Figure 5.21. Because of the calibration issues discussed in the previous section, it is difficult to draw any conclusions from these measurements regarding the wavelength of the UV cutoff.

Figure 5.22 shows the relative transmission of the atmosphere between 300-350nm. The relative transmission, R(%), was obtained by taking the ratio:

R(%) = (S(>..)object) X 100 (5.5) S(>..)BB

where S(>..)object is the AFOS spectrum of {3 Centauri, S(>..)BB is the synthetic blackbody continuum at T=25,500K. The transmission was normalised to the peak flux value of the {3 Centauri spectrum at 354nm such that the relative transmission, R(%) at this wavelength is 100%. At 350nm, the relative transmission is 93%. At 335nm, the transmission decreases to 61%, and finally at 310nm, the relative transmission has dropped to 6% of the stellar flux. This relation for R(%) assumes that the transmission properties of the telescope optical components and the responsivity of the COD are constant across this wavelength range.

Figure 5.22 indicates that the absolute UV cutoff of the Antarctic atmosphere 5-2. Data Reduction 137 from the AFOS calibration occurs at A = 305nm. This value is lower than the 310nm quoted at a standard site such as Lick Observatory, though no comparative spectral data was available for sites such as Mauna Kea, where a 20% reduction in ozone absorption may show similar or better transmission (Sarazin 2003). However, because of the calibration issues discussed in the previous section, it is difficult to draw any conclusions from these measurements regarding the wavelength of the UV cutoff. Further work will include a continuation of the search for comparable spectra from other sites. The analysis here indicates that South Pole shows comparable blue and near-ultraviolet transmission to other good optical observing sites.

The quartz-halogen lamp emits almost no flux at such short wavelengths. An alternative method of fiat-fielding, or use of a hotter calibration source, is necessary to more accurately determine the spectral response of the instrument in the blue and UV. It is very hard to draw conclusions about the UV-cutoff of the atmosphere without a much better understanding of the absolute instrument transmission in the blue/UV region of the spectrum. Secondly, calibration lines bluer than 380nm would be desirable to rule out any non-linearity in the spectrograph grating dispersion between 300-350nm.

5-2.5 Second order leakage

A close inspection of Figure 5.15 shows that between 300 and 500nm the dichroic beamsplitter allows a small amount of blue light into the red fibre. The imperfect performance of the beamsplitter (these peaks can also be seen in the manufacturer's transmission curves in Chapter 3) is a potential problem. Second-order radiation dispersed by the spectrometer has the effect of increasing the flux in the red spectrum and decreasing the depth of any absorption features at the wavelength

.A2 = 2.A1 (Palmer 2002). In the case of the AFOS waveband, blue light transmitted through the red fibres between 300 and 450nm will be dispersed in the second order 5-2. Data Reduction 138 to wavelengths between 600 and 900nm.

100 ~--~----~----~----~----~----~----.---~

80 Jobin Yvon CP200 Spectrometer Concave holographic grating 200g/mm

0 L---~----~-----L----~----~----~----~--~ 200 400 600 800 Wavelength (nm) Figure 5.23. Grating efficiency of the Jobin Yvon CP200 spectrometer (Jobin Yvon Inc. 1993).

The quartz-halogen source does not emit enough intensity in the blue/UV for this to be apparent in the red QH spectrum (Figure 5.17) but most of the stellar sources emit primarily in the blue and UV and so a small percentage leak at these wavelengths could still have a significant effect. The grating efficiency of a spectrometer is defined as the percentage of the incident beam, as a function of wavelength, that is diffracted into the order being measured. The grating in the AFOS spectrometer is an unblazed concave holographic grating and a plot of the efficiency curve for the instrument is shown in Figure 5.23.

The grating efficiency is dependent on a number of factors, including the polarisation state and angle of the incident beam, the order of diffraction, groove spacing and grating material (Palmer 2002). An in-depth discussion of these effects can be found in Loewen et al. (1977) and Palmer (2002).

f3 Centauri is the hottest, bluest source in the list of objects observed by the 5-2. Data Reduction 139

1st order 800 2nd order

~ 600 .;..>'"" g- r rJJ .;..> { 1=1 400 ::l 0 -<..> r ~ E: 200 l ~vJ

3000 4000 5000 6000 7000 8000 9000 Wavelength {angstroms)

Figure 5.24. Light leakage at first order recalculated to the second order for the spec trum of Beta Centauri.

{ Total red flux (including leakA~l\.J 15000 I ~

.;..>~ / ,r' r i 10000 { rJJ f 1:1.... ~ 0 f <..> I .a:< 5000 i t::::.. I -..__ __,~/

3000 4000 5000 6000 7000 8000 9000 Wavelength {angstroms)

Figure 5.25. Light leakage at first order recalculated to the second order for the spec trum of Beta Centauri and compared with the relative flux from the star in the red fibre. 5-2. Data Reduction 140

.1 ~ ~..... ~ :p0 r:: .05 .§ C::..

6000 6500 7000 7500 8000 8500 9000 Wavelength (angstroms}

Figure 5.26. Ratio, R, taken by dividing the leakage flux, 1(2>.) by the total red flux, (Istar + 1(2>.)).

AFOS and so a spectrum taken of this star was used to calculate an upper limit of this effect on the AFOS data. The spectrum shown in Figure 5.15 provided a simple means of measuring the maximum flux from the blue region of the spectrum that would be diffracted in the second order into the red. The stellar leakage flux, Izeak(:>..), in the region from 300-450nm is the amount of radiation leaked through the beamsplitter after transmission through the fibre and diffraction off the spectrometer grating. The intensity of the second order, Izeak(2:>..), will be less than that of the first order, though the calculation of the intensity is complex and has not been presented here.

To establish an upper limit on the intensity of the leakage, we shall consider a case whereby flux transmitted in the second order is 100% of that transmitted in the first order, such that 1(2:>..) =I(:>..). The leaked light in the blue region of Figure 5.15 was doubled in wavelength (see Figure 5.24) and compared quantitatively with the genuine flux of Beta Centauri in the red fibre between 600-900nm, given by Istar, 5-2. Data Reduction 141 by taking the ratio Rleak = Ileak(2A.)/Istar· Figure 5.25 shows the relative flux levels of Istar and Ileak(2'A). Figure 5.26 shows the ratio Rleak in the wavelength region from 600-900nm. The IRAF spectrum arithmetic package SARITH was used to calculate this ratio.

Statistics of the ratio Rleak were measured in the IRAF analysis program SPLOT. This analysis showed that over the entire spectral range from 600-900nm, the mean leakage intensity 1(2>..) was 3% of the stellar flux, Istar, in the waveband. Between 600-850nm, this mean intensity was 1.8% of the stellar flux, and the peak flux in this region does not exceed 5% of the incident stellar flux.

The regions of highest leakage transmission are peaks at 635nm, where the intensity is 4.9% of the stellar flux, and at 865nm, where the peak intensity of the leakage is 13% of the stellar signal. This peak at wavelengths beyound 850nm is a result of the decreasing stellar flux as well as a peak in the leakage flux at this point.

For any data analysis at wavelengths longer than 850nm, this leakage flux would certainly need to be considered and, if possible, subtracted. A second filter in front of the red fibre bundle would be the best way to remove the problem. For the modeling in the following chapters the spectral regions of most interest lie between 600-850nm and a second order leakage of< 5% lies within the errors of the analysis.

5-2.6 Normalisation

The reduced images shown in Figures 5.18 and 5.19 were cropped to ihe wave length regions, Blue: 300-600nm and Red: 600-900nm using the SCOPY routine in IRAF. The final stellar spectra are shown in Figures 5.27 and 5.29. For the modeling analysis detailed in the next chapter, the data needed to be normalised. The spectral analysis IRAF program SPLOT was used to normalise 5-2. Data Reduction 142 the data, which implements a function fitting subroutine ICFIT. The routine fits a user-selected function, again a cubic spline (though the user can also select Legendre or Gaussian functions for fitting absorption features), to the data range defined by the parameter "sample", in this case the full region of the spectrum.

Points are excluded from the fit if they are at a certain distance from the fit: this distance (in pixels) is set by the "grow" parameter. The fit is iterated a number of times, specified by the "niterate" parameter, set here to 10. The user can also manually exclude points from the fit and change the order of the cubic spline fit to the data. The normalisation functions fit using ICFIT, to the two spectra of {3 Centauri, are shown in Figures 5.28 and 5.30, where ovals indicated the rejected points and the parameters of the fit are listed in the header of the image. 5-2. Data Reduction 143

6500 7000 7500 8000 8500 Wavelength (angstroms} Figure 5.27. Final reduced red-fibre spectrum of Beta Centauri, 11:02pm (UT), 27th of July 2003.

NOAO/IRAF V2,11EXPORT Jtd@phantom Mon 20:39:56 14-Jun-2004 func=spline3, order=10, low_reJ=2, high_reJ=4, niterate=10, grow=1 total=359, sample=359, reJected=83, deleted=O, RMS= 8,926 Cnewred.fitsJ: unknown 30. ap:1 beam:2

6500 7000 7500 8000 8500 Wavelength (angstroms) Figure 5.28. ICFIT cubic spline fit to the red stellar spectra of Beta Centauri, with a fitting order of 10. The ovals indicate points excluded from the fit. 5-2. Data Reduction 144

4000 4500 5000 5500 6000 Wavelength (angstroms) Figure 5.29. Final reduced blue-fibre spectrum of Beta Centauri, 10:49pm (UT), 27th of July 2003.

NOAO/IRAF V2.11EXPORT jtd@phantom Mon 20:46:42 14-Jun-2004 func=spline3, order=14, low_rej=2, high_rej=4, niterate=10, grow=! total=372, sample=372, rejected=95, deleted=O, RMS= 110,9 Enb,fitsJ: unknown 30, ap:2 beam:!

4000 4500 5000 5500 6000 Wavelength (angstroms) Figure 5.30. ICFIT cubic spline fit to the blue stellar spectra of Beta Centauri, with a fitting order of 14. The ovals indicate points excluded from the fit. 5-3. Summary 145

5-3 Summary

After normalisation, the general reduction processes on the data were complete. Each day of uninterrupted observation produced a 'blue' and 'red' set of normalised absorption spectra of the nine object stars, as well as bracketing sets of background and quartz-halogen spectra for daily calibration. The limitations on observing were primarily weather and satellite connectivity problems, though a persistent fault with the mount allowed only a select number of observing runs late in the winter of 2003.

A series of scripts was written to automate the majority of the observing, and these were successfully implemented before the end of the 2003 winter season. The data reduction process was designed and optimised after a detailed study of the instrument response. A combination of IRAF routines was used for all of the AFOS data reduction presented here.

The UV cutoff wavelength of the atmosphere above South Pole was calculated from the reduced blue-fibre AFOS spectra. The cutoff was measured at 305nm, which is comparable to other good optical observing sites.

Second-order light leakage from the blue into the red part of the spectrum, as a result of imperfections in the beamsplitter, was analysed. It was shown the peak leakage :flux was less than 5% of the stellar :flux in the 600-900nm region, with a mean :flux of 3%. No attempt was made to remove this additional :flux for the modelling presented in the next chapter, as the peak leakage contribution was within the errors of that analysis. Chapter 6

Atmospheric Modelling

6-1 The Antarctic Atmosphere

The atmosphere above the high Antarctic plateau has been studied for nearly fifty years to better understand its unique properties. Weather balloons are launched twice daily to monitor meteorological quantities above South Pole station. The still, cold, dry atmosphere offers the best conditions in the world for observing in the infrared and submillimeter wavebands (Storey et al. 2003). A submillimeter telescope, AST/RO, and a submillimeter tipping radiometer continuously monitor the submillimeter opacity of the South Pole atmosphere and these measurements can be used to calculate the precipitable water vapour content above the site.

The AFOS was designed to probe the optical astronomical potential of the plateau. The spectra collected by the AFOS in the 2002/2003 seasons were of bright, hot stars with relatively smooth, featureless spectral signatures. A range of meteo rological and atmospheric properties are measured daily by a number of experiments at South Pole station. This provided the opportunity to test the methodology for extracting atmospheric information from the stellar spectra observed with AFOS. 6-1. The Antarctic Atmosphere 147

6-1.1 Radiosonde Data

The Antarctic Meteorological Research Center (AMRC) of the University of Wis consin has launched weather balloons twice-daily at South Pole station since 1956. The radiosonde data for each day of AFOS observations were obtained from their data archive (Antarctic Meteorological Research Centre 2004). An example of a typical profile is shown in Figure 6.1. The radiosondes measure the temperature, pressure and relative humidity from the ground to the upper atmosphere.

Relative humidity is the ratio of the amount of water vapour actually in the air compared to the amount of water vapour required for saturation at the current temperature and pressure. Relative humidity is measured as a percentage, therefore 50% humidity means that the air has half the water vapor it can possibly hold, while 100% relative humidity means that the air is saturated with water vapor.

At extremely low temperatures, the amount of water vapour that the air can hold is very small. The relative humidity measured at each height through out the atmosphere by the radiosonde can be converted into a value of the total integrated precipitable water vapour (PWV). This number is the actual

2 amount of water (in grams/cm ) if the total water vapour of a vertical column of air extending from the ground to top of the atmosphere was condensed into liquid.

The relative humidity can be defined as the ratio, expressed as a percentage, between the partial pressure of water vapour and the value such pressure would have if the air was saturated at the given ambient temperature. If T a is the ambient temperature and Td is the dewpoint temperature (the temperature at which the liquid and gaseous phase of a material are in equilibrium at a given gas pressure) then the relative humidity (in percent) can be written as: 6-1. The Antarctic Atmosphere 148

3rd JuJy 14:00UTC 2003 AMRC Kadiosonde data

5000

-20 -40 -60 -80 -1 0~ 200 400 600 0 20 40 60 Temperature( degC) Pressure\mB) Relative Humidity(%) Figure 6.1. Temperature, pressure and relative humidity profile above South Pole sta tion as a function of altitude, 14:00hrs(UTC), 3rd July 2003. Courtesy of the Antarctic Meteorological Research Centre (Antarctic Meteorological Research Centre 2004). 6-1. The Antarctic Atmosphere 149

RH = 100 ( P(Td)) (6.1) P(Ta) where P(Ta) is the pressure at ambient temperature, and is called the saturation pressure. P(Td) is the pressure at the dewpoint temperature, and is the water vapour partial pressure.

The Bolton equation (Bolton, D. 1980) provides a means of computing the saturation pressure, P(Ta), of water vapour from the known ambient temperature in degrees Celsius:

17.67Ta P(Ta) = 6.1121eTa.+243.s (6.2)

The water vapour partial pressure, P(Td) can then be derived from:

P(T. ) = P(Ta)RH (6.3) d 100 The derived P (T d) can then be used with the ideal gas law to determine the mass density of water vapour at each reported height. This calculation was undertaken on each radiosonde data set from the AMRC archive. A full derivation of the precipitable water vapour density Pwv is given elsewhere (Schwerdtfeger 1984). This Pwv is given by:

_ 100P(Td) 6 (6.4) Pwv = 2.167 X 10 Ta + 273_2

2 and the precipitable water vapour overhead, PWV, in g/cm , is then given by:

PWV = j Pwvdz (6.5) 6-1. The Antarctic Atmosphere 150

where each reading of the radiosonde data corresponds to an increment ozi in the altitude and so the PWV is written:

PWV = 1000 L Pwv,i + Pwv,i-l.D.(zi) (6.6) i 2 where the PWV is in g/cm 2 and the altitude interval is in metres. The radiosonde measurements for each day of AFOS observations were used to calculate the daily PWV overhead for comparison with the results from the modelling of the AFOS data.

6-1.2 Millimeter and Submillimeter PWV Measurements

Astronomical observations at millimeter and submillimeter wavelengths are difficult from most ground-based sites. The dominant absorber of submillimeter radiation in the atmosphere is water vapour, and the precipitable water vapour content above most sites is so high that all extra-terrestrial sub-mm radiation is absorbed before reaching the ground. Only the highest, driest sites in the world, such as Mauna Kea, Hawaii, the highest peaks in Chile and the central plateau of Antarctica offer transparent windows for the transmission of submillimeter radiation from outside our atmosphere.

AST /RO is the Antarctic Submillimeter Telescope and Remote Observatory (Stark et al. 2001). AST/RO has a 2 metre primary mirror and five heterodyne receivers and has been run each winter since 1995. Narrow-band sky opacity measurements were made over the entire austral winter of 1995 using a heterodyne receiver tuned to a band centred at 610,um (Chamberlin et al. 1997). A submil limeter tipping photometer situated on the roof of the AST /RO telescope measures the sky opacity in the 350,um window (Peterson et al. 2003).

