Assessing the feasibility of a New Solid State Instrument for Remote Sensing of Stratospheric and Other Atmospheric Species.

A thesis submitted to the University of Manchester for the degree of Master of Philosophy (MPhil) in the Faculty of Faculty of Engineering and Physical Sciences

YEAR OF SUBMISION

2016

Eric Gonzalez Peralta

The School of Earth Atmospheric and Environmental Sciences

Total number of words 37540

Table of contents

TABLE OF CONTENTS ...... 2

TABLE OF FIGURES ...... 4

LIST OF TABLES ...... 10

ABSTRACT ...... 11

DECLARATION ...... 12

COPYRIGHT STATEMENT ...... 12

ACKNOWLEDGEMENTS ...... 13

CHAPTER 1. INTRODUCTION ...... 14

1.1. OVERVIEW ...... 14

1.2. STRATOSPHERIC OZONE ...... 15

1.3. AIMS AND OBJECTIVES OF THE RESEARCH ...... 22

CHAPTER 2. CURRENT METHODS OF MEASURING STRATOSPHERIC OZONE ...... 23

2.1. INTRODUCTION ...... 23

2.2. BACKGROUND THEORY OF THE MEASUREMENTS ...... 23 2.2.1. position ...... 24 2.2.2. Scattering ...... 29

2.3. ESTIMATION OF STRATOSPHERIC OZONE ...... 33 2.3.1. Relative optical ...... 33 2.3.2. Beer - Lambert Law...... 37 2.3.3. Differential optical absorption spectroscopy (DOAS) technique...... 42 2.3.4. DOAS technique applied to Brewer spectrophotometer calculations ...... 45

2.4. SPECTROPHOTOMETER ...... 47 2.4.1. Monochromator ...... 47 2.4.2. Plane diffraction grating...... 51 2.4.3. Prism as a dispersion element ...... 54 2.4.4. Photomultiplier tube (PMT) ...... 55 2.4.5. Multichannel detectors: Arrays (PDA) and Charge-Couple Device (CCD) .... 57

2.5. GROUND-BASED MEASUREMENTS ...... 58 2.5.1. Dobson spectrophotometer ...... 59 2.5.2. Brewer spectrophotometer ...... 62 2.5.3. Other ground-based instruments ...... 77

2.6. SATELLITE MEASUREMENTS ...... 82 2.6.1. DOAS technique applied to satellite measurements ...... 83

2 2.6.2. Satellite spectrometers mapping ozone ...... 84

CHAPTER 3. SPECIFICATIONS FOR A NEW INSTRUMENT TO MEASURE STRATOSPHERIC OZONE ...... 89

3.1. GENERAL REQUIREMENTS FOR A GROUND-BASED INSTRUMENT TO MEASURE TOTAL COLUMN OZONE ...... 89

3.2. DESIRED IMPROVEMENTS TO EXISTING SYSTEMS ...... 90

CHAPTER 4. POTENTIAL APPLICATION OF SOLID-STATE TECHNOLOGY ...... 91

4.1. LABORATORY VIRTUAL INSTRUMENT ENGINEERING WORKBENCH (LABVIEW) ...... 91

4.2. DESCRIPTION OF ACOUSTO-OPTICAL TUNABLE FILTER (AOTF) ...... 92

4.3. AVALANCHE PHOTO-DIODE (APD) ...... 95

4.4. LOCK-IN AMPLIFIER ...... 96 4.4.1. Lock-in amplifier algorithm ...... 97 4.4.2. Simulation of a dual channel lock-in amplifier using LabVIEW ...... 101

4.5. PERFORMANCE OF THE LOCK-IN AMPLIFIER ALGORITHM WITH REAL DATA COMING FROM THE APD

MODULE APD110A2/M ...... 105

4.6. OUTLINE FEASIBILITY STUDY ...... 112

CHAPTER 5. CONCLUSION AND FURTHER WORK ...... 113

5.1. CONCLUSION ...... 113

5.2. FURTHER WORK ...... 114 5.2.1. Optical system of the spectrometer ...... 114 5.2.2. Improvements to the LabVIEW program ...... 115

REFERENCES ...... 117

3 Table of figures

Figure 1.1. This figure shows that the vertical distribution of ozone in the atmospheric profile of ozone throughout atmosphere is not uniform. There is a higher concentration in the stratosphere compared with any other region of the atmosphere (WMO, 2010a)...... 15 Figure 1.2 Comparison of the extra-terrestrial solar irradiance measured by satellites to the irradiance measured at ground level, under clear- conditions and solar zenith angle of 60° at the mean Sun-Earth distance (Kondratyev, 1969)...... 16 Figure 1.3 Average of the global atmospheric temperature profile. Image taken from: www.physicalgeography.net ...... 18 Figure 1.4 Chapman cycle of ozone formation. Image taken from: www.scienceinschool.org ...... 18 Figure 1.5 Ozone destruction cycle 1 may start with Cl or ClO, where both substances are re-formed when the reaction cycle is completed, destroying a considerable amount of ozone in the process (WMO, 2010a)...... 20 Figure 1.6 Polar ozone destruction cycles 2 and 3. A large amount of ozone is destroyed in early spring in the polar regions because a significant amount of chlorine is created on PSCs during winter, where the Polar Vortex obstructs the entrance of air rich in ozone from lower latitudes. Image taken from http://ozone.unep.org ...... 21 Figure 1.7 Arctic and Antarctic average temperatures during winter. During winter, Arctic and Antarctic temperatures reach values below ˗ 78oC, which is the temperature required for the formation of PSCs (WMO, 2010a)...... 21 Figure 2.1 Solar declination angle during summer and winter solstice. Image taken from www.itacanet.org ...... 24 Figure 2.2 Representation of the solar declination angle over one-year period for the northern hemisphere. Image taken from www.reuk.co.uk ...... 24 Figure 2.3 Values of the EOT in a period of one year (Stine and Geyer, 2001) ...... 25 Figure 2.4 Representation of the solar hour angle (Stine and Geyer, 2001) ...... 27 Figure 2.5 Graphic representation of variables to estimate the solar zenith angle (Stine and Geyer, 2001)...... 27 Figure 2.6 Representation of the solar zenith angle, solar elevation angle (α) and azimuth angle at a specific point on Earth Q, where S represents a solar ray from the centre of the Sun (Stine and Geyer, 2001)...... 28 Figure 2.7 Scattering of solar radiation by different particles in the atmosphere. Image taken from: http://hyperphysics.phy-astr.gsu.edu ...... 29

4 Figure 2.8 Comparison of different scattering patterns depending on the ratio between the size of the particle Φ and the wavelength of the incident radiation (Trager, 2007)...... 32 Figure 2.9 Diagram representing the energy transition for Rayleigh and Raman scattering. Image taken from: http://bwtek.com ...... 32 Figure 2.10 The portion of the direct solar radiation that reaches the surface of the Earth is approximately half a degree. Image taken from: http://newport.com ...... 33 Figure 2.11 A model of the Earth showing the solar zenith angle at the equinox. As the sun moves north to south through the year, the sun is directly overhead at other latitudes. Image taken from: http://bxhorn.com/extraterrestrial- irradiance/ ...... 34 Figure 2.12 Description of the relative optical air mass taking into account attitudes where stratospheric ozone is concentrated. Solar beam A and B are no

refracted and refracted rays, respectively. Where θ0 is the apparent solar

zenith angle, Zi is the effective height of the layer and Zi-1 is the effective height of the adjacent lower layer (Thomason et al., 1983)...... 35

Figure 2.13 Graph that compares relative optical air mass functions for air (ma), water

vapour (mw), ozone (mo) and nitrogen dioxide (mn) with apparent zenith angles from 68° to 87° (Tomasi et al., 1998)...... 36 Figure 2.14 Portion of the sample analysed by the Beer-Lambert law...... 37 Figure 2.15 Solar irradiance at 324nm measured during clear sky on May 7, during variable clouds on May 6 and during heavy thunder clouds on May 2. Crosses represent measurements made by Brewer #014, while the solid line denotes an estimated model (McArthur et al., 1999)...... 39 Figure 2.16 Example of Langley extrapolation of irradiance measurements at 320nm on the 217th Julian Day of 2000, using a Neutral-density filter 3. The irradiance is considered as photon counts of the Brewer’s photomultiplier tube (Marenco et al., 2002)...... 41 Figure 2.17 Absorption cross-section of ozone at 293K from 231 to 794nm measured with the satellite GOME FM spectrometer in the visible and UV spectral ranges (Burrows et al., 1999a)...... 42 Figure 2.18 Absorption coefficient of ozone within the Huggins bands from 3000Å to 3500 Å. Where Å is a unit equal to 1x10-10m or 0.1nm (Inn and Tanaka, 1953)...... 43 Figure 2.19 Illustrative graph that shows the two components of the absorption cross- section (휎). Panel A displays the smooth slope 휎푏휆, while Panel B shows the faster absorption variations of the trace gas 휎′휆 using 휎푏휆, as a reference. Image taken from:https://www.wmo.int...... 43 Figure 2.20 Principle of DOAS where 휎′ represents the strongest wavelength dependent term and 퐼0′ and 휎푏 is the weaker one (Platt, 2008)...... 45

5 Figure 2.21 Czerny-Turner monochromator. The first concave mirror (M1) collimates the incident beam that passes throught the input-slit, then it is dispersed by the

diffraction grating, then it is focused using the second concave mirror (M2), and finally the selected wavelength passes through the exit slit. Image taken from: www.rp-photonics.com/spectrometers.html ...... 48 Figure 2.22 Ebert-Fastie monochromator design (Mielenz, 1964)...... 48 Figure 2.23 Left diagram displays the Brewer monochromator design and the right picture shows exposed parts of the Brewer’s monochromator (taken by the INTA, Spain during intercomparison 2015)...... 49 Figure 2.24 Representation of spherical aberration with six incident parallel rays...... 50 Figure 2.25 Representation of coma aberration...... 50 Figure 2.26 Effect of coma aberration in the photon count process...... 50 Figure 2.27 Representation of astigmatism aberration...... 51

Figure 2.28 Example of a monochromator using a concave mirror. S1 is the entrance

slit, S2 is the exit slit, G represents the grating, M is a reflective mirror and C is the centre of the Rowland circle (Murty, 1974)...... 51 Figure 2.29 Differences between ruled and holographic diffraction grating. Image taken from: http://ssioptics.com ...... 52 Figure 2.30 Graphical description of grating equation (2.54). Image taken from www.intl- lighttech.com ...... 53 Figure 2.31 A portion of the intensity of the incident ray will be difracted at different diffraction angles, according to m ...... 54 Figure 2.32 Overlapping of spectral orders. Values in parentheses indicates wavelengths absorbed by the atmosphere. Image taken from: http://www.horiba.com ...... 54 Figure 2.33 Designs of a monochromator using a prism as a diffraction element, where L0, L1 and L2 are lens to focus and collimate the incident beam...... 55 Figure 2.34 General schematic of a photomultiplier tube using a linear focusing configuration of dynodes (Hamamatsu Photonics, 2006)...... 55 Figure 2.35 Diagram of a photodiode array spectrometer with concave grating...... 57 Figure 2.36 General diagram of the Dobson spectrophotometer. (R. & J. Beck LTD) ...... 59 Figure 2.37 Horizontal diagram of the Dobson’s monochromator, where M, P and L are the mirrors, prisms and lenses (Komhyr, 1980)...... 60 Figure 2.38 General light path into the monochromator of Dobson spectrophotometer. Image taken from: http://www.o3soft.eu ...... 60 Figure 2.39 Brewer spectrophotometer model MKIII with a double monochromator...... 63 Figure 2.40 Zenith positioning system of the Brewer spectophotometer (right diagram)...... 64 Figure 2.41 Standard azimuth positioning system of the Brewer spectrophotometer ...... 65 Figure 2.42 Simplified design of the single monochromator of a Brewer spectrophotometer (Josefsson, 1986)...... 65

6 Figure 2.43 The optical path of the three sources in Brewer spectrophotometer (left diagram) (Kipp & Zonen, 2014)...... 66 Figure 2.44 Optical elements of the double monochromator Brewer spectrophotometer (MkIII) (Kipp & Zonen, 2014) ...... 67 Figure 2.45 Slit selector mask in the Brewer spectrophotometer (Kipp & Zonen, 2014)...... 67 Figure 2.46 Micrometers in a opened Brewer MkIII...... 68 Figure 2.47 Scan of a 296.728nm mercury line, which is fitted into an isosceles triangle. The circles and crosses are fitted points (Gröbner et al., 1998)...... 69 Figure 2.48 Slit function comparison of a single monochromator against a double one, using a HeCd Laser at 325nm. Image taken from: http://kippzonen- brewer.com ...... 69 Figure 2.49 Temperature dependence of Brewer #107 (blue crosses) and Brewer #037 (green circles) (Lakkala et al., 2008)...... 71 Figure 2.50 Temperature dependence graph of Brewer #014 using an internal lamp over temperatures from 1°C to 39°C (Kerr, 2002) ...... 71 Figure 2.51 Picture of the Brewer spectrophotometers involved in the calibration campaign at El Arenosillo, Spain on May 2015 ...... 75 Figure 2.52 General schematic of the Pandora instrument using a CMOS detector (Herman et al., 2009), as well as an actual instrument performing field measurements (right diagram) ...... 77 Figure 2.53 Diagram of Phaethon spectrometer (Kouremeti et al., 2013) ...... 78 Figure 2.54 Comparison of total column ozone between Phaethon spectrometer and Brewer spectrophotometer #005 (Kouremeti et al., 2013)...... 79 Figure 2.55 Precision Spectroradiometer PSR-003 (Physikalisch-Meteorologisches Observatorium und Weltstrahlungszentrum Davos, 2014) ...... 79 Figure 2.56 Line spread function (LSF) for PSR-003 (Physikalisch-Meteorologisches Observatorium und Weltstrahlungszentrum Davos, 2014; Gröbner et al., 2014)...... 80 Figure 2.57 Photograph of the SAOZ spectrophotometer ...... 80 Figure 2.58 METCON spectrometer at Manchester City (Smedley et al., 2015) (a) displays the temperature stabilised enclosure, the global irradiance entrance (located in the upper right corner), as well as the Brewer #172. (b) Close-up image that shows the direct entrance optics mounted on the solar tracker ...... 81 Figure 2.59 Footprint of the Meteor-3 TOMS projected onto Earth’s surface during a day of each measured sample at 1050km of altitude. The right samples (1 to 18) are obtained from two consecutive scans, displaying the overlap of consecutive orbits at the equator (NASA, 1996a). X-axis represents latitude and Y-axis represents longitude...... 82 Figure 2.60 Different view geometries used to perform satellite measurements of global ozone: limb, nadir and occultation view-geometries (Burrows, 2011) ...... 83

7 Figure 2.61 Fields of view of the Earth Probe TOMS projected onto Earth’s surface of each measured sample at 50km and 750km of altitude. Two consecutive scans are displayed in the bottom of the graph (NASA, 1998b). The X-axis represents latitude and Y-axis represents longitude...... 85 Figure 2.62 Chart that shows the operation of SCIAMACHY at different altitudes (Bovensmann et al., 1999)...... 87 Figure 4.1 Visual interface to manipulate the RF controller of the Acousto-Optical Tunable Filter (AOTF), as well as a section of the block diagram using LabVIEW...... 91

Figure 4.2 Diagram of a non-collinear AOTF (TeO2 crystal) with an unpolarized incident beam [Image taken from http://www.brimrose.com/pdfandwordfiles/tunable_filter.pdf] ...... 92 Figure 4.3 Diagram of the two AOTF-configurations: A) represents collinear configuration and B) represents the non-colinear configuration with incident polarized light (Kaur and Kaur, 2015)...... 93 Figure 4.4 Collinear configuration of the AOTF used in this thesis (Brimrose, 2014)...... 94 Figure 4.5 Virtual interfaz to control the RF generator...... 94 Figure 4.6 Silicon APD detector (image taken from (Hamamatsu, 2004)...... 96 Figure 4.7 Phase-sensitive detection using a chopper wheel and a lock-in amplifier instrument (PerkinElmer Instruments, 2000)...... 97 Figure 4.8 Phase-sensitive detection using an AOTF and a double channel lock-in amplifier algorithm in LabVIEW...... 97 Figure 4.9 Comparison of the spectrums 푋(푡) in logarithmic scale and the filtered outputs with and without offset voltage. The cut-off frequency of the low-pass filter is less than 15Hzand aims to let only the DC component pass. The modulation frequency and the reference frequency are in phase and aqual to 50Hz...... 98 Figure 4.10 Block diagram of the lock-in amplifier detector ...... 99 Figure 4.11 Spectrum of X(t) on a logarithmic scale, before being attenuated by the low- pass filter, with square modulation of 50Hz and reference frequency equal to 50Hz (in phase). Right image shows a close-up of the left image...... 100 Figure 4.12 Comparison of the spectrums X(t) (logarithmic scale) and the filtered outputs (cut-off frequency of the low-pass filter equals 15Hz), where the modulation frequency and the reference frequency are not the same. 푓1 = fsignal − fref = 10퐻푧, and 푓2 = fsignal + fref = 110...... 101 Figure 4.13 Graphic interface developed by the Autor using LabVIEW 2003. It illustrates the output signal of the lock-in amplifier, which uses a clean square intput signal with an amplitude of 0.1V (low-pass filter order = 4 and cut-off frequency = 10Hz) ...... 102

8 Figure 4.14 Display of the output signal of the lock-in amplifier with white noise (amplitude = 1V) 10 greater than the input signal (amplitude = 0.1V), using a low-pass filter order = 4 and a cut-off frequency = 4Hz...... 103 Figure 4.15 Display of the output signal of the lock-in amplifier with white noise (amplitude = 100V) 1000 greater than the input signal (amplitude = 0.1V), using a low-pass filter order = 4 and a cut-off frequency = 2Hz...... 104 Figure 4.16 The graphs shows that the output starts becoming less steady if the noise starts becoming close to the modulation frequency. Amplitude square signal = 0.1V, amplitude noise = 1V, settling time = 0.8 seconds...... 105 Figure 4.17 ADC instrument NI-USB4431 (National Instruments, 2009) ...... 106 Figure 4.18 Avalanche photodetector module APD11A02/M (THORLABS, 2011), as well as its quantum efficiency (Hamamatsu, 2004)...... 107 Figure 4.19 Measured spectrum of a UV LED with substracted dark-current measurements, using a TE cooled CCD array spectrometer (model BTC112E) (central wavelength is near 365nm)...... 107 Figure 4.20 AOTF with the water cooling system...... 108 Figure 4.21 Field of view of polarizers (THORLABS, 2015) ...... 108 Figure 4.22 The plot displays the measured spectrum of a UV LED with substracted dark-current measurements, after being filtered by the AOTF. The spectrometer used was a TE cooled CCD array model BTC112E...... 109 Figure 4.23 Block diagram of the optical system used to test the lock-in amplifier...... 109 Figure 4.24 Picture of the optical system used in the experimentation stage...... 110 Figure 4.25 Graphic interface designed in LabVIEW to manupulate the RF control of the AOTF and visualize the input and output signal of the lock-in amplifier (modulation frequency equal to 80Hz)...... 111 Figure 4.26 Left panel of the graphic interface designed in LabVIEW, which displays the input and output signal of the lock-in amplifier when the intensity of the light was reduced using the UV LED driver...... 111 Figure 5.1 Block diagram of the proposed design of the spectrometer prototype...... 115

9 List of tables

Table 2.1. Coefficients used in the equation (2.4) ...... 26 Table 2.2 Constants used in equation (2.15) for standard air at 101.325kPa and 288.15K (Bucholtz, 1995) ...... 30 Table 2.3 Ozone absorption and scattering coefficients for the selected pair of wavelengths ...... 61 Table 2.4 Comparison among the TOMS instruments (NASA, 1996b; NASA, 1996a; NASA, 1998b; NASA, 1998a)...... 86 Table 4.1 List of modulation frequencies generated by the electronic circuit...... 106

10 Abstract

Assessing the feasibility of a New Solid State Instrument for Remote Sensing of Stratospheric Ozone and Other Atmospheric Species.

Eric Gonzalez Peralta

A thesis submitted to the University of Manchester for the degree of Master of Philosophy (MPhil), 2016.

Column ozone (predominantly located in the stratosphere) has been measured at distinct ground sites since the 1920s, and globally from satellites from the late 1970s. The importance of ozone was highlighted by recognition of the Antarctic ozone hole in 1985, leading to international action (Montreal Protocol) to protect the ozone layer. Monitoring the success of this action and the future ozone layer is currently reliant on the “gold standard” represented by Dobson and Brewer spectrophotometers. The manually operated Dobson instrument is no longer manufactured, although it is still in operation, leaving the Brewer as the only ground-reference spectrophotometer commercially available. While these instruments provide the standard for column ozone measurement, improvements in technology over the past 40 years provide possibilities for alternative approaches to column ozone measurement with advantages that might include increased speed of data acquisition, reduced size and weight, improved portability and lower cost. The key elements of any spectrophotometer are a dispersion system to isolate different wavelengths (traditionally diffraction gratings), and a suitably sensitive detector for those wavelengths (traditionally a photomultiplier tube PMT). Any replacement technology should achieve an accuracy comparable to measurements obtained with Dobson and Brewer spectrophotometers. Most of the recent alternative spectrometers found in the literature use CCD or photodiode arrays to replace the PMT, but retain the diffraction grating. They have to overcome stray-light and sensitivity issues to achieve an accuracy similar to the Brewer spectrophotometer measuring ozone. Here a radically different design is proposed for a UV ground-based spectrometer to measure column ozone. An Acousto-Optical Tuneable Filter (AOTF) replaces the traditional diffraction grating, and an avalanche photo-diode detector (APD) is used instead of the photomultiplier tube, with a lock-in amplifier algorithm to perform the digital signal processing. This new spectrometer design has the potential to diminish moving parts in the instrument (improving reproducibility and decreasing maintenance), increase the measurement rate and reduce production costs without affecting the reliability of the measurements. The aim of this thesis is to prove the feasibility of a new spectrometer that uses this solid state technology and the lock-in amplifier algorithm. Experimental results suggest that the AOTF and APD are feasible options to replace the diffraction grating and the photomultiplier tube respectively in the traditional spectrometer design.

11 Declaration

No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning;

Copyright statement

i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=487), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.manchester.ac.uk/library/aboutus/regulations) and in The University’s policy on Presentation of Theses.

12 Acknowledgements

The financial support for the work found into this Master of Philosophy was given by the Mexican government through an awarded scholarship given by CONACyT (Mexican National Council for Science and Technology). In addition, Kipp and Zonen provided all the necessary resources and facilities, such as specialized equipment and material, to complete this initial stage of the project.

First and foremost, I would like to express my deepest gratitude to my main supervisor Prof. Ann Webb, as well as my co-supervisors Dr. John Rimmer and Prof. Matthew Halsall, for providing me extraordinary guidance in successfully finishing my Master of Philosophy degree at the University of Manchester. They steered me in the right direction, being always willing to support my research and being concerned about my academic, professional and personal development.

I would also like to thank Dr. Andrew Smedley and Dr. Richard Kift, who were involved in the process of preparing my thesis from the beginning, whose advices, suggestions and feedback were extremely useful in my research. In addition, I would like to thank Eric Whittaker, whose wise advice and skills in the laboratory were extremely helpful during the experimental stage.

Finally, I would like to express my gratitude to my family, professors and friends, who have always supported and encouraged my academic trajectory through the good times and the bad ones. This accomplishment would not have been possible without them

13 Chapter 1. Introduction

1.1. Overview

Ozone molecules are formed by three oxygen atoms and exist as a gas in the troposphere and the stratosphere. The stratospheric ozone is destroyed and regenerated in a natural cycle by ultraviolet radiation, acting as a barrier to this dangerous radiation that damages any kind of life on Earth. For this reason, it has been monitored since scientists developed the tools to estimate the amount of stratospheric ozone. Those instruments have been undergoing changes in their physical and internal design, technology and algorithms over the years, allowing significant improvements in accuracy, time of response, the number of samples per second and temperature dependence.

Developments in ozone instrumentation, have allowed scientists to study tropospheric and stratospheric ozone as well as its chemical reactions in the atmosphere. Measurements such as ozone vertical profiles and total column ozone are obtained by using remote sensing and in-situ instruments. While in-situ instruments are used to measure ground-level ozone and ozone profiles by analysing chemically or optically a sample of air, satellites and ground-based spectrophotometers use differential absorption techniques to estimate the total column amount of ozone as well as ozone vertical profile at a specific location on Earth.

By comparing the intensities of two or more ultraviolet wavelengths present in the solar spectrum, the thickness of the ozone layer can be measured indirectly after differential absorption in the atmosphere. At the end of the 1920s, the Dobson instrument was the first spectrophotometer designated and dedicated to measuring stratospheric ozone. Consequently in the late 1950s, the global network of Dobson spectrophotometers, in charge of monitoring the ozone layer, was created. Nowadays, it is still active with approximate 60 instruments used in 2012 (WMO, 2015). The Dobson spectrophotometer is a large, heavy and manually operated device that needs to be placed in a thermally insulated location. For this reason, the necessity of improving this instrument and adding the ability to measure other atmospheric variables pushed scientists to develop the Brewer ozone spectrophotometer.

The Brewer instrument is an automated outdoor spectrophotometer which measures UV radiation at five wavelengths in almost real-time. By solving weighted Beer-Lambert equations for each wavelength, the spectrophotometer obtains the , allowing an accurate estimation of the total column ozone, SO2 (MKII and MKIII) and NO2 (MKIV model) as well as the vertical ozone profile. The Brewer spectrophotometer is also able to measure UV spectral radiation over a limited spectral range at ground level. Nonetheless, new technologies continue to emerge with features that could be used to develop new instruments capable of measuring ozone and other atmospheric species, whilst at the same time reducing the number of moving parts, decreasing the size, maintenance and production costs of new spectrophotometers, without losing the quality of the measurement. In order to assess the capabilities of such new

14 technologies, it is first necessary to understand both the importance of ozone and the existing instrumentation used in its measurement.

1.2. Stratospheric ozone

The average concentration of ozone molecules in the atmosphere is about one in ten million air molecules, but is not uniformly distributed in latitude or altitude. The greatest concentration of ozone occurs in the stratosphere, in a zone called the ozone layer approximately from 18km to 35km altitude, where it is naturally generated by solar radiation. While stratospheric ozone is responsible for protecting living beings from harmful ultraviolet radiation, the remaining ozone that resides in the troposphere (about 10%) is considered a pollutant gas with toxic effects in humans and vegetation. Figure 1.1 shows the average ozone distribution over the Earth.

Figure 1.1. This figure shows that the vertical distribution of ozone in the atmospheric profile of ozone throughout atmosphere is not uniform. There is a higher concentration in the stratosphere compared with any other region of the atmosphere (WMO, 2010a).

The global average thickness of stratospheric ozone is about 3 millimetres or 300 Dobson Units (DU) if all the ozone is brought to STP (standard temperature and pressure at the ground), varying geographically between 230 DU and 500 DU. The Dobson unit, in honour of Gordon Miller Bourne Dobson (inventor of the Dobson spectrophotometer), is used as a standard for reporting the total column amount of ozone in the atmosphere, between the surface of the Earth and the sun. One Dobson unit is defined as the number of ozone molecules that would be required to create a 0.01 millimetres thick layer of pure ozone at 0 degrees Celsius with a pressure of 1 atmosphere.

It is important to mention that data records from radiometers, which are instruments dedicated to measuring electromagnetic radiation, has been used for many years to estimate the amount of incident solar radiation that reaches the atmosphere before being attenuated, absorbed or scattered by molecules and at different attitudes. According to the World Meteorological

15 Organization (WMO, 2010b), the total electromagnetic radiation of the Sun received at the outer atmosphere per unit of time and per unit of area is approximately similar to a blackbody at a temperature of 6000 K, with about 97 percent of all the energy in a range of wavelengths from 290nm to 3μm. The extra-terrestrial solar radiation on a horizontal surface at the top of the Earth’s atmosphere oscillates from 1412 Wm-2 to 1324 Wm-2 during a year according to the distance between the Earth and the Sun (Paulescu, 2013). However, the average solar constant accepted by WMO (2010b) is 1367 Wm-2. Figure 1.2 makes a comparison among the irradiance emitted by a blackbody at 6000K, the extra-terrestrial solar irradiance, and the irradiance at ground level, as well as the most important compounds that attenuate solar radiation such as O3, O2, H2O and

CO2.

Figure 1.2 Comparison of the extra-terrestrial solar irradiance measured by satellites to the irradiance measured at ground level, under clear-sky conditions and solar zenith angle of 60° at the mean Sun-Earth distance (Kondratyev, 1969).

The ozone layer plays an important role protecting living organisms by blocking harmful ultraviolet radiation from the solar spectrum. Ultraviolet radiation is a region of the electromagnetic spectrum with wavelengths from 100nm to 400nm, according to the BS ISO 21348:2007 standard (Ace, 2007). And it is subdivided into three subcategories: UV-A (315 - 400nm), most of which reaches the Earth’s surface; UV-B (280 - 315nm), which is significantly attenuated; and UV-C (100 - 280nm), which is almost all absorbed by the ozone layer. The effects of UV radiation on biological systems are based on the photon energy which in turn relates to the wavelength. For example, non-prolonged exposure to UV-A plays a necessary role in some living creatures, but on the other hand, the UV-B exposure can generate molecular damage or in the worst case DNA damage.

Humans, animals and plants are adversely affected in their health or their growth by over- exposure to UV radiation. Although small amounts of UV exposure in humans are necessary to produce vitamin D, scientific studies have found that the most important consequences of

16 excessive exposure to UV radiation are skin diseases, skin cancer, cataracts and immune system suppression; where ultraviolet radiation photons are directly absorbed on a molecular level (Morison, 1989; Ravanat et al., 2001; Matsumura and Ananthaswamy, 2004).

The amount of atmospheric ozone determines the intensity of the ultraviolet radiation that reaches the surface of the Earth. However, other factors must be taken into account in UV measurements, such as the zenith angle of the Sun and particles in the atmosphere, better known as aerosols, which absorb or scatter solar radiation according to the ratio between their size and the wavelength of the incident radiation. It is possible to accurately predict the elevation angle of the Sun at any specific location on Earth at a specific time by knowing the latitude, longitude, date and time. The Sun’s location is usually expressed as the zenith and azimuth angle relative to the observer’s position. The concepts, equations and the algorithms used to obtain the scattering, solar zenith angle and solar azimuth angle will be more fully explained in Chapter 2.

Most of the stratospheric ozone is generated in the tropics, because of the lower solar zenith angle during midday. However, the natural stratospheric phenomenon called the Brewer-Dobson circulation carries ozone from the equator to the poles at stratospheric altitudes (Miyazaki et al., 2005), a process that causes the amount of ozone to be a minimum at the equator and a maximum around the poles. Moreover, there is a meteorological correlation of ozone with the upper troposphere and the lower stratosphere, having a seasonal variation where the maximum total column ozone is in spring and the minimum total column ozone is in autumn at mid-high latitudes.

High-energy solar radiation is responsible for the photochemical reactions that generate and destroy ozone in the stratosphere. The heat produced during the cycle of ozone formation is dissipated in the air, playing an important role in the temperature structure of the atmosphere. Figure 1.3 describes the temperature profile of the atmosphere, where a higher temperature in the stratosphere than in the troposphere can be noticed due to the ozone layer.

17 Figure 1.3 Average of the global atmospheric temperature profile. Image taken from: www.physicalgeography.net

Sydney Chapman (1930) proposed the cycle of photochemical reactions that explain the natural production and destruction of ozone in the stratosphere, which was called the Chapman cycle in his honour (see Figure 1.4). This process will be described in the following paragraphs.

First, a molecule of oxygen is hit by a high-energy photon from the ultraviolet spectrum (ℎ푣) with 휆 < 110푛푚, and then it is dissociated into two oxygen atoms.

∎ 푂2 + ℎ푣 → 2푂 (1.1)

The resulting oxygen atoms (2푂∎) are highly reactive, and they combine with oxygen molecules in an exothermic reaction creating ozone. Therefore, three molecules of oxygen are transformed into two molecules of ozone.

∎ 푂 + 푂2 + 푀 → 푂3 + 푀 (1.2)

Where M is an energy-absorbing molecule.

After a high-energy ultraviolet photon (푎푝푝푟표푥𝑖푚푎푡푒푙푦 휆 < 300푛푚) interacts with an O3 molecule, it is dissociated into an oxygen molecule and a free oxygen atom. Nonetheless, an excess of kinetic energy is released in the reaction, transforming UV light radiation into heat in the stratosphere. Then the free atom joins with an O2 molecule to re-form another ozone molecule, and the cycle starts again.

∎ 푂3 +ℎ푣 → 푂2 + 푂 (1.3)

When an oxygen atom and an ozone molecule react, they form two oxygen molecules.

∎ 푂3 + 푂 → 2푂2 (1.4)

Figure 1.4 Chapman cycle of ozone formation. Image taken from: www.scienceinschool.org

18 However, ozone is also produced in small amounts at ground level when solar radiation reacts with oxides of nitrogen and volatile organic compounds, which are generated by natural processes such as volcanic eruptions and human activities. See equations (1.5) and (1.6).

푁푂2 + ℎ푣(휆 ≤ 400푛푚) → 푁푂 + 푂 (1.5)

Where ℎ푣 is a photon of UV radiation and λ is the wavelength of the photon.

푂 + 푂2 → 푂3 (1.6)

Tropospheric ozone, also called ground ozone, is the principal component of smog commonly found in polluted environments. However, it can be transported to rural areas by the wind, affecting a bigger population. Because of the smog, the air quality may reach unhealthy levels, generating health problems such as eye and respiratory infections. On the other hand, even when ground-level ozone is able to absorb UV radiation, its contribution is insignificant when it is compared with the amount of solar radiation attenuated by the stratospheric ozone, where the ozone layer resides.

Since the late 1970’s, there has existed a remarkable local depletion of the ozone layer, as can be seen from the record of the ozone hole over Antarctica since 1979 (NASA). The ‘ozone hole’ is a seasonal phenomenon that covers almost the entire southern continent, where the amount of ozone in the stratosphere reduces dramatically. In the outstanding depletions of the mid-1990s, the column of ozone remained close to 100 DU (NASA), the lowest values in ozone records. However, the ozone depletion is not only limited to the South Pole, it has affected North America, Europe, Russia, Australia, New Zealand and a considerable part of South America with no significant depletion in the tropics, allowing higher levels of harmful UV reach a larger surface area of the Earth.

In 1985, the ozone hole was identified by British Antarctic Survey staff in the Antarctic (Farman et al., 1985) and validated by satellite measurements, resulting in a significant interest in the measurement of stratospheric ozone to know its impact on the environment. Although the Dobson spectrophotometers (Dobson, 1931) had been collecting information about ozone in the stratosphere since the mid-1920s, with the majority being deployed during the International Geophysical Year (IGY) in 1957, it was not until the 1980s that an increased number of scientists around the world began to focus on coordinating the data and taking a global view, helped by the advent of satellite measurements.

The catalytic destruction of ozone by chlorofluoromethanes was first described by Molina and Rowland (1974). Chlorofluorocarbons (CFCs), first synthesized in the 1920s, were massively produced and consumed particularly in developed countries after 1960’s. They were commonly used to manufacture plastic materials such as Styrofoam, used as industrial solvents, as propellants in sprays, and used in refrigerators and air conditioners. One of the main reasons for their initial success was that CFCs are non-toxic and non-carcinogenic substances to human beings. Because of their chemical inertness, CFCs don’t react with other molecules in the lower

19 atmosphere. For this reason, they are able to reach the stratosphere, where chlorine is produced by the photolysis reaction.

Since free oxygen and ozone molecules are extremely unstable; they react easily with other atmospheric compounds, such as bromine, chlorine, nitrogen and hydrogen, in the presence of UV radiation. Figure 1.5 describes the cycle 1 of ozone depletion due to chlorine, which is more noticeable at tropical and middle attitudes, where the ultraviolet reaction is more intense.

Figure 1.5 Ozone destruction cycle 1 may start with Cl or ClO, where both substances are re-formed when the reaction cycle is completed, destroying a considerable amount of ozone in the process (WMO, 2010a).

The cycle previously described is called catalytic, because bromine and chlorine remain after reactioning. Both substances are regenerated each time the cycle is completed, destroying a significant amount of stratospheric ozone before they are transformed into a relatively unreactive substance. Whereas ozone is depleted and naturally regenerated by UV radiation below 242nm in the stratosphere, chlorine is produced by visible light radiation (Farman et al., 1985). Thus, this catalytic chain reaction depletes stratospheric ozone at a faster rate than the photochemical Chapman cycle. Although there are natural sources of chlorine in the stratosphere, the majority of these halogen substances come from man-made compounds (CFCs or HCFCs), consequently resulting in the destabilization of the natural cycle.

A decrease in temperature in the atmosphere significantly contributes to ozone depletion, where lower temperatures at the poles in winter create the required conditions to generate polar stratospheric clouds (PSCs). In the polar winter, there are no photochemical reactions due to the absolute darkness. However, a quantity of ClO gas is generated when HCl and ClONO2 react on solid and liquid particles of PSCs. As soon as the solar radiation reaches the Antarctic or Artic in the late winter, chlorine is obtained from the photochemical dissociation of ClO, and the chain reaction starts destroying the ozone layer (see Figure 1.6). Consequently, the Antarctic ozone hole occurs during the southern spring between September and November until the polar vortex disappears and ozone depletion in the Artic takes place between March and April.

20

Figure 1.6 Polar ozone destruction cycles 2 and 3. A large amount of ozone is destroyed in early spring in the polar regions because a significant amount of chlorine is created on PSCs during winter, where the Polar Vortex obstructs the entrance of air rich in ozone from lower latitudes. Image taken from http://ozone.unep.org

The polar vortex is a winter phenomenon in which the jet stream acts as a barrier between the polar regions and the mid-latitude warmer circulation, not allowing air rich in ozone to move to the poles. The reduced amount of land in the southern hemisphere and the amount of water that surrounds the continent allow Antarctic Polar vortex to be isolated, stable and extremely cold. The stratospheric temperatures inside the polar vortex fall below ˗78°C for about 1 to 2 months over the Arctic and 5 to 6 months over the Antarctic during the winter season, where temperatures below ˗78°C allows polar stratospheric clouds to form inside the vortex, and they remain there until the photochemical reaction of ozone once again begins to heat the atmosphere and transporting air from lower latitudes takes place.

Figure 1.7 Arctic and Antarctic average temperatures during winter. During winter, Arctic and Antarctic temperatures reach values below ˗ 78oC, which is the temperature required for the formation of PSCs (WMO, 2010a).

In order to protect the ozone layer by regulating and phasing out substances that destroy stratospheric ozone, the Montreal Protocol was put in place in 1987. Although production of CFCs has been almost phased out and hydro-chlorofluorocarbons (HCFCs) have been significantly reduced (90% in 2015) by the Montreal Protocol (UNEP, 2012), ozone destruction will still continue, due to the fact that CFC molecules can stay in the atmosphere for several decades. It

21 is estimated that the ozone layer will return to 1980 levels between 2050 and 2070 (Douglass et al., 2014).

The steady ozone depletion extending over the world as well as the seasonal ozone depletion at the poles, most severely at the Antarctic, has important consequences on Earth’s surface. The increased intensity of solar UV radiation has implications for the sustainability of ecosystems and it directly or indirectly affects all the organisms on Earth. Its strong relationship with the chemistry in the atmosphere allows an increasing photochemical smog produced in the troposphere, which rises when ultraviolet radiation is more intense. The photolysis of the pollutant NO2 produces ozone at ground level, which is a hazardous compound for all living beings. Thus, an increase in UV radiation has a considerable impact on human and animal health, affecting primary human activities such as fisheries, forestry and agriculture mainly in polluted cities and their surroundings (Solomon, 2008).

By recording the long term variation in stratospheric ozone, it is possible to study a major influence on the variation of solar UV radiation that reaches the surface of the Earth, as well as studying the chemistry in the atmosphere (Kerr et al., 1981). Moreover, these values are used to observe the recovering of the ozone layer and mitigate any further depletion.

1.3. Aims and objectives of the research

The aim of this research is to determine the feasibility of developing a spectrophotometer based on the use of an Acoustic Optical Tunable Filter (AOTF), as a replacement of the diffraction grating; an avalanche photodiode (APD), instead of the photomultiplier tube; and the implementation of digital signal processing using the program LabVIEW. Each part of the process has to be evaluated and analysed separately to determine if this new device will be capable of accurate measurements of total column ozone to the same standards as existing instruments.

The first task is to thoroughly understand the performance and characteristics of the current commercially available Brewer spectrophotometer, used worldwide to measure column ozone. This provides the benchmark that any new instrument must at least match.

Following that the characteristics of the AOTF must be assessed, which requires software control and a detection system to be developed. To prove that the AOTF is a feasible core technology on which to base a spectrophotometer, it has to be able to separate a polychromatic source into its monochromatic components, providing a monochromatic signal that can be clearly identified by the detection system. If it can do this at ultraviolet wavelengths then it shows promise as the core of an ozone detection system. The practical objectives are, therefore, to provide control software for the AOTF, to develop digital signal processing for the output of the APD detector, and to show that the system works at a range of wavelengths.

22 Chapter 2. Current Methods of Measuring Stratospheric Ozone

2.1. Introduction

By routinely measuring and computing the quantity of ozone over different countries around the world, it is possible to have a record of ozone’s behaviour with time. These measurements also help to assess the effectiveness of the Montreal Protocol and help to model estimations of the photochemistry of the atmosphere. In order to have reliable data, it is necessary to obtain high-quality, long-term records from well-maintained and calibrated instruments.

Nowadays, there are two methods used to measure the ozone: in-situ and remote sensing measurements. The first method (in-situ) analyses a sample of the air to determine the density of ozone in a particular region by a chemical reaction or analysing its UV radiation absorption. Some instruments are installed a few meters above the ground in large cities to measure ozone created by industrial and urban pollution in the lower troposphere. This technique is also used by ozone- sondes to obtain a vertical profile of ozone. The second method (remote-sensing) estimates the total amount of ozone contained in a vertical column in the atmosphere by measuring irradiance, usually in the UV spectrum. Dobson and Brewer spectrophotometers are ground-based instruments that measure ultraviolet light absorption of sunlight, in order to estimate the total amount of ozone at a specific point over the Earth. Some satellites use the same principle to map the ozone layer, providing a global view. However, ground references around the world are necessary to validate the data coming from satellites. Brewer and Dobson spectrophotometers are considered as ground reference for satellite measurements.

2.2. Background theory of the measurements

The accuracy of the spectrophotometers analysing the direct beam resides in positioning the solar tracker in the correct zenith and azimuth angle because it is necessary to align the entrance slit of the spectrophotometer with the centre of the solar disc. This dependence is accentuated at high solar zenith angles, such as measurements performed near twilight or measurements performed by stations at high latitudes during winter (Josefsson, 1992). For this reason, this chapter will mention the most important variables used by spectrophotometers measuring the direct solar beam.

23 2.2.1. Sun position

The following paragraphs explain how to obtain the solar zenith angle (θ) and the solar azimuth angle (A), which are essential to measure or compute directly or indirectly many variables related to solar radiation.

Declination angle of the Sun

The axis of rotation of the Earth, which is always pointing towards the Pole Star, is currently tilted 23.45° compared with the normal plane of the Earth’s orbit around the Sun. This angle is called the inclination of the Sun, and it is defined as the angle between the equatorial plane and the imaginary line that pass through the centre of the Earth towards the centre of the Sun. As a consequence of this tilt, the number of hours of daylight during the day is affected by the latitude of the observer, reaching maximum and minimum values in winter and summer respectively. Figure 2.1 shows both cases.

Figure 2.1 Solar declination angle during summer and winter solstice. Image taken from www.itacanet.org

The declination angle varies seasonally and periodically over a one-year cycle, having different values in the northern and southern hemispheres, but having zero values at Autumnal and Vernal equinox. The declination angle reaches its maximum and minimum values during June and December respectively (see Figure 2.2).

Figure 2.2 Representation of the solar declination angle over one-year period for the northern hemisphere. Image taken from www.reuk.co.uk

24 The values of the declination angle of the Sun (훿) simulate approximately a sine wave, which can be conveniently estimated by expression (2.1) in terms of the day of the year (푑) starting on 1 of January (Cooper, 1969).

284 + 푑 훿 = (23.45) ∙ 푠𝑖푛 [ (2휋)] 365 (2.1)

However, the formula (2.1) does not take into account the asymmetry generated by the elliptical movement of the Earth around the Sun. The equation (2.2) gives a more accurate approximation over a cycle of four years using the values of the Julian day (푁) (Scharmer and Greif, 2000).

360푁 360푁 훿 = 푎푟푐푠𝑖푛 {0.3978 ∙ 푠𝑖푛 [ − 80.2° + 1.92° ∙ 푠𝑖푛 ( − 2.8°)]} 365.25 365.25 (2.2)

Local solar time

The local solar time, also known as sun time, is computed based on the latitude and the longitude of a particular place. On the other hand, the term called local time or mean time refers to the system used in our daily life according to the time zone to which we belong.

In order to calculate the local solar time, first it is necessary to obtain the local standard time meridian (LSTM) and the equation of time (EOT).

The LSTM is a reference meridian situated in the location where the estimation of hour angle in solar time is going to take place. See equation (2.3).

퐿푆푇푀 = (15°) ∙ (difference in hours between the local time and the GMT) (2.3)

The EOT is used to correct the solar time, which is affected by the Earth’s axial tilt, varying minutes or seconds from day to day during the entire year. It is defined as the difference in minutes between the local solar time and the true local solar time at a specific point on Earth. Figure 2.3 estimates the periodical behaviour of the EOT in 365 days, indicating the minimum and maximum values obtained.

Figure 2.3 Values of the EOT in a period of one year (Stine and Geyer, 2001)

25 The mathematical expression (2.4) presented by (Lamm, 1981), considers the coefficients 퐴푘 and 퐵푘 given in Table 2.1 to model the periodic values of the EOT.

5 2휋푘푁 2휋푘푁 퐸푂푇(푁) = ∑ [퐴 푐표푠 ( ) + 퐵 푠𝑖푛 ( )] 푘 365.25 푘 365.25 (2.4) 푘=0

N is defined as the number of the day of interest from 1 to 1461.