Data from the 350,um tipping radiometer, taken concurrently with the AFOS 6-1. The Antarctic Atmosphere 151 observations, was obtained courtesy of the Harvard Smithsonian Center for Astro physics (NRAO/CFA 2004). Radiation enters the radiometer through an expanded PTFE window, where it is focused by an off-axis parabolic mirror. A chopper wheel at 0. 75Hz provides a reference for the collected beam. The beamwidth is 6 degrees, defined by the focal length of the parabolic mirror and the entrance aperture diameter of the compound parabolic concentrator (CPO). The CPO focuses the radiation through a multilayer resonant mesh filter and then onto the pyroelectric detector. Further details of the instrument and measurements can be seen in Peterson et al. (2003).

The tipper undertakes a set of observations five times a day where it measures the sky radiation at several angles from zenith to the horizon. It then observes two internal blackbody calibrators, one at a temperature of 300K and the other at 340K. These calibration measurements are used to convert the measured detection voltages to sky brightness temperatures. A simple single-slab model of the atmosphere is assumed when calculating the sky opacity from the measurements. The data for each tip is least-squares fitted using the equation:

(6.7)

where T is the effective temperature of the atmosphere, Ai = csc(fh) are the air mass values, Tis the zenith optical depth. Ti are sky brightness temperatures, where the index i counts through the seven elevation angles used in the tip

(Peterson et al. 2003). Daily measurements of the sky opacity, T, have shown a strong correlation with the PWV measurements by radiosonde (Chamberlin 2001). The tipper sky opacities obtained concurrently with the AFOS measurements will be compared with the PWV fits determined from the AFOS spectra. 6-2. MODTRAN 152

6-2 MODTRAN

MODTRAN is a MODerate resolution radiative TRANsmission atmospheric mod elling program. It can model the emittance or transmission of the atmosphere in a user-specified spectral region from the submillimeter to the ultraviolet. The full description of the parameters and assumptions upon which the model is derived can be seen in the MODTRAN report (Berk et al. 1993). A brief description of the geometry and method will be discussed here.

Figure 6.2. Slant path through the atmosphere for MODTRAN transmittance and radiance modes.

Figure 6.2 shows the spherical geometry assumed by the MODTRAN model. A path through the atmosphere is defined by the initial and final altitudes, Za and

zb and by the zenith angle 80 at Za. The transmittance and radiance along a path through the atmosphere is dependent on the total amount and the distribution of the absorbing or scattering species along the path. While a single-slab model of the atmosphere is suitable for some calculations (such as the derivation of sky opacity in a single narrow band, as with the tipper results), the calculation of air mass for realistic atmospheric paths requires that the earth's curvature and refraction be taken into account. 6-2. MODTRAN 153

The atmosphere is modelled as a set of spherically symmetric shells, where the temperature, pressure and absorber densities are specified at the layer boundaries. This requires an adequate description of the local thermal and constituent environ ment. The MODTRAN program provides a database of realistic vertical profiles for temperature and gas mixing ratios for most standard atmospheric conditions. It also allows user input profiles, which was necessary for modelling the unique Antarctic atmosphere.

To compute the molecular absorption that incoming radiation experiences on the defined path, a database of the absorption properties of the molecular species in the atmosphere is required. MODTRAN uses the HITRAN 1996 line atlas (Rothman et al. 2003), which contains all lines from the near-ultraviolet (442nm) to the millimeter (10mm) spectral" region that have significant absorption for atmospheric paths. The resulting maximum spectral resolution from the calculation of the band model absorption is 0.05nm (FWHM).

6-2.1 Input Parameters

The tape 5 input file for the MODTRAN program specifies all the variable parameters for the creation of a MODTRAN spectrum. One of the tape 5 input files is shown in Appendix D. MODTRAN offers a range of empirically derived atmospheric profiles to generate the models which can be called from this input file. Alternatively, it allows the user to define their own atmospheric profiles and conditions, including temperature, pressure and relative humidity profiles collected from radiosondes. The maximum number of user input data lines is eighty-nine, and so the radiosonde data (up to 250 lines) was interpolated over the full height of the sonde measurement to reduce the file to the input requirements. 6-2. MODTRAN 154

The model was computed over a spectral range from 600-850nm (roughly the range of the red fibre spectrum of the AFOS). The model allows the user to define

2 the value of the precipitable water vapour column (PWV, in gm/cm ). The line-of sight geometry requires the user to specify the observer height above sea-level, path length and zenith angle. Finally, the wavelength increment, the FWHM of the slit function and degradation of spectral resolution parameters must be defined.

6-2.2 Constructing the Model

The input parameters also allow the user to specify the type of output the model should produce. The standard output files, tape 6 and tape 7, output the record of the model calculations and the standard results file respectively. Tape 6 serves primarily as a trouble-shooting file, where any miscalculations or errors in the input file can be detected and corrected. For example if the precipitable water vapour column is set too high (above the calculated 100% relative humidity given the ambient temperature of the model) the PWV is automatically set to the maximum allowable value for the particular model. The tape 7 output produces the calculated transmittance and radiance of the model as a function of frequency. An example of a standard tape 7 output can be seen in Appendix D.

MODTRAN offers additional output files where the atmospheric transmission is listed as a function of wavelength in nanometers (instead of inverse centimeters) and this was the output format used for all the models discussed here. An example output file is shown in Appendix D. The output file, output.plt, was entered into the IRAF program WSPECTEXT which produces a IRAF FITS file spectrum from an ASCII file of wavelengths and transmittances. An example of the initial MODTRAN spectrum is shown in Figure 6.3. This model was computed for a zenith angle of 30° and a ground temperature of -60°0. The model was computed with

2 the maximum PWV content. At -60°0, PWV(MAX) = 0.76 gm/cm .

The model spectrum was then normalised with a simple spline fit in SPLOT (us- 6-2. MODTRAN 155

T r a n s M i s s i 0 n

600 650 700 750 800 850 Wavelen~th

Figure 6.3. MODTRAN model, for zenith angle 30 degrees, ground temperature -60°0, PWV(MAX) = 0.076 gmfcm2•

N 0

M a 1 i

e d r 1 .a u X

1!1 r b i t r a r !::1 6000 6500 7000 7500 8000 8500 Wavelength

Figure 6.4. Visual comparison of an AFOS spectrum of a Crux, 26th June 2003 (black), and the MODTRAN model (red). 6-3. Fitting the data 156 ing the same method as for the AFOS data normalisation, described in Chapter 5). At this point the MODTRAN and AFOS spectra could be compared visually to ensure that the model was satisfactory and ready for fitting, as in Figure 6.4. The normalised MODTRAN spectrum is shown in red in Figure 6.4 compared to the spectrum of a Crux from June 26th, 2003.

6-3 Fitting the data

The IRAF program RSPECTEXT was used to return the normalised MODTRAN spectrum to ASCII format. The AFOS spectra were likewise converted to text format in the same way. The first subroutine in the FORTRAN (F90) fitting code, chisquare.f90, linearly interpolates the model transmission values to match the exact wavelength bins of the AFOS data. The core program chisquare.f90 can be seen in Appendix E. Once the model and data were binned identically, the program compared each set of data and model values, using a least-squares method where x2 is given by:

(6.8)

where d(yi) is the data point, m(yi) is the model point and CTi is the uncer tainty in the observed data point, d(yi)· The program returns the chi-squared value for the particular model-data spectral comparison. A program subroutine calculates a fit for any number of models, with varying parameters, to each data spectrum. Once the MODTRAN program and fitting routine were debugged, the temperature, resolution and water vapour content variables in the MODTRAN model were set as free parameters, and chi-squared minimisation of the fits to the AFOS spectra allowed a determination of the best model for each day of observation. 6-3. Fitting the data 157

6-3.1 Determining the Resolution

MODTRAN can produce an absorption spectrum at higher resolution than the AFOS data. It was first necessary to determine the appropriate slit function value for the model to best approximate the AFOS resolution. The MOD TRAN program convolves the computed model with a slit function, where the user can select the full-width-half-maximum (FWHM) resolution of the resulting spectra.

N 0 I I r 1.75 - m a l i 1.5 - s ~: ~ e ~: d 1.25 - t: - f l u 1 r , X ~ .75 i- - a r b .5 i- - i t r .25 i- - a r I I I I I I !:l 600 650 700 750 800 850 Wavelength (nanometers}

Figure 6.5. MODTRAN modt>l with three different resolutions, slit width lOnm(blark), 25urn(bluc) and 50mn(rcd). Offset iu y-axis is for case of comparison ouly. ,

The maximum FWHM for the MODTRAN slit function is limited to 50 times the calculation bin size, which is 1 cm-1. The resolution (in cm-1 , which is the fixed unit for the MODTRAN input file) was set as a free-parameter in the MODTRAN input file, and a series of models with decreasing resolution were fit to the AFOS data. An example of the effect of decreasing the resolution of the spectra is seen in 1 Figures 6.5 and 6.6. The lowest possible resolution is 50cm- , which is the black curve in Figure 6.6. 6-3. Fitting the data 158

H 0 r 1,75- m a l i 1.5- s e d

f l u X

.75 a r b .5- i t r .25- a r !,1 755 760 765 770 775 Wavelength (nanometers) Figure 6.6. MODTRAN A-band atmospheric oxygen absorption at three different res olutions: slit width lOnm(black), 25nm(blue) and 50nm(red). Offset in y-axis is for ease of compaxison only.

200 I ' I ' I I ' PolynomiQI fit porornetcre: I 'Q I P[O] = 55.::51.944824 +- 240.728666 I ' I F(1) = -215.617584 +- 10.257103 I 180 '' I ' P(2) = 2.139673 +- 0.109099 I ' \ ' I Chi-::quorod 0.234468 I ' I ' I ' I -o 'Q { ~160 Best Res: FWHM=50nm I 0 I ' ::::> ' { cr ' ' I Cfl { ' { I ' { ' ' I :.c { U140 hi I I ' \ \ ' \ \ ' I \ ' \ I ' CJ.. ' 120 ' ' ' ' ' '

~ ' 100 45 50 55 Slit Function (FWHM in nm) Figure 6. 7. Chi-squaxed minimisation of MODTRAN spectral resolution. The best resolution for fit to AFOS data was with a slit function FWHM of 50nm. There axe no data points to the right of the curve as this is the maximum slit function value allowed in the MODTRAN program. 6-3. Fitting the data 159

The chi-squared values returned by the chisquare.f90 program were then plotted and a polynomial fit determined the best slit-width value for the model to be 50cm-I, as seen in Figure 6.7. No points are plotted beyond 50cm-1 as this was the minimum resolution that could be calculated with the MODTRAN program. A resolution of 50cm-1 corresponds to a resolution of 2.8nm at 750nm (in the centre of the red fibre wavelength range), which is very close to the AFOS's spectral resolution of 2.4nm. This result confirmed that the model was closely approximating the atmospheric spectrum observed with the AFOS, and that the chisquare.f90 program was working.

6-3.2 Modelling the temperature

The deep atmospheric absorption band of molecular oxygen, centred at 760nm, provides a good test of the fitting method and model parameters. The depth of this line is independent of the relative and absolute PWV in the atmosphere, and is solely a function of the vertical temperature profile.

Both the radiosonde profiles and NOAA meteorological records provide the daily ground temperature, T a, at South Pole. In the following discussion, each radiosonde profile is referred to by its ground temperature measurement, T 0 . An average South Pole radiosonde profile was modified to create seven different models, with temperature profiles shifted by 6.5°0 from Tamin=-77°0, through to Tamax=-42°0. The temperature at each altitude increment in the profile was altered by 6.5°0, therefore the shape of the radiosonde profile was identical for each model. The seven models can be seen compared with an AFOS spectrum in Figure 6.8 and Figure 6.9.

Figure 6.9 shows the oxygen A-band for the seven MODTRAN temperature models compared to an AFOS spectrum of a Crux from the 26th of June, 2003. 6-3. Fitting the data 160

H 0 1 T' m a 1 .9- 1 s e d

F 1 u X

a r b 1 t r a T' 1,1 6000 6500 7000 7500 8000 8500 Wavelength

Figure 6.8. AFOS data(red) and seven MODTRAN models with temperature profiles varying from TGmin=-77°0 up to TGmaa:=-42°0.

The oxygen band depth is less dependent on the temperature variation between the models than the water vapour bands. The agreement between the MODTRAN A-band and the AFOS data is excellent, indicating that the resolution and zenith angle parameters of the synthetic spectrum accurately model the AFOS observations.

The relative humidity, zenith angle and resolution were all fixed for this fit. The effect of decreasing the temperature on the water vapour content of the atmosphere is observed in Figure 6.10 and 6.11. These figures show a normalised AFOS spectrum of a Crux, plotted against the seven MODTRAN temperature models

for the 720nm and 820nm H2 0 bands respectively. The relative humidity remains constant but the absolute humidity decreases as a function of ambient temperature. This is observed in the plots as the depth of the absorption band can be seen to decrease as the ambient temperature of the model is lowered.

The AFOS dataset was fit with the seven models using the chisquare.f90 program, and the resulting chi-squared values were minimised with a polynomial 6-3. Fitting the data 161

H 0 r m

1" i s e d

F 1 u X

a r b 1 t r a r .4~-~--~------~----~----~------~----= !,! 7550 7575 7600 7625 7650 7675 7700 Wavelength

Figure 6.9. The oxygen absorption A-band: AFOS data(red) and seven MODTRAN models with temperature profiles varying from Tamin=-77°0 up to Tam==-42°0. fit, as seen in Figure 6.12. This set of plots shows the polynomial fit to the red spectrum of each star at 760nm. The green (10 seconds), red (30 seconds) and blue (60 seconds) curves are fits to spectra with different exposure-times. For each day of observation the best fit ground temperature, T a, was obtained for each object.

The four days of stellar spectra analysed here were collected over time periods between three and six hours (dependent on conditions and satellite connectivity). The NOAA database provides hourly meteoro logical data from South Pole for each day between 1977 and 2004 at http://www.cmdl.noaa.gov /info/ftpdata.htmL The NOAA ground tempera ture measurements (National Oceanic and Atmospheric Administration 2004), were averaged over the time period of observations for each individual day. This average is plotted in Figure 6.13 and 6.14 in green. Figure 6.13 plots To calculated by the fits to each stellar spectrum as a function of the object zenith angle. The average of the To fits for each day is plotted in Figure 6.13 and 6.14 in red. No systematic fitting error is observed. Good agreement was found with the NOAA met data. 6-3. Fitting the data 162

.9~----~------~----~------._ ____ ~ ______._ ____ ~ 7050 7100 7150 7200 7250 7300 7350 7400 Wavelength {angstroms)

Figure 6.10. The H20 absorption band at 720nm: AFOS data (red) of a Crux from the 26th of June 2003, and seven MODTRAN models with temperature profiles varying from Tcmin=-77°C up to Tcmax=-42°0.

1.02

s;: .98 ..-:::: .$. ::; a .96 "E $ ~s :IE .92

.9~----~------~----~------~----~------~----~ 8050 8100 8150 8200 8250 8300 8350 8400 Wavelen~th (angstroms>

Figure 6.11. The H20 absorption band at 820nm: AFOS data (red) of a Crux from the 26th of June 2003, and seven MODTRAN models with temperature profiles varying from Tcmin=-77°C up to Tcmax=-42°0. 6-3. Fitting the data 163

~300 L 0 J 0" \\ f200 \o, _[ 0 ~\ 100 \\~. 0

3rd July Bet Cen i 600 Tmin = -65.17+1-, .3 de 00 ,' "0 ill L 0 J

0"400[)) 400 400 I ·-_[ 0 200 200 200

400 400 \ 3rd July Alp Psa 3rd July Bet Car Tmin = -61.84+/-0.5 deg -o Tmin = -65.17+/-0.5 deg ~ 300 300 0 J fJ \ m 200 ' 200 I ~ * _[ _[ 100 u ~. u 100 ' 100

-40 -60 -80 -40 -60 -80 -40 -60 -80 Temperature Temperature Temperature Figure 6.12. Chi-squared minimisation of the MODTRAN models fit to the A-band oxygen absorption at 760nm for the AFOS observations on the 27th of June, 2003. Each set of curves is a different stellar object. The individual curves are exposures of different lengths. 6-3. Fitting the data 164

-60 .-----~----.-----~----.-----. .---~-----.----~----.----. -60 26TH JUNE 27TH JUN 0 -62 -sf> Q) Q) -o"" -o"" -----64 0 -64 e 0 e ....::> 0 Qslto AVE: -65.14 degC ::> E'-66 NOM AVE: Q66.5 degC NOAA"'AVE: -65.5 degC -6~ Q) ~ ~~0=-=0~==~~======~==~==~ 0.. E 0 E ~- Dolo AVE: -66.7 degC 68 0 0 -6~

0 -70 ~----~----~----~----~--~ ~--~----~----~----~--~ -70 20 40 60 20 40 60 Zenith Angle Zenith Angle -60 .-----.-----.-----~----.-----. .---~-----.----~----.----. -60 28TH JUNE. 3RD JULY u :l u o>62 -6~ Q) Daia..AV.E..:..._=63..0_de_gc_ ____,:t __~--~· ~· ~ Q) -o ~ '--"'-64 e 0 -6~ ::> ) 0.... 0 ::> ~-66 fNOM AVE: -67.4 degC 1-~-0-AA--P:...<.VE':-:---6-5-.7---d-eg-C------::1 -6~ Q) 0.. 0 0.. ~68 0~ 0 1- 0 -6£1- !octo AV'?.: -67.37 degC -70 -70 20 40 60 20 40 60 Zenith Angle Zenith Angle Figure 6.13. Plots of Tmin resulting from each chi-squared minimisation of the MOD TRAN model to the AFOS data for four separate days of observation in June/July 2003, as a function of zenith angle of the observed object. The average Tmin from the data is in red, compared to the NOAA measured ground temperatures for that day in green.

-60 -60

0 -62 o> -sf>o> Q) Q) -o 0 -o ';-64 0 -6\;' '- 0 0 0 __ '- ....::> Dote A~: -§?.14 ~- ::> E'-66 NOAA AYit.: -66.5 degC NOAA AVE: -65:5 degC"' -6(1§ Q) Ql 0.. =···· ~ . - 0.. 0 0 0 E Dote AVE: -66.7(;5legC E ~68 0 -6~ 0 -70 -70 D 2 4 6 8 10 0 2 4 6 8 10 Obs number Obs number -60 -60

u 0 u o>-62 -6~ Ql ..Dola..AILE.:.....:.fi3 0 degC 0 v -o -o ';-64 -64 ,_ 0 e ::> 0 0 0 ::> .... (") ..... ::'-66 NOAA AVE: -67.4 degC -61:2 Q) "NOAA AVE: -65.7 degC Q) 0.. 0 0. 0 g_68 0 -s& 1- 0 0 1- Dolo AVE: -67.37 degC -70 ~~--~~--L-~--L-~--~~~ -70 0 2 4 6 8 10 0 2 4 6 8 10 Obs Number Obs number Figure 6.14. As in Figure 6.13 but plotted in order of time of observation. 6-3. Fitting the data 165

The stellar fits were replotted in Figure 6.14 in the order the stars were observed. The fits are again compared with the averaged NOAA temperatures. The stellar fits did not appear to be sensitive to hourly changes observed in the NOAA data, mainly a result of the fact that the variations were very small on clear observing days. On the 27th and 28th of June, when conditions were very clear, the NOAA ground temperature did not vary by more than a degree over the time of observations. On the 3rd of July the conditions were poor, with patchy cloud and blowing snow. This day of observation shows the largest disparity between the NOAA met data and the AFOS derived temperature, which is likely a result of the poorer conditions.

These fits were encouraging as they showed that the MODTRAN model parameters for each object were accurate (eg. the line-of-sight geometry, calculated for each object as a function of zenith angle) and that the chi-squared fitting program was working as expected. Finally, the good agreement of the fitted results with measured temperatures at South Pole showed that the model was a good approximation to the conditions above the plateau and ~hould be sufficiently accurate to determine other properties of the atmosphere, such as the precipitable water vapour content.

6-3.3 Water Vapour Bands

Once the method was tested, individual MODTRAN model files were created for each object, as each star had different zenith angles and required a different line of-sight geometry calculation. Each object therefore had a different MODTRAN input file for each day of observation. Once these input files were constructed all input parameters in the model were fixed, with the exception of the PWV content which was varied from a minimum value to the maximum PWV content possible (100% RH for the ambient temperature of each model). 6-3. Fitting the data 166

H 0 1.01 r "'a 1 ~ "'e d

F 1 u X

"r b i t r a r !,J 7100 7200 7300 7400 Wavelen~th (an~~tromc)

Figure 6.15. 720nm. H20 band fit (red) of Epsilon Centauri (black), 27th June 2003.

H 0 1.01 r "'a l i "e d

F l. u X

a r b i t r "r !,J 7100 7200 7300 7400 Wovelon~th (on&strom~>

Figure 6.16. 720nm H20 band fit (red) of Alpha Piscis Austrini (black), 28th June 2003.

Two examples of the fits to the 720nm H20 band are seen in Figure 6.15 and

6.16. An example of the fit to the 820nm H20 band is seen in Figure 6.17. The main problem was slight variations in the stellar spectra which had features near to the water vapour bands being fitted. Alpha Centauri, for example, had a number of small stellar absorption features near both water vapour bands, and the fits to the spectra were degraded as a result.

The resultant chi-squared values for each fit was plotted as a function of the PWV in each model, and each plot was fitted with a polynomial. An example of 6-3. Fitting the data 167

1'1 3..01 0 r "'.. 3. i .." d f' l u X

r" b i t r

"r .96 '----i---l----''-----'-----'-----'----' >l 9050 8100 8150 9200 9250 9300 8350 9400

Figure 6.17. 820nm H20 band fit (red) of Alpha Piscis Austrini (black), 28th June 2003. the fits can be seen in Figure 6.18. The object is Epsilon Centauri, fitted for four separate days of observations. The results of each fit were then piotted for each day of observation, as in Figure 6.19. The depth of the H20 lines is just sufficient for detection of the lines, and the fits for some objects are likely affected, as with Alpha Centauri, by small stellar spectral features near the atmospheric absorption bands.

In addition. tho polarisation effects discussed in Chapter 4 arc of the order (5%) of the absorption band depths, and this oscillation may be apparent in the fits. This may cause an overestimation in the depth of the absorption bands, and therefore an overestimation of the PWV. Figure 6.19 shows that the 720nm fit consistently estimates a higher PWV than the 820nm line. One reason for this may be the polarisation-dependent oscillation in the red spectra when the object spectra are misaligned in the red fibre. A trough in the oscillation is observed at 720nm, which may account for the deeper absorption band in this region.

A second possible reason for the shallower band at 820nm could be the light leakage from the dichroic beamsplitter, discussed in the last section of Chapter 5. The largest peak in the light leakage (over 5%) is observed between 830-900nm. 6-3. Fitting the data 168

100 ..,..----,-.--~--...,---~---. 100 * 26th June Eps~ 27th June EpsCen \W; 0.03865 gm/~- ..-/. PIW: 0.02834 gm/cm2 "D 80 ...... ____~---- -~ BOo ~ ~ 0 0 ::::> :J r::r r::r (/) Ui I 60 SQ.!. :c ..r::: (..) (..)

40 40

0.02 0.04 0.06 100 28th June EpsCen 3rd July EpsCen P\W: 0.03579 gm/cm2 P\W: 0.03661 gm/cm2 80 "' 80~ -g1... 0 0 ::::> :::> 0" 0" (/) (/) 601 :cI 60 :c (.) (.)

40 40

0.02 0.04 0.06 0.02 0.04/ 0.06 H20 gm/cm2 H20 gm cm2 Figure 6.18. Chi-squared minimisation fits of 720nm H20 band for Epsilon Centauri 2 for four different days of observation, and resultant PWV(gmjcm ) for each day from the fits. The green (10 seconds), red (30 seconds) and blue (60 seconds) curves are fits to spectra with different exposure-times.

0.05 .--~-.--~-.--~::-:-:--r::-::-:-::1 0 26th JUNE 27th JUN o.os

720nm fit 0 ...... N" '§o.04 kv::::e=ro-=ge;------.:.:.x_____ ~ 720nm fit 0 o.o4E 0 tJ ...... 0 k~~e~rC-Q0------4~------·~ ...... E -a20nm fit X E 3 X .o-.tunm 1 ,__.o> ~0.03 o.o:§ D.. D..

0.02 '---~--'---~---'--~--'----' 0 2 4 6 0 2 4 6 Obs number 0.05 Obs number 0.05 28th JUNE 3rd JULY

N' p20nm fit §o.o4 Co..----~-Q.--..c-- •.0 ~20nrrQfit ...... :---~--<>--o------E 11.veroqe ..:!: veroge

)( ~20nm fit X X ~0.03 X 0.0~ D.. 0.. is20,m fit

X 0.02 0 0.02 2 4 6 0 2 4 6 Obs number Obs number Figure 6.19. Resultant PWV(gmjcm2) for each object plotted in order of observation, for for different days of observation. The average of the 720nm fit(red), 820nm fit(green) and average(blue) for each day is shown. The circles (10 sec), stars (30 sec) and crosses (60 sec) are fits to spectra with different exposure-times. 6-3. Fitting the data 169

This leakage flux will decrease the depth of absorption lines in the region, and could result in a lower PWV value fit for this absorption band. Objects at higher zenith angles (and so deeper absorption profiles) were better estimated than those at low zenith angles, with smaller absorption features.

6-3.4 Comparison with 350 micron data

The South Pole 350J.Lm tipper data is available via the web (NRAO /CFA 2004). The plot of the sky opacities from the 26th of June 2003 is shown in Figure 6.20. The data was collected from the exact time periods during which AFOS obser vations were carried out. The data from this time period was averaged and this

2 sky opacity value plotted against the AFOS averaged PWV, in gm/cm . The plot of this comparison can be seen in Figure 6.21. The error in the tipper opacity is the standard deviation of T during the AFOS observing period for each day of observation. The error in the AFOS data is the standard deviation of the fitted AFOS data from the average PWV calculated and shown in Figure 6.19. The water vapour levels measured by the AFOS observe the same trend as the tipper data for each respective day of observation, confirming that the AFOS can detect daily variations in the PWV of the atmosphere.

2 Figure 6.22 plots the calculated AFOS PWV(gm/cm ) versus the total integrated 2 PWV(gmjcm ) determined directly from the radiosonde profiles using the method described in Section 6.1. This plot shows a strong correlation between the daily PWV measured by the AFOS and that measured by the radiosonde balloons. Though the exact PWV returned by the AFOS measurements is approximately 10% larger than that determined by integrating the radiosonde profile, there is still very good agreement between the two methods.

The slight overestimation by the AFOS data fits is possibly a result of the periodic variation in the red fibre spectra discussed in Chapter 4. This effect was not able to 6-4. Conclusions 170

South Pole 350 micron Tipper ~-4.------r------~.------r------.

1..3-

~ 1.1-

1..0

0-~L------~------L------~------~ 20CJ3 Jun 26 2003 Jun 26 7.00:_, .Jun ;-._:; 7003 JUrJ 26 :2003 JI.Jn '27 O(l :ooz c)6 :O(JZ 1..2: :O:oz :1.s =·~·oz ·~o =·~~oz Date .~nd T Lnc Figure 6.20. South Pole 350p,m tipper sky opacity(T) data for 26th of June 2003. Courtesy of Harvard Smithsonian Centre for Astrophysics webpage (NRAO/CFA 2004). be completely removed in the normalization of the data, and could be deepening the absorption bands. The excellent agreement in Figure 6.22 shows that despite this, the AFOS data can still detect the variations in daily PWV with an accuracy of 10%.

6-4 Conclusions

The reduced and normalised AFOS spectra from the winter season of 2003 were analysed to determine if the observations were sufficiently sensitive to daily variations in the precipitable water vapour (PWV). The PWV above the Antarctic plateau is very low, and is monitored at South Pole by a number of methods as it is of importance for a range of astronomical observations, particularly in the submillimeter waveband. The daily measurements of the water vapour above South Pole, and the ready access to this data for the dates of the AFOS observations, allowed a direct comparison between results obtained by the AFOS and other, well-tested techniques. 6-4. Conclusions 171

1.4 AFOS PWV(ave) vs 350 micron tipper opacity (tau)

...... y = 23.8062x + 0.19106 :::l 0 c 1.3 ..0 ·c::; 0 0.. 0 c 1.2 0 '-- .E(,) 0 1.1 I.{) I"')

0.9 L-~~~~~~~~~~~~~~--~~~~~~~~-L~--~~~~ 0.03 0.035 0.04 0.045 0.05 AFOS PWV(gm/cm2) Figure 6.21. 350f.Lm sky opacity(T) vs PWV(AFOS) in gmjcm2 for the averaged fit to the stellar spectra.

0.045 AFOS PWV(ave) vs Radiosonde PWV ...... N E y = 0.8677x - 0.00351 (,) ...... E ~0.04

o_~

Q) "0 c 00.035 C/) 0 :.:0 0 0:::

0.03

0.03 0.035 0.04 0.045 0.05 AFOS PWV(gm/cm2) Figure 6.22. Radiosonde integrated PWV vs PWV(AFOS) in gmjcm2 for the averaged fit to the stellar spectra. 6-4. Conclusions 172

The MODTRAN atmospheric modelling program was used to model the absorp tion spectrum of the Antarctic atmosphere, using tailored models incorporating balloon-launched radiosonde profiles for each respective day of observation. These models were compared to the AFOS spectra with a chi-squared fitting algorithm. This program was first tested by fitting the data for spectral resolution and ground temperature, both of which showed good agreement with the known value of the AFOS resolution, and the daily ground temperatures measured and recorded by South Pole meteorological instruments.

The models were then generated for a variety of PWV values and the best-fit model determined by the chi-squared program. The PWV for four days of complete observations between the end of June and the beginning of July, 2003, were analysed using this method. To test the validity of the method, these results were compared with the opacity values from a submillimeter tipping radiometer, and the integrated PWV calculated from the daily balloon-launched radiosondes. Both comparisons showed good agreement with the PWV determined from the AFOS data.

The water vapour bands in the AFOS spectra are only just at the limits of detection of the method, and the PWV were overestimated in comparison to the radiosonde results by approximately 10% on each day of observation. This is possibly a result of the periodic variation observed in some of the AFOS stellar spectra and discussed in Chapter 4.

Three of the observed objects (selected for both their brightness and range of zenith angles) had stellar absorption features close to the atmospheric bands of interest, which affected the accuracy of the modelling in these cases. The spectra of stars at low zenith angles, which looked through a shorter optical path in the atmosphere, were not modelled as well as objects close to the horizon, whose atmospheric absorption lines were deeper. Chapter 7

Earth Albedo Measurements

In early August and early September, a series of experiments were conducted with the AFOS to measure the earth albedo by observing the dark side of the moon. The project was undertaken for two reasons: the science was unique and had not been attempted previously from South Pole, and secondly, continuous observations of the moon over periods exceeding 24 hours were ideal to develop and test the automated observing routines designed for the AFOS.

7-1 Earth's signature

Earthshine is the term used to describe the illumination of the dark side of the moon by reflection from the sunlit earth. In essence, it is the same process as looking in a mirror. When a person looks at their reflection, their face is illuminated by a light source and they observe the portion of that light that is reflected to their eyes in the glass. When the dayside of the earth is illuminated by the sun, a percentage of this light is reflected onto the lunar surface and then reflected back again through the earth's atmosphere to the observer.

The geometry of earthshine measurements is illustrated in Figure 7.1. The 7-1. Earth's signature 174 reflected light of the earth illuminates both the bright and dark sides of the moon, but on the sunlit side, the sun's reflection greatly exceeds the much fainter signal from the earth. The dark side of the moon reflects only the earthshine, though any observation of the moon will also be affected by scattered light from the bright side of the moon, and this signal must be removed to recover the spectrum of the earthshine.

Why observe earthshine? The earthshine spectrum, as previously mentioned, is the sun's emission after having traveled through the earth's atmosphere to the surface, and then back through the atmosphere to the moon. As such, it differs from the sunlit lunar spectrum as it has undergone absorption and scattering by the constituents in the earth's atmosphere as well as the land, ocean and vegetation on the earth's surface. Satellite observations can also observe this signal but they can only observe a small percentage of the earth's surface at a time. The moon, however, acts as a diffuse reflector of the entire sunlit portion of the earth in its field of view. The ratio of the earthshine spectrum to the bright moon spectrum thus produces a spectrum of the integrated global albedo of the earth at one instant in time. This spectral ratio is known as the earth albedo.