Table 2.1. Coefficients used in the equation (2.4)

k Ak (hr) Bk (hr)

0 2.0870x10-4 0

1 9.2869x10-3 ˗ 1.2229x10-1

2 ˗ 5.2258x10-2 ˗ 1.5698x10-1

3 ˗ 1.3077x10-3 ˗ 5.1602x10-3

4 ˗ 2.1867x10-3 ˗ 2.9823x10-3

5 ˗ 1.5100x10-4 ˗ 2.3463x10-4

The time correction equation (2.5) takes into account: longitude variations, the EOT and the fact that it takes the Earth four minutes to rotate one degree of longitude.

푇𝑖푚푒 푐표푟푟푒푐푡𝑖표푛 = 4 ∙ (퐿표푐푎푙 푙표푛푔𝑖푡푢푑푒 − 퐿푆푇푀) + 퐸푂푇 (2.5)

Applying the time correction adjustment, the local solar time is determined by the equation (2.6) using a format of 24 hours.

푇𝑖푚푒 푐표푟푟푒푐푡𝑖표푛 퐿표푐푎푙 푠표푙푎푟 푡𝑖푚푒 = 퐿표푐푎푙 푡𝑖푚푒 + + 퐷 60 (2.6)

Note that 퐷 is equal to 1hr if the daylight savings time is used in the location and it is equal to zero otherwise.

Hour angle

The hour angle indicates the angular difference between the meridian of the observer and the solar noon meridian, with negative values during the morning and positive values after noon.

The hour angle (ω) in local solar time is obtained by equation (2.7).

ω = 15°(퐿표푐푎푙 푠표푙푎푟 푡𝑖푚푒 − 12ℎ푟푠) (2.7)

26

Figure 2.4 Representation of the solar hour angle (Stine and Geyer, 2001)

Solar zenith angle

The solar zenith angle is defined as the angle between the normal axis to the surface of the Earth above the observer (Q), and the direct solar ray (S) in Figure 2.5.

Figure 2.5 Graphic representation of variables to estimate the solar zenith angle (Stine and Geyer, 2001).

The solar zenith angle (see Figure 2.5) varies according to time, day number, local longitude and local latitude. As a consequence, the solar zenith angle at midday of solar time will not be constant through the year. The formula (2.8) computes the solar zenith angle taking into account all the parameters described above.

푐표푠(휃) = 푠𝑖푛 휑 푠𝑖푛 훿 + 푐표푠 휑 푐표푠 훿 푐표푠 ω (2.8)

Where:

θ is the solar zenith angle. φ is the local latitude.

27 훿 is the declination angle of the Sun. ω is the hour angle in the local solar time.

Solar azimuth angle

The solar azimuth angle (퐴) is the angle from the local meridian pointing to the geographic North to the horizontal projection of the solar ray in a clockwise direction (WMO, 2010b). It is measured in a plane that is tangential to the Earth’s surface at a specific point Q (see Figure 2.6), having a range of values from 180° to -180° with negatives values when the hour angle is positive and positive values otherwise.

Figure 2.6 Representation of the solar zenith angle, solar elevation angle (α) and azimuth angle at a specific point on Earth Q, where S represents a solar ray from the centre of the Sun (Stine and Geyer, 2001).

The solar azimuth angle is obtained using the equation (2.9) (WMO, 2010b):

− 푐표푠 훿 ∙ 푠𝑖푛 ω 푠𝑖푛(퐴) = 푠𝑖푛 휃 (2.9)

However, after solving for (2.9), it is necessary to correct the values depending on the solar zenith angle. See equation (2.10).

푠𝑖푛 훿 푐표푠(훿) ∙ 푠𝑖푛(ω) 푐표푠 휃 < → 퐴 = 푎푟푐푠𝑖푛 (− ) 푠𝑖푛 휙 푠𝑖푛(휃) 𝑖푓 (2.10) 푠𝑖푛 훿 푐표푠(훿) ∙ 푠𝑖푛(ω) 푐표푠 휃 ≥ → 퐴 = 180 − 푎푟푐푠𝑖푛 (− ) { 푠𝑖푛 휙 푠𝑖푛(휃)

Where 휙 is the local latitude.

However, (2.11) expresses a second method to estimate the solar azimuth angle using the South as a reference (Duffie, 1980).

푠𝑖푛(훿) − 푐표푠(휃) ∙ 푠𝑖푛(휙) 푐표푠(퐴) = 푠𝑖푛(휃) ∙ 푐표푠(휙) (2.11)

28 Solving for A, the resulting equation (2.12) automatically shows negatives values before local noon and positives after local noon, such that

푠𝑖푛(훿) − 푐표푠(휃) ∙ 푠𝑖푛(휙) ω < 0 → 퐴 = − |푎푟푐푐표푠 ( )| 푠𝑖푛(휃) ∙ 푐표푠(휙) 𝑖푓 (2.12) 푠𝑖푛(훿) − 푐표푠(휃) ∙ 푠𝑖푛(휙) ω ≥ 0 → 퐴 = + |푎푟푐푐표푠 ( )| { 푠𝑖푛(휃) ∙ 푐표푠(휙)

While this does not require any post-correction to the sign, it has the opposite sign convention to equation (2.11). Therefore, it is important to know and understand exactly which convention is being followed.

2.2.2. Scattering

The particles in the atmosphere, such as aerosol particles, water droplets, ice particles and air molecules, interact with the Sun’s radiation by absorbing photons, changing the photon’s trajectory or modifying its frequency with each collision. The relationship between the characteristics of the particle (size, shape and composition) and the wavelength of the incident radiation determines the type of collision produced as well as the scattering pattern (see Figure 2.7). Other optical phenomena such as polarization, phase function and coherence are also taken into account in the estimation of the scattering coefficients. The following paragraphs will describe the three different scatterings regimes: Rayleigh scattering, Mie scattering and Raman scattering.

Figure 2.7 Scattering of solar radiation by different particles in the atmosphere. Image taken from: http://hyperphysics.phy-astr.gsu.edu

Rayleigh scattering

Lord Rayleigh discovered that the radiation scattering produced by the collision between photons from solar radiation and gas molecules has a wavelength dependence of 휆−4. It is considered an elastic collision, where the scattering pattern is symmetric and the energy of the scattered photon is not affected, keeping the initial wavelength.

In order to be classified as Rayleigh scattering, the diameter of the gas molecule must be much smaller than the wavelength of the incident radiation, which for solar radiation implies the

29 gas molecules of the air in the atmosphere. Most molecules are concentrated in the troposphere, where the short wavelengths from the solar spectrum are effectively scattered. Thus, the visual perception of the blue sky during the day in cloud-free conditions comes from Rayleigh scattering of short wavelengths from the solar spectrum.

The gas molecules that contributes to Rayleigh scattering are considered spherical and stationary (with no rotational or vibrational movements) in the further mathematical analysis. The formula (2.13) solves for the total Rayleigh scattering cross-section per molecule (Bucholtz, 1995), indicating the effective area of the molecules that disperse the direct solar radiation at a specific wavelength:

3 2 2 24휋 (푛 + 1) 6 + 3휌푛 훼(휆) = 4 2 2 2 ( ) (2.13) 휆 푁 (푛 + 2) 6 − 7휌푛

Where

훼(휆) is the total Rayleigh scattering cross-section per molecule λ is the wavelength of the incident photon. n is the refractive index of the particle at λ. N is the molecular number of density air.

휌푛is depolarization factor (the air particles do not uniformly disperse the light). The equation (2.14) represent the Rayleigh scattering extinction coefficient, which is the product of the total absorption cross-section (휎푅푠) and the molecular number density at established attitude 푁(푧).

훽푅푠(휆, 푧) = 휎푅푠(휆)푁(푧) (2.14)

The mathematical model (2.15) used by Bucholtz (1995) estimates the Rayleigh cross- section for the terrestrial atmosphere. The coefficients A, B, C, and D are wavelength dependent, as illustrated for two broad wavebands in Table 2.2 (notice that wavelength unit is μm).

퐷 −(퐵+퐶휆+ ) 휎(휆) = 퐴휆 휆 (2.15)

Table 2.2 Constants used in equation (2.15) for standard air at 101.325kPa and 288.15K (Bucholtz, 1995)

Coefficients 0.2μm ≤ λ ≤ 0.5μm λ ≥ 0.5μm

A [cross-section] 3.015577x10-28 4.01061x10-28

A’ [volume scattering] 7.68246x10-4 10.21675x10-4

B 3.55212 3.99668

C 1.35579 1.10298x10-3

D 0.11563 2.71393x10-2

Substituting the equation (2.15) into (2.14).

30 퐷 −(퐵+퐶휆+ ) 훽푠(휆) = 퐴휆 휆 푁푠 (2.16)

Since the product of the molecular number density 푁푠 and A is a constant value, it is possible to simplify the formula (2.16) by using a new coefficient 퐴′, which is calculated at

Ps=1013.25mbars and Ts=288.15K for standard air (see Table 2.2). The 퐵, 퐶 and 퐷 coefficients remain without changes.

퐷 −(퐵+퐶휆+ ) 훽푠(휆) = 퐴′휆 휆 (2.17)

However, it is necessary to correct Rayleigh scattering values, due to the temperature 푇 and pressure 푃 at the point of measurement, by using the equation (2.18).

푃푇푠 훽 = 훽푠 (2.18) 푃푠푇

Rayleigh scattering is the most predominant scattering in the UV spectrum, because of the strong wavelength dependence. Therefore, the Rayleigh scattering extinction coefficient is the only one specifically accounted in deriving the total amount of stratospheric ozone.

Mie scattering

Mie scattering, like Rayleigh scattering, is generated by the elastic collision between photons of incoming radiation and particles or molecules of a sample, but for particles with a diameter similar to the wavelength of the incident radiation. It does not have a strong wavelength dependence and it can be exactly estimated by the addition of an infinite series of terms instead of a unique mathematical expression. Mie theory is a collection of methods and solutions that describe the dispersion of electromagnetic waves when they interact with a homogeneous sphere. Water droplets in clouds are perfect examples of these particles, whose diameter is comparable to the wavelengths of visible light, dispersing uniformly over a range of wavelengths. For this reason, multiple scattering in clouds results in their neutral colouration, generally taking colours from grey to white without predominant wavelengths at low zenith angles. Nonetheless, analysing Mie scattering becomes much more complex if particles of random shape are considered, such as dust particles, aerosols or ice particles in high cirrus clouds.

The intensity of Mie scattering is affected by the size of the particles, producing a larger forward lobe in its scattering pattern as the particle of interest increases in size (see Figure 2.8).

Despite the contributions of Mie scattering to measuring solar irradiance within the visible spectrum, the Mie scattering extinction coefficient makes insignificant contributions when measuring direct solar irradiance in the UV spectrum during blue sky conditions. Therefore, it is not considered in Beer-Lambert equations if the measurements of interest use wavelengths below 400nm.

31

Figure 2.8 Comparison of different scattering patterns depending on the ratio between the size of the particle Φ and the wavelength of the incident radiation (Trager, 2007).

Raman scattering

Raman scattering is the inelastic scattering of light, where photons impact molecules with a diameter significantly greater than the wavelength of the incident radiation. The resulting photon differs in frequency from the incident beam due to the exchange of energy with the molecule. The wavelength of the photon after colliding with the molecule is shifted from the original, called Stokes scattering if the frequency diminishes and anti-Stokes scattering if it increases (see Figure 2.9). This phenomenon possesses the same wavelength dependence as Rayleigh scattering, but its effect when estimating stratospheric ozone through UV measurements is generally negligible. However, it is widely used in infrared spectroscopy computing tropospheric ozone.

Figure 2.9 Diagram representing the energy transition for Rayleigh and Raman scattering. Image taken from: http://bwtek.com

32 2.3. Estimation of stratospheric ozone

Estimating the total column ozone requires computing a set of variables first, such as the relative optical air mass for O3, SO2, aerosols and scattering; extinction coefficients of O3, SO2, aerosols and scattering, as well as extra-terrestrial constants at different wavelengths. All those parameters will be described below.

2.3.1. Relative optical air mass

The solar radiation that is measured using ground-based instruments has two components: the direct solar radiation that directly reaches the surface of the Earth; and the diffuse solar radiation produced mainly by the Rayleigh scattering as well as Mie scattering over clouds and aerosols. However, Dobson and Brewer spectrophotometers only evaluate sets of monochromatic measurements from the collimated direct solar beam to estimate the total column amount of ozone.

The diameter of the Sun observed from the Earth (considering the distance between each other is around 149.6 x106 kilometres) is approximately 9.3x10-3 radians or 0.53 degrees (see Figure 2.10), varying according to the day of the year (Bhattacharya et al., 2008). The field of view of the direct input optic is 3 degrees for the Brewer and 8 degrees for Dobson spectrophotometers (Staehelin et al., 2003). One reason to have a field of view bigger than the solar disk angle is to compensate for irregularities in the solar tracker and the resolution of the stepping motors that control the azimuth and zenith movements. Thus, it is particularly important to position accurately the UV detector pointing toward the centre of the Sun to obtain reliable measurements from the direct solar beam; otherwise, the spectrophotometer may obtain incorrect measurements from diffuse solar radiation.

Figure 2.10 The portion of the direct solar radiation that reaches the surface of the Earth is approximately half a degree. Image taken from: http://newport.com

In order to estimate the amount of light removed by absorption or scattering from the edge of the outer atmosphere to the ground, first it is necessary to compute the ability of the medium to

33 block electromagnetic radiation at a given wavelength (2.19), better known as the optical depth (휏).

∞ 휏휆 = ∫ 훽휆(푧)푑푧 (2.19) 푧0

Where:

τλ= Optical depth at a given wavelength. βλ(z)= Volume extinction coefficient of the attenuating species at height z. z= Distance from the centre of the Earth to the height of interest.

z0= Distance from the centre of the Earth to the observing point.

The absorption cross-section 휎휆 is a property of all substances that can be measured in the laboratory, having different values according to the wavelength of the incident radiation.

Air mass (푚) is a relative measure of the optical length of the path through the atmosphere as a function of the solar zenith angle (see Figure 2.11), approaching its maximum value as the zenith angle is close to 90°. The term AM1 is used to define the minimum value of air mass (푚=1), where the solar beam is perpendicular to the surface of the Earth (휃 = 0).

Figure 2.11 A model of the Earth showing the solar zenith angle at the equinox. As the sun moves north to south through the year, the sun is directly overhead at other latitudes. Image taken from: http://bxhorn.com/extraterrestrial-irradiance/

A basic approximation is shown in the equation (2.20) for small θ.

1 푚 = 푐표푠 휃 (2.20)

Nonetheless, the sphericity effect and atmospheric refraction due to different compounds (see Figure 2.12) cause a significant error computing the air mass value if it is considered large zenith angles ( usually 휃 >60°). Therefore, it is necessary to analyse the atmosphere as a group of thin layers with different optical depth in order to obtain an accurate air mass value.

34

Figure 2.12 Description of the relative optical air mass taking into account attitudes where stratospheric ozone is concentrated. Solar beam A and B are no refracted and refracted rays, respectively. Where θ0 is the apparent solar zenith angle, Zi is the effective height of the layer and Zi-1 is the effective height of the adjacent lower layer (Thomason et al., 1983).

The formula (2.21) described by Thomason et al. (1983) estimates the relative optical air mass at a specific wavelength in the zenith direction. This mathematical expression applies Snell’s law.as a sum of several thin layers of different compounds with different diffraction index.

∞ 1 휎휆(푧)푑푧 푚휆(휃) = [ ] ∫ 1 휏 2 ⁄2 휆 푧0 푛 푧 푠𝑖푛(휃 ) (2.21) [1 − 0 ( 0 0 ) ] 푛(푧) 푧 Where:

mλ(θ)=Relative optical air mass as a function of θ at a specific wavelength. τλ= Optical depth at a given wavelength. θ0=Apparent solar zenith angle. n(z)=Refractive index of the atmosphere at altitude z.

n0=Refractive index at the Earth’s surface. Air mass calculations are sensitive to vertical uniformity and temperature. For this reason, Thomason et al. (1983) suggests using separate air mass values for those particles that do not have a constant vertical distribution, such as ozone, nitrogen dioxide and water vapour (see Figure 2.13).

35

Figure 2.13 Graph that compares relative optical air mass functions for air (ma), water vapour (mw), ozone (mo) and nitrogen dioxide (mn) with apparent zenith angles from 68° to 87° (Tomasi et al., 1998).

Relative optical air mass value used in Dobson and Brewer spectrophotometers

Bernhard et al. (2005) simplify the equation (2.21) by considering insignificant the wavelength dependence of 푚휆(휃) in the spectral range where Dobson and Brewer spectrophotometers operate. In order to use this approximation, the attenuation layer must also be considered a delta function at height h above the sea.

푛 (푅 + ℎ) 푚(휃) = ℎ 2 2 2 2 2 1⁄ (2.22) [푛ℎ (푅 + ℎ) − 푛0 (푅 + 푟) ∙ 푠𝑖푛 (휃0)] 2

Where:

푛ℎ = The refractive index at height h. R= Radius of the Earth (6370km). r= Altitude of the spectrophotometer. h= Height above the sea level where radiation extinction occurs. The equation (2.23) is a practical approximation used to estimate the relative optical air mass at the stratospheric altitudes. It is achieved by considering the refractive index of the air at zero altitude (푛0) equals to the refractive index (푛ℎ), 푛0 = 푛ℎ = 1 (Thompson et al., 1997)

1 μ(휃) = (푅 + 푟)2 √1 − 푠𝑖푛2 휃 (2.23) (푅 + ℎ)2

Where μ is the relative optical air mass related to ozone absorption.

36 2.3.2. Beer - Lambert Law

The principle of the Beer-Lambert Law was published in the book Photometria by Johann Heinrich Lambert in 1760. It stated that the absorbance of light in the materials is proportionally related to the path length. But it was not until 1852 when August Beer completed the law by indicating that the concentration of molecules and the path length of the sample are correlated to the absorbance of the substance.

After a beam of light interacts with a sample, the photons can be transmitted in the same direction of the incident beam, scattered into different directions or absorbed by the molecules of the sample. The concept of transmittance (푇) links these terms in a single equation, which describes the ratio between the transmitted light (퐼) once it has passed through the sample, and the original intensity of light (퐼0).

퐼 푇 = (2.24) 퐼0

The equation (2.25) represents the logarithmic relationship between the absorbance (퐴) and the transmittance.

퐼 퐴 = 퐼푛 ( 0) 퐼 (2.25)

In order to measure the amount of atmospheric trace gases, firstly, it is necessary to determine in the laboratory the photon absorption profile of the compound as a function of wavelength (cm2 per molecule), better known as absorption cross-section. It is a unique chemical signature of each element that indicates the effective area of the particle blocking the electromagnetic radiation.

Figure 2.14 represents the portion of the sample to be analysed by the Beer-Lambert law:

Figure 2.14 Portion of the sample analysed by the Beer-Lambert law.

In order to apply the Beer-Lambert equations, it is necessary that all the molecules must be homogeneously distributed in the medium, as well as choosing a small dx, so that none of the absorber elements should shadow another. Also, the beam of the incident radiation must be collimated, allowing each photon to pass through the same length in the medium, which can be solid, liquid or gas. The following analysis is shown for gas samples.

37 The next equations estimate the proportion of light that is absorbed or scattered by particles. Where the term 푁푆푑푥 is the number of molecules in the path length of the light, 휎푁푆푑푥 represents the effective area where light interacts with the molecules, and 푑퐼푥.denotes the fraction of light not 퐼푥 transmitted.

푑퐼 휎푁푆푑푥 − 푥 = 퐼 푆 (2.26)

Where:

푑퐼푥 = Change in intensity across x (the variation is always negative). 퐼 = Intensity of the incident light. 휎 = Absorption cross-section or extinction coefficient. 푁 = Number density of absorbers in 1cm3. 푆 = Total cross-section area of the sample to be analysed. 푑푥 = Thickness. 퐿 = Path length. Integrating both sides:

퐼 푑퐼 퐿 ∫ 푥 = − ∫ 휎푁푑푥 퐼 (2.27) 퐼0 0

퐼푛(퐼) − 퐼푛(퐼0) = − 휎푁퐿 (2.28)

퐼 퐼푛 ( ) = − 휎푁퐿 (2.29) 퐼0

Solving for 퐼

−휎푁퐿 퐼 = 퐼0푒 (2.30)

The total extinction coefficient 훽(휆), also known as total extinction cross-section, measures the capacity of a sample to block the light at a particular wavelength:

훽(휆) = 휎(휆)푁 (2.31)

By replacing the equation of the optical depth 휏 (2.19) into (2.30), the resulted expression

−휏 퐼 = 퐼0푒 (2.32)

Solving for 휏:

퐼 휏 = 퐼푛 ( 0) 퐼 (2.33)

훽(휆) can be represented by the sum of absorption coefficient 훽푠(휆) and the scattering coefficient 훽푎(휆):

훽(휆) = 훽푎(휆) + 훽푠(휆) (2.34)

38 Nonetheless, if the sample is defined as a mixture of components with different absorption cross-section, it is possible to separately evaluate each component, and then calculate 훽푎(휆) as the sum of all the absorption and scattering coefficients.

훽푎(휆) = ∑ 휎𝑖푁𝑖 (2.35) 𝑖=1

Where 푀 is the number of elements considered in the sample.

By substituting the equation (2.34) and (2.35) in (2.30), the Beer-Lambert law is obtained as a function of wavelength:

−퐿 ∑푖 휎푖(휆)푁푖−퐿훽푠(휆) 퐼(휆) = 퐼0(휆)푒 (2.36)

Where 퐼(휆) is the direct normal irradiance over the sensor of the instrument at a given wavelength.

Beer-Lambert law applied to estimate the total column of stratospheric ozone

In order to get reliable values of stratospheric ozone, it is necessary to perform direct solar measurements during blue sky conditions, or at least when the solar disc and surrounding are free of cloud, fog, snow and rain. Otherwise, ozone must be estimated using diffuse radiation data from zenith sky measurements, inducing new variables in the analysis such as a layer of clouds (although in reality this is often non-homogenous) and the single-scattering albedo. Figure 2.15 displays the solar irradiance measured by Brewer 14 at 324nm during three different atmospheric conditions: clear sky on May 7, variable clouds on May 6 and heavy thunder clouds on May 2 at Toronto (McArthur et al., 1999).

Figure 2.15 Solar irradiance at 324nm measured during clear sky on May 7, during variable clouds on May 6 and during heavy thunder clouds on May 2. Crosses represent measurements made by Brewer #014, while the solid line denotes an estimated model (McArthur et al., 1999).

By measuring UV radiation of wavelengths strongly and weakly absorbed by ozone in the atmosphere, passive ground-based spectrophotometers estimate the total column amount of

39 stratospheric ozone. The equation (2.37) is obtained by applying the concepts of extinction cross- section, relative optical air mass and scattering to the Beer-Lambert law.

퐼푛[퐼(휆)] = 퐼푛[퐼 (휆)] − ∑ 푚 (휆, 휃)휎(휆) 푁 − 푚 훽 0 𝑖 𝑖 𝑖 푠 푠 (2.37) 𝑖

Where 푚푠 is the relative optical air mass corresponding to the scattering and 훽푠 is the scattering extinction coefficient.