7-1.1 Previous Studies

Measurements of the earth albedo have been conducted for over seventy five years. Danjon (1936, 1954) collected over two hundred measurements of the earthshine between 1926 and 1930 and these measurements were continued by his colleague, Dubois until1960 (Dubois 1947). Danjon and Dubois used a double aperture 'eat's eye' photometer which simultaneously observed the dark side of the moon and the sunlit portion (with the light stopped down to match the intensity of the dark side). Danjon's measurements provided the first description of the earth's radiation budget on a global scale, and though he found no long-term trend in the earth 7-1. Earth s signature 175

Solar radiation

Solar radiation ' • 4

' \ , ..,...... - ·-·-·-· \ ,. ; ", .. , · ,. \ i \ / I I .1 I 1 • Bright side obs I ! I \ ! I .I 2 '" Dark 1 obs ~ ! I I 3 =Dark 2 obs 1 I I I 4 =Scattered light obs j I I j I I i I I \ I I I \ \ I I \ \ .' \ \ I ' i ~ ,. ~ ,. ·, ' ... , ...... ·"· I ' · ~ . .... -- ·-·-·-_._...... ,' Lunar orbit Atmosphere I I I Figure 7.1. Diagram (not to cale) showing how the earthshine is observed from the South Pole. Viewpoint is directly downwards over the South Pole. The labels 1-4 indi cate t he four positions of the AFOS beam in ach seri s of lunar ob ervations. A set of ob ervations, (three exposures at each position) was conducted every twenty minut in a 24 hour period. albedo over a twenty year period the data still recorded some ignificant result including detecting a large change in the earth albedo during 1941-42 as a result of El ino (Qiu et al. 2003).

In the 1960 atellite observations of the earth atmo phere uperseded this method but in recent year a number of groups have revived the tech nique (Arnold et al. 2002; Woolf et al. 2002· Goode et al. 2000) to achieve a variety of cientific goals. Goode et al. (2000 2001) and Qiu et al. (2003) ob erved the earth albedo with accurate ( < 1%) photometric measurements of the moon between 199 and 2002 at Big Bear Solar Ob ervatory in Colorado. Combined 7-1. Earth's signature 176 with a detailed theoretical simulation of the earth's radiation budget, Goode et. al. derive a method of measuring the instantaneous large-scale reflectance of the earth. These measurements can be compared and contrasted with existing satellite and ground-based data to observe large-scale, and long-term trends in the earth's climate.

Spectroscopic measurements of the earth albedo have now become the focus of interest in the search for life on extrasolar planets (Arnold et al. 2002). Though detection of earth-like planets is still some years in the future, several studies are underway to isolate the features in a planet's spectrum that may indicate the presence of life. One such feature is the surface colour of a planet produced by vegetation. The spectral signature produced by terrestrial vegetation is known as the 'vegetation red edge'(VRE) which is an increase in the earth albedo spectrum beyond 720nm as a result of reflection at these wavelengths by bacteria and plants, which absorb strongly at shorter wavelengths (Arnold et al. 2002; Woolf et al. 2002). A useful test of the method is to observe the integrated light from the earth and examine the earth albedo for the VRE produced by our own planet.

Measuring the earthshine allows the observer to view the earth's radiation diffused over the entire lit surface, which simulates the spectrum one would observe of an extrasolar planet. Arnold et. al. observe a VRE of between 5 and 7%,

comparing favourably to theoretical predictions (~ 5% in Schneider et al. (2000)). Woolf et. al. detect a VRE of approximately 6%, also indicating that the method shows promise for future spectroscopic observations of extrasolar planets. The experiments undertaken by Arnold et. al. and Woolf and Traub formed the basis for the observations taken with the AFOS and are described in detail in Arnold et al. (2002) and Woolf et al. (2002).

All of the previous observations of earthshine discussed here were limited to individual observing periods of six hours or less as a result of their mid-latitude 7-1. Earth's signature 177 locations. Goode et. al. (Goode et al. 2001) propose observations of the earthshine from the South Pole to study the variation in the earth albedo over a continuous 24 hour period. In a continuous observation of this length, changes in the earth albedo can be correlated to the amount of ocean, cloud and vegetation illuminated by the sun and reflected on the moon's surface as the earth rotates. A detection of the variation in these features in the earthshine spectrum is significant for studies of the planet's radiation budget, and the potential for detecting life on extrasolar planets. The AFOS was therefore uniquely placed to undertake a topi cal scientific goal, and the observations and results of this project are presented here.

7-1.2 Method

A description of the method for arriving at the earth albedo is taken from Arnold et al. (2002). If the solar spectrum as observed outside of the earth's at mosphere is given by S(>..), earth's atmospheric transmittance as AT(>..), Moonlight MS(>..), Earthshine ES(>..), Moon albedo MA(>..) and Earth albedo EA(>..) then the Moonlight, MS(>..) can be given by:

MS(>..) = S(>..) x MA(>..) x AT(>..) x 91 (7.1)

The Earthshine, ES(>..) can be written:

ES(>..) = S(>..) x EA(>..) x MA(>..) x AT(>..) x 92 (7.2)

The Earth albedo, EA(>..) is then given by the ratio of these two equations:

EA(>..) = ES(>..) x 91 (7.3) MS(>..) X 92

In these equations g1 and g2 are geometric factors that pertain to the relative sun-earth-moon positions. Arnold et al. (2002) set these factors to 1, equivalent to a spectrum normalisation, and this assumption is used for the measurements presented here. 7-2. Observations and Reduction 178

Arnold et. al. quantify the vegetation signature, VRE as follows:

(7.4)

where rR and r1 are the mean refiectances in the 600-670nm and 740-800nm windows respectively. The AFOS was not designed for accurate spectroscopic measurements such as those accomplished by Arnold and Woolf, however it was reasoned that while an absolute measurement of the VRE might not be possible, a relative variation in the earth albedo over a 24 hour period could be observed, provided the local conditions at the South Pole remained clear during this time.

Figure 7.2 (Walker 2004) shows the simulated image of the earth as seen from the moon during the 24 hour period of observations on September 4th and 5th, 2003. The land/ocean ratio in the illuminated portion of the earth is continually changing, and a comparison of the earth albedo over a 24 hour period should reflect the change in this ratio.

7-2 Observations and Reduction

A peculiarity of the Sun-Earth-Moon geometry at the geographic South Pole is the fact that during the austral winter the phase of the observable moon never decreases below half-full (¢ = ±90°). Throughout the year, the moon rises and remains up for two weeks at a time, and then sets for two weeks. Therefore, during the winter the observable lunar phase only varies from -90° to +90°. Most earthshine measurements are taken during a crescent moon to minimise the amount of light from the sunlit lunar surface that is scattered onto the dark side of the moon during observations.

This reduced the potential observing periods for 24 hour observations to the first 7-2. Observations and Reduction 179

Sept 04, 2003, UT 20:00 Sept 05, 2003, UT 02:00

Sept 05, 2003, UT 08:00 Sept 05, 2003, UT 14:00 Figure 7.2. Four imulations of the illuminated eru·th as viewed from the moon for the AFOS observing period 4th/5th of September 2003. These simulations images are available at http:/ /www.fourmilab.ch/earthviewfvplanet.html (Walker 2004). weeks of August and September of 2003 when the moon rose at approximately half-full and increased to three-quarters full whereupon the sunlit lunar signal was too bright for the earthshine spectrum to be detected.

The automated ob erving cripts de igned to run the AFOS and G-mount for these observation are described in Chapter 5. Each set of observations consisted of an automated program that firstly ob erved the bright limb of the sunlit moon (position 1 in Figure 7.1) then two separate patche on the dark side of the lunar surface (position 2 and 3, offset from ea h other by approximately two arcminutcs the third po ition about 30" from the edge of the moon) and finally a cattered light pectrum (po ition 4 blank sky an arcminute from the dark edge of the moon). 7-2. Observations and Reduction 180

At each position, three measurements of 0.1, 1 and 10 second exposure time lengths were taken. At the beginning and end of each observing period a set of quartz-halogen and bias exposures were taken. In this way, each set of observations took approximately six minutes and the script was run once every twenty minutes via a cronjob.

Unlike the stellar observations detailed in previous chapters, the moon illumi nated all five operational fibres and all data were reduced separately to examine the comparative response of the fibres, as well as to increase the data pool.

No data was taken during the first observing window in August, 2003, as a result of a recurring mechanical fault in the telescope mount drive. During the second window, a five day period from September 1st to September 5th, 2003, the mount drive was working sufficiently well to run the automated programs for the following periods

• 20:00 UT 1/9/03 - 06:00 UT 2/9/03

• 19:00 UT 3/9/03- 9:00 UT 4/9/03

• 20:00 UT 4/9/03- 21:00 UT 5/9/03

The first two periods were unfortunately affected by cloud-cover at South Pole. The third period, the only full 24 hour set of observations, consisted of 600 CCD images (3000 individual spectra).

Data from each fibre were extracted and reduced using the methods detailed in Chapter 5. Following this reduction, two scripts in IRAF were written to calculate the ratio of the spectra and produce the earth albedo signal described in the previous section. The scripts, divvy.cl and ratio.cl are shown in Appendix F. The first script, divvy.cl placed each set of observations (consisting of 12 spectra in each fibre) into separate directories, and then created three subdirectories 7-3. Results 181 separating the spectra by exposure time length. Thus, the spectra were divided into directory listing eg. /fibre4/obs31/10s/a402.fits, where each subdirectory had four observations, a bright limb spectrum, two dark moon spectra and a scattered light spectrum, all taken at most four minutes apart.

The second script, ratio.cl operated in each of these subdirectories as follows. The scattered light spectrum was subtracted from all three lunar observations using the spectral arithmetic program SARITH, such that:

MS(>-.) - Bright(>-.)- Scat(>-.) (7.5) ES(>-.) - Dark(>-.)- Scat(>-.) (7.6)

The script then used SARITH again to take the ratio of ES(>-.)/MS(>-.) and produce a spectral ratio EA(>-.) for each observation.

7-3 Results

The first step required in analysis of the data was to ascertain if the local conditions at South Pole remained consistent during the 24 hours of the observing period. As described in previous chapters, this was non-trivial as there was no pointing camera on the telescope, and the automated observations had not been able to be monitored in real-time as a result of the short period of satellite connectivity.

The moonlight spectra, MS(>-.) were analysed to assess this concern. A single, red fibre was selected, and each bright moon spectrum was plotted in IRAF using the interactive plotting program, SPLOT. A feature of SPLOT allows the user to mark a region of the spectrum and calculate the total integrated flux in the wavelength band selected. This calculation was performed on each MS(>-.) and the integrated flux plotted as a function of time. This plot is shown in Figure 7.3. 7-3. Results 182

+ + + 5 3.5X10 + + + + + +

+ ++ + + + + + + + + + + +++++++++ ++ + + + +

1.5x1o5 L.....~~-'-~-~....__._~~-'-~-~....__._~~_.__~-~....__._~~_.__~-~,__j 0 100 200 300 400 500 600 700 800 Tirne (rninutes) Figure 7.3. Plot of the total integrated flux (arbitrary units) of the moonshine, MS(.A) in the central red fibre as a function of time from 20:00 UTC, 4th September 2003. As can be seen, cloud/blowing snow affect the observations severely beyond the third hour of observations.

As can be seen, cloud and/or blowing snow severely affect the spectra beyond the third hour of observations. The gradual increase in MS(-\) prior to this point

is a result of the increasing phase of the moon (the factor g1 in Equation 7.1). Though this was disappointing, analysis was conducted on the short section of data taken during clear conditions to determine if the method (and AFOS sensitivity) was sufficient to detect the earth albedo.

7-3.1 Scattering

As mentioned in the previous section, the script ratio.cl took the ratio ES(-\)/MS(-\) as a function of exposure time, ie, the 10 second ES(-\) spectrum was divided by the 10 second MS(-\) spectrum in each observation set. An example of the resulting ratio in the red fibre can be seen in Figure 7.4. It was dismaying to observe a periodic variation in the spectra of identical phase and peak position to the effect produced by object misalignment discussed in detail in Chapter 4. 7-3. Results 183

I I I I I -

.2-1 I I I 6000 6500 7000 7500 BOOO 6500 9000 Wuvclcnglh (angstro,..t,.)

Figure 7.4. Ratio EA(A.) in the red fibre, obtained by dividing each 10 second ES(.A) spectrum by its corresponding 10 second MS(.A) spectrum. The peak positions and period of the large-scale variation seen are identical to those observed in misaligned stellar images, calculated in Chapter 3.

A stellar object is a point source, and the periodic variation in the spectra was observed when the source was positioned near the edge of the fibre core, instead of in the centre. The moon is an extended source, and it is therefore not possible to 'shift' the lunar image across the fibre face in this way.

Secondly, when observing a star this variation changed in phase as the image was shifted from one side of the fibre core to the other. Though a change in amplitude of the oscillation was observed in the ratios of the lunar spectra, this was a function of the exposure time length and no change in the phase of the variations was seen.

The identical shape of the periodic variation to that observed in the stellar spectra however, indicated that the cause may still be a polarisation effect, recalling that the period and peak position of the variation in the stellar spectra, discussed in Chapter 4, was a result of the different transmission function of the S state to the P state of polarised light as it was transmitted through the dichroic beamsplitter. 7-3. Results 184

In Chapter 4 it was proposed that the spectral attenuation observed when a stellar image was positioned on the edge of the fibre core was a result of slightly higher transmission of the P-state of polarised light to that of the S-state, at either the fibre face or in the propagation of light through the fibre to the spectrograph.

In the case of the lunar spectra, it seemed that the periodic variation was dependent on the exposure time length, and the position on the moon that was observed. The amplitude of the oscillation increased as the exposure time length increased. The amplitude was smaller when observing the dark side of the moon than when the bright limb was observed. It was therefore most likely that the effect was dependent on the amount of incident light in each observation.

The AFOS was not designed to observe an extremely bright source such as the moon. It is certain that each observation of the bright limb of the moon would produce more scattering within the telescope barrel and optics than the observation of the dark side of the moon taken moments later. It is also likely that light glancing off the edges of the optics - most particularly the entrance aperture to the dichroic beamsplitter - could cause a partial polarisation of the incident light prior to transmission through the beamsplitter.

Consider the case of a 10 second exposure of the bright lunar limb, where the edge of the entrance aperture polarises an additional 5% of the incident light, into the P-state, in comparison with the successive observation of the dark side of the moon. As the two polarisation states are transmitted differently through the dichroic as a function of wavelength, the ratio of these two spectra - like the ratios of the edge-aligned spectra to the centred spectra in Chapter 4 - would display an oscillation such as the one observed in Figure 7.4.

To test this hypothesis, the data were reanalysed but with a modified ratio.cl whereby each 10 second dark moon spectrum was divided through by the shortest 7-3. Results 185

exposure length (0.1sec) bright moon spectrum in each observation set. If th~ scattering theory was correct, the oscillation in the ratio EA(.A) should at the very least be reduced in amplitude. The same dark moon spectrum as shown in Figure 7.4 is shown again in Figure 7.5 when divided through by a 0.1 second bright moon spectrum.

5500 ()000 6500 7000 7500 13000 0500 9000 Wnvelcnglh (angstroms) Figure 7.5. EA(.A) recalculated using a shorter time length bright moon exposure. The oscillations in the ratio are almost completely removed, and features of the earth albedo, prominently the oxygen A-band, are observed.

7-3.2 Earth Albedo Features

The reanalysis of the spectra was successful in retrieving the earth albedo from the earthshine spectrum. The first ten observations sets, spanning ap proximately three hours from UT 20:00 to 23:00 on September 4th, 2003, were reanalysed using this method in fibres 1 through to 5 (three blue and two red fibres).

As can be seen from Figure 7.5, even with the new analysis the large-scale features of the beamsplitter shape can still be seen somewhat in the red spec trum. For this reason it was concluded that resolving the VRE from the spectra 7-3. Results 186 would not be possible with this data. However, a study of the results to determine if this was truly an earth albedo spectrum was still necessary to validate the method.

5000 /,_--" ~ckbody spectrum (6000K)

11-000 ~~:~~~ HI) Mgb ~ ~ .b Ha ~ 3000 I ( i rf:'u Hy B band (atm) (I} ...... 2000 jHS § Ozone (attn) u0 liV'Vlf ~. 1000 c::: 0 / ·1000 5000 6000 7000 11000 0000 Wuvelcngtb (~•ngsb·oms) Figure 7.6. Combined earthshine spectrum, ES(.\), showing both the blue and red fibre spectra, taken during a single 10 second exposure and reduced using the methods detailed in Chapter 4. The blackbody curve, T = 6000K, provides an indication of the continuum emission of the sun as observed outside the earth's atmosphere.

Figure 7.6 shows the reduced earthshine spectrum ES(.\) in the blue and red fibres combined using the spectral adding tool SCOMBINE, which allows a new spectrum to be created from two spectra with different wavelength ranges. The blue spectrum (300-550nm) is an average of Fibres 1, 3 and 5 and the red spectrum (550-900nm) is an average of Fibres 2 and 4, taken in a single expo sure at position three on the dark side of the moon at 20:40 UT, September 5th 2003.