The natural logarithm is used in the mathematical analysis of this thesis, nonetheless, there are still some spectrophotometers using log10 because of their historic record of data (Bernhard et al., 2005).

The equation (2.38) calculates the spectral irradiance 퐼(휆) of the incident solar beam in the UV spectrum at the surface of the Earth. This mathematical expression only considers elements that play an important role computing the transmittance such as ozone, SO2, aerosols and Rayleigh scattering.

푝푠 푂3 ′ 푆푂2 퐼푛[퐼(휆)] = 퐼푛[퐼0(휆)] − 휇푋 훼(휆) − 휇 푋 훾(휆) − 푚푅 ( ) 훽(휆) − 푚푎훿(휆) (2.38) 푝0

Where:

I0(λ)=Total irradiance measured outside the atmosphere at a given wavelength, better known as extra-terrestrial constant (ETC). α(λ)=Ozone extinction coefficient.

μ=Relative optical air mass corresponding to O3 absorption. XO3=Total column ozone.

μ'=Relative optical air mass corresponding to SO2 absorption. γ(λ)=SO2 extinction coefficient. 푆푂2 푋 = 푇표푡푎푙 푐표푙푢푚푛 표푓 푠푢푙푓푢푟 푑𝑖표푥𝑖푑푒푒 (푆푂2). mR=Relative optical air mass corresponding to Rayleigh scattering. β(λ)=Rayleigh scattering extinction coefficient at 1013.25 hPa.

ps=Observing point pressure. p0=Sea level pressure (1013.25 hPa). δ(λ)=Aerosol extinction coefficient at wavelength λ.

ma=Relative optical air mass corresponding to aerosols. The effective attitude where the UV extinction due to ozone and sulphur dioxide occurs is at approximately 22km, the value used in Dobson and Brewer retrievals of ozone, although the Dobson has a latitude dependence (Savastiouk and McElroy, 2005). Then, the known altitude h is substituted into (2.23) to estimate the air mass factor at stratospheric altitudes (μ and μ'). On the other hand, a significant number of gasses, particles, aerosols and dust that strongly contribute Rayleigh scattering and aerosol extinction are present in the troposphere. The centre of mass of the atmosphere is considered at a height around h=5km, which is commonly used to estimate the air mass factors mR and m푎, though other heights such as 6km have been suggested as a better estimate.

The Rayleigh extinction coefficient is obtained by substituting the values of wavelength, temperature and pressure at the spectrometer’s location in equation (2.18). However, the aerosol

40 extinction coefficient δ(λ) can be approximated by using the Angstrom exponent expressed in the equation (2.39) (Ångström, 1964), where C represents the optical depth of aerosols in the atmosphere.

−1 δ(λ) = 퐶휆 (2.39)

Substituting (2.39) into (2.38) leaves three unknown variables: 푋푂3, 푋푆푂2 and 퐶; and only one equation (2.40). In order to compute the total column amount of ozone in the atmosphere, this difficulty is solved by measuring at different wavelengths and applying the DOAS technique (explained in section 2.3.3),

푝푠 푂3 ′ 푆푂2 −1 퐼푛[퐼(휆)] = 퐼푛[퐼0(휆)] − 휇푋 훼(휆) − 휇 푋 훾(휆) − 푚푅 ( ) 훽(휆) − 퐶휆 (2.40) 푝0

Langley extrapolation

One challenge of applying the Beer-Lambert Law in the atmosphere is obtaining the real value of the initial extraterrestrial intensity of the solar beam 퐼0, since the vast majority of all measurements are made beneath the atmosphere. Also, the extraterrestrial irradiance is not constant, it is a function of latitude, date, time and it can be affected by other solar variations such as the 11-year solar cycle and solar rotation (Paulescu, 2013). But it can be estimated by using Langley extrapolation ideally at high altitudes to avoid the influence of aerosols. Brewer spectrophotometers located at Mauna Loa Observatory and at Izaña Atmospheric Research Centre are used to obtain the ETC values used to estimate the total column amount of ozone (note this is not the extraterrestrial radiation in absolute units, but is related to it). These instruments are considered world references, and they are used to transfer calibrations and ETC to other spectrophotometers.

By getting direct measurements in clear-sky conditions, where the amount of ozone, SO2 and aerosols are approximately constant throughout half a day, the relation between the measured irradiance 퐼(휆) versus optical air mass is nearly a linear function. Langley extrapolation plots this relation estimating the intersection at 0 air mass factor, in order to estimate the value that would be measured with no ozone in the atmosphere, better known as extra-terrestrial constant.

Figure 2.16 Example of Langley extrapolation of irradiance measurements at 320nm on the 217th Julian Day of 2000, using a Neutral-density filter 3. The irradiance is

41 considered as photon counts of the Brewer’s photomultiplier tube (Marenco et al., 2002).

2.3.3. Differential optical absorption spectroscopy (DOAS) technique

Spectroscopy enables the study of the interaction between electromagnetic spectra and molecules or atoms. It is able to identify and quantify different chemical elements by measuring the electromagnetic spectrum that is emitted, absorbed or scattered by the molecules of solids, gases or solutions. The high sensitivity of optical spectroscopy has made it a suitable technique to remotely measure atmospheric compounds in the broad solar spectrum.

Applying spectroscopy techniques to analyse the solar irradiance that reaches the Earth’s surface, Cornu (1881) suggested that the attenuation of solar radiation in the UV spectrum is produced by the atmosphere. Hautefeuille and Chappuis (1880) revealed the absorption in the visible light spectrum of liquid ozone, and Hartley (1881) discovered that the atmospheric absorption of UV radiation was performed by ozone molecules. Some years later, Sir William Huggins made the first measurement of atmospheric absorption from 319nm to 333nm during astronomic observations of the star Sirius (Huggins, 1890). Those events became into the starting point of ozone absorption analysis. The major ozone absorption bands are thus named the Hartley, Huggins and Chappuis bands (see Figure 2.17).

Figure 2.17 Absorption cross-section of ozone at 293K from 231 to 794nm measured with the satellite GOME FM spectrometer in the visible and UV spectral ranges (Burrows et al., 1999a).

Nonetheless, the range of wavelengths from around 310nm to 340nm, called Huggins band, is widely used by Dobson and Brewer spectrophotometers to estimate the total column amount

42 of ozone (see Figure 2.18), while other wavelengths are used for other gases such as NO2

(Brewer et al., 1973).

Figure 2.18 Absorption coefficient of ozone within the Huggins bands from 3000Å to 3500 Å. Where Å is a unit equal to 1x10-10m or 0.1nm (Inn and Tanaka, 1953).

Differential optical absorption spectroscopy is used in conjunction with the Beer-Lambert Law to estimate the total column density of trace gases in the atmosphere from solar radiation measurements (trace gasses are compounds that occupy less than 1% of the atmosphere’s volume). By relying on the unique absorption spectra of the gases and the relative irradiance measured by the spectrophotometer, DOAS is able to isolate the trace gas of interest from many of the compounds in the atmosphere to give an accurate analysis. Due to its efficiency, DOAS is broadly used by ground-based spectrophotometers, satellite spectrometers, ozone-sondes and spectrometers mounted on aircrafts to measure atmospheric compounds.

In order to use the DOAS technique, the absorption values of the measured trace gas needs to vary rapidly with wavelength in comparison with the low or smooth absorption values of aerosols and others atmospheric elements. Total 휎(휆) is expressed as a sum of its two ′ components: one possesses a smooth slope 휎푏(휆) and the other one 휎 (휆) shows rapid variations with wavelength compared with 휎푏(휆) (see Figure 2.19). The equation (2.41) represents the mathematical relationship between both components of the absorption cross-section.

휎(휆) = 휎푏(휆) + 휎′(휆) (2.41)

λ[nm]

Figure 2.19 Illustrative graph that shows the two components of the absorption cross- section (휎). Panel A displays the smooth slope 휎푏(휆), while Panel B shows the faster

43 absorption variations of the trace gas 휎′(휆) using 휎푏(휆), as a reference. Image taken from:https://www.wmo.int

Applying DOAS to the Beer- Lambert law equation by substituting (2.41) into (2.36), gives the following:

′ −퐿 ∑푖 휎 (휆)푁푖 −퐿 ∑푖 휎푏 (휆)푁푖−퐿훽푠 퐼(휆) = 퐼0(휆) ∙ 푒 푖 ∙ 푒 푖 (2.42)

Where 훽푠 represents the scattering extinction coefficient.

“The first exponential function describes the effect of the structured ‘differential’ absorption of a trace species, while the second exponential constitutes the slowly varying absorptions as well as the influence of Rayleigh and Mie scattering” (Platt, 2008).

If the analysis is focused on one single trace gas, the equation (2.42) can be simplified:

′ −퐿휎 (휆)푋 −퐿 ∑푖 휎푏 (휆)푁푖−퐿훽푠 퐼(휆) = 퐼0(휆) ∙ 푒 ∙ 푒 푖 (2.43)

Where 푥 represents the total column amount of the trace gas of interest in the atmosphere.

′ Consequently the value 퐼0 is defined in (2.44) as the irradiance without differential absorption. This low-frequency component, which includes Rayleigh and Mie scattering, can be approximated by low order polynomials or Fourier series. The most common fitting baseline method is the Levenberg-Marquardt method, which is a recursive non-linear least square technique that guarantees stability since it always converges (Platt, 2008). The differential absorption cross- section 휎′(휆) is acquired from absorption coefficients obtained in the laboratory.

′ −퐿 ∑푖 휎푏 (휆)푁푖−퐿훽푠 퐼0(휆) = 퐼0(휆) ∙ 푒 푖 (2.44)

Substituting (2.44) into (2.42)

′ −퐿휎′(휆)푋 퐼(휆) = 퐼0(휆) ∙ 푒 (2.45)

The differential optical density 퐷′ (see Figure 2.20) is defined as the logarithm of the ratio ′ between 퐼0(휆) and 퐼0(휆) from eq. (2.45)

퐼′ (휆) 퐷′ = 퐼푛 [ 0 ] = 퐿휎′(휆)푋 퐼(휆) (2.46)

44

Figure 2.20 Principle of DOAS where 휎′ represents the strongest wavelength ′ dependent term and 퐼0 and 휎푏 is the weaker one (Platt, 2008).

Therefore, the number of particles can be obtained solving for 푋 in equation (2.46).

퐷′ 푋 = 퐿휎′(휆) (2.47)

2.3.4. DOAS technique applied to Brewer spectrophotometer calculations

Note that the next mathematical process is obtained from Savastiouk and McElroy (2005).

푂3 ′ 푆푂2 −1 퐼푛[퐼(휆)] = 퐼푛[퐼0(휆)] − 휇푋 훼(휆) − 휇 푋 훾(휆) − 푚푅훽(휆) − 퐶휆 (2.48)

All the known values of the equation (2.40) are added together into the constant S

푝푠 푆 = 퐼푛[퐼0(휆)] − 퐼푛[퐼(휆)] − 푚푅 ( ) 훽(휆) (2.49) 푝0

Substituting (2.49) into (2.40) and measuring at five different wavelengths in the Huggins band:

−1 훼1 훾1 휆 푆1 1 훼 훾 −1 2 2 휆2 푆2

푂3 훼 ′ 푆푂2 훾 −1 휇푋 3 + 휇 푋 3 + 푚푎퐶 휆3 = 푆3 (2.50) −1 훼4 훾4 휆 푆4 4 [훼 ] [훾 ] −1 5 5 [휆5 ] [푆5]

푂3 푂3 푂3 푂3 푂3 The equation (2.50) is then multiplied by a row vector [푘1 푘2 푘3 푘4 푘5 ]

5 5 5 5

푂3 푂3 ′ 푆푂2 푂3 푂3 −1 푂3 휇푋 ∑ 푘𝑖 훼𝑖 + 휇 푋 ∑ 푘𝑖 훾𝑖 + 푚푎퐶 ∑ 푘𝑖 휆𝑖 = ∑ 푘𝑖 푆𝑖 (2.51) 𝑖=1 𝑖=1 𝑖=1 𝑖=1

Where the coefficients 푘𝑖 are known as the weighting coefficients obtained using a fitting technique. They help in the linear fitting process of the DOAS, as long as the next conditions are valid:

45 5

∑ 푘푂3훾 = 0 𝑖 𝑖 𝑖=1 5 푂3 −1 ∑ 푘 휆𝑖 = 0 (2.52) 𝑖 𝑖=1 5

푂3 ∑ 푘𝑖 = 0 { 𝑖=1

Substituting (2.52) into (2.51) and solving for 푋푂3

5 푂3 ∑ 푘 푆𝑖 푂3 𝑖=1 𝑖 푋 = 5 푂3 (2.53) 휇 ∑𝑖=1 푘𝑖 훼𝑖

Brewer spectrophotometers make measurements at 306.3nm, 310.1nm, 313.5nm, 316.8nm and 320.1nm; but at 306.3nm the irradiance is highly absorbed by ozone and strongly affected by 푂3 SO2. Thus, the constant 푘1 is set to zero in order to avoid SO2 influence. The remaining weighting 푂3 푂3 푂3 푂3 coefficients are commonly fixed to 푘2 = 1, 푘3 = −0.5, 푘4 = −2.2 and 푘5 =1.7, nonetheless, they can be customised ’to optimize measurements for particular spectrophotometers (Savastiouk and McElroy, 2005).

The low-frequency component of the ozone absorption cross-section is obtained from tables of measurements, which are made at a laboratory under specific conditions of temperature, pressure and optical depth. Examples of tables with ozone absorption cross-section values are given by Molina and Molina (1986), Bass and Paur (1985), and other datasets based on Bass and Paur such as Komhyr et al. (1993). Also, there are high-resolution ozone cross-section values that complement the values mentioned above such as those proposed by Daumont et al. (1992), Brion et al. (1993) and Malicet et al. (1995).

Data coming from the Middle Atmosphere Nitrogen Trend Assessment (MANTRA) campaign in 1998, were analysed using ozone cross-section values from the Global Ozone Monitoring Experiment (GOME) (Burrows et al., 1999a) by Savastiouk and McElroy (2005). However, newly released sets of ozone cross-section values are compared with the old ones in order to evaluate their performance, for example, the coefficients from Serdyuchenko et al. (2014) were compared with the existing ones by Redondas et al. (2014), in order to determine the advantages and disadvantages of using it in Brewer calculations. Thus far, no changes have been made to the standard coefficients in use across the Brewer community, as it is important that consistent coefficients are used by all, and changing them is a non-trivial undertaking.

Once 푋푂3 is computed, it is used in the same procedure to estimate 푋푆푂2 from (2.51), but 푆푂2 considering that 푘1 needs a value different from zero. Finally, it is solved for the amount of aerosols (퐶) in equation (2.40).

The key to estimating the total column ozone is that the absorption coefficient of trace gasses in the atmosphere possesses a unique pattern of the solar spectrum. However, spectrophotometers are usually faced with measuring irradiance in a narrow range of

46 wavelengths where the absorbance of the trace gas of interest is overlapped by the absorbance of another gas. This problem can be solved by increasing the wavelength range of measurement ′ in order to obtain a larger profile of I0, but it reduces the resolution and sensitivity of the instrument. Nevertheless, further technology will allow increasing the number of trace gasses to be analysed by a single instrument, without decreasing the accuracy.

2.4. Spectrophotometer

By measuring the spectral character of an electromagnetic radiation source, often after passing through an absorbing sample, spectrophotometers perform an essential role in spectroscopy. They are widely used in different scientific fields, such as chemistry, biology, microbiology, , materials science, atmospheric sciences and geology. However, depending on the technology used to design the spectrophotometer, their range of wavelengths, time of exposure, resolution, level of stray light rejection and precision may vary from one spectrophotometer to another. This section focuses on the design and construction of the Brewer spectrometer, as well as a brief review of other instruments used to measure stratospheric ozone.

The operation of Dobson and Brewer spectrophotometer is divided into two stages. First, the monochromator disperses the incident beam, using a prism in the Dobson and a diffraction grating in the Brewer, in order to select the desired wavelength. And secondly, the photometer counts the number of photons, whose value then is converted into a more measurable variable such as current or voltage. However, optical aberrations, internal reflections, imperfections in the optics and the finite width of the slits will produce undesired stray light that has to be considered in interpreting the measurements.

This chapter will mainly focus on the monochromator, the diffraction grating and photomultiplier tube (PMT), which are the most important elements in the Brewer spectrophotometer.

2.4.1. Monochromator

The monochromator is considered the core of the spectrophotometer because it is here where the selection of the desired wavelength takes place by collimating, dispersing and focusing the incident beam. Almost all monochromators share the same components such as entrance and exit slits; a dispersion element (prism or diffraction grating) and further optics according to the geometry of the design. However, their design is often customised at a specific range of wavelengths for use in different applications; for example, the Brewer monochromator works in the UV spectrum.

The first design of a monochromator using one concave spherical mirror and a plane diffraction grating was developed by H. Ebert (Ebert, 1889). However, after being severely criticized by Kayser (1910), no further research was developed in Ebert’s design during the following years

47 (Fastie, 1991). Some years later, Czerny and Turner (1930) developed the Czerny-Turner monochromator, whose geometry, plane grating and two symmetrical off-axis spherical mirrors minimize the coma aberration in its measurements (see Figure 2.21).

Figure 2.21 Czerny-Turner monochromator. The first concave mirror (M1) collimates the incident beam that passes throught the input-slit, then it is dispersed by the diffraction grating, then it is focused using the second concave mirror (M2), and finally the selected wavelength passes through the exit slit. Image taken from: www.rp- photonics.com/spectrometers.html

It was not until 1952 that William G. Fastie, seeking to develop a small, efficient and simple monochromator capable of measuring in the UV redesigned Ebert’s monochromator. His significant contribution was curving the slits in order to correct optical aberrations, mostly astigmatism (Fastie, 1952). The new design was called the of Ebert-Fastie monochromator, which uses a large spherical mirror and one plane diffraction grating (see Figure 2.22). After the light enters through the entrance slit, the first half of the mirror collimates and directs the beam of light to the diffraction grating, while the second half of the mirror focuses the dispersed light from the grating at the desired wavelength onto the exit slit, whose width is proportional to the narrow spectrum that the monochromator will let pass through. The position of the grating and the location of the exit slit are strongly related to the wavelength exiting the monochromator (Palmer and Loewen, 2005).

Frontal view

Figure 2.22 Ebert-Fastie monochromator design (Mielenz, 1964).

The advantage of using the Ebert-Fastie monochromator over the Czerny-Turner monochromator is the simplicity in its design, as it is only necessary to align one single mirror. Therefore, the Ebert-Fastie monochromator was widely accepted for use in ground and satellite spectrophotometers (Fastie, 1991). The monochromator used in Brewer spectrophotometers is based on the Ebert-Fastie design, which uses a concave mirror and straight slits (see Figure 2.29), correcting critical aberrations such as coma, spherical aberration and astigmatism by using optical lenses between the entrance slit and the concave mirror. The diffraction grating and the

48 slits were designed to take measurements at a specific wavelength; either fixing the diffraction grating while the slit mask rotates or rotating the diffraction grating maintaining a fixed exit slit.

Correction lens

Figure 2.23 Left diagram displays the Brewer monochromator design and the right picture shows exposed parts of the Brewer’s monochromator (taken by the INTA, Spain during intercomparison 2015).

Double monochromators

Undesired reflections from optical elements and the walls of the monochromator, incorrect alignments of the optics, optical aberrations, wavelengths of different orders, light scattered by the lens, grating and mirrors; as well as Rayleigh and Mie scattering of the air inside the monochromator, are the sources of stray light reaching the photodetector. Moreover, the transfer function of a monochromator usually is based on a “sampling function”, which adds adjacent wavelengths to the desired one, even without stray light. Recent spectrophotometers, as the proposed in this thesis, use new techniques and technologies in hardware and software to minimize errors in their measurements.

One method to increase the stray light rejection is connecting two monochromators in series, reducing stray light by “approximately the product of ratios of stray light intensity” of each single monochromator (Palmer and Loewen, 2005). Therefore, the double monochromator system is how the standard used to produce Brewer spectrophotometers (known as the MkIII Brewer), which produces more accurate measurements than its predecessors with a single monochromator (MkII and MkIV versions).

Optical aberrations

Spherical aberration is an intrinsic defect of almost all spherical lenses which focuses incoming rays from a collimated source at different points along the optical axis after passing through spherical lens (see Figure 2.24). However, spherical aberration can be minimized either by adding correcting lenses or bending the lens.

49

Figure 2.24 Representation of spherical aberration with six incident parallel rays.

Coma aberration occurs when the incident wave-front is tilted or not centred with the optical axis (see Figure 2.25).

Figure 2.25 Representation of coma aberration.

Correcting this aberration is critical to obtaining accurate measurements. Figure 2.26 shows the effect of coma aberration (hatched area) over the slit function, after being processed by a photometer.

Figure 2.26 Effect of coma aberration in the photon count process.

Astigmatism aberration is produced by the difference in the curvature of the lens at different planes, such that the transversal and longitudinal incident rays present two focal points (see Figure 2.27).

50

Figure 2.27 Representation of astigmatism aberration.

2.4.2. Plane diffraction grating

The diffraction grating is an optical device capable of dispersing the incident electromagnetic spectrum in a defined range of wavelengths. They first became a substitute for prisms as a dispersing element in spectrometers, due to their better resolution and low-cost production by replication. While a prism must be fabricated with special materials to disperse UV and IR wavelengths, changes in the coating and number of grooves of the diffraction grating will determine its characteristics of dispersion.

Monochromators can use plane or concave diffraction gratings according to the geometry of their design. The use of concave gratings reduces the number of reflecting elements in the design of the monochromator (see Figure 2.28), which is commonly used in instruments measuring below 200nm (Murty, 1974). This section will focus on plane diffractions gratings, which are used in Brewer spectrophotometers.

Figure 2.28 Example of a monochromator using a concave mirror. S1 is the entrance slit, S2 is the exit slit, G represents the grating, M is a reflective mirror and C is the centre of the Rowland circle (Murty, 1974).

A master grating can be generated by ruled or holographic processes, using glass or silica substrates. However, the benefits of fabricating a replica grating from an original saves time and money, because one expensive master grating is able to produce hundreds of replicas before becoming useless. Moreover, a sub-master or first replica generation can produce a second generation of replicas with little degradation in performance.

51 The ruled master grating is ruled mechanically with a diamond tool, producing parallel grooves in the reflective metal coating (usually aluminium or gold) of the grating substrate. By changing the shape of the diamond tool, the groove profile can be blazed or trapezoidal. This was the first production method invented and can take up to six weeks to fabricate a master diffraction grating, depending on the density of grooves per millimetre (Palmer and Loewen, 2005).

The replication process is the same for ruled or holographic gratings. First, a thin film with a pattern of grooves identical to the master grating or ‘original’ is created. Then it is strongly cemented to the substrate material, and finally, the optical surface is coated with a reflective material. The resultant replica is almost as efficient as the master grating and their differences in quality are insignificant.

On the other hand, the holographic master grating is generated by a process called photolithographic technique. First, the substrate is coated by a photo-resistant material. Then an equally spaced fringe pattern is recorded by the interaction of two laser beams, where the exposure light strengthens or weakens the intermolecular bonds of the material. Finally, the sinusoidal grooves are produced by immersing it in a chemical developer. This method takes less time to produce a master grating compared with ruled technique, because the grooves are generated simultaneously. Even if this process only produces sinusoidal grooves, the groove profile can become blazed by performing a process called ion bombardment (ion etching). Figure 2.29 shows the visual difference between a sinusoidal and a blazed profile in diffraction gratings.

Figure 2.29 Differences between ruled and holographic diffraction grating. Image taken from: http://ssioptics.com

While ruled gratings contain periodic errors and spacing errors produced by mechanical irregularities, holographic diffraction gratings are free of these imperfections by generating all the grooves simultaneously, reducing the false spectral lines that increases stray light (Palmer and Loewen, 2005).

The advantage of using blazed gratings over holographic ones is a higher efficiency over a defined spectral section, usually in the UV spectrum (Sharpe, 1984), because the relationship between the angle of the grooves and the incidence angle decides the wavelength where the diffraction grating has a maximum efficiency. Nonetheless, the ‘blaze wavelength’ becomes an important disadvantage when the grating rotates to focus a specific wavelength at the exit slit, because the efficiency will not remain constant for all the wavelengths. Thus, due to its wide

52 spectral range and for better performance against stray light, most Brewer spectrophotometers use plane holographic diffraction gratings with sinusoidal grooves.

Grating equation

The dispersion of the incident beam by the diffraction grating depends on the distance between each groove (d) and the angle from the normal of the plane diffraction grating to the incident beam 훼, see equation (2.54).

푚휆 = 푑(sin 훼 ± sin 훽) (2.54)

Where m is the order of the diffracted ray, λ is the diffracted wavelength, 훽is the angle between from the normal to the diffracted ray, see Figure 2.30.

Figure 2.30 Graphical description of grating equation (2.54). Image taken from www.intl-lighttech.com

The diffraction grating is designed in conjunction with the focusing mirror to direct a very narrow waveband (ideally monochromatic) onto the exit slit plane. However, stray light from other wavelengths may also pass through the exit slit since the equation (2.57) can be satisfied by more than one diffraction angle 휃푑, one for each value of m as long as |푚휆| < 2푑 (Palmer and Loewen, 2005). The diffraction order can have positive or negative values depending on the direction of the diffracted ray, but if 훼 = ˗훽 (zero order), the diffraction grating reflects the incident beam as if it were a mirror (see Figure 2.31). Note that the intensity of the diffracted ray decreases if the magnitude of the order increases

53

Figure 2.31 A portion of the intensity of the incident ray will be difracted at different diffraction angles, according to m

By operating the grating in the first order, the spectral overlapping of low-order wavelengths is avoided (see Figure 2.32).

Figure 2.32 Overlapping of spectral orders. Values in parentheses indicates wavelengths absorbed by the atmosphere. Image taken from: http://www.horiba.com

2.4.3. Prism as a dispersion element

The prism was the first dispersing element used in monochromators to refract the direction of the incident beam according to its wavelength (see Figure 2.33). They are made of transparent materials such as glass, which commonly absorb UV wavelengths below 350nm; quartz, plastic or a specific material according to the range of wavelengths where the prism will be used. The linear dispersion, as well as the absence of overlapping orders, are still important advantages of prisms over diffraction gratings. Spectrophotometers such as the Dobson, designed when gratings had low line densities, use a prism as a dispersing element. However, after the production of high-resolution diffraction gratings (with more grooves per millimetre than the early ones), further designs of spectrophotometers have replaced the prism because of its limited resolution, bulky shape and expensive production as well as larger straylight contributions.

54

Figure 2.33 Designs of a monochromator using a prism as a diffraction element, where L0, L1 and L2 are lens to focus and collimate the incident beam.

2.4.4. Photomultiplier tube (PMT)

The photomultiplier tubes (PMT) are used in spectrometry as a sensitive photodetector, which produces a current signal proportional to the intensity of the incident beam. Its high accuracy transforming photons into a proportional number of electrons is the main reason why specialized spectrophotometers such as the Dobson and the Brewer have used the PMT as a light detector.

The PMT (see Figure 2.34) is based on a detection stage (photocathode), focusing and accelerating electrodes, an electron multiplier stage, which contains a series of secondary electron emissions better known as dynodes; and the electron collection stage (anode). First, the energy transference between the incident photons and the electrons of photocathode material produces the emission of photoelectrons due to the photoelectric effect. Then, those photoelectrons are focused and accelerated into the first electron multiplier (dynode), and the secondary emitted electrons are repeatedly focused and accelerated into the next dynode up to the last one. Finally, the anode collects the total number of electrons from the last dynode obtaining a proportional output current.

Figure 2.34 General schematic of a photomultiplier tube using a linear focusing configuration of dynodes (Hamamatsu Photonics, 2006).

In order to estimate the gain of the PMT, first, it is necessary to estimate parameters such as the quantum efficiency and the relative collection efficiency. The quantum efficiency, usually expressed in percentage terms, is the ratio between the incident photons on the cathode material

55 and the average photoelectric yield. Major energy losses during the photoemission process are because of internal or external reflections in the cathode, which are affected by the wavelength of the light, the cathode’s thickness and the compound used in the construction of the cathode, which has a characteristic spectral response. The ratio used to compare the number of electrons emitted by the cathode and the electrons reaching the first dynode is called relative collection efficiency. Notice that both efficiencies describe a non-uniform sensitivity over the entire anode or cathode surface (Photonis, 2002).

Each time an electron hits the surface of a dynode, a certain number of secondary electrons are emitted, whose ratio is controlled by the high voltage on each dynode by the voltage divider network. Thus, the number of dynodes in the PMT and the high voltage used are directly related to the gain of the PMT, If the collection efficiency of cathode, dynodes and anode are close to 100%, the equation estimate the gain of the PMT, which is the ratio between the anode current

(퐼푎) due to a cathode photocurrent (퐼푘) for a N number of dynodes.

푁 퐼푎 퐺 = = 휂(푔1)(푔2) … (푔푁) = 휂 ∏ 푔𝑖 (2.55) 퐼푘 𝑖=1

Where 휂 is the collection efficiency of the first dynode and 푔𝑖 is the gain of the ith multiplier stage.

Expressing the gain in terms of interdynode voltage per stage (푉𝑖), the equation (2.55) can be written as:

푁 훼 퐺 = ∏ 푘𝑖푉𝑖 (2.56) 𝑖=1

Where 푘𝑖 is a proportionality constant and 훼 is a value between 0.6 and 0.8.

Dark current

However, when voltage is supplied to the PMT, an undesired current is measured by the anode in total darkness, better known as dark current when the output is a DC signal or dark count when the output estimate the number of photons reaching the cathode of the PMT (ET Enterprises Limited, 2011). The unwanted current has a continuous component plus a pulsed component. The DC component is generated almost entirely by leakage currents due to imperfections in the PMT base, such as the surface conductivity of the PMT leads and socket; while the pulsed component usually is produced by thermal electrons mostly emitted by the cathode and amplified by the dynodes (thermionic emission of electrons), which is temperature dependent; electron emission because of field effect, which is voltage dependent and external radiation, such vibration and inductive noise. However, dark measurements can be reduced by operating the PMT at the correct voltage, using an efficient cooling system and applying proper isolation, in order to obtain a more accurate radiation measurement.

56 An advantage of using the PMT in the counting mode over the analogue mode, is an improvement in the signal-to-noise ratio due to reductions of variations in the signal. For this reason it is commonly used in spectrophotometry, even if the processing data is more complex. The algorithm used by Brewer spectrophotometers to estimate the photon counting is the Poisson statistic distribution. However, it is not the only method used by all spectrometers.

There are not two PMTs alike, even if both are the same model from the same manufacturer (Hamamatsu, 2001). Thus, all devices using a PMT must be carefully characterised in order to obtain an accurate measurement, as Brewer spectrophotometers do.

2.4.5. Multichannel detectors: Photodiode Arrays (PDA) and Charge-Couple Device (CCD)

The PDA, as well as CCD, is a photodetector made from semiconductor materials capable of measuring a complete spectral range of wavelengths from a polychromatic source, including UV, visible light and near infrared. Note that Silicon semiconductor devices are less sensitive to UV light, being more expensive those which are sensitive in the UV spectrum. PDA and CCD technologies are commonly used in spectrophotometry due to their ability to take simultaneous measurements across the full wavelength range, instead of using scanning or stepping wavelength techniques used with photomultiplier detectors which measure each wavelength sequentially.

PDA detectors use one-dimensional arrays (a row) or two-dimensional arrays (a 2D matrix) of individual (each photodiode is considered a pixel), which simultaneously measure the complete range of wavelengths coming from a diffracting element (see Figure 2.35). On the other hand, CCD detectors usually use a monolithic grid of sensing elements, whose semiconductor substrate can be modified to measure across different wavelength bands. Thus, the total resolution of the spectrometer depends on the entrance slit width, the properties of the dispersion element, the size of the photodetector and the number of pixels in each row of the array.

Figure 2.35 Diagram of a photodiode array spectrometer with concave grating.

Each row of pixels in two-dimensional, solid-state sensors is used as a single channel detector, which generally uses a different diffraction grating for each channel, allowing the analysis of more

57 than one spectral range of wavelengths at a time. However, pixels in solid state detectors are not individually adjusted to detect specific wavelengths. For this reason, the integration time of the whole detector must be regulated in order to avoid the saturation of any pixel. Otherwise, it may saturate or modify the value of neighbouring pixels.

The principal advantages of using this technology over the PMT are the ability to design compact spectrophotometers with fewer moving parts and solid state design; and the simultaneous multi-wavelength measurement. Nonetheless, measurements from solid state detectors at low photon count rates cannot compete against those measurements obtained by a PMT, because of its internal gain, higher sensitivity, lower response time, and lower noise levels. Moreover, the use of a cooler is a common requirement in solid state devices to avoid the noise produced by an increase in temperature.

New spectrophotometers based on CCD or PDA detectors have been developed as experimental instruments to measure atmospheric trace gases, including the total column ozone. These include the Pandora spectrometer (Tzortziou et al., 2012), Phaethon spectrometer (Kouremeti et al., 2013), METCON spectrometer (Smedley et al., 2015) as well as the Precision Solar Spectroradiometer (PSR) (Gröbner et al., 2014). However, they are still in an experimental phase.

2.5. Ground-based measurements

Atmospheric compounds can be measured using natural or artificial light sources using ground-based spectrophotometers, which estimate concentrations with direct measurements or scattered light measurements. However, this chapter will focus on spectrophotometers which perform direct ozone measurements.

The passive DOAS technique employs the sun, the moon or stars as a light source to estimate the column concentration of the atmospheric element of interest throughout the whole depth of the atmosphere, while active DOAS perform measurements which can reach into the lower stratosphere, using external light sources such as lasers and xenon-arc lamps (Platt, 2008).

The most recognized and widely used instruments for ground-based remote sensing of column ozone through UV measurements are Dobson spectrophotometer, Brewer spectrophotometer the M˗124 filter ozonometer used in Russia (Gushchin et al., 1985). Although, some other spectrophotometers such as SAOZ (Système D'Analyse par Observations Zénithales) (Kyrö, 1993; Green et al., 2001) and Pandora (Tzortziou et al., 2012; Herman et al., 2015) also perform UV measurements to estimate the total column amount of stratospheric ozone their use has thus far been limited to experimental applications rather than the routine data reporting (WMO, 2010b).

58 2.5.1. Dobson spectrophotometer

In 1913 Fabry and Buisson made the first measurements of ozone absorption coefficients, demonstrating that the high concentration of ozone in the stratosphere absorbs ultraviolet solar radiation (Fabry and Buisson, 1913). Then, in the early 1920s they built the Fabry-Buisson double spectrograph, which was the first instrument capable of estimating the total amount of ozone from the observing point to the edge of the atmosphere by comparing two different wavelengths (Fabry and Buisson, 1921). This complex instrument used two prisms at right-angle and a detection plate as a detection device.

In the mid-1920s, G.M.B. Dobson and his colleagues at Oxford developed the Fèry spectrograph that was simpler and more portable than its predecessor. Due to scientific interest in ozone measurements, the Royal Society supported the production of more spectrographs in order to study the ozone distribution over Europe. Although Fèry spectrographs provided the distribution and seasonal variation of ozone, they presented serious downsides. Firstly, it took weeks bringing the detection plates to the laboratory in Oxford to extract the data. Secondly, the solar zenith angle had to be high enough to obtain a reliable measurement. And finally, this spectrograph only could be operated by a trained professional.

In the late 1920s, Dobson designed a new spectrophotometer, which is able to directly estimate the total column ozone and be operated on a routine basis by an assistant with a minimum scientific background. The new design used two prisms as a double monochromator and a sodium cell photoelectric detector, which eliminated the necessity of using detection plates. By using two monochromators, the accuracy of the spectrophotometer increases with an error less than 1% (Dobson, 1968), rejecting a greater amount of stray light. It was achieved by dispersing the incident beam with the first prism, and focusing the selected monochromatic beam onto the exit slit with the second one; instead of dispersing twice the wavelengths as some designs of that time (see Figure 2.37).

Figure 2.36 General diagram of the Dobson spectrophotometer. (R. & J. Beck LTD)

59

Figure 2.37 Horizontal diagram of the Dobson’s monochromator, where M, P and L are the mirrors, prisms and lenses (Komhyr, 1980).

Dobson spectrophotometer estimates the amount of stratospheric ozone by applying the wavelength pair technique to two wavelengths, one strongly and one weakly absorbed by ozone. The two wavelengths are separated by the first prism and then selected by passing through two slits. The more intense wavelength (less absorbed by ozone) passes through an attenuating wedge which is adjusted until the resulting signal matches that from the less intense (more strongly absorbed) wavelength. The two monochromatic beams are chopped so that one and then the other continue through the spectrometer. The second prism focuses both wavelengths onto the detector and a measurement of the position of the optical wedge is made when there is a constant signal (i.e. the two exit beams are the same. The position of the optical wedge indicates the amount of attenuation required to balance the two signals and is used to determine the ozone from tables. A photomultiplier tube was added later to the spectrophotometer design. Figure 2.38 shows the component structure of the monochromator in the Dobson spectrophotometer.

Figure 2.38 General light path into the monochromator of Dobson spectrophotometer. Image taken from: http://www.o3soft.eu

In 1957, during the International Geophysical Year (IGY), due to the success of its measurements, a network of Dobson spectrophotometers was designated to monitor atmospheric total ozone values. Some years later, in 1980, the Dobson spectrophotometer #083 was selected by WMO as the ground-based spectrophotometer reference for the world, only used for direct or indirect calibrations of other spectrophotometers and not for routine measurements (Komhyr et al., 1989).

60 Komhyr and Grass (1972) performed relevant improvements by replacing the vacuum tube electronics by solid state circuitry and then incorporating an optically activated electromechanical commutator. Those improvements minimize errors and increase the sensibility of the Dobson spectrophotometer, allowing measurements of ozone from the almost full moon when it is high enough (Komhyr et al., 1989). Nonetheless, even when Dobson spectrometers can be configured to measure at different pairs of wavelengths, depending on the atmospheric conditions (see Table 2.3), the accuracy of the measurements decreases significantly under cloudy and at high solar zenith angles (Komhyr, 1980).

Table 2.3 Ozone absorption and scattering coefficients for the selected pair of wavelengths

Pair Wavelength λ1 305.5 nm A λ2 325.4 nm λ1 308.9 nm B λ2 329.1 nm λ1 311.5 nm C λ2 332.4 nm λ1 317.5 nm D λ2 339.9 nm

The wavelength pair technique used by Dobson spectrophotometer estimate the amount of ozone based on DOAS, where the ratio of measurements 퐼(휆) at two different wavelengths (weak 퐼(휆′) and strong ozone absorption) allows the cancellation of the cross-section coefficient component that slowly varies with wavelength, see equation (2.57) (Brönnimann et al., 2003).

퐼 (휆) 퐼(휆) 푙표푔 [ 0 ] − 푙표푔 [ ] − 푑 − [훽(휆) − 훽(휆′)]푚 − [훿(휆) − 훿(휆′)] sec(푧) 퐼 (휆′) 퐼(휆′) 푋푂3 = 0 (2.57) [훼(휆) − 훼(휆′)]휇

Where 휆 푎푛푑 휆′ represent the short and long wavelengths respectively, 훽 is the Rayleigh scattering coefficient, 훿 is the aerosol extinction coefficient, z is the apparent solar zenith angle, 훼 is the ozone absorption coefficient, m is the aerosol optical depth related to Rayleigh scattering, 휇 is the relative optical depth related to ozone absorption, and d is the extinction coefficient of the

퐼0(휆) instrument. The term ′ , which represents the ratio of irradiance outside the atmosphere, is 퐼0(휆 ) estimated during the calibration of the Dobson reference by applying Langley extrapolation to its measurements at Mauna Loa, Hawaii (Staehelin et al., 2003), whose calibration will be transferred to other Dobson spectrophotometers during calibration campaigns.

Although, many Dobson spectrophotometers are still operating around the world, they are no longer manufactured.

61 2.5.2. Brewer spectrophotometer

While Dobson instrument offers accurate ozone measurements, it presents serious disadvantages such as the inevitable calibration of the wedge, the limited conditions to take accurate measurements, its bulky and heavy design (around 50kg); its manual operation and the restriction of measuring only one pair of wavelengths at a time (Brewer, 1973).

The Brewer is an automated outdoor spectrophotometer remotely controlled by a computer, which can perform unattended measurements for several days without any intervention, although, an operator is needed to check that the programmed schedule is running properly. The control system (a computer) is usually located in a nearby building, thus, the user does not need to suffer the weather conditions, excepting during physical maintenance, such as replacing of the desiccant or cleaning the dome and the window.

Even though Dobson and Brewer instruments use essentially the same principle to measure ozone (DOAS technique), their construction is different as are the technologies used. The Brewer spectrophotometer is smaller and lighter than in its predecessor, with better weather isolation and lower temperature dependence. Also, by measuring five wavelengths at once and using a double diffraction grating (MkIII model), Brewer spectrophotometers obtain a better performance rejecting stray light measuring total column ozone at a greater range of zenith angles than the Dobson, optimizing the Umkehr ozone profile as well as allowing an estimation of the aerosol optical depth and SO2 amounts.

Thus, due to the accuracy of the Brewer spectrophotometer measuring stratospheric ozone over a wide range of conditions compared with that obtained by Dobson spectrophotometer, and the lack of spare parts necessary to repair Dobson instruments, the Brewer spectrophotometer has become the instrument of choice for measuring column ozone. However, the necessity of preserving the continuity of the measurements allows the remnant Dobson spectrophotometers to keep working.

History

The Brewer spectrophotometer was developed in Canada by Alan Brewer, Jim Kerr, Tom McElroy and David Wardle (Kerr et al., 1981) to overcome some of Dobson’s downsides and supplement ozone measurements made by the Dobson spectrophotometer and the M˗83 filter ozonometer.

The model Mark II (MkII) was the first Brewer produced and commercialised in the early 1980s by Sci-Tec Instruments Inc. (Kerr, 2010). It has a single monochromator based in the Ebert-Fastie architecture with a plane diffraction grating of 1800 lines per mm, operating in the second order. Therefore it requires optics to separate the UV radiation and block wavelengths around 600nm (due to the second order). The initial aim of the Brewer MkII was only performing ozone measurements. However, the increased concern about UV radiation exposure after the discovery of the hole in the ozone layer (Farman et al., 1985) triggered the modification in Brewer design to

62 also measure absolute, spectral UV radiation at ground levels. Those modifications to the MkII spectrophotometer were the additions of a dome, a diffuser and the internal optics for the radiation path from the diffuser.

Concerns over atmospheric pollution led to requirements for the measurement of nitrogen dioxide (NO2) in the atmosphere, which is a chemical involved in photochemical reactions of ozone mainly in the troposphere. NO2 is usually generated by natural processes in the stratosphere and by manmade pollution in the troposphere, and its radiation absorption occurs in UV and visible spectrum. However, its absorption has practically linear wavelength dependence in the UV region, and strong wavelength dependence in the visible spectrum, allowing the use of the DOAS technique to estimate the column amount of NO2. Therefore, in 1985 the grating of the Brewer MkII was substituted by a diffraction grating of 1200 lines per mm in the model Mark IV

(MkIV) to enable NO2 measurements at 426.4nm, 431.4nm, 437.3nm, 442.8nm, 448.1 and 453.2nm (SCI-TEC Instruments Inc, 1999). This modification allows the monochromator to take measurements of NO2 in the second order (visible spectrum) while the grating operates in the third order during ozone measurements. Thus, an increased ratio of stray light measuring the total column of O3 and SO2 diminishes its accuracy compared with MKII and MKIII spectrophotometers (Thompson et al., 1997). A nickel sulphate/UG˗11 filter combination is required to block the wavelengths of undesired orders, such as the first order in the MKII monochromator, and the first and second order in MKIV (Kerr, 2010).

The Brewer Mark III (MkIII) (see Figure 2.39) released in 1992 has an increased stray light rejection measuring UV irradiance from 286.5nm to 363nm in 0.5nm steps, compared with the MKII and MKIV models (Kipp & Zonen, 2014). It was designed using a double spectrophotometer with diffraction gratings of 3600 lines per mm operating in the first order, reaching an accuracy of ±1% in its measurement of the total column amount of ozone. Its reliability measuring stratospheric ozone during sunrises or sunsets was a noticeable improvement because the stray light becomes a significant problem for single monochromators with large optical depth. Thus, because of the high accuracy measuring the total column amount of ozone and UV radiation, nowadays the Brewer Mk III is the only model produced by the company Kipp & Zonen Inc., after production was transferred in 1996 from Sci-Tec Instruments Inc.

Figure 2.39 Brewer spectrophotometer model MKIII with a double monochromator.

63 Solar tracking

The zenith prism is the core of zenith positioning system. During direct solar measurements, the prism has 90 degrees of free rotation, between being vertically positioned at zenith angle equal to zero (see Figure 2.40), and horizontal. The position of the prism is determined by the solar zenith angle which is calculated in the control software according to section 2.2.1.

Figure 2.40 Zenith positioning system of the Brewer spectophotometer (right diagram).

The azimuth tracker, which is mounted on a tripod with the monochromators fixed on top, is in charge of positioning the Brewer at the correct azimuth angle, following the trajectory of the Sun. It has a robust weatherproof case, which protects its inside elements, such as the electronics, the gear system and its power supply (see Figure 2.41), mainly from rain, dust and snow. Since it acts as the positioning base of the spectrophotometer, it is extremely important to adjust each tripod leg to completely level the tracker, pointing the indicated leg away from the equator (it is highly recommended to do this before mounting the spectrometer). Once it is levelled, the spectrometer is mounted and minor corrections to the position (azimuth and zenith) should be made using the iris to determine when the solar disc is focussed in the field of view. The Brewer will then move to track the sun under software control. Since the cables that connect the azimuth tracker with the spectrometer are external, the Brewer should be monitored the first time that is turned on, to prevent the cables becoming entangled.

64

Figure 2.41 Standard azimuth positioning system of the Brewer spectrophotometer

Brewer’s monochromator description

The dispersing prism used in Dobson instruments was substituted in the Brewer by a plane diffraction grating, whose groove density is defined by the model of Brewer used. The next analysis only considers direct ozone measurements, with no clouds occulting the solar disc and its surroundings.