Plotted with the spectrum is a blackbody curve, temperature T=6000K, to approximate the continuum emission from the sun prior to incidence upon the earth's atmosphere. A mixture of stellar features (Ha, H,B and the Mgb line) and atmospheric features (Oxygen A and B band and strong ozone absorption at wavelengths shorter than 550nm) can be seen in the earthshine spectrum. 7-3. Results 187

To determine if the method detailed in Section 7-1.2 was actually producing a detection of the earthshine, the spectra in the two red fibres were analysed in the following way. An average was taken of the bright limb lunar spectrum, MS(.X) in the first ten observations sets using SARITH. An average was then taken of the dark moon spectrum, ES(.X) at the position 3, furthest from the sunlit portion of the moon.

These two spectra were normalised in SPLOT, as described in the last section of Chapter 5. The two averaged, normalised moonshine and earthshine spectra are plotted in Figures 7.7 and 7.8. The ratio of these two spectra, EA(.X), is shown in Figure 7.9.

From Figure 7.9 we can conclude that the sensitivity of the AFOS is sufficient to detect the earth albedo on the dark side of the moon. Atmospheric absorption features, predominantly the molecular oxygen absorption band at 760nm, can be observed in the earthshine spectrum in Figure 7.9, while the stellar absorption features present in both MS(.X) and ES(.X) have been removed. MS(.X), ES(.X) and EA(.X) as shown in Figures 7.7 to 7.9 compare well to a similar data set and analysis presented by Arnold et al. (2002).

The red fibre EA(.X) prior to normalisation is shown in Figure 7.5, and the periodic variation produced by the polarisation of scattered light in the telescope can still be identified. The reanalysis reduced the intensity of the effect so that the key earthshine absorption features can be observed. The scattering effect was strong enough to overshadow variations in the continuum produced by the rotation of the earth, and prevent the observation of the Vegetation Red Edge in the extracted earthshine spectrum.

Figure 7.10 shows the earth albedo EA(.X) calculated by dividing a 10 second ES(.X) to a 10 second MS(.X) in the blue fibre. The earth albedo in the blue should 7-3. Results 188

ra· .9 .5 ~ .8 ~ c:;:; "<::! .7 <1> =aen E .6 z0 .. 5

6000 6500 7000 7500 6000 6500 9000 Wn.Vt"lcngtb (ungst.romN) Figure 7.7. Averaged, normalised bright limb moonshine spectrum, MS(.\).

Wuvc]C"ngtb (nngslroras) Figure 7.8. Averaged, normalised dark moon spectrum, ES(.\).

GOOO GGOO 7000 7500 8000 nsoo !lOOO Wt.lvclcns:th {ang~lroms) Figure 7.9. EA(.\)=ES(.\)/MS(.\). The stellar Ha line at 656.3nm is removed. Re maining are the features produced by the radiation's absorption and scattering in the atmosphere and earth's surface. The deep A-band oxygen line is the strongest feature observed. 7-3. Results 189

.2 1-

.ai.... .15 1- ..s -

I I 0 It I I I 3500 4000 500() 5500 GOOD 1favclcngt.h (angstroms)

Figure 7.10. Earth Albedo ratio, EA(A) calculated using a 10sec bright moon exposure. The data are contaminated by scattering of the bright moonlight within the telescope tube. exhibit a Rayleigh scattering dependence of 1/>..4 (Arnold et al. 2002), which is clearly not observed in this plot.

The larger flux of incident light in the telescope during a bright moon obser vation, in comparison to the dark moon measurement, produces more scattering within the telescope and thus a larger attenuation of the bright moon signal as a function of wavelength than during an observation of the dark moon. As MS(>..) is the denominator when calculating the earth albedo, the resulting ratio EA(>..), as shown in Figure 7.10, shows a rapid decrease at short wavelengths.

As with the red fibre, the blue fibre spectral ratios were recalculated using the 0.1sec exposures to decrease the comparative instrument scattering in the moonshine and earthshine spectra.

Figure 7.11 shows the EA(>..) in the blue, averaged over Fibre 1, 3 and 5 for the observation set collected at 20:40 UT, September 4th, 2003. Observed from space, earth is a 'bluish' color not only because of the large percentage of ocean, but 7-4. Conclusions 190

.65

.825 i .(I ..0 a ~5?5 Cl)

:2l .. (){) Cl) r::::l -0 :p .525 & .5

.4?5

3500 4000 4500 5000 5500 Wavr:lcngth (angsLro:cHI)

Figure 7.11. Earth Albedo ratio, EA(.X) averaged over the three blue fibres in a single exposure at 20:40 UT, Sept 5th 2003. The effect of Rayleigh scattering of the atmosphere can be seen. as a result of Rayleigh scattering in the earth's atmosphere (Arnold et al. 2002; Woolf et al. 2002). This Rayleigh scattering is observed in Figure 7.11.

7-4 Conclusions

Observations of the dark side of the moon were successfully completed in two observing periods in August and September of 2003. These observations firstly confirmed the viability of software programs designed, as part of this thesis, to automate the AFOS observing and allow continuous observations to run outside of times when satellite access was available to the observer.

Secondly, the dark moon observations successfully detected features of the earth albedo in the earthshine spectra obtained. Initial analysis revealed that strong scattering within the telescope barrel and optics produced a polarisation-dependent oscillation in the earth albedo spectra. A method was devised to reduce this effect in the analysis, and the subsequent recalculation of the earth albedo showed 7-4. Conclusions 191 spectral features consistent with those obtained by other research groups (Danjon 1936; Arnold et al. 2002; Goode et al. 2000).

Though the telescope successfully collected lunar spectra for continuous periods of up to 24 hours, all but a three hour period of time on the final observation day was strongly affected by poor weather at South Pole. Thus it was not possible to conduct an analysis of variations in the earth albedo as a function of the changing illumination of the earth as it rotated.

A data set taken during improved weather conditions is required to improve the small set of results presented here. However, this preliminary analysis of a short set of moon observations show that the 30cm AFOS telescope is sufficient to detect the earth albedo. Modifications to the instrument to reduce the internal scattered light would be highly desirable to improve the quality of the observations to the point where variations in the albedo over a 24 hour period would be detectable. Chapter 8

Conclusions

The South Pole and Dome C have been shown to have great potential as sites for infrared and submillimetre observing. Though ground level seeing at South Pole is comparatively poor, new measurements at Dome C have shown the best seeing of any site in the world. The work of this thesis endeavoured to further the understanding of the optical properties of the atmosphere above the high Antarctic plateau.

With the increasing interest in Dome C as a site for optical astronomy, the effect of auroral emission on observations is a question of high importance. A detailed study of the effect of auroral emission on optical observations was conducted. This analysis and results have been submitted for publication in Dempsey & Storey (2004). Analysis of auroral measurements at South Pole have shown that in an average winter season, the B band sky brightness is 21.9 magnitudes per square arcsecond for 50% of the observing time. In V band, the median sky brightness contribution is 20.8 magnitudes per square arcsecond in an average winter. Calcu lations were used to show that at Dome C, the contribution to sky background in B and Vis up to 3.1 magnitudes less than at South Pole. The use of notch filters to reduce the contribution of the strongest emission lines and bands was also calculated. 193

This thesis work included the construction of an infrared cloud-observing instru ment, COBBER (Cloud OBserver), which directs radiation from the sky through a hemispherical ZnSe lens onto a thermopile detector optimised at lOJ.Lm. COBBER collected data at Dome C from February to June of that year. In 71 observing days, only four days of cloud were recorded. The results of this experiment are presented in Dempsey et al. (2003).

The first optical stellar spectra observed from the high Antarctic plateau were taken at South Pole station with the Antarctic Fibre Optic Spectrometer (AFOS). The AFOS was installed at South Pole station on a dual-telescope alt-az mount in January of 2003. Instrument and data analysis from the 2002 and 2003 AFOS observing seasons were presented. The improvement in pointing accuracy as a result of this study allowed a streamlining of the observation techniques to the point where automated observing sessions (up to twenty-four hours in length) were undertaken during August and September of 2003.

A wavelength-dependent spectral attenuation was observed in both the blue and red fibres, and a detailed analysis of this effect was necessary to understand the cause of these variations. This revealed a Rayleigh-scattering effect in the blue fibre when the object is offset from the centre of the fibre. In the red fibre, a position dependent periodic variation was observed in the ratio of a misaligned spectrum to a centred spectrum. This attenuation, which changed in phase as it moved from one edge of the fibre to the other, was shown to be polarisation-dependent, though further investigation is required to ascertain the exact source of the effect. Understanding of this effect led to a tightening of the pointing accuracy constraints for the 2003 observing season and the data quality and consistency was greatly improved as a result. To better understand these instrument effects, tests of the entire system at a temperate site would be desirable.

Two years of AFOS observations were described, including the selection of 194 sources, the design of observing scripts and the creation of a data reduction method for the stellar spectra collected with the AFOS. An attempt to estimate the UV cutoff wavelength of the atmosphere above South Pole was conducted using data from the reduced blue-fibre AFOS spectra. Calibration issues made it difficult to draw any conclusions from these measurements regarding the wavelength of the UV cutoff. Second-order light leakage from the blue into the red part of the spectrum, as a result of imperfections in the beamsplitter, was analysed. It was shown the peak leakage flux was less than 5% of the stellar flux in the 600-900nm region, with a mean flux of 3%.

AFOS data from the winter season of 2003 were analysed to determine if the observations were sufficiently sensitive to daily variations in the precipitable water vapour (PWV). The MODTRAN atmospheric modelling program was used to model the absorption spectrum of the Antarctic atmosphere, using tailored models incorporating balloon-launched radiosonde profiles for each respective day of observation. To test the validity of the method, these results were compared with the opacity values from a submillimeter tipping radiometer, and the integrated PWV calculated from the daily balloon-launched radiosondes. Both comparisons showed good agreement with the PWV determined from the AFOS data.

The dark moon observations successfully detected features of the earth albedo in the earthshine spectra obtained. Strong scattering within the telescope barrel and optics produced a polarisation-dependent oscillation in the earth albedo spectra. A method was devised to reduce this effect and the subsequent recalculation of 8-1. Future Work 195 the earth albedo showed spectral features consistent with those obtained by other research groups (Danjon 1936; Arnold et al. 2002; Goode et al. 2000). A data set taken during improved weather conditions is required to improve the small set of results presented here. Modifications to the instrument to reduce the internal scattered light would be highly desirable to improve the quality of the observations to the point where variations in the albedo over a 24 hour period would be detectable.

8-1 Future Work

Work continues to build a new COBBER instrument. The new version of the cloud detector would be installed with an Automated Weather Station planned for Dome A. A traverse to the site is proposed for the coming austral summer 2004/2005. Cloud cover statistics obtained from such an instrument are vital to characterise the highest region of the Plateau for future astronomical projects.

A continuation of the initial auroral studies presented in this thesis is vital to complete the optical site testing at Dome C. An auroral monitoring instrument consisting of a spectrograph and COD detector is being designed for deployment at Dome 0 in January of 2005. The primary objectives of this auroral monitor are:

• Assessment of the Dome 0 sky background in the B, v; R and I wavebands over (several) winter seasons.

• Identification and intensity measurements of dominant auroral emission lines and bands, particularly in the red and near-infrared spectral regions.

• Evaluation of the spatial distribution and frequency of strong auroral displays at Dome C.

If the results of measurements with this auroral monitor are consistent with the initial estimates of sky brightness at Dome 0, they will further strengthen the case 8-1. Future Work 196 for optical astronomy at the site. In addition, the work of this thesis has shown that automated, unheated instrumentation can be successfully operated in Antarctic conditions. The skies above the vast Antarctic plateau are clear, cold, dry and still. The ice at the bottom the world offers the darkest ground-based window on the universe. It is time to enjoy the view. Appendix A

Auroral Intensity Calculations

A-1 Rayleighs to Magnitudes

An intensity of 1 Rayleigh is defined by the equation (Baker & Romick 1976):

(A.1)

The energy of a single photon is expressed as:

he Ephot = T J (A.2)

where A is the wavelength of light in metres, h is Plancks constant and is equal to 6.626x10-34 Js, and cis the speed of light, equal to 2.99x108 mjs. Substituting these constants into this equation, we can calculate the energy in Watts per metre

2 squared per steradian ( wm- ster-1) for an intensity of 1 Rayleigh:

1.5808 X 10-16 1 Rayleigh= A Wm- 2ster-1 (A.3)

where A is the wavelength of light in nanometres. Now 1 steradian is equal to (32400)/7r2 square degrees, or (4.199 x 1011 )/7r2 square arcseconds so we can express the above equation as:

3.718 X 10-27 1 Rayleigh = A Wm- 2 arcsec-2 (A.4) A-1. Rayleighs to Magnitudes 198

We now consider the contribution an emission line would make to the sky bright ness in a photometric band. To do this we first convert R/nm to Janskys. To convert this value to a flux in W m-2 Hz-1 arcsec-2, we can use the following substitution:

(A.5)

where A is in nanometers, and v is in Hz. Therefore:

27 9 2 3.718 ~ 10- ) X (10cA ) 1 Rayleigh/nm ( A (A.6) 26 2 1 2 - (1.24 x 10- ) x A [W m- Hz- arcsec- ] (A.7) where A is in nanometers, and vis in Hz. Now we can arrive at an expression that can convert the intensity in Rayleighs to a flux in j.tJanskys per square arcsecond as the definition of a j.tJ ansky is given by:

(A.8)

so we can arrive at:

6 2 1 Rayleigh/nm = (1.24 x 10 ) x A [~tJanskys arcsec- ] (A.9)

where A is the wavelength in metres. Let us consider a standard filter band, with a central wavelength of Ac (in metres). Given an emission intensity, I in Rayleighs/nm, averaged across that band, this is equivalent to a flux in j.tJansky per square arcsecond of:

6 2 F = I x (1.24 x 10 ) x Ac [ j.tJanskys arcsec- ] (A.10)

where Ac is the wavelength in metres. This can then be converted to an intensity in magnitudes/ arcsec2 by the expression:

2 Magnitudesjarcsec = 20- 2.5log10 (:) (A.ll)

where F is the average flux in the band in ~tJanskys per square arcsecond, and K is 42.6, 36.4 and 30.8 for the B, V and R bands respectively (Bessel 1979; Benn & Ellison 1998). Appendix B

Observation Logs and Scripts

B-1 G-mount Command Logs and Scripts

B-1.1 Standard G-mount command log

A standard logfile of G-mount command line control of a manual observing run is shown in this example from June 26th 2003:

Escape character is 'A]'. cgmount afos Command ok axes on Gmount axes are ON at 03:47:35(UT) cfile cfile c020621 Pointing correction set to c020621.cfile Pointing correction updated at 03:48:56(UT) cfile coord astars Coordinate file set to astars.coord aperture blue Aperture blue selected: Xa = 342.44" Ya = 607.68" Phi = 1.00 track 1 Acquiring object at 03:49:45(UT) Object= ALPCAR (1) RA = 06:23:57.12 Dec= -52:41:44.5 Aperture=blue gstatus 14 Gmount status: 14 Time_To_Acq 32 sees gstatus 3 Gmount status: 3 Motion_State SLEWING_TO_TRACK gstatus 14 Gmount status: 14 Time_To_Acq 14 sees B-1. G-mount Command Logs and Scripts 200

gstatus 3 Gmount status: 3 Motion_State TRACKING gstatus 16 gstatGmount status: 16 Delta_X 0.0000000000 radians us 17 Gmount status: 17 Delta_Y 0.0000000000 radians gstatus 16 gGmount status: 16 Delta_X 0.0000000000 radians status 17 Gmount status: 17 Delta_Y 0.0000000000 radians aoffset 0 2 Focal-place offset by X = 0.00" Y = 2.00" relative to base 2 Jun 25 04:14:22 AFOS > f -49617 10 20107 1098 9 4 8508.0 aoffset 0 2 Focal-place offset by X = 0.00" Y = 2.00" relative to base gstatus 14 Gmount status: 14 Time_To_Acq 6 sees gstatus 3 Gmount status: 3 Motion_State TRACKING gstatus 4 Gmount status: 4 Az_Status NORMAL gstatus 20 Gmount status: 20 Coordinate_File - astars.coord gstatus 3 Gmount status: 3 Motion_State TRACKING halt tGmount halted at 06:07:44(UT)

~R track 7 Acquiring object at 06:08:07(UT) Object = ALPCRU (7) RA = 12:26:35.87 Dec = -63:05:56.6 Aperture=afos gstatus 14 Gmount status: 14 Time_To_Acq 46 sees gstatus 3 Gmount status: 3 Motion_State TRACKING gstatus 3 Gmount status: 3 Motion_State TRACKING gstatus 3 Gmount status: 3 Motion_State TRACKING halt Gmount halted at 06:18:05(UT) track 1 Acquiring object at 06:18:17(UT) Object = ALPCAR (1) RA = 06:23:57.12 Dec= -52:41:44.5 Aperture=afos gstatus 14 Gmount status: 14 Time_To_Acq 88 sees gstatus 3 B-1. G-mount Command Logs and Scripts 201

Gmount status: 3 Motion_State SLEWING_TO_TRACK gstatus 14 Gmount status: 14 Time_To_Acq 63 sees gstatus 24 Gmount status: 24 Azimuth 417.520 spiral start Spiral commenced at 06:34:12(UT) Increment 10.0 arcsecs Speed 4.0 (arcs ec/sec) gstatus 18 Gmount status: 18 Spiral_State SPIRAL_IN_PROGRESS gstatus 24 Gmount status: 24 Azimuth 522.897 gstatus 25 Gmount status: 25 Altitude 52.710 gstatus 3 Gmount status: 3 Motion_State TRACKING gstatus 25 Gmount status: 25 Altitude 52.710 spiral stop Spiral stopped at 06:44:09(UT) spiral cancel Spiral cancelled at 06:44:24(UT) gstatus 3 Gmount status: 3 Motion_State TRACKING gstatus 24 Gmount status: 24 Azimuth 521.453 gstatus 25 Gmount status: 25 Altitude 52.710 axes park Slewing to park preset position: azimuth: 427.000 altitude: 0.000 gstatus 3 Gmount status: 3 Motion_State SLEWING_TO_HALT gstatus 3 Gmount status: 3 Motion_State SLEWING_TO_HALT gstatus 14 Gmount status: 14 Time_To_Acq 10 sees gstatus 3 Gmount status: 3 Motion_State STATIONARY axes off Gmount axes are OFF cgmount off Disconnect ok logout B-1. G-mount Command Logs and Scripts 202

B-1.2 Gmount Pointing and Co-ordinate files

A standard pointing correction file, c030108. cfile is given as:

IA -303.90 IE -27.49 CA -41.0 AN +265.60 AW -607.62 NPAE +190.0 TX -16.2 ACEC +2.19 ACES +1.25

The terms of correction pertain to the parameters defined in the pointing software TPOINT (Wallace 1998).