A complete diagram with the most important parts of the Brewer spectrophotometer is shown in Figure 2.42.

Figure 2.42 Simplified design of the single monochromator of a Brewer spectrophotometer (Josefsson, 1986).

65 First, a right-angle zenith prism (ZP1) selects the source of light from the three available sources: quartz window, which is exclusively to perform direct sun measurements; UV diffuser, which is used in zenith sky measurements of ozone due to cloudy condition or UV irradiance measurements; or a test lamp, which performs internal calibration as well as stability checks (see Figure 2.43). Once the light source has been selected, it is directed and collimated by lenses to pass through an iris diaphragm (IR1), which is used to manually align the image of the sun at the centre of the iris; and two filter wheels (FW1 and FW2), which polarize and attenuate the light with neutral density (ND) filters if necessary. Then the beam is focused onto the entrance slit of the monochromator.

Figure 2.43 The optical path of the three sources in Brewer spectrophotometer (left diagram) (Kipp & Zonen, 2014).

After the light comes through the entrance slit (ES1), the optical lens (LE5 and LE6) corrects coma and astigmatism aberrations of the monochromator system. A large concave mirror collimates and directs the incident bean to a diffraction grating, which disperses the light into the different wavelengths of its spectrum. Then, the dispersed light is focused by the same concave mirror, and the desired wavelength passes through the designated slit of the cylindrical slit mask (EX1). Depending on whether the Brewer uses a single or double spectrophotometer the beam is directed straight to the photomultiplier tube or to the second monochromator (Brewer MKIII) with identical characteristics than the first monochromator (see Figure 2.44).“The second monochromator recombines the spectrum dispersed by the first, and hence does not increase the spectral resolution” (Bais et al., 1996). However, undesired reflected photons at the end of the middle slit (slit mask) will not be focused on the exit slit by the second grating, significantly reducing the amount of stray light that reaches the photomultiplier tube (PM1).

66

Figure 2.44 Optical elements of the double monochromator Brewer spectrophotometer (MkIII) (Kipp & Zonen, 2014)

The cylindrical slit mask is designed with eight positions, six of them only match with one exit slit, another blocks all the slits to perform dark count measurements, and the last one, which opens two exit slits together (slit #2 and slit #4), is used to measure the dead time of the photomultiplier tube (see Figure 2.45). The six slits are designed to enter one selected wavelength to the photomultiplier, but one slit is dedicated to the mercury lamp test (303.2nm) while the other five are used in ozone measurements (306.3nm, 310.1nm, 313.5nm, 316.8nm and 320.1nm) (Kipp & Zonen, 2014). During UV irradiance measurements, the slit mask remains in a fixed position while the diffraction grating rotates, but during direct ozone measurements the grating is fixed and the slit mask rotates.

Figure 2.45 Slit selector mask in the Brewer spectrophotometer (Kipp & Zonen, 2014).

A tungsten-halogen lamp, which is used to perform stability checks in all Brewers, has an important additional role in the Brewer MkIII double monochromator, aligning the gratings with respect to each other. The rotation of both diffraction grating and slit mask uses stepper motors, also called micrometers, which must be routinely calibrated. A mercury lamp calibrates the micrometers that move the diffraction grating to a specific wavelength using the Hg slit, aligning

67 the wavelength with the centre of the slit (in the MkIII spectrophotometers both diffraction gratings rotate simultaneously). During standard lamp tests, the micrometer that controls the slit mask is calibrated ensuring that the desired wavelength passes through the centre of the slit. Figure 2.46 displays the micrometers in a Brewer MkIII.

Figure 2.46 Micrometers in a opened Brewer MkIII.

Finally, the photomultiplier tube detects the number of photons and then the photon counts (RAW data) counted in a measurement interval are sent to a computer using the serial communication port. From there the data can be processed and stored.

The slit function in a Brewer spectrophotometer

The purpose of the monochromator (diffraction grating and slits) in spectrophotometers is to direct only photons of the desired wavelength to the photomultiplier tube (PMT) so that it measures only monochromatic photons. However, because of limits on dispersion and the finite width of the exit and entrance slits, light photons from adjacent wavelengths are also measured by the PMT.

The slit functions of Brewer spectrophotometers are evaluated during calibration, obtaining the dispersion profile of each slit, which represents the relationship between the step number of the micrometer that rotates the diffraction grating and the photon count rate when the spectrophotometer scans across the wavelength region of a monochromatic source (usually a laser). Then, the measurements are fitted into a triangular or trapezoidal profile (see Figure 2.47), in order to determine the central wavelength and the full width at half maximum (FWHM) of each slit function (Redondas et al., 2014).

68

Figure 2.47 Scan of a 296.728nm mercury line, which is fitted into an isosceles triangle. The circles and crosses are fitted points (Gröbner et al., 1998).

Since a step number is related to a particular wavelength, the slit function represents the spectral response of the final transmittance after each exit slit, as long as the diffraction grating micrometer is set to measure a fixed wavelength (Lakkala et al., 2008). By comparing the slit function of two Brewers with single and double monochromator optical layouts (see Figure 2.48), it can be appreciated that more stray light is rejected when a double monochromator is used, allowing Brewer MKIII to take measurements at larger ozone slant paths.

Figure 2.48 Slit function comparison of a single monochromator against a double one, using a HeCd Laser at 325nm. Image taken from: http://kippzonen-brewer.com

The approximation method used by Bernhard et al. (2005) to estimate the effective ozone cross-section coefficients, 훼𝑖(휆), for Dobson instruments, is essentially the same as that used in Brewer measurements (2.58) (Redondas et al., 2014).

69 ∫ 휎(휆)푆 (휆)푑(휆) 훼 (휆) = 𝑖 𝑖 ∫ 푆 (휆) 푑(휆) 𝑖 (2.58)

Where 푆𝑖(휆) is the slit function and 휎(휆) is the ozone cross-section coefficients obtained at ˗45°C in a laboratory.

By narrowing the width of slit, the spectral resolution increases, but the transmitted irradiance is reduced. This leads to using longer acquisition times to overcome the lower photon count rate at UV wavelengths.

Temperature dependence

Double monochromator Brewers are thermally regulated by an internal heater to keep the temperature above a minimum temperature of 10°C or 20°C. A fan distributes the air inside the instrument trying to maintain a uniform temperature (Kipp & Zonen, 2014). Nonetheless, the spectrophotometer does not have an internal active cooler to keep a stable temperature in case the instrument is exposed to higher air temperatures or direct solar heating.

Since the Brewers are not designed with full temperature stabilisation, the temperature dependence of the optics, the order blocking filter (MKII and MKIV) and diffuser in the monochromator may affect the measurements. To minimise this, the component materials are chosen to have very similar coefficients of expansion/contraction. Moreover, the photomultiplier tube, which is also temperature dependent, increases the dark current count if the temperature raises, hence the need for a dark current measurement with every ozone or UV measurement.

The temperature dependence test performed by Lakkala et al. (2008) on Brewers #107 (MKIII) and #037 (MKII) suggest that the measurements can be corrected using the following linear mathematical approach (2.59) (see Figure 2.49).

퐼(휆) = 퐼0 × [1 + 푐(휆) × 푑푇] (2.59)

Where 푑푇 is the temperature difference (푇 − 푇0) between the Brewer’s internal temperature 푇 and the reference temperature 푇0=23°C; I0 is the reference irradiance at 푇0 in counts per cycle, 퐼(휆) is the measured irradiance in counts per cycle and 푐(휆) is the temperature dependence factor.

70

Figure 2.49 Temperature dependence of Brewer #107 (blue crosses) and Brewer #037 (green circles) (Lakkala et al., 2008).

Kerr (2002) also states that the temperature dependence in Brewer spectrophotometers is linear, but it becomes non-linear at higher wavelengths (see Figure 2.50).

Figure 2.50 Temperature dependence graph of Brewer #014 using an internal lamp over temperatures from 1°C to 39°C (Kerr, 2002)

Temperature dependence tests are usually performed in a laboratory under controlled conditions. The test performed by Kerr (2002) shows a temperature dependence around 4.8 × 10−3/°C at 310nm in the Brewer #014 (MKII) and Lakkala et al. (2008) found a lower value (around 1× 10−3/°C) in the Brewer #037 (MKII) at 310nm. However, because of the low temperature dependence, it is unusual to monitor the internal temperature of the Brewer and then correct the data during field measurements.

71 Dark count

This is performed by measuring the number of photons detected when the slit mask blocks the light entrance to the photomultiplier tube. This offset is then subtracted from the photon count when the Brewer measures either direct or diffuse irradiance. Each photomultiplier tube has to be monitored and characterized since the dark count is temperature dependent and it varies from tube to tube.

Dead time

The pulse generated each time that the photomultiplier tube detects a photon has a finite width of approximately 40ns (Kipp & Zonen, 2014), which is called the dead time. Since, the photodetector can measure one photon a time, if more than one photon reaches the detector during the dead time, they will only be counted as a single photon. This dilemma is solved by correcting the photon count using equation (2.60), which estimates the ratio between the true and the measured photon count using the Poisson statistical distribution (Kipp & Zonen, 2014).

1 푁0 휏 = 푙표푔 ( ) (2.60) 푁0 푁

Where 휏 is the dead time in seconds, 푁0 is the true count-rate (counts per second) and N is the observed count-rate. However, this approximation becomes less accurate if the photon count- rate is too high, so, a neutral density filter is used to adjust the level of light entering to the monochromator to minimize its effect (if the count rate exceeds an established limit, the filter will be automatically replaced by another with higher attenuation, and vice versa).

Angular response

The angular response function 푓푏(휃) is related to the performance of the Teflon diffuser and the covering dome used in measuring the absolute UV irradiance. Thus, the cosine error, which is wavelength-dependent and increases at large solar zenith angles, represents the deviation between Brewer measurements and those values considering that the dome and diffuser were ideal (proportional to the cosine of the solar zenith angle) (Fioletov et al., 2002).

The characterization of the angular response is necessary to correct the measurements of absolute UV irradiance. The method used by Lakkala et al. (2008) to determine the angular response was removing the dome, then covering the diffuser with a black screen except for a hole in the most sensitive part of the diffuser, and finally uniformly illuminating the selected area from different angles with respect to the normal. They found that the maximum cosine error of Brewers MKII #037 and MKIII #107 was about 10% and 9% respectively for solar zenith angles up to 85° with no azimuth angle dependency.

72 Calibrations and maintenance

Regular maintenance (e.g. daily) includes cleaning the quartz dome and the quartz window usually with isopropyl alcohol, as well as verifying that the schedule is running as desired. Also, the humidity sensor must be frequently checked to maintain the humidity lower than 15% inside the spectrophotometer; otherwise, the desiccant must be replaced to protect the interior elements from moisture. Less frequent maintenance includes performing alignment checks (sun sighting) ideally every two weeks, and cleaning the drive-plate of the azimuth solar tracker every two months to remove particles generated by friction during the rotation of the spectrometer.

The stability of ozone measurements is checked on a daily basis by reference to the internal lamp of the Brewer, which does not have a known irradiance, but it is spectrally stable. On the other hand, the UV stability is checked by performing diffuse measurements using the 50W calibrated lamps with a known irradiance, and then their agreement is compared with the measurements obtained during the last UV calibration. Consequently, changes to the parameters of the calibration file may be made to account for optical, mechanical or electronic adjustments in the internal elements of the spectrophotometer. By checking the records of data measurements, as well as the calibration tests, any need for corrective maintenance or calibration can be determined.

Advantages and limitations of Brewer spectrophotometer

By replacing the prism of the Dobson spectrophotometer by a plane holographic diffraction grating as the dispersion element of the monochromator, the design of the Brewer spectrophotometer was simplified and an optimization of the slit function for a desired wavelength range was achieved. Also, the narrow field of view (3° instead of 8° in Dobson) for direct sun measurements, the use of a slit mask, which automatically rotates allowing the measurement of five wavelengths almost simultaneously; as well as the use of the DOAS technique enable Brewer spectrophotometers to make faster and more accurate ozone measurements at low zenith angles.

Other improvements of the Brewer compared with the Dobson are its relatively compact and light design and its ability to measure other trace gases (SO2 and NO2) and absolute UV irradiance. Nonetheless, its autonomy taking measurements within any environment, with a lower temperature dependence than the Dobson (Staehelin et al., 2003), is its greatest asset.

However, the Brewer spectrophotometer is not perfect. The temperature dependence of the optical parts, in addition to the potential for optical misalignment of the monochromator due to moving parts, are its inherent weaknesses. Therefore, the use of new technologies in optics and electronics can improve the speed of ozone measurements, obtaining accurate values in locations where the atmospheric conditions rapidly change, as well as taking advantage of short periods of time where the solar disc is free of clouds to take measurements.

The Brewer network, with more than 180 spectrophotometers distributed around the world, relies on the long-term stability of three well maintained MkII Brewers (#008, #014 and #015) settled at Toronto, Canada, which were designated as a calibration “triad” reference by

73 Environment Canada in 1984 (Fioletov et al., 2005). They are independently calibrated on a regular basis at Mauna Loa Observatory, Hawaii, where the ETC values are obtained by Langley extrapolation.

The triad transfers the calibration to the travelling reference (Brewer #017), which is used to compare and verify measurements or adjust other spectrophotometers around the world, commonly during calibration campaigns, avoiding the movement of the triad. This selected Brewer is calibrated against the Canadian triad before being transported and after returning from its journey, and its stability is monitored during the campaign by performing tests with internal and external calibrated lamps (Kerr et al., 1985; Fioletov et al., 2005).

In 2003, the Regional Brewer Calibration Centre for Europe (RBCC-Europe) was located at Izaña Observatory, Spain, where the conditions of clear skies and low ozone variability are similar to those at Mauna Loa Observatory. Thus, there resides a second triad of Brewer spectrophotometers (MkIII), designated as European reference: the Brewer #157 as regional primary reference, the Brewer #183 as regional secondary reference and the Brewer #185 as regional traveling reference (Redondas and Rodriguez-Franco, 2014). Generally a comparison of both triads is performed during calibration campaigns by the comparison of their respective traveling Brewer.

During a campaign the Brewers take measurements side by side, while different characterization tests are performed on them, such as the slit function dispersion, cosine response of the diffuser, polarization of the window, dead time, dark count and ND filters tests. Also, preventive and/or corrective maintenance is performed, commonly to the track plate of the azimuth tracker, the electronics, the communication system, the power supply, the cables, the sealing of the enclosure and the micrometers of the monochromator.

The historic record of all Brewers is evaluated, as well as their performance during the campaign to compare their stability and the dispersion ratio with the reference instrument. The traveling reference transfers its calibration to others spectrophotometers participating in the campaign, with customized calibration parameters for each Brewer. And finally, the spectrophotometers undergo reference lamp calibration for UV irradiance to preserve their accurate and stable measurements until the next calibration.

The European Brewer Network (EUBREWNET) coordinates a network of Brewer spectrophotometers in Austria, Belgium, Czech Republic, Denmark, Finland, Germany, Greece, Hungary, Ireland, Italy, Netherlands, Norway, Poland, Portugal, Slovakia, Spain, Sweden, Switzerland, Turkey and United Kingdom (European Cooperation in Science and Technology, 2005). The EUBREWNET aims to establish a solid network of Brewer spectrophotometers and provide consistent and accurate records of ozone measurements, ozone profiles, global UV irradiance, and aerosol optical depth of every participating site, in order to have an international dataset of measurements available to use in further analysis.

74 RBCCE Intercomparison campaign and EUBREWNET campaign at El Arenosillo, Spain on May 2015

The 2015 RBBC-E intercomparison campaign was held at El Arenosillo, Spain (latitude 37.10N, longitude 6.73W and altitude of 50 a.m.s.l) and attended by the author. A total of 21 Brewers participating in the campaign were placed together on the roof of “Estacion de Sondeos Atmosfericos” of the “Instituto Nacional de Tecnica Aeroespacial (INTA)”, where measurements of direct, global irradiance and Umkehr profile were performed by all the Brewers under the same atmospheric conditions (see Figure 2.51).

Figure 2.51 Picture of the Brewer spectrophotometers involved in the calibration campaign at El Arenosillo, Spain on May 2015

Brewers #172 (MKIII) and #126 (MKII) located at the University of Manchester, UK, as well as Brewer #075 (MKIV), which is normally sited in Reading, UK, participated in the calibration campaign. In order to check their stability after the journey to/from the UK, measurements were performed using a set of 50W calibrated lamps positioned above the diffuser before being moved and after being installed at the campaign site (or home site on return).

The dedicated computer systems used to remotely manipulate the Brewers were located below the measurement platform, ensuring the operators did not interfere with the measurements. During the first hours of the morning, all the Brewers were programed with a specific schedule, depending on the day and the model of the spectrophotometer. Due to blue skies and cloud-free conditions during the whole campaign, the spectrophotometers could take measurements without disruption unless a schedule was interrupted to perform maintenance or other tests.

The data acquired in the first days were analysed and compared with the Canadian and European traveling references, Brewer #017 and Brewer #185 respectively. The characterization of each Brewer was obtained through different tests, such as the angular response of the diffuser, the polarization of the window and the slit function dispersion test. The measurements obtained in the campaign, as well as the measurements obtained from the 1000W calibration lamp, were analysed in combination with the historic records of measurements of each Brewer to obtain the new calibration parameters. Thus, those values are completely customized for each Brewer.

Corrective and preventive maintenance were performed on the spectrophotometers, usually without the need to disassemble the entire instrument. The azimuth tracker, sealing of the enclosure, micrometers, gears, power supply, the communication module, cables, desiccant and electronics were the most common elements that required maintenance.

75 The intercomparison of near 600 simultaneous direct sun measurements during the 8 days of the campaign allowed the ozone calibrations of all participating Brewers. The comparison between the actual and the previous ozone calibration of Brewers #172, #126 and #075, which were also made in 2013 also at El Arenosillo, Spain; shows changes of ˗1%, ˗1.5% and ~˗5% respectively. The UV calibrations, which were obtained by using two 1000W calibrated lamps, had changed by ˗1% to 0.5% for #172, +15% to +5% for #126, and ˗6% to ˗4% for #075 with respect those obtained in 2013, according to the report presented by Lamb (2015). This shows that the Mark III Brewer in Manchester has been very stable for the past 2 years, for both ozone and spectral UV measurements. The single monochromator instruments (#126 and #075) are not maintained for routine UV monitoring and the lack of stability is more apparent here. They are also slightly less stable in ozone measurement, with standard lamp records not as steady as those obtained by the Brewer #172 (also Brewer #075 presented an abrupt decrease of ˗2% associated with updated temperature coefficients in 2014).

The maintenance performed on the Brewers #172, #126 and #075 mainly consisted of cleaning the gear system of micrometers and the tracking plate of the azimuth tracker, as well as replacing the seals and improving the sealing of the enclosures to reduce the replacement intervals of desiccant. The test used to verify the optimum sealing of the Brewer consisted of pulling air inside the instrument and measuring the interior pressure with a manometer. Then, any significant gas leak was repaired until achieving the desired sealing.

The software that remotely control the Brewer was updated in all the instruments and a script in the program of Brewer #172 was modified to enable it to display interior humidity measurements. However, the calibration of the humidity sensor to obtain reliable values was postponed until its return to Manchester.

In order to ensure that the Brewer #172, #126 and #075 do not lose their calibration after being returned to their monitoring locations, a new set of measurements using the 50W calibrated lamps was performed before packing the instruments, and also after installing them at their respective places in the UK. The manual states that if the differences between UV stability test are greater than 5%, an absolute calibration using a 1000W calibrated lamp should be required (Kipp & Zonen, 2014). Nonetheless, differences of 1% using 50W calibrated lamps suggest that the instruments had not been disturbed by the journey

Then, a reanalysis and correction of the ozone data records from the last calibration (2013) to the new calibration (2015) was performed for each Brewer instrument, assuming a linear variation in calibration during this two-year period. Only Brewer #172 (Mark III) is relied upon for spectral UV monitoring in the UK. Although, the change in UV response between 2013 and 2015 was less than 2% (which is the UK threshold for reprocessing UV data records), strictly a new calibration was not necessary after this intercomparison. Nonetheless, the new calibration obtained in El Arenosillo was put into operation and applied to all new data from 2015 onwards.

Data coming from the Brewer #172 located in Manchester, UK is sent to the World Ozone and Ultraviolet Radiation Data Centre (WOUDC), while the measurements from Brewer #126 are used as a backup. However, measurements of global irradiance from the Brewer #075 in Reading, UK,

76 are not used for routine data reporting, due to its co-location with a Bentham DM150 spectroradiometer, a double monochromator with a wavelength range from 280-500 nm. Ozone measurements from both Brewer #172 in Manchester and Brewer #075 are reported to WOUDC.

2.5.3. Other ground-based instruments

Pandora spectrometer

Pandora is a spectrometer developed at NASA's Goddard Space Flight Centre to perform direct measurements from sun, moon or sky to estimate the total column amount of atmospheric gases, such as NO2, HCHO, H2O, SO2, including ozone, as well as their vertical profiles.

The entrance optics of the Pandora instrument for direct and diffuse measurements are connected through a fibre optic to the spectrometer, which is stored in a temperature stabilised enclosure (see Figure 2.52). Its monochromator is based on Czerny-Turner geometry with a diffraction grating of 1200 lines per mm, and it uses a CCD or CMOS sensor as a photodetector, depending on the model. The version using a CMOS detector only performs direct sun measurements from 280nm to 500nm while the CCD version is also able to obtain direct sun and sky measurements from 280nm to 525nm with 0.6nm of resolution (Tzortziou et al., 2012).

Figure 2.52 General schematic of the Pandora instrument using a CMOS detector (Herman et al., 2009), as well as an actual instrument performing field measurements (right diagram)

Tzortziou et al. (2012) compared the total column ozone estimated by Pandora #09 and Pandora #02 with the Brewer spectrophotometer #171 and the Aura Ozone Monitoring Instrument (OMI) over more than one year. Pandora-Brewer comparisons showed a seasonal dependence with an average deviation of +3.6DU in summer, ˗4.74DU in winter, 1.3DU in spring and 2.4 in fall from a range of solar zenith angles from 15° to 80°.

The 1-year comparison made by Herman et al. (2015) (limited to free-cloud sky conditions) contrasted the temperature-corrected ozone estimations from Pandora #34 with the ozone estimations from Dobson #061 (average deviation of 1.1 ± 5.8DU), and satellites such as Aura- OMI (average deviation of 1.1 ± 8DU) and NPP-OMPS (average deviation of 3.8 ± 8DU).

77 The advantages of Pandora over the Dobson and Brewer spectrophotometer are its compact design, its temperature stability, its ability to also measure the total column of other trace gases besides ozone, as well as its narrow field of view of 2.2° for sky measurements with a diffuser and 1.6° for direct-sun measurements without a diffuser. Also, the average deviations of Pandora spectrometer compared with those from ground-based instruments and satellite spectrometers suggest that further improvements in technology will make Pandora a reliable source measuring total column ozone.

Phaethon spectrometer

The Phaethon spectrometer is a portable ground-based instrument (see Figure 2.53) developed at the Laboratory of Atmospheric Physics, Aristotle University of Thessaloniki, Greece. It performs direct sun measurements or sky spectral irradiance measurements from 310nm to 1000nm, using a CCD detector, to estimate the aerosol optical depth, as well as the total column amount of trace gases as NO2, HCHO, SO2, O4 and BrO, including ozone. However, it is still only used in experimental applications, and no literature is available related to the technical features of the spectrometer.

Figure 2.53 Diagram of Phaethon spectrometer (Kouremeti et al., 2013)

The daily comparisons of total column ozone measurements between Phaethon spectrometer and Brewer MKII #005, made by Kouremeti et al. (2013) (see Figure 2.54), suggest that Phaeton spectrometer has the potential to provide reliable measurements of atmospheric ozone. Nonetheless, it is necessary to perform at least a one-year comparison to obtain a better characterization of the instrument, such as its relative seasonal dependence.

78

Figure 2.54 Comparison of total column ozone between Phaethon spectrometer and Brewer spectrophotometer #005 (Kouremeti et al., 2013).

Precision spectroradiometer (PSR)

PSR is a compact solar radiometer (see Figure 2.55) developed at the Physikalisch- Meteorologisches Observatorium Davos / World Radiation Center (PMOD/WRC), which was designed to perform direct measurements from 300nm to 1020nm with FWHM from 1.5nm to 6nm. The main aim of the PSR is to obtain the spectral aerosol optical depth; however, it is also able to perform measurements of other atmospheric parameters including total column ozone.

The main components of the PSR are two entrance optics, a temperature-stabilised plane grating of 200 lines/mm; an NMOS linear image detector with a wavelength step of 0.7nm per pixel, an Analogue to Digital Converter (ADC) of 18 bits and a sealed nitrogen flushed enclosure (Gröbner et al., 2014)

Figure 2.55 Precision Spectroradiometer PSR-003 (Physikalisch-Meteorologisches Observatorium und Weltstrahlungszentrum Davos, 2014)

In order to minimize the temperature dependence, the PSR uses a temperature regulation system based on Peltier technology to stabilise the temperature of the detector and a heating system in the rest of the instrument to prevent lower temperatures than the minimum desired. The

79 performance of the PSR against the stray light was characterised by a tunable laser at the national metrology institute of Germany (PTB) (see Figure 2.56).

Figure 2.56 Line spread function (LSF) for PSR-003 (Physikalisch-Meteorologisches Observatorium und Weltstrahlungszentrum Davos, 2014; Gröbner et al., 2014).

After the PSR was fully tested and characterized, it was made available on a commercial basis from March 2014. The temperature stabilised system is its most outstanding feature. Since the main purpose of the instrument is to measure spectral optical depth no long term comparison of ozone measurements with Dobson or Brewer spectrophotometer are available in the literature.

SAOZ (Système D'Analyse par Observations Zénithales)

SAOZ is a broad-band spectrometer developed at CNRS, France, dedicated to measuring sunlight at zenith sky from 300nm to around 600nm with 1nm of resolution (Kyrö, 1993). The SAOZ spectrometer uses a holographic diffraction grating with 360 grooves per mm in conjunction with a diode array as a detector to estimate the total column of trace gases such as nitrogen dioxide, bromine oxide and chlorine oxide, as well as ozone, which is obtained by measuring irradiance in the visible Chappuis band. If the solar zenith angle is lower than 85°, the instrument performs measurements of the solar spectrum each hour. However, it takes measurements every five minutes during sunrises and sunsets at solar zenith angles up to 94° (Hendrick et al., 2011) to retrieve ozone.

Figure 2.57 Photograph of the SAOZ spectrophotometer

The network of SAOZ instruments, which was initially deployed in 1988 in the polar regions, consists of about 20 SAOZ located at different latitudes in both hemispheres. They contribute data to the Network for the Detection of Atmospheric Composition Change (NDACC), and are

80 periodically compared with ozone estimations from satellite instruments (TOMS, GOME-GDP4, SCIAMACHY-TOSOMI, SCIAMACHY-OL3, OMI-TOMS and TOMI-DOAS) and ground-based spectrophotometers (Dobson and Brewer) (Hendrick et al., 2011; Kyrö, 1993; Green et al., 2001).

The SAOZ spectrometer is a simple and robust instrument, without moving parts and low maintenance, mainly designed to measure ozone at high solar zenith angles in polar regions. However, it is not the most suitable option to use at other latitudes where solar zenith angles reach lower values, there is more UV, and shorter twilight periods give a reduced number of SAOZ ozone measurements. A SAOZ instrument is located at Aberystwyth; data has recently been reprocessed back to 1998, the processing algorithm automated, and is now being submitted to NDACC and WOUDC data repositories.

METCON spectrometer

The METCON (Smedley et al., 2015) is a double channel spectrometer, which is able to perform spectral measurements of global irradiance and direct spectral measurements from 280nm to 700nm with a resolution of 2.2nm. Each channel of the spectrometer has a set of external entrance optics connected by optic fibre to a temperature stabilised and weatherproof enclosure (for this application), which contains two fixed diffraction gratings, two 512-pixel diode array detectors and one 15-bit data acquisition device allow quasi-simultaneous measurements of both channels (see Figure 2.58). One potential disadvantage is that both detectors use the same integration time, set to avoid the saturation of either sensor. This can mean that one sensor is less sensitive than it otherwise might be.

Figure 2.58 METCON spectrometer at Manchester City (Smedley et al., 2015) (a) displays the temperature stabilised enclosure, the global irradiance entrance (located in the upper right corner), as well as the Brewer #172. (b) Close-up image that shows the direct entrance optics mounted on the solar tracker

The entrance optics used to perform direct sun measurements possesses a field of view of 2°±0.1°. They are mounted on a solar tracker, which uses an external photodiode detector to optimize its pointing accuracy (<0.01°).

The comparison of total column ozone measurements between the METCON spectrometer and the Brewer #172 was performed in Manchester, where the METCON was placed next to the

81 Brewer to obtain measurements under the same atmospheric conditions. The comparisons showed a dependence on air mass of 6.3DU per air mass unit with an overall bias of 1.1DU and scatter of ±6.7DU for measurements at μ<4 (Smedley et al., 2015).

Although the primary aim of the METCON spectrometer in this configuration (Smedley et al., 2015) was to obtain the total column ozone and the spectral aerosol optical depth, the instrument design should, in principle, also permit the estimation of other trace gases from the same measurements. However, experimentally this capability was limited because of the need of further characterizations and improvements.

2.6. Satellite measurements

Spectrometers mounted in satellites, such as the Total Ozone Mapping Spectrometer (TOMS), are widely used to measure global irradiance and estimate the column density of atmospheric compounds. These spectrometers are carefully designed and characterized in a laboratory considering all the possible issues to overcome, such as different atmospheric conditions, albedo effect and pollution. However, once the satellite has been launched and it is in orbit, the only way to validate its measurements is to compare them with those obtained by ground-based instruments. Thus, the accuracy of satellite spectrometers measurements such as total column ozone measurements relies on ozone ground-references: Dobson and Brewer spectrophotometers, as well as the filter ozonometer M-83 (Fioletov et al., 1999).

Satellite measurements and corresponding ground-based data provide a mutual validation system. However, multiple comparisons are performed with a range of ground-based instruments at varying latitudes and satellite instruments. If the satellite disagrees with multiple ground-based stations then the satellite is suspect. If a single ground-based instrument differs from the satellite then that instrument must be further investigated.

Satellite global view versus point measurements from ground spectrometers

The size of the area represented by each sample of the satellite, better known as the footprint, depends on the altitude of its orbit and on the characteristics of the instrument. However, the shape of the footprint will vary at different latitudes, being small and square-shape in polar regions and asymmetrical near the equator because of the rotation of the Earth, see Figure 2.59.

Figure 2.59 Footprint of the Meteor-3 TOMS projected onto Earth’s surface during a day of each measured sample at 1050km of altitude. The right samples (1 to 18) are obtained from two consecutive scans, displaying the overlap of consecutive orbits at the equator (NASA, 1996a). X-axis represents latitude and Y-axis represents longitude.

82 The resolution achieved by a satellite spectrometer plays an important role during intercomparison with ground-based spectrophotometers. The satellite instrument averages the total column ozone and vertical profile of ozone over several square kilometres depending the field of view (see Table 2.4), while ground spectrometers perform measurements at a specific location. Although, ozone measurements do not often have abrupt changes day by day, there is still a need to match the ground-based measurement to the overpass time for the best intercomparison. Thus, a more reliable comparison between ground-based spectrometers and satellite spectrometers is obtained from satellites with high spatial resolution and frequent global coverage.

Satellite measurements are also cloud dependent, with partly cloudy conditions, giving a non- homogenous view of the pixel from above, being most difficult to deal with. However, an increase in resolution, reducing the possibility of inhomogenous areas, commonly is related to an increase of the global coverage time, unless improved technology is used.

2.6.1. DOAS technique applied to satellite measurements

Satellite spectrometers are able to globally monitor atmospheric compounds by applying DOAS technique to the backscattered radiation from the Earth’s surface and the scattered radiation from the atmosphere. Thus, ozone satellite measurements are estimated by analysing wavelengths where ozone is strongly absorbed; and depending on the technology of the satellite, they can be obtained either from a specific number of wavelengths or from a complete spectral range analysis of the incoming radiation.

The most common view geometries to perform measurements of total column of trace gases as well as their vertical profiles are nadir, limb and occultation, with limb and occultation mainly used to obtain vertical profiles (see Figure 2.60).

Figure 2.60 Different view geometries used to perform satellite measurements of global ozone: limb, nadir and occultation view-geometries (Burrows, 2011)

83 Satellite spectrometers using the nadir view geometry provide a global mapping of the total column ozone and vertical profile of ozone, performing measurements at wavelengths that are strongly absorbed by ozone. Ozone measurements are obtained by measuring the solar irradiance entering to the atmosphere, the backscattered radiation from the atmosphere and the reflected radiation from the Earth’s surface (considering that solar radiation passes twice through the atmosphere). This technique allows satellite spectrometers to achieve high horizontal resolution in measuring the total column ozone as well as other trace gases such as NO2, CO,

CH4, H2O, N2O, SO2 and aerosol particles when passing over sunlit areas of the globe. However, the accuracy of tropospheric measurements may be highly affected by multiple scattering of clouds and pollution, thus, producing some uncertainty in column ozone measurements (dependent on region), greater uncertainties in the lower part of the vertical profile and limited resolution.

Satellite spectrometers use the occultation view-geometry to measure ozone vertical profiles from the moon and stars, as well as the sun. They measure the direct radiation from those celestial bodies before and after being attenuated through the limb of the atmosphere (see Figure 2.60), and for this reason they are considered self-calibrated devices. Most satellite spectrometers using this technique analyse a wide spectral range of wavelengths, from UV to near infrared, in order to obtain an accurate composition of the atmosphere. Even when most satellite spectrometers perform sun and star occultation measurements, they have low horizontal resolution. Furthermore, the time to obtain global coverage is much larger than for those instruments using nadir view-geometry.

Limb scattering is a technique similar to that of occultation that allows satellite spectrometers to provide high vertical resolution measurements of stratospheric and tropospheric compounds. It measures the scattered solar radiation coming from the limb of the atmosphere, with a line of sight tangential to layers of the atmosphere. However, unlike the occultation view-geometry, instruments using limb scattering perform more reliable measurements of total column of trace gases, such as ozone, NO2, CO, CH4, H2O, N2O, SO2 and aerosol particles, depending on the satellite model. The reason why occultation scattering is not commonly used to perform total column measurements as limb and nadir does, resides on the inherent footprint associated with its view-geometry. In nadir scattering and limb scattering, the footprint is considered a rectangular area, while the occultation scattering region is more like a line.

2.6.2. Satellite spectrometers mapping ozone

The satellite spectrometers explained below are the most recognised and widely used for global ozone data. Other countries have launched satellites with ozone measuring capability, but these data are not widely available. Of note for this thesis, the spectrometer SPICAM (Spectroscopy for Investigation of Characteristics of the Atmosphere of Mars) on board of the Russian mission Mars 96 was the first documented satellite spectrometer using the Acousto- Optical Tunable Filter (AOTF) technology to analyse the Mars atmosphere in the near infrared

84 spectrum. However, after its failure in 1996, another SPICAM spectrometer was on board the Mars Express which was launched in 2003 (Korablev et al., 2002; Korablev et al., 2006; Bertaux et al., 2006).

Total Ozone Mapping Spectrometer (TOMS)

In order to monitor atmospheric ozone, the National Aeronautics and Space Administration (NASA) designed TOMS, which was able to retrieve data used to produce a global ozone map in a one-day period. It used a single monochromator based on the Ebert-Fastie design with a plane diffraction grating and six exit-slits to measures the backscattered Earth irradiance in the UV spectrum. By using the nadir geometry and DOAS technique, TOMS estimated total column ozone, as well as ultraviolet irradiance, backscattering from clouds, reflectivity of Earth’s surface, volcanic SO2 and aerosols. However, cloudy atmospheric conditions would increase the uncertainty in measurements mostly from tropospheric ozone.

The Nimbus˗7 TOMS was the first satellite spectrometer to provide a long-term record of ozone retrieval beginning in 1978 (NASA, 1996b). Then, the Russian satellite Meteor˗3 was equipped with a TOMS (NASA, 1996a), recording data from 1991 to 1994. Some years later, ADEOS TOMS (NASA, 1998a) and Earth Probe TOMS (NASA, 1998b) were designed to continue ozone global mapping (see Table 2.4). However, after the failed mission of the ADEOS satellite, the Earth Probe TOMS, which was initially intended to obtain outstanding spatial resolution from an altitude of 500km, was moved to an altitude of 750km to provide almost complete global coverage (84% at equator and 100% at 30° latitude) on a daily basis (see Figure 2.61).

Figure 2.61 Fields of view of the Earth Probe TOMS projected onto Earth’s surface of each measured sample at 50km and 750km of altitude. Two consecutive scans are displayed in the bottom of the graph (NASA, 1998b). The X-axis represents latitude and Y-axis represents longitude.

All TOMS instruments possessed the same optics, taking 35 samples points with 3° intervals over a perpendicular axis to the orbit of the Earth. Regardless of the altitude of their orbits, each instrument performed a complete global coverage of Earth in a daily routine; excepting Earth Probe TOMS.

85 The final TOMS, Earth Probe TOMS, stopped mapping global ozone in 2006. Since then, the Ozone Monitoring Instrument (OMI) has continued the long-term record of the TOMS instruments.

Table 2.4 Comparison among the TOMS instruments (NASA, 1996b; NASA, 1996a; NASA, 1998b; NASA, 1998a).

TOMS Operational Nominal Field of view Wavelengths instrument years altitude 1 degree latitude 312.34nm, 317.35nm, Nimbus-7 1978 - 1993 935 km by 1.25 degree 331.06nm, 339.66nm, longitude 359.88nm and 379.95nm 1 degree latitude 312.35nm, 317.40m, Meteor-3 1991 - 1994 1050 km by 1.25 degree 331.13nm, 339.73nm, longitude 360.00nm and 380.16nm 1 degree latitude 308.68nm, 312.59nm, ADEOS 1996 - 1997 800 km by 1.25 degree 317.61nm 322.40nm longitude 331.31nm, and 360.11nm 3 degrees latitude 308.65nm, 312.56nm, Earth Probe 1996 - 2006 750 km by 3 degree 317.57nm 322.37nm longitude 331.29nm, and 360.40nm

Global Ozone Monitoring Experiment (GOME)

The GOME spectrometer (Burrows et al., 1999b), supported by the European Space Agency (ESA), was mounted on the European Remote-sensing Satellite - 2 (ERS˗2) and launched in 1995 (it continued operating until 2011). GOME was a four-channel high-resolution spectrometer, whose design was based on a small modified version of the SCIAMACHY instrument, that used a double monochromator with one prism and one holographic diffraction grating for each channel, as well as four photodiode arrays of 1024 pixels.

GOME measured the backscattered Earth irradiance using the nadir observation mode from 240nm to 790nm with a resolution from 0.2nm to 0.4nm (depending on the range of wavelengths) at 785km altitude, achieving global coverage in 3 days after 43 orbits. Although, the main purpose of this instrument was monitoring total column ozone and its vertical profile, the GOME spectrometer was also able to study different tropospheric and stratospheric compounds, such as NO2, BrO, H2O, O4 and O2; as well as OClO at polar regions and SO2 and HCHO under polluted conditions (Burrows et al., 1999b).

Due to the success of the GOME spectrometer recording atmospheric ozone values, a second version of this instrument called GOME-2 (Callies et al., 2000) was launched in 2007 on board the Metop-A satellite, aiming to extend atmospheric measurements up to 2024 on board Metop satellites (Metop-A, Metop-B and Metop-C) (WMO).

86 Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY)

SCIAMACHY (Bovensmann et al., 1999; Burrows et al., 1995), whose design was used to develop the GOME spectrometer, is an eight-channel double instrument able to measure wavelengths from 240nm to 2380nm. It combines a prism, 8 diffraction gratings (one per each channel) and further optics to diffract the incident beam twice and focus the light into each 1024- pixel diode array, allowing a resolution from 0.24nm to 1.5nm, depending on the channel.

Figure 2.62 Chart that shows the operation of SCIAMACHY at different altitudes (Bovensmann et al., 1999).

SCIAMACHY was able to perform nadir, limb and occultation measurements, obtaining more accurate values of tropospheric total column of traces gasses than its predecessors (see Figure 2.62). For this reason, an optimum sequence of these three different measurements was scheduled during each of the 14 orbits per day (Bovensmann et al., 1999), providing a global coverage in 6 days, alternating between nadir and limb measurements, and in three days if it only performed one type of scan.

In 2002, SCIAMACHY instrument was finally launched on the Environmental Satellite (Envisat), which contained other satellite spectrometers monitoring ozone such as the Global Ozone Monitoring by Occultation of Stars (GOMOS) (Bhartia et al., 2013) and the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) (Laeng et al., 2015) plus a range of other instruments to measure other atmospheric compounds. However, ESA announced the end of life of the Envisat in 2012 when communication with the satellite was lost (ESA, 2012).

Ozone Monitoring Instrument (OMI)

OMI (Levelt et al., 2006) is a nadir-spectrometer mounted on board the NASA Aura satellite, which was launched in 2004 and it is still operating. It uses a dual-channel CCD detector to perform measurements of backscattered Earth irradiance on a daily basis from 270nm to 500nm,

87 having one channel for UV measurements (270nm ˗ 365nm), and the other for measurements in the visible spectrum (365nm ˗ 500nm). The resolution and sensitivity achieved by OMI is higher than obtained by its predecessors (TOMS and GOME).

Even though that the OMI spectrometer is mainly dedicated to measuring total column ozone, other trace gasses such as NO2, SO2, HCHO, BrO and OClO as well as volcanic ashes can be measured. Nonetheless, the end of life of the entire satellite is estimated to occur in 2016 (WMO).

88 Chapter 3. Specifications for a New Instrument to Measure Stratospheric Ozone

3.1. General requirements for a ground- based instrument to measure total column ozone

This discussion relates to the measurement of total column ozone on a regular and long-term basis i.e. a monitoring spectrometer. While such an instrument could also be used for short-term research projects, attributes such as stability and robustness are particularly important for a monitoring instrument intended for widespread use. The Global Atmospheric Watch (GAW) is a programme of the WMO, which aims to provide reliable long-term observations of the chemical composition and physical properties of the atmosphere. This programme coordinates most of the total column ozone measurements performed in different locations around the world, and data has been stored since 1961 by the World Ozone and Ultraviolet Data Centre (WOUDC), which is located in Toronto, Canada. However, the global coverage of ground-based ozone observations still has gaps due the large areas in the world that remain without monitoring. This includes, but is not limited to, the oceans. Thus, the importance of developing a reliable and commercial spectrometer able to perform total column ozone measurements simply and cheaply is a relevant topic, as long as it meets all the requirements proposed by WMO GAW in order to report measurements on a daily basis.

The general requirements proposed by GAW WMO programme for a ground-based instrument dedicated to measuring total column ozone are an accuracy of at least ±1%, which is comparable to the existing monitoring systems (Brewer, Dobson and Russian ozonometers); as well as a weatherproof housing able to minimize temperature variations in the instrument (achieving a steady accuracy). If the instrument is manually operated, it is necessary that a trained operator is available to perform and record ozone measurements (as for the Dobson spectrophotometer). On the other hand, if the spectrometer is autonomous, a periodic supervision may still be required for periodic checks and maintenance (as for the Brewer spectrophotometer) in order to provide reliable data (WMO, 2001). The WMO GAW programme suggest a minimum of three direct sun measurements to obtain a daily average of total column of ozone, which should be reported to WOUDC on a monthly basis (Environment Canada, 2013).

89 3.2. Desired improvements to existing systems

In general, an improved spectrometer should offer advantages such as increased sensitivity, lower stray light, low power requirements, less mechanical parts as well as being a robust and portable instrument with less temperature dependence than its predecessors (in this case Dobson and Brewer spectrophotometers). Most of the new ground-based spectrophotometer designs found in the literature have focused on replacing the photomultiplier tube with a CCD, CMOS or photodiode array sensor instead of changing the design or the technology inside the monochromator. This reduces power requirements, makes the instrument more robust and enables it to measure a determined spectral range of wavelengths almost instantaneously (rather than a wavelength at a time). However, by using an array of sensors (pixels) the stray-light rejection is usually reduced and the light-saturation of pixels becomes an important issue, since different pixels receive very different signal levels at the same time because the sensor has to measure wavelengths that are strongly absorbed, weakly absorbed or not absorbed by ozone. Thus, the integration time for the measurement using sensor arrays has to be adjusted to optimise the combined response of all the channels involved in the measurement to avoid saturation.

The main improvement to Dobson and Brewer spectrophotometers suggested in this thesis is the replacement of moving parts in the monochromator, such as the diffraction grating, mirrors and slit mask, for a device able to obtain high speed measurements at one wavelength at a time using solid state technology (see sections 4.2 and 4.3). By measuring one wavelength at a time, it is expected a high stray-light rejection with a resolution at least comparable to existing instruments (Dobson and Brewer spectrophotometers). In conclusion, this instrument is planned to be an autonomous or semi-automatic instrument mainly controlled by a control system (PC), which will provide the possibility of performing the digital data processing in-situ.

The design of the new spectrometer is expected to be weatherproof and temperature stabilized with minimal moving parts, able to perform direct sun measurements with high time resolution. Other practical and commercial advantages over current systems would be low power requirements, small size and weight; and a lower cost compared with Brewer spectrophotometer.

90 Chapter 4. Potential Application of Solid-State Technology

4.1. Laboratory Virtual Instrument Engineering Workbench (LabVIEW)

NI LabVIEW is a programing environment designed by National Instruments (NI), which provides a variety of tools for signal processing, simulating or hardware controlling of devices (usually peripheral products manufactured by NI), such as sensors, actuators, data acquisition instruments, digital to analogue or analogue to digital convertors, multimeters, oscilloscopes, cameras, etc. This software provides sophisticated and powerful tools commonly used in engineering and scientific areas.

Programming in LabVIEW does not require writing commands as C++ or other text based language do. Instead, it focuses on the flow of data, displaying a graphic environment that uses lines to link inputs, outputs and processes to create a block diagram (see Figure 4.1). Although it is not a requirement having previous experience programming, it was an advantage for the author having prior knowledge in structured programming to design more elaborate programs.