The coordinate file used for all AFOS observations astars. coord is shown here:

"ALPCAR" 06 23 57.119 -52 41 44.50 HD: 45348 -0.72 FO CANOPUS "ALPCEN" 14 39 35.885 -60 50 07.44 HD: 128620 0.00 GO+K5 RIGEL KENTAURUS "ALPER!" 01 37 42.852 -57 14 12.18 HD: 10144 0.46 B5 ACHERNAR "BETCEN" 14 03 49.408 -60 22 22.79 HD: 122451 0.61 B1 BETA CENTURI "ALPPSA" 22 57 39.055 -29 37 20.10 HD: 216956 1.16 A3 FOMALHAUT "BETCRU" 12 47 43.237 -59 41 19.46 HD: 111123 1.25 B1 BETA CRUCIS "ALPCRU" 12 26 35.871 -63 05 56.58 HD: 108248 1.33 B1 ACRUX "GAMCRU" 12 31 09.929 -57 06 47.50 HD: 108903 1.63 M3 GACRUX "LAMSCO" 17 33 36.534 -37 06 13.72 HD: 158926 1.63 B2 SHAULA "BETCAR" 09 13 11.957 -69 43 01.95 HD: 80007 1.68 AO MIAPLACIDUS "ALPGRU" 22 08 14.000 -46 57 39.59 HD: 209952 1.74 B5 AL NA'IR "GAMVEL" 08 09 31.965 -47 20 11.91 HD: 68273 1. 78 Oap "EPSSGR" 18 24 10.327 -34 23 04.73 HD: 169022 1.85 AO KAUS AUSTRALIS "EPSCAR" 08 22 30.833 -59 30 34.51 HD: 71129 1.86 KO+B AVIOR "THESCO" 17 37 19.151 -42 59 52.21 HD: 159532 1. 87 FO "ALPTRA" 16 48 39.869 -69 01 39.82 HD: 150798 1.92 K2 ATRIA "ALPPAV" 20 25 38.852 -56 44 06.38 HD: 193924 1.94 B3 PEACOCK "THECEN" 14 06 40.951 -36 22 12.03 HD: 123139 2,06 KO MENKENT "BETGRU" 22 42 40.063 -46 53 04.69 HD: 214952 2.11 M3 "LAMVEL" 09 07 59.776 -43 25 57.38 HD: 78647 2.21 K5 "IOTCAR" 09 17 05.404 -59 16 31.04 HD: 80404 2.25 FO ASPIDISKE "ZETPUP" 08 03 35.052 -40 00 11.64 HD: 66811 2.25 03 "EPSSCO" 16 50 09.820 -34 17 35.72 HD: 151680 2.29 KO "EPSCEN" 13 39 53.246 -53 27 58.99 HD: 118716 2.30 B1 "ALPLUP" 14 41 55.768 -47 23 17.51 HD: 129056 2.30 B2 "ETACEN" 14 35 30.429 -42 09 28.39 HD: 127972 2.31 B3p+A2p B-2. AFOS Command logs 203

"ALPPHE" 00 26 17.030 -42 18 21.81 HD: 2261 2.39 KO ANKAA "KAPSCO" 17 42 29.276 -39 01 48.09 HD: 160578 2.41 B2 "KAPVEL" 09 22 06.828 -55 00 38.60 liD: 81188 2.50 B3 "ZETCEN" 13 55 32.388 -47 17 18.12 liD: 121263 2.55 B2p "DELCEN" 12 08 21.515 -50 43 20.74 liD: 105435 2.60 B3p "ALPCOL" 05 39 38.947 -34 04 27.01 liD: 37795 2.64 B5p PHACT "BETLUP" 14 58 31.929 -43 08 02.40 liD: 132058 2.68 B2p "ALPMUS" 12 37 10.958 -69 08 07.96 liD: 109668 2.69 B3 "UPSSCO" 17 30 45.843 -37 17 45.03 liD: 158408 2.69 B3 LESATH "PI_PUP" 07 17 08.554 -37 05 51.08 liD: 56855 2.70 K5 "IOTCEN" 13 20 35.822 -36 42 44.32 liD: 115892 2. 75 A2 "THECAR" 10 42 57.368 -64 23 40.08 liD: 93030 2.76 BO "BETHYI" 00 25 45.056 -77 15 15.40 liD: 2151 2.80 GO "DELCRU" 12 15 08.683 -58 44 56.08 liD: 106490 2.80 B3 "BETTRA" 15 55 08.547 -63 25 50.37 liD: 141891 2.85 FO "BETARA" 17 25 17.999 -55 31 47.61 liD: 157244 2.85 K2 "ALPHYI" 01 58 46.201 -61 34 11.43 HD: 12311 2.86 FO "ALPTUC" 22 18 30.099 -60 15 34.59 liD: 211416 2.86 K2 "GAMTRA" 15 18 54.551 -68 40 46.38 liD: 135382 2. 89 AO "TAUPUP" 06 49 56.173 -50 36 52.73 liD: 50310 2.93 KO "ALPARA" 17 31 50.509 -49 52 34.29 liD: 158427 2.95 B3p "GAMSGR" 18 05 48.491 -30 25 26.69 liD: 165135 2.99 KO ALNASL

B-2 AFOS Command logs

A standard AFOS observing log, from June 25th 2003, is shown below.

Trying 199.4.251.133 ... Connected to 199.4.251.133. Escape character is •-]•. > 02 Jun 25 03:45:02 SET_QUIET_MODE delaying 15 seconds; hit key to abort; time remaining: 12 02 Jun 25 03:45:06 SAFEAFOS > unsafe Connection closed by foreign host. pharlap'l. pharlap'l. afos Trying 199.4.251.133 ... Connected to 199.4.251.133. Escape character is •-]•. > 02 Jun 25 03:45:19 READ_FILE can't open file < •. \run\AFO.ini> Error #128 - CANT_OPEN_FILE (3) 02 Jun 25 03:45:19 AFOS > powerup B-2. AFOS Command logs 204

02 Jun 25 03:45:24 BATTERY_MANAGEMENT 02 Jun 25 03:45:24 AFOS > initccd 02 Jun 25 03:45:28 INIT_CCD_CAMERA initialising the Oriel camera ... in oriel_initialise globals, clocks, timer control, adc_reset, speed up, insints, done, EMS OK 02 Jun 25 03:45:31 AFOS > status 02 Jun 25 03:45:37 STATUS time 1.000000 (exposure time in seconds) gain lo (camera preamp gain) speed lo (camera readout speed) shift 0 (number of bits to shift data to the right) chip 1024 256 (x,y dimensions of the CCD) sra 1 1 1024 256 (readout 1024x256; from (1,1) to (1024,256)) ssa 1 1 1024 256 (stats 1024x256; from (1,1) to (1024,256)) bin 1 1 (CCD binning in x and y) observer unknown object unknown filter unknown imagetyp unknown equinox 2000.000 close (shutter status) no exposure in progress 02 Jun 25 03:45:37 AFOS > gettemp 02 Jun 25 03:45:45 GET_CCD_TEMPERATURE CCD temperature is 14.8 degrees C; requested 0 degrees C; raw= 29.0; check = 15 02 Jun 25 03:45:45 AFOS > settemp -40 02 Jun 25 03:46:04 SET_CCD_TEMPERATURE 02 Jun 25 03:46:04 AFOS > g=gettemp 02 Jun 25 03:46:09 AFOS > f=fibres 02 Jun 25 03:46:14 AFOS > quiet yes 02 Jun 25 03:46:18 SET_QUIET_MODE 02 Jun 25 03:46:18 AFOS > g CCD temperature is 4.1 degrees C; requested -40 degrees C; raw= 47.3; check= 16 02 Jun 25 03:46:31 AFOS > battstat battery voltage 22.99 V; battery current -837.9 mA; boost OFF power management: loops to shutdown 5 delay between loops 60 seconds maximum uptime 7200 seconds minimum valid voltage 24.00 V battery threshold 12.00 V count errors yes 02 Jun 25 03:46:35 AFOS > batt 5 60 15000 24 12 yes 02 Jun 25 03:47:03 AFOS > expose 02 Jun 25 03:47:15 AFOS >readout B-2. AFOS Command logs 205

02 Jun 25 03:47:20 AFOS > rm d:\aOOO 02 Jun 25 03:47:28 AFOS > ls d:

Volume in drive D is MS-RAMDRIVE Directory of D:\

BOOT LOG 440 06-24-02 6:52a AFOS LOG 42 06-25-02 3:47a A001 0 06-25-02 3:47a 3 file(s) 482 bytes 1,221,120 bytes free 02 Jun 25 03:47:32 AFOS > setfibres 142 168 189 213 234 254 2 5 02 Jun 25 03:47:46 AFOS > f -49638 0 0 0 0 1 8469.0 02 Jun 25 03:47:52 AFOS > time 10 02 Jun 25 03:51:07 AFOS > 10(f) -49610 10 22336 4253 9 5 8517.2 -49611 10 22570 4075 9 5 8517.5 -49611 10 22712 3982 9 5 8518.0 -49610 10 22703 4082 9 5 8517.8 -49611 10 22598 4099 9 5 8517.7 -49611 10 22310 4142 9 5 8516.8 -49611 10 22364 3997 9 5 8516.9 -49611 10 22200 4174 9 5 8516.7 -49610 10 22378 4561 8 5 8517.5 -49610 9 22327 4715 8 5 8517.4 02 Jun 25 03:53:08 AFOS > f -49614 10 22402 1537 9 4 8514.1

02 Jun 20 02:39:08 AFOS > time 20 02 Jun 20 02:39:13 AFOS > expose 02 Jun 20 02:39:39 AFOS > readout 02 Jun 20 02:39:43 AFOS > ls d: 02 Jun 20 02:40:24 AFOS > f

-53244 11 3995 16208 6 24 9107.9 02 Jun 20 02:41:15 AFOS > f -53246 9 7819 11791 6 27 9108.3 02 Jun 20 02:41:46 AFOS > time 10 02 Jun 20 02:41:52 AFOS > f -53265 4 3424 2980 4 19 9087.3

telnet> get d:\a003 receiving 269952 bytes as file d:a003

02 Jun 20 02:42:10 AFOS >time 1 B-3. Automated Observing Scripts 206

02 Jun 20 02:42:17 AFOS > expose 02 Jun 20 02:42:23 AFOS > readout 02 Jun 20 02:42:26 AFOS > ls d:

02 Jun 20 02:42:51 AFOS > time 10 02 Jun 20 02:42:56 AFOS > expose 02 Jun 20 02:43:12 AFOS > ls d:

B-3 Automated Observing Scripts

A series of scripts in TCL EXPECT were designed to automate the observations.

B-3.1 Instrument control script

The EXPECT script, runafos to control the AFOS observations, and rungrnount were run from the following script as follows:

#!lbinlsh # # A program for running an AASTO instrument. # # Michael Ashley I UNSW I 02-Apr-1997 # Jessica Dempsey I UNSWI 15-Aug-2003 # The following allows this program to work regardless of where the # expect program is stored, provided it is in the path. See "Exploring # Expect" by Don Libes, page 216. #

set kludge { ${1+"$~"} shift shift

exec expect -f $0 -- ${1+"$~"} }

proc Usage {} { puts "usage: connect \ [ -nomail\] \ [aasto lnismlmismlgmount I sodarl afos I test\] \[file I commands\] \r" exit 1 }

# We set the "TERM" environment variable, since we may be running from B-3. Automated Observing Scripts 207

#a cron script, which may not set "TERM". See "Exploring Expect" page 373. set env(TERM) vt100 set switch [lindex $argv 0] set mail_results [string compare $switch "-nomail"] if {[string first "-" $switch] == 0} { set argv [lrange $argv 1 end] incr argc -1 }

set instrument [lindex $argv 0] set aasto_pid 0 set instrument_pid 0

# Parse the remaining command-line switches, and set the "mode" # variable as follows: # # mode meaning #

# 0 open interactive connection to the instrument # 1 send a file to the instrument # 2 send a command to the instrument

if {$argc > 1} { set commands [lrange $argv 1 end] if [catch {open $commands "r"} input] { set mode 2 } else { set file $commands set mode 1 close $input } } else { set mode 0 }

set hostname [exec hostname] puts "running on host $hostname\r" if {[string compare $hostname "newt.phys.unsw.edu.au"] == 0} { set telnet "/usr/bin/telnet" set gzip "/lenz/usr/local/bin/gzip" } else if {[string compare $hostname "acbar. spole. gov"] 0} { set telnet "/home/aasto03/mcba/./telnet" set gzip "/usr/bin/gzip" } elseif {[string compare $hostname "mcba"] 0} { B-3. Automated Observing Scripts 208

set telnet "/usr/bin/telnet" set gzip "/bin/gzip" } else { puts "the host computer is not known to this program\n" exit 1 }

# Now find out which instrument we will communication with, and set # various variables appropriately.

} if {[string compare $instrument "gmount"] 0} {

set remote_ip "199.4.251.132" set instrument_ip "199.4.251.132" set logfile "gmount-log"

set aasto_prompt "((SUPER) I (ISM)I(AASTO)).* > II set instrument_start "\r" set instrument_prompt II II set aasto_port 2000 set instrument_port 200B set port_setup "baud COMB 9600 2000 yes\r" set max_ delay 300 set initialisation_file "gmount.ini" set direct 0

} ·elseif {[string compare $instrument "afos"] 0} {

set remote_ip "199.4. 251.133" set instrument_ip "199.4.251.133" set logfile "afos-log"

set aasto_prompt "AFOS.* > II set aasto_port 2000 set max_delay BOO set initialisation_file "afos.ini" set direct 0

} elseif {[string compare $instrument "aasto"] 0} {

set remote_ip "199.4. 251.132" set instrument_ip "199.4.251.132" set logfile "aasto-log" set aasto_port 2000

set aasto_prompt II ((SUPER) I (ISM) I (AASTO)). * > II set max_delay 300 set direct 1 B-3. Automated Observing Scripts 209

} else {

Usage

}

# This exit handler ensures that any telnet sessions are closed #properly, and then handles any required a-mailing of logfiles.

exit -onexit { global logfile global mail_results global aasto_pid global instrument_pid global aasto_id global instrument_id

# If a telnet session is still open, we use the telnet escape # sequence and quit command to terminate the session. We then # explicitly kill the session to be doubly sure.

if {$aasto_pid != 0} { set spawn_id $aasto_id if [catch {send -- "\035quit\r"} junk] { } else { close catch {wait -nowait} sleep 1 catch {exec kill $aasto_pid} } }

if {$instrument_pid != 0} { set spawn_id $instrument_id if [catch {send -- "\035quit\r"} junk] { } else { close catch {wait -nowait} sleep 1 catch {exec kill $instrument_pid} }

}

if {$mail_results != 0} { puts "mailing results\r"

B-3. Automated Observing Scripts 211

while 1 { send -- "flush\r" sleep 2 expect { "> " break } set elapsed_time [expr $elapsed_time + $timeout] if {$elapsed_time > $max_delay} { HandleTimeout }

} expect -re "·*" set elapsed_time 0 while 1 { send -- "quiet 0 beep\r" expect { -re "BEEP.*> " break } set elapsed_time [expr $elapsed_time + $timeout] if {$elapsed_time > $max_delay} { Handle Timeout } }

} proc MakeContactSodar {} { global max_delay

set timeout 10 expect -re ". *" set elapsed_time 0 while 1 { send -- "\r" expect { ": " break } set elapsed_time [expr $elapsed_time + $timeout] if {$elapsed_time > $max_delay} { Handle Timeout }

}

} proc Telnet {ip port id} { global telnet B-3. Automated Observing Scripts 212

spawn $telnet $ip $port set id $spawn_id }

proc Send {string} { global max_delay global prompt

set timeout $max_delay expect -re "·*" send -- $string while 1 { expect { -re $prompt {break} timeout {HandleTimeout} "\r" {} }

} } proc SendNoWait {string} { global prompt

send -- $string } proc SendFile {filename} {

if [catch {open $filename} input] { puts "\r\n$input\r" return }

while {[gets $input line] != -1} { if {[string first "#" $line] ! = 0} { if {[string first "EXEC" $line] ! = 0} { Send "$line\r" } else { eval [string range $line 5 end] }

}

} close $input } proc SetinstrumentTime {} { B-3. Automated Observing Scripts 213

# Set the time.

set time [timestamp] set settime [timestamp -format "setdate r,d i'.m r,y settime ••. r,H roM r.s\r" -seconds [expr $time - 3600*12]]

Send $settime

}

proc TakeAfosExposure {exposure} { global prompt global gzip

# Find the current FITS filename.

send -- "loginfo\r" expect { -re "The current file is ( ...... ) ( • *)$prompt" { set file $expect_out(1,string) } timeout {HandleTimeout} }

Send "sra 1 128 1024 256 time $exposure expose readout\r" sleep 1 Send "\035get $file\r\r" Send "rm $file\r" system "$gzip --force --best $file" # system "uuencode $file.gz $file.gz > $file.uu" # system "/usr/ucb/mail -s \"$file from AASTO\" jtd0newt.phys.unsw.edu.au < $file.uu > /dev/null >2 /dev/null # sleep 10 # system "/bin/rm $file. uu" } log_file $logfile set timeout $max_delay set aasto_id 0 set instrument_id 0 puts "test3\r" set aasto_pid [spawn $telnet $remote_ip $aasto_port] set aasto_id $spawn_id puts "test4\r" B-3. Automated Observing Scripts 214

WaitForTelnetConnection MakeContact set prompt $aasto_prompt puts "test5\r"

if {[string compare $instrument "afos"] 0} {

# Make sure we are running the "unsafe" version of the AFOS software.

expect -re ". *" send -- "\r" while 1 { expect { -re " AFOS >" {break} -re " SAFEAFOS >" { send -- "unsafe\r" expect { "closed" {} timeout {HandleTimeout} } close wait sleep 30 set aasto_id 0 set instrument_id 0 set aasto_pid [spawn $telnet $remote_ip $aasto_port] set aasto_id $spawn_id WaitForTelnetConnection MakeContact send -- "\r" } timeout {HandleTimeout} "\r" {} }

}

SetinstrumentTime

# Make sure that the CCD is initialised, that the temperature is set to # -40C, and that the CCD is down to at least -35C.

expect -re ". *" send -- "gettemp\r" # CCD temperature is 13.1 degrees C; requested 0 degrees C; raw= 31.9; check= 15 while 1 { B-3. Automated Observing Scripts 215

expect { -re "CCD temperature is (.*) degrees C; requested (.*) degrees(.*)$prompt" { puts "Temp $expect_out(1,string)" if {$expect_out(2,string) != -40} { Send "settemp -40\r" } elseif {$expect_out(1,string) < -35} { break } puts "well?\r" sleep 10 send -- "gettemp\r" } puts "well2?\r" -re "CCD_NOT_INITIALISED(.*)$prompt" {send -- "initccd\r"} -re $prompt {send -- "gettemp\r"} timeout {HandleTimeout} "\r" {} }

}

TakeAfosExposure 0.1 TakeAfosExposure 1 TakeAfosExposure 10

exit 0 }

if {[string compare $instrument "gmount"] 0} { send -- "\r" sleep 5 puts "step one ok\r" send -- "cgmount afos\r" expect { "Command ok" {} timeout {HandleTimeout} }

send -- "aperture dark2\r" expect { "Aperture dark2" {} timeout {HandleTimeout} }

send -- "axes on\r" expect { B-3. Automated Observing Scripts 216

"Gmount axes" {} "axis servos" {} puts "step two ok\r" timeout {HandleTimeout} }

expect -re " . *"

send -- "axes recover\r" expect { "axis servo not tripped" {} "axis" {} timeout {} }

expect -re " • *"

send -- "slew 600 21\r" expect { "Slewing to" {} timeout {HandleTimeout} }

sleep 600

send -- "track moon\r" expect { "Acquiring moon" {} timeout {HandleTimeout} } exit 0 exit 0$ Appendix C

Reduction Routines

C-1 IRAF

The following are the programs in IRAF used to reduce the AFOS data presented in this thesis.

C-1.1 CCD Reduction

1 HEDIT routine to label imagetype.The first reduction step is carried out on all images. Using CCDproc in IRAF (noao/imred/ccdred packages) first all dark or sky images used for zeroing the remaining images are given imagetypes 'zero' using HEDIT. :

cc> hedit images to be edited (*.fits): dark*.fits fields to be edited (DATASEC): imagetype value expression ([1:1024,1:128]): zero dark0.1.fits,imagetype (unknown-> zero): dark0.1.fits,imagetype: unknown-> zero update dark0.1.fits? (yes):

2 ZERO COMBINE to create bias/ dark frame

PACKAGE = ccdred C-1. IRAF 218

TASK zerocombine

input dark*.fits List of zero level images to combine (output Zero) Output zero level name (combine= average) Type of combine operation (reject = minmax) Type of rejection (ccdtype= images *.fits List of CCD images to correct (output = List of output CCD images (ccdtype= CCD image type to correct (max_cac= 0) Maximum image caching memory (in Mbytes) (noproc = no) List processing steps only?

(fixpix = no) Fix bad CCD lines and columns? (oversea= no) Apply overscan strip correction? (trim yes) Trim the image? (zero cor= no) Apply zero level correction? (darkcor= no) Apply dark count correction? (flat cor= no) Apply flat field correction? (illumco= no) Apply illumination correction? (fringec= no) Apply fringe correction? (readcor= no) Convert zero level image to readout correction? (scancor= no) Convert flat field image to scan correction?

(readaxi= line) Read out axis (columnlline) (fixfile= File describing the bad lines and columns (biassec= Overscan strip image section (trimsec= [10:1000,1:128]) Trim data section (zero Zero.fits) Zero level calibration image (dark Dark count calibration image (flat Flat field images (illum Illumination correction images (fringe Fringe correction images (minrepl= 1.) Minimum flat field value (scantyp= shortscan) Scan type (shortscanllongscan) (nscan = 1) Number of short scan lines

3 JD and LJD Dates on Image Headers

images to be edited (dark*.fits): *.fits fields to be edited (imagetype): jd value expression (zero): 2000 add ai.fits,jd = 2000 update a1.fits? (yes): ai.fits updated C-1. IRAF 219

Likewise for field 'ljd'.

4 Dispersion axis on image headers This field must be set to '1' so that IRAF reads across the rows for the spectral data, instead of down the columns.

images to be edited (a13.fits): *.fits fields to be edited (jd): dispaxis value expression (2000): 1 add Zero.fits,dispaxis = 1 update Zero.fits? (yes): Zero.fits updated add a1.fits,dispaxis = 1 update a1.fits? (yes): al.fits updated

5 Datasection; cropping to data area To align the row and column numbers to allow the trimsec parameter to work.

images to be edited (*.fits): fields to be edited (dispaxis): datasec value expression (1): [1:1024,1:128] Zero.fits,datasec ([1:1024,129:256] -> [1:1024,1:128]): Zero.fits,datasec: [1:1024,129:256] -> [1:1024,1:128] update Zero.fits? (yes): Zero.fits updated

6 CCDPROC to subtract bias fram Zero.fits, and to trim all data frames. Then the package ccdproc can be run on all the image and sky fits images twice, first to zero the images, and then to trim them.

images *.fits List of CCD images to correct (output = List of output CCD images (ccdtype= CCD image type to correct (max_cac= 0) Maximum image caching memory (in Mbytes) (noproc = no) List processing steps only?

(illumco= no) Apply illumination correction? (fringec= no) Apply fringe correction? (readcor= no) Convert zero level image to readout correction? (scancor= no) Convert flat field image to scan correction?

(interac= no) Fit overscan interactively? (functio= legendre) Fitting function (order = 1) Number of polynomial terms or spline pieces

C-1.2 Aperture extraction and Flat fielding

1 DOFIBERS, used to determine dispersion on COD, collapse spectral lines from COD image, and wavelength calibrate individual spectra.

objects a*.fits,b*.fits,c*.fits,d*.fits List of object spectra (apref a10.fits) Aperture reference spectrum (flat Flat field spectrum (through= Throughput file or image (optional) (arcs1 a10.fits) List of arc spectra (arcs2 List of shift arc spectra (arctabl= Arc assignment table (optional)

(readnoi= o.) Read out noise sigma (photons) (gain 1.) Photon gain (photons/data number) (datamax= INDEF) Max data value I cosmic ray threshold (fibers 2) Number of fibers (width 5.) Width of profiles (pixels) (minsep = 8.) Minimum separation between fibers (pixels) (maxsep = 20.) Maximum separation between fibers (pixels) (apidtab= ) Aperture identifications C-1. IRAF 221

(crval INDEF) Approximate central wavelength (cdelt INDEF) Approximate dispersion (objaps = 1,2) Object apertures (skyaps = ) Sky apertures (arcaps = Arc apertures (objbeam= 1,2) Object beam numbers (skybeam= ) Sky beam numbers (arcbeam= Arc beam numbers

(scatter= no) Subtract scattered light? (fitflat= no) Fit and ratio flat field spectrum? (clean = no) Detect and replace bad pixels? (dispcor= yes) Dispersion correct spectra? (skyalig= no) Align sky lines? (savearc= no) Save simultaneous arc apertures? (skysubt= no) Subtract sky? (skyedit= no) Edit the sky spectra? (savesky= no) Save sky spectra? (splot yes) Plot the final spectrum? (redo no) Redo operations if previously done? (update = no) Update spectra if cal data changes? (batch = no) Extract objects in batch? (listonl= no) List steps but don't process?

(params = Algorithm parameters (flpar no) flush pfile on assign? (mode ql)

2 SARITH used to manually subtract the Quartz-halogen lamp spectrum from all star images in first step of flat fielding. SARITH then used again to times this result by a black-body at the same temperature as the q-h lamp.

PACKAGE = specred TASK = sarith

input! ®input.txt List of input spectra op I Operation input2 qh2606.1.ms.fits List of input spectra or constants output ®output.txt List of output spectra (w1 INDEF) Starting wavelength (w2 INDEF) Ending wavelength (apertur= 1,2) List of input apertures or columns/lines (bands ) List of input bands or lines/bands (beams List of input beams or echelle orders C-1. IRAF 222

(apmodul= 0) Input aperture modulus (O=none)

(reverse= no) Reverse order of operands in binary operation? (ignore a= no) Ignore second operand aperture numbers?

(format = multispec) Output spectral format (renumbe= yes) Renumber output apertures? (offset = 0) Output aperture number offset (clobber= no) Modify existing output images? (merge no) Merge with existing output images? (rabin yes) Rabin to exact wavelength region? (errval 0.) Arithmetic error replacement value (verbose= no) Print operations? (flpar no) flush pfile on assign? (mode ql)

3 SCOPY to crop reduced spectra to relevant spectral range (3500-6500 for blue fibres, 6500-8500 for red fibres)

PACKAGE = specred TASK = scopy

input ~input.txt List of input spectra

output ~output.txt List of output spectra (w1 INDEF) Starting wavelength (w2 6300.) Ending wavelength (apertur= 1) List of input apertures or columns/lines (bands ) List of input bands or lines/bands (beams List of beams or echelle orders (apmodul= 0) Input aperture modulus (O=none)

(format = multispec) Output spectra format (renumbe= no) Renumber output apertures? (offset = 0) Output aperture number offset (clobber= no) Modify existing output images? (merge no) Merge with existing output images? (rabin yes) Rabin to exact wavelength region? (verbose= no) Print operations? (flpar no) flush pfile on assign? (mode ql) Appendix D

MODTRAN input and output files

D-1 Input File

A typical tape 5 MODTRAN input file, using interpolated radiosonde balloon data from the 3rd of July is shown here: t 7 2 1 0 0 0 0 5 6 5 1 1 0 .000 -1 f Of 0 365.000 3.000 0.970 f t f DATA/BMP98_01.BIN 9 2 0 0 0 0 5.00 .000 .00 .000 2.835 81 0 0 anto 2.835EO 6.818E2 2.033E2 5.300E1 o.oo 2.33E-5 AAH E 2.888EO 6.758E2 2.107E2 2.931E1 0.00 2.29E-5 AAH E 2.942EO 6.700E2 2.175E2 3.877E1 o.oo 2.24E-5 AAH E 2.995EO 6.646E2 2.242E2 6.150E1 0.00 2.20E-5 AAH E 3.049EO 6.592E2 2.308E2 5.891E1 0.00 2.18E-5 AAH E 3.103EO 6.540E2 2.339E2 5.900E1 0.00 2.17E-5 AAH E 3.156EO 6.489E2 2.348E2 4.839E1 o.oo 2.16E-5 AAH E 3.210EO 6.439E2 2.350E2 3.970E1 0.00 2.14E-5 'AAH E 3.264EO 6.388E2 2.350E2 3.636E1 0.00 2.13E-5 AAH E 3.317EO 6.339E2 2.353E2 2.757E1 0.00 2.12E-5 AAH E 3.371EO 6.290E2 2.356E2 1.950E1 0.00 2.11E-5 AAH E 3.425EO 6.241E2 2.357E2 2.043E1 0.00 2.10E-5 AAH E 3.478EO 6.193E2 2.356E2 2.100E1 0.00 2.09E-6 AAH E 3.532EO 6.145E2 2.353E2 2.018E1 o.oo 2.07E-6 AAH E 3.586EO 6.097E2 2.353E2 1.800E1 0.00 2.06E-5 AAH E 3.639EO 6.060E2 2.351E2 1.900E1 0.00 2.05E-5 AAH E 3.693EO 6.003E2 2.349E2 1.900E1 0.00 2.03E-5 AAH E 3.746EO 6.956E2 2.346E2 2.000E1 o.oo 2.01E-6 AAH E D-1. Input File 224

3.800EO 5.911E2 2.344E2 2.200E1 0.00 1.98E-5 AAH E 3.854EO 5.864E2 2.341E2 2.400E1 0.00 1.95E-5 AAH E 3.907EO 5.818E2 2.336E2 2.600E1 0.00 1.93E-5 AAH E 3.961EO 5.773E2 2.332E2 2.800E1 0.00 1.91E-5 AAH E 4.015EO 5.727E2 2.331E2 2.629E1 0.00 1.89E-5 AAH E 4.068EO 5.683E2 2.326E2 2.600E1 0.00 1.88E-5 AAH E 4.122EO 5.638E2 2.324E2 2.500E1 0.00 1.87E-5 AAH E 4.176EO 5.594E2 2.323E2 2.800E1 0.00 1.87E-5 AAH E 4.229EO 5.551E2 2.323E2 2.700E1 0.00 1.86E-5 AAH E 4.283EO 5.506E2 2.323E2 2.600E1 0.00 1.86E-5 AAH E 4.337EO 5.463E2 2.321E2 2.500E1 0.00 1.85E-5 AAH E 4.390EO 5.421E2 2.320E2 2.500E1 o.oo 1.84E-5 AAH E 4.444EO 5.378E2 2.317E2 2.500E1 o.oo 1.84E-5 AAH E 4.497EO 5.336E2 2.314E2 2.400E1 0.00 1.83E-5 AAH E 4.551EO 5.293E2 2.313E2 2.282E1 0.00 1.82E-5 AAH E 4.605EO 5.252E2 2.310E2 2.550E1 0.00 1.82E-5 AAH E 4.658EO 5.210E2 2.309E2 2.443E1 0.00 1.81E-5 AAH E 4.712EO 5.169E2 2.307E2 2.500E1 0.00 1.80E-5 AAH E 4.766EO 5.128E2 2.304E2 2.820E1 o.oo 1.78E-5 AAHE 4.819EO 5.087E2 2.303E2 3.844E1 0.00 1.76E-5 AAH E 4.873EO 5.047E2 2.301E2 5.250E1 o.oo 1. 74E-5 AAH E 4.927EO 5.007E2 2.297E2 6.100E1 0.00 1.73E-5 AAH E 4.980EO 4.967E2 2.293E2 5.900E1 0.00 1.71E-5 AAH E 5.034EO 4.927E2 2.291E2 5.700E1 o.oo 1.70E-5 AAH E 5.088EO 4.888E2 2.289E2 6.000E1 0.00 1.69E-5 AAH E 5.141EO 4.850E2 2.288E2 5.420E1 o.oo 1.67E-5 AAH E 5.195EO 4.811E2 2.285E2 5.780E1 0.00 1.66E-5 AAH E 5.248EO 4.772E2 2.282E2 5.200E1 0.00 1.65E-5 AAH E 5.302EO 4.734E2 2.279E2 4.688E1 0.00 1.64E-5 AAH E 5.356EO 4.697E2 2.276E2 4.300E1 0.00 1.64E-5 AAH E 5.409EO 4.659E2 2.271E2 4.108E1 0.00 1.63E-5 AAH E 5.463EO 4.622E2 2.268E2 4.190E1 0.00 1.62E-5 AAH E 5.517EO 4.584E2 2.264E2 4.200E1 0.00 1.61E-5 AAH E 5.570EO 4.547E2 2.261E2 4.200E1 0.00 1.61E-5 AAH E 5.624EO 4.511E2 2.259E2 4.300E1 0.00 1.60E-5 AAH E 5.678EO 4.474E2 2.256E2 4.700E1 0.00 1.59E-5 AAH E 5.731EO 4.438E2 2.253E2 4.600E1 0.00 1.59E-5 AAH E 5.785EO 4.402E2 2.249E2 4.500E1 0.00 1.58E-5 AAH E 5.839EO 4.366E2 2.246E2 4.300E1 0.00 1.57E-5 AAH E 5.892EO 4.330E2 2.244E2 3.918E1 0.00 1.56E-5 AAH E 5.946EO 4.296E2 2.240E2 3.550E1 0.00 1.55E-5 AAH E 6.000EO 4.260E2 2.235E2 3.200E1 0.00 1.54E-5 AAH E 6.896EO 3.708E2 2.171E2 4.000E1 o.oo 1.46E-5 AAH E 7.793EO 3.213E2 2.111E2 3.020E1 o.oo 1.31E-5 AAH E 8.689EO 2.773E2 2.054E2 2.600E1 0.00 1.52E-5 AAH E 9.586EO 2.386E2 2.013E2 2.673E1 0.00 2.17E-5 AAH E D-2. Output file 225

1.048E1 2.047E2 1.997E2 2.700E1 0.00 3.77E-5 AAH E 1.138E1 1.756E2 1.982E2 2.600E1 o.oo 6.14E-5 AAH E 1.228E1 1.510E2 2.001E2 2.700E1 0.00 6.12E-5 AAH E 1.317E1 1.294E2 1.967E2 2.500E1 o.oo 8.47E-5 AAH E 1.407E1 1.107E2 1.943E2 2.300E1 0.00 1.27E-4 AAH E 1.497E1 9.446E1 1.920E2 2.100E1 0.00 1.67E-4 AAH E 1.586E1 8.049E1 1.893E2 2.000E1 0.00 1.70E-4 AAH E 1.676E1 6.838E1 1.871E2 1.800E1 0.00 1.68E-4 AAH E 1.766E1 5.795E1 1.852E2 1.700E1 0.00 1.55E-4 AAH E 1.855E1 4.911E1 1.833E2 1.600E1 0.00 1.36E-4 AAH E 1.945E1 4.150E1 1.829E2 1.500E1 o.oo 1.14E-4 AAH E 2.034E1 3.505E1 1.806E2 1.400E1 0.00 9.89E-5 AAH E 2.124E1 2.960E1 1.806E2 1.400E1 o.oo 8.64E-5 AAH E 2.214E1 2.498E1 1.805E2 1.400E1 0.00 7 .18E-5 AAH E 2.303E1 2.110E1 1.822E2 1.500E1 o.oo 6.23E-5 AAH E 2.393E1 1.780E1 1.806E2 1.400E1 0.00 5.53E-5 AAH E 2.483E1 1.508E1 1.818E2 1.500E1 o.oo 5.01E-5 AAH E 0.000 24.83 20.283 0000.00 6356.79 .000 0 11765 16666 20 50 TN

D-2 Output file

D-2.1 Tape 7

The standard tape 7 MODTRAN output file is as follows:

t 7 2 1 0 0 0 0 5 5 5 1 1 0 0.000-1 9 2 1 0 0 0 5.00000 0.00000 0.00000 0.00000 2.83500 -99.000 -99.000 -99.000 -99 -99.00000 -99.00000 -99.00000 0.052715 0.000480 H20 & 03 COLUMNS [GM/CM2] 81 anto 2.83500 24.83000 20.28300 23.44413 0.072986356.79004 0 -99 -99 -99 -99 -99.00000 -99.00000 -99.00000 -99.00000 -99.00000 -99.00000 -99.00000 -99.00000 11765 16666 20 50 0 FREQ TOT TRANS PTH THRML THRML SCT SURF EMIS GRND RFLT TOTAL RAD DEPTH 11765. 0.52256602 2.9293E-32 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 2.9293E-32 0.649 11785. 0.52258974 2.6045E-32 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 2.6045E-32 0.649 11805. 0.52260387 2.3158E-32 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 2.3158E-32 0.649 11825. 0.52258039 2.0591E-32 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 2.0591E-32 0.649 11845. 0.52252305 1.8309E-32 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 1.8309E-32 0.649 D-2. Output file 226

data omitted here for brevity ...... ··········· ...... 16485. 0.48706546 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.719 16505. 0.48676747 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.720 16525. 0.48649409 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.721 16545. 0.48624489 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.721 16565. 0.48601183 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.722 16585. 0.48579898 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.722 16605. 0.48561952 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.722 16625. 0.48547333 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.723 16645. 0.48535344 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.723 16665. 0.48525196 O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO O.OOOOE+OO 0.723 -9999.

D-2.2 output.plt

The user-defined output.plt output file, as a function of nanometers is shown here:

849.979 0.52256602 848.536 0.52258974 847.099 0.52260387 845.666 0.52258039 844.238 0.52252305 842.815 0.52246255 841.397 0.62240205 839.983 0.52234972 838.574 0.62226186 837.170 0.52214104 data omitted here for brevity

606.877 0.48676747 605.144 0.48649409 604.412 0.48624489 603.682 0.48601183 602.954 0.48579898 602.228 0.48561952 601.504 0.48547333 600.781 0.48636344 600.060 0.48626196 Appendix E

Fitting Programs

E-1 chisquare.f90

The core program used to fit the MODTRAN model to the AFOS data is shown below.

program j essything implicit none integer res,zen,zenstep,resstep,resmin,zenmin,resmax,zenmax integer, parameter .• nobs=39 integer, parameter .. nmod=44 real*B obs(3,nobs),model(2,nmod),chisq,chisqmin real*B resminmin,zenminmin

common /chistuff/ chisq

resmin=1 resmax=1 resstep=1 zenmin=1 zenmax=1 zenstep=1

open(BO,file='output.dat',status='unknown')

chisq=O.dO chisqmin=999999.d0 E-1. chisquare.f90 228

resminmin=999999.d0 zenminmin=999999.d0 do res = resmin,resmax,resstep do zen = zenmin,zenmax,zenstep call readobs(res,zen,nobs,obs)

call readmod(res,zen,nmod,model) model(1, :)=model(1,:)*10.d0 print *, chisq call chi2(nobs,obs,nmod,model) print *,chisq write(BO,*)res,zen,chisq if (chisq.lt.chisqmin) then chisqmin=chisq resminmin=res zenminmin=zen end if enddo enddo close(BO) print *•'minimum chisq ..• ' print *,'zen ',zenminmin print *,'res •,resminmin print *,'chisq = ',chisqmin

stop end program jessything

subroutine readobs(x,y,n,data)

implicit none integer, intent(in) :: x,y,n real*B, intent(inout) :: data(3,n) character*127 fname,fbase logical isthere

fbase = 1 /home/jtd/datafile/' ! base path of data files write(fname,'(2a,i1.1,a,i2.2,a)') & TRIM(fbase), 'res' ,x, '_zen' ,y,' .dat' fname='Atestdata.dat'

inquire(file=TRIM(fname),exist=isthere) E-1. chisquare.f90 229

if (.not.isthere) then write(*,'(3a)') 'file ',TRIM(fname),' does not exist.' stop endif

open(60,file=TRIM(fname),status='old') read(60,*)data(:,:) close(60) return end subroutine readobs

subroutine readmod(x,y,n,data)

implicit none integer, intent(in) :: x,y,n real*8, intent(inout) :: data(2,n) character*127 fname,fbase logical isthere

fbase = '/home/jtd/datafile/' ! base path of data files write(fname,'(2a,i1.1,a,i2.2,a)') & TRIM(fbase), 'res' ,x, '_zen' ,y, '.dat' fname="Atestmodel.dat"

inquire(file=TRIM(fname),exist=isthere) if (.not.isthere) then write(*,'(3a)') 'file ',TRIM(fname),' does not exist.' stop endif

open(60,file=TRIM(fname),status='old') read(60,*)data(:,:) close(60) return end subroutine readmod

subroutine chi2(nobs,obs,nmod,model)

implicit none integer, intent(in) :: nobs,nmod real*8, intent(in) :: model(2,nmod),obs(3,nobs) real*8 nmodel(nobs) E-1. chisquare.f90 230

real*B, intent(inout) .. chisq real*8 chisq integer j common /chistuff/ chisq

call interp(nmod,model(1,:),model(2,:),nobs,obs(1,:),nmodel(:))

chisq=O.dO do j=1,nobs print *,nmodel(j),obs(2,j),obs(3,j) chisq=chisq+((nmodel(j)-obs(2,j))**2)/obs(3,j)**2 print *,((nmodel(j)-obs(2,j))**2)/obs(3,j)**2,chisq end do

print *•'in routine',chisq return end subroutine chi2

subroutine interp(nold,x,y,nnew,newx,newy) implicit none integer, intent(in) .. nold,nnew real*B, intent(in) .. x(nold),y(nold),newx(nnew) real*B, intent(out) .. newy(nnew) real*B m,c,xa,xb,ya,yb integer i,j

do i=1,nnew

do j=1,nold print *•'old/new',x(j),newx(i) if (x(j).eq.newx(i)) then newy(i)=y(j) goto 30 else if (x(j).gt.newx(i)) then xb=x(j) xa=x(j+1) yb=y(j) ya=y(j+1) endif E-1. chisquare.f90 231

if (x(j).lt.newx(i)) goto 20 enddo !20 print *,xb,xa,yb,ya

20 m=(yb-ya)/(xb-xa) c=yb-(yb-ya)*xb/(xb-xa)

newy(i)=newx(i)*m+c stop 30 end do

return end subroutine interp Appendix F

Earth Albedo IRAF scripts

F-1 divvy.cl

The first script, divvy.cl placed each set of observations (consisting of 12 spectra in each fibre) into separate directories, and then created three subdirectories separating the spectra by exposure time length.

string tempi, temp2 string tbeg,name,dirname struct >tEPOCH", expr="yes", >> "tempi") !sort -k4,4 tempi > temp2 listi="temp2" i=O j=O while (fscan(listi, exp, tbeg, name, epoc) != EOF){ i=i+1 if (exp==0.1 && i==1){ j=j+1 dirname="obs"//j mkdir (dirname) cd (dirname) mkdir 0.1s mkdir is F-1. divvy.cl 233

mkdir 10s cd . . I } else flpr if (exp==0.1 && i<=12){

11 11 imrename (oldnames=name, newnames=dirnameii I0.1s ) } else flpr if (exp==1.0 && i<=12){

11 11 imrename (oldnames=name, newnames=dirnamell 11s ) } else flpr if (exp==10.0 && i<=12){

11 11 imrename ( oldnames=name, newnames=dirnamel I I 10s ) } else flpr if (i==12){ i=O } else flpr } F-2. ratio.cl 234

F-2 ratio.cl

MS(>-.) - Bright(>-.)- Scat(>-.) (F.l) ES(>-.) - Dark(>-.)- Scat(>-.) (F.2)

The script then used SARITH again to take the ratio of ES(>-.)/MS(>-.) and produce a spectral ratio EA(>-.) for each observation.

string tempi, temp2, temp3, temp4 string dirac, image struct *list1 struct *list2

!ls -d ob* > tempi list1="temp1" while (fscan(list1, dirac) != EOF){ cd (dirac) cd ("0.1s") # !ls *fits > temp2 list2="temp2" i=O while (fscan(list2, image) != EOF){ i=i+1 if (i<=3){ print (image,>> "temp3") } else flpr if (i==4){ print (image,» "temp4") } else flpr } # print ("dark1. fits", » "results") F-2. ratio.cl 235

print ("bright.fits", >>"results") print ("dark2.fits", >> "results") sarith (input1="®temp3", op="-", input2="®temp4", output="®results") # sarith (input1="dark1.fits", op="/", input2="bright.fits", output="ratioi.fits") sarith (input1="dark2.fits", op="/", input2="bright.fits", output="ratio2.fits") del temp* del results # # # cd .• / cd ("is") lls *fits > temp2 list2="temp2" i=O while (fscan(list2, image) != EOF){ i=i+i if (i<=3){ print (image,>> "temp3") } else flpr if (i==4){

print (image,>> "temp4 11 ) } else flpr } # print ("darki.fits", >> "results") print ("bright.fits", >>"results") print ("dark2.fits", >> "results")

sarith (input1="®temp3 10 , op="-", input2="®temp4", output="®results") # sarith (input1="dark1.fits", op=" /", input2="bright .fits", output="ratioi.fits") sarith (input1="dark2.fits", op=" /", input2="bright .fits", output="ratio2.fits") del temp* del results # # # cd •. / cd ("10s") !ls *fits > temp2 list2="temp2" F-2. ratio.cl 236

i=O while (fscan(list2, image) != EOF){ i=i+1 if (i<=3){ print (image,>> "temp3") } else flpr if (i==4){ print (image,>> "temp4") } else flpr } # print ("dark1.fits", >> "results") print ("bright.fits", >> "results") print ("dark2.fits", >> "results") sarith (input1="~temp3", op="-", input2="~temp4", output="~results") # sarith (input1="dark1.fits", op="/", input2="bright.fits", output="ratio1.fits") sarith (input1="dark2.fits", op="/", input2="bright.fits", output="ratio2.fits") del temp>!< del results cd .. / .. / del tempi

} Appendix G

Published Works

G-1 Published Papers

The following papers are included in their published form in this Appendix:

1. COBBER: A Pocket Cloud Detector for Dome C Dempsey, J.T., Storey, J.W., Ashley, Michael C., 2002, Mem. S.A. It., 2, 70-72.

2. AFOS: Probing the optical potential of the Antarctic Plateau Dempsey, J.T. and Storey, J.W.V, 2004, Proc.SPIE, Astronomical Instrumentation, Glas gow June 2004, in press.

3. Auroral contribution to sky brightness for optical astronomy on the Antarctic plateau Dempsey, J.T. and Storey, J.W.V, 2004, Pub. Astron. Soc. Aust., submitted, August 2004.

The following articles have been removed from the digital copy of this thesis. Please see the print copy of the thesis for a complete manuscript

Title: COBBER: A Pocket Cloud Detector for Dome C Authors: Dempsey, J.T., Storey, J.W., Ashley, Michael C. Journal: Mem. S.A

Title: AFOS: Probing the optical potential of the Antarctic Plateau Authors: Dempsey, J.T. and Storey, J.W.V, Journal: Proc.SPIE, Astronomical Instrumentation

Title: Auroral contribution to sky brightness for optical astronomy on the Antarctic plateau Authors: Dempsey, J.T. and Storey, J.W.V, Journal: Astronomical Society of Australia