Figure 4.1 Visual interface to manipulate the RF controller of the Acousto-Optical Tunable Filter (AOTF), as well as a section of the block diagram using LabVIEW.

LabVIEW 2013 was the version used to design the digital lock-in amplifier and the interface to manipulate the RF controller of the Acousto-Optical Tunable Filter (AOTF). No other additional library or module was needed because all the necessary data processing algorithms were available in this version. However, to improve the performance of the program, some express VI tools were taken advantage of instead of newly developing customised functions. Further improvements to the software are included in section 5.2.

91 4.2. Description of acousto-optical tunable filter (AOTF)

The AOTF is an electronic tunable filter, based on the diffraction of light due to the interaction between acoustic waves and light over an anisotropic crystal. The diffracted narrow band of light (less than 1nm according to Brimrose (2014)) is selected by tuning the frequency of acoustic waves generated by radio frequency (RF) signals. As a result, the AOTF technology is able to filter a chosen wavelength using no moving parts with a fast filtering response (in the order of nanoseconds).

The anisotropic diffraction of light in a birefringent crystal is the core of the AOTF-design. After an unpolarised beam hits the surface of a birefringent material, it splits into ordinary and extraordinary rays. An ordinary ray (o-ray) is optically characterized by having a polarization perpendicular to the optical axis of the crystal, while an extraordinary ray (e-ray) possesses a polarization in the direction of its optical axis. Thus, each ray has different refractive index, whose maximum difference is called the birefringence of the crystal.

The architecture of the AOTF consists of a piezoelectric transducer and an acoustic absorber bonded to the birefringent crystal. Once an RF signal is applied to the transducer, acoustic waves travel through the crystal (see Figure 4.2). Consequently, because of the acoustic waves, refractive index variations are created by compressing and relaxing the crystal, producing the diffraction of the narrow band of light centred at a desired wavelength. Therefore, it is possible to achieve fast wavelength tuning, limited by the time taken of the acoustic wave to cross the crystal, due to the variation of the diffracting properties of the crystal due to the acoustic wave,

Figure 4.2 Diagram of a non-collinear AOTF (TeO2 crystal) with an unpolarized incident beam [Image taken from http://www.brimrose.com/pdfandwordfiles/tunable_filter.pdf]

The material used in the AOTF crystal is selected according to its optical properties, optimum for the desired spectrum of light to analyse. For example, Tellurium Dioxide (TeO2) is usually used for visible and near infrared spectra with wavelengths higher than 360nm while quartz is commonly used in UV and visible applications (Chang, 1977; Young and Shi-Kay, 1981). The

92 tendency of the crystal to vibrate due to the acoustic waves is taken into account to decide the configuration of operation of the AOTF, collinear if it has low acoustic figure of merit and non- collinear otherwise. For example, if the crystal used is TeO2, the usual configuration is non- collinear, where acoustic waves and light travel in perpendicular directions while if the crystal is made of quartz, the common configuration is collinear, where acoustic waves and light propagates in the same direction (see Figure 4.3). In collinear AOTF it is necessary to use a polarizer to split e-rays from o-rays, because they travel on the same direction. On the other hand, in non-collinear AOTF the rays have different directions without it being necessary to use a polarizer. Notice that the input beam was previously polarized in order to avoid unwanted stray light at the output of the AOTF.

Figure 4.3 Diagram of the two AOTF-configurations: A) represents collinear configuration and B) represents the non-colinear configuration with incident polarized light (Kaur and Kaur, 2015).

The most important highlights of the AOTF over a traditional diffraction grating are the solid state technology without moving parts, a fast transition from one wavelength to another, the smaller size and the reduction in power consumption. For this reason, it is the core of the new spectrometer design. Due to the range of wavelengths of operation in UV spectrum, from 250nm to 400nm, the substrate of the used AOTF is quartz which is mounted in a collinear configuration (see Figure 4.4). Notice that the design of the spectrometer uses two UV polarizers, one to polarize the collimated incident beam and other with orthogonal polarization to only let pass through the desired wavelength of light, blocking the zero order polarization.

The heat produced by the acoustic transducer and the acoustic absorber is dispersed by using a water cooling system, whose purpose is to maintain a homogenous temperature inside the AOTF to achieve accurate measurements.

93

Figure 4.4 Collinear configuration of the AOTF used in this thesis (Brimrose, 2014)

The AOTF module used to test the feasibility of use this solid-state technology in UV spectrometers includes an RF controller as well as its default virtual interface, which sends a set of instructions allowing a sequential selection of RF frequencies or specific ones in order to isolate the desired wavelengths one at a time. However, it was necessary to develop a customizable computer interface to fill all the experimental requirements (see Figure 4.5). This interface is in its initial stage, being able to vary the RF frequency (from 80MHz to 165MHz for this particular application) and the RF power (from 0% to 100%). The purpose of varying the RF power is to regulate the intensity of the beam at the exit of the AOTF, allowing an electronic attenuation without using a filter wheel with ND (Neutral Density) filters as traditional spectrophotometers do.

Figure 4.5 Virtual interfaz to control the RF generator.

94 The AOTF is a tested technology used in near infrared measurements, however there is not a documented UV spectrometer based on AOTF with an avalanche photodiode as a detector. The AOTF technology provides attractive features compared with the Brewer spectrophotometer such as shorter time of measurement, a larger optical spectrum of operation (250nm to 400nm) versus the Brewer operating range (286.5nm to 363nm), a smaller, lighter and cheaper architecture with no mechanical adjustments, as well as low power consumption. Although, the exit beam is wavelength dependent, the correct characterization of the AOTF plus efficient algorithms will minimize its impact on the obtained measurements, hoping to achieve an accuracy similar to the Brewer. Thus, the benefits replacing the traditional diffraction grating for the AOTF led to suggest the feasibility of developing a new spectrometer using such technology to perform total column of ozone measurements.

4.3. Avalanche photo-diode (APD)

The APD is a high-sensitivity detector able to transform light into electricity based on the photoelectric effect. Its internal high current gain is achieved by an avalanche multiplication effect (chain reaction) generated by applying a high reverse voltage of a few hundred volts (Dunai et al., 2010), obtaining a higher gain if the applied voltage is increased. This solid-state technology started during the 1960s and early 1970s, when scientists studied alternatives to measure photons (Renker, 2006). Nowadays, the APD detector is used as a cheaper and smaller replacement for the PMT, and according to its configuration can be used either as a linear detector of light or as a single-photon count, depending on the application.

The APD gain is directly related to the reverse bias voltage. However, there is a maximum gain achieved by this device, having non-proportional relation between incident light and the output current if the voltage goes further. After the bias voltage achieves this breakdown value, the APD is able to operate in a single-photon count mode, also called Geiger-mode (Akiba et al., 2010; Cavalcanti et al., 2011; Williams and Huntington, 2011). The gain in this mode is increased and a recovery time without detecting photons is needed, generating intense current pulses at the output. Nevertheless, the work in this thesis is focused in the linear mode of the APD detector, where the reverse voltage bias is below the breakdown voltage and the output signal is proportional to the number of photons.

The material used in the APD is selected according to the range of wavelengths of the beam to be measured, Silicon being the most common material used in APDs (see Figure 4.6), due to its spectral response range from 200nm to 1000nm (Hamamatsu, 2004). Regardless the material used in the APD, the quantum efficiency is higher than the PMT, achieving values of 85%. However, an increased dark noise is generated by gain instabilities and temperature dependence. For these reasons, it is important to have an effective algorithm to reduce noise in the digital signal processing stage.

95

Figure 4.6 Silicon APD detector (image taken from (Hamamatsu, 2004).

The electronic driver of the APD usually consists of a high-voltage source, a current to voltage converter and an amplifier stage. But, because the APD is temperature dependent and the gain decreases while the temperature rises, sophisticated APD drivers adjust the reverse voltage if the temperature changes in order to maintain a constant output. The provisional APD module used in this thesis provides a power supply without automatic adjustments due to temperature variations, as well as a current to voltage converter. Thus, the output of the module was connected directly to an ADC device in order to be processed by the lock-in amplifier algorithm.

The pros of the APD detector, such as high quantum efficiency, high gain, fast operation, voltage supply of hundreds of volts instead of kV, compact size, as well as an accessible price are the main reasons why the APD was selected as a replacement of the traditional PMT in the new spectrometer design.

4.4. Lock-in amplifier

Stray light rejection is a common issue that faces most spectrophotometers measuring solar UV irradiance. Additionally, noise inherent to the photodetector, such as dark current that increases with temperature, affects the accuracy in measurements. The approach to minimize noise of the signal in the proposed new spectrophotometer is digital data processing, which will attenuate the components generated by undesired sources other than the signal of interest. Because of the ability of the AOTF to be electronically modulated, it is possible to apply the technique used in lock-in amplifier instruments, better known as phase sensitive detection.

Phase-sensitive detection is a widely used technique to recover weak signals in a noisy environment, even when the signals are buried in noise hundreds of times larger. This detection is performed by either a dedicated lock-in amplifier instrument or a lock-in amplifier algorithm in a PC. The only requirement is that the signal of interest must be modulated at a known frequency and phase in order to use this signal processing tool.

Applying this technique to optical systems, it is possible to measure the intensity of a desired wavelength coming from a specific source, even when there is a wide spectral range of wavelengths in the background (including the wavelength of interest). First, the incoming beam needs to be modulated either with a mechanical device such as a chopper wheel or with a solid state device, such as an AOTF. Then, if the phase-sensitive detection is performed by a dedicated lock-in amplifier instrument, the modulation signal and the output of the photodetector need to be

96 monitored (see Figure 4.7). Alternatively, those signals may be analysed by a digital data processing algorithm in a computer. According to the complexity of the lock-in amplifier, it may have one or two channels, depending on the number of internal reference signals used. A single- channel instrument needs a manual adjustment of the phase of the modulation signal in order to be synchronized with its internal reference. On the other hand, a dual-channel lock-in amplifier will perform the phase-sensitive detection without knowledge of the phase of the modulation signal (see Figure 4.8).

Figure 4.7 Phase-sensitive detection using a chopper wheel and a lock-in amplifier instrument (PerkinElmer Instruments, 2000).

Incident AOTF Photodetector ADC beam

LabVIEW Modulation interface (PC)

Figure 4.8 Phase-sensitive detection using an AOTF and a double channel lock-in amplifier algorithm in LabVIEW.

However, since the signal from the AOTF is electronically chopped at a customizable frequency, it enables the use of a control system (a computer) to run a lock-in amplifier algorithm using digital data processing tools in LabVIEW.

4.4.1. Lock-in amplifier algorithm

The data acquisition instrument measures the voltage coming from the photodetector. 휔 However, this correlates to a modulated beam of light from the AOTF at a frequency 푓 = , so 2휋 that:

푉𝑖(푡) = 푉𝑖 푠𝑖푛(휔𝑖푡 + 휙) + 푉표푓푓푠푒푡 (4.1)

Where 푉𝑖(푡) represent the input signal, 푉𝑖 is the peak-voltage of the signal, 휔𝑖 is the modulation frequency, t is the time in seconds and 휙 is the phase. This signal has low SNR (signal-to-noise ratio), because the measurements include stray light reaching the photodetector.

97 LabVIEW will simulate two identical sinusoidal signals, but phase shifted 휋, to use as 2 references, 푉푟푒푓1(푡) and 푉푟푒푓2(푡), in the lock-in algorithm.

푉푟푒푓1(푡) = 푉푟푒푓 푠𝑖푛(휔푟푒푓푡 + 휙푟푒푓) (4.2)

휋 푉 (푡) = 푉 푠𝑖푛 (휔 푡 + 휙 + ) 푟푒푓2 푟푒푓 푟푒푓 푟푒푓 2 (4.3)

Where 푉푟푒푓 is the amplitude, 휔푟푒푓 is their modulation frequency and 휙푟푒푓 is the phase of the reference signals.

Then, each reference signal 푉푟푒푓1(푡) and 푉푟푒푓2(푡) multiplies the input signal 푉𝑖(푡). The following analysis considers the same frequency for both the reference signals and modulation (휔 = 휔𝑖 =

휔푟푒푓).

푋(푡) = 푉𝑖푉푟푒푓 푠𝑖푛(휔푡 + 휙) 푠𝑖푛(휔푡 + 휙푟푒푓) + 푉표푓푓푠푒푡푉푟푒푓 푠𝑖푛(휔푡 + 휙푟푒푓) (4.4)

휋 휋 푌(푡) = 푉 푉 푠𝑖푛(휔푡 + 휙) 푠𝑖푛 (휔푡 + 휙 + ) +푉 푉 푠𝑖푛 (휔푡 + 휙 + ) 𝑖 푟푒푓 푟푒푓 2 표푓푓푠푒푡 푟푒푓 푟푒푓 2 (4.5)

Where 푋(푡) is the product of 푉𝑖(푡) and 푉푟푒푓1(푡), and 푌(푡) is the product of 푉𝑖(푡) and 푉푟푒푓2(푡).

The presence of an offset voltage in the input signal will add a sinusoidal component to X(t) and Y(t) with a frequency equal to the modulation frequency (see Figure 4.9).

Input signal without offset voltage, and Vpk=0.1V

Input signal with offset voltage =0.5V, and Vpk=0.1V

Figure 4.9 Comparison of the spectrums 푋(푡) in logarithmic scale and the filtered outputs with and without offset voltage. The cut-off frequency of the low-pass filter is less than 15Hzand aims to let only the DC component pass. The modulation frequency and the reference frequency are in phase and aqual to 50Hz.

98 Although, the lock-in amplifier algorithm can reject the undesired frequency produced by the presence of an offset voltage, a capacitor, acting as a DC filter, will be placed between the photodetector and the data acquisition instrument in order to eliminate the offset (푉표푓푓푠푒푡 = 0)

(see Figure 4.10). Thus, equations (4.4) and (4.5) are simplified by substituting 푉표푓푓푠푒푡 = 0 and applying the trigonometric product identity described in (4.6).

푐표푠(훼 − 훽) − 푐표푠(훼 + 훽) 푠𝑖푛 훼 푠𝑖푛 훽 = 2 (4.6)

푽풊푽풓풆풇 푿(풕) = [풄풐풔(흓 − 흓 ) −풄풐풔(ퟐ흎풕 + 흓 + 흓 )] ퟐ 풓풆풇 풓풆풇 (4.7)

Similarly

푉 푉 휋 휋 푌(푡) = 𝑖 푟푒푓 [푐표푠 (휙 − 휙 − ) − 푐표푠 (2휔푡 + 휙 + 휙 + )] 2 푟푒푓 2 푟푒푓 2 (4.8)

푽풊푽풓풆풇 풀(풕) = [풔풊풏(흓 − 흓 ) +풔풊풏(ퟐ흎풕 + 흓 + 흓 )] ퟐ 풓풆풇 풓풆풇 (4.9)

Note that two components are present in the output signals, one is a DC component (since

휙 − 휙푟푒푓 is constant) and the other is a sinusoidal component with twice the modulation frequency. Therefore, a low-pass filter will be placed after the multiplication stage, in order to let pass through the DC component and attenuate all the frequencies higher than the cut-off frequency of the filter (see Figure 4.10).

푉𝑖 푠𝑖푛(휔푡 + 휙) 푋(푡) Low-pass ADC Photodetector filter

푉 푠𝑖푛(휔푡 + 휙) + 푉 푉푟푒푓 sin(휔푡) 𝑖 표푓푓푠푒푡 Reference signal

휋 shift 2

Low-pass 푉 cos(휔푡) 푌(푡) 푟푒푓 filter

Figure 4.10 Block diagram of the lock-in amplifier detector

Once (4.7) and (4.9) are filtered, only the DC component of the signals will remain:

푉 푉 푋(푡) = 𝑖 푟푒푓 푐표푠(휙 − 휙 ) 2 푟푒푓 (4.10)

푉 푉 푌(푡) = 𝑖 푟푒푓 푠𝑖푛(휙 − 휙 ) 2 푟푒푓 (4.11)

99 Note that 푋(푡) reaches the maximum value and 푌(푡) is equal to zero if the signals are in phase

(휙 − 휙푟푒푓 = 0°), while 푌(푡) reaches the maximum value and 푋(푡) is equal to zero if 휙 − 휙푟푒푓 = 90°.

Solving for the amplitude of the input signal 푉𝑖, the result does not depend on the phase difference between the signal and the reference of the lock-in amplifier.

2 2 2 푉𝑖 = ( ) √푋 (푡) + 푌 (푡) (4.12) 푉푟푒푓

The phase shift is estimated using the next equation:

푌 (휙 − 휙 ) = 푎푟푐 푡푎푛 ( ) 푟푒푓 푋 (4.13)

The advantage of using two references is that it is not necessary to adjust 휙푟푒푓 in order to make the phase difference zero, as in some lock-in amplifiers instruments that only use a single reference. The benefit of this is that it allows the modulation signal to be independent of the lock- in amplifier.

The algorithm described before considered a sinusoidal input signal. However, the modulation performed by the AOTF is almost a square signal. Hence, by considering a periodic square wave as an infinite summation of sinusoidal waves (Fourier series) (see Figure 4.11), the lock-in amplifier algorithm is applied to all the harmonics. However, the low-pass filter will reject all components with frequencies higher than the cut-off frequency of the filter, excepting the DC value obtained from the multiplication of its fundamental harmonic and the reference signals. The only necessary correction is made to the amplitude of the DC output signal, as the peak-voltage (푉푝푘) of a square signal is not the same as the peak-voltage of its sinusoidal fundamental harmonic.

4 ∙ 푉푝푘 푉푝푘 = 푠푞푢푎푟푒 푠𝑖푔푛푎푙 푓𝑖푟푠푡 ℎ푎푟푚표푛𝑖푐 휋 (4.14)

Figure 4.11 Spectrum of X(t) on a logarithmic scale, before being attenuated by the low-pass filter, with square modulation of 50Hz and reference frequency equal to 50Hz (in phase). Right image shows a close-up of the left image.

Note that, if the frequency of the reference is not the same as the modulation frequency (Δ푓 =

|fsignal − fref| ≠ 0), mathematical expressions (4.7) and (4.9) will have two sinusoidal components

100 with frequencies 푓1 = |fsignal − fref| and 푓2 = fsignal + fref without any DC value. In that case, the spectral response of the low-pass filter will remove the component with frequency 푓2, and it may attenuate the amplitude of the sinusoidal component with frequency 푓1 (see Figure 4.9).

Input signal:

Vpk=0.1v Modulation frequency=50Hz Reference signal:

Vpk=1v Modulation frequency=60Hz

Input signal:

Vpk=0.1v Modulation frequency=50Hz Reference signal:

Vpk=1v Modulation frequency=50Hz

Figure 4.12 Comparison of the spectrums X(t) (logarithmic scale) and the filtered outputs (cut-off frequency of the low-pass filter equals 15Hz), where the modulation frequency and the reference frequency are not the same.

푓1 = |fsignal − fref| = 10퐻푧, and 푓2 = fsignal + fref = 110.

The presence of a difference in frequency between the modulation signal and the reference signal will mean that the filtered output signal may not be completely steady and it may present attenuation and/or undesired noise. Thus, it is highly recommended to use identical frequencies in the modulation signal and the reference signal.

4.4.2. Simulation of a dual channel lock-in amplifier using LabVIEW

The aim of the new spectrometer is to extract a small signal at a desired wavelength from an extensive solar spectrum. However, due to the extensive background noise, the digital phase sensitive detection technique is key to achieve the desired accuracy in the measurements. Thus, the feasibility of the current instrument lies both in hardware and software.

This section will focus on the characterization of the digital data processing, exploring the limits of the lock-in amplifier algorithm in a controlled environment, before directly applying the technique to the measurements. Thereby, it would be easier to identify in future experiments if a problem relies on the optical system or the digital data processing.

The program LabVIEW was used to develop a graphic interface to simulate the performance of the phase-sensitive detection technique described before. The input signal of the lock-in amplifier equation (4.1) was generated using a square wave generator with configurable

101 amplitude, frequency, offset voltage and phase, with an option of enabling the addition of white noise with frequencies up to 5kHz, which is the highest white noise frequency achieved by the LabVIEW function. However, a limiting filter was added to the noise generator to block frequencies near the modulation frequency. The reference signals, equations (4.2) and (4.3), were produced by a sine generator, with variable amplitude and adjustable frequency difference with respect the input frequency (Δ푓). Also, a generic Butterworth low-pass filter was used, because of its smooth and flat response.

The purpose of including virtual noise as an available option was to test the performance of the lock-in amplifier algorithm under controlled conditions. Additionally, by plotting inputs and outputs, as well as their spectral response, the program provides a better interpretation of the results obtained. Figure 4.13 shows the graphic interface developed by the author using LabVIEW, estimating the output signal of the lock-in amplifier having a square input signal of 0.1V of amplitude.

Figure 4.13 Graphic interface developed by the Autor using LabVIEW 2003. It illustrates the output signal of the lock-in amplifier, which uses a clean square intput signal with an amplitude of 0.1V (low-pass filter order = 4 and cut-off frequency = 10Hz)

The order and cut-off frequency of the low-pass filter are configured according to the expected level of noise and the desired ripple (see bottom left panel of Figure 4.13). However, it is necessary to consider that a high-order value combined with a low cut-off frequency will increase the ripple in the filter response, requiring more time for the output signal to become steady. For this reason, the algorithm was designed to not consider the values during the selected settling time (see bottom panel of Figure 4.13). Note that its maximum value of the settling time depends on the ratio between the number of samples and the sample frequency.

102 The input signal, the two internal references signals, as well as the output signal are stored in double-precision floating-point 64-bit arrays. The number of elements of each array is determined by the number of samples selected in the upper-right panel of Figure 4.13. Note that the DC output will take into account the entire array or only a portion of the output array, depending on the selected settling time. Currently the simulation uses two techniques to estimate the output DC voltage, and the one that proves the most robust will be used in the final version. The first method uses the tool called DC estimator from LabVIEW (method 1), while the second method estimates an average of the selected output array (method 2). Further work contemplates a deeper analysis of both results to determine which method would be the best option to use.

Having determined the performance of the lock-in amplifier with a clean signal input, noise is then added to the system. Figure 4.14 shows a square-wave signal with an added white noise of amplitude equal to 1V, but excluding the region from 35Hz to 65Hz (±15Hz modulation frequency) to demonstrate the accuracy of the algorithm without frequencies near the modulation frequency. The low-pass filter used is 4th order with a cut-off frequency equal to 6Hz, but these parameters can be modified to either reduce the settling time or decrease noise in the settled signal. The modulation frequency was fixed to 50Hz for all the simulations. However, in practice this frequency would be carefully chosen to avoid expected frequencies of any noise, such as from fluorescent lamps.

Figure 4.14 Display of the output signal of the lock-in amplifier with white noise (amplitude = 1V) 10 greater than the input signal (amplitude = 0.1V), using a low-pass filter order = 4 and a cut-off frequency = 4Hz.

Figure 4.15 shows what happens when the noise is increased to 100V – now 1000 times greater than the underlying signal to be extracted. Note that with the original settings, the output

103 started becoming noisy, requiring a lower cut-off frequency (changed from 6Hz to 2Hz) and a longer settling time (from 0.4 to 0.8 seconds).

Figure 4.15 Display of the output signal of the lock-in amplifier with white noise (amplitude = 100V) 1000 greater than the input signal (amplitude = 0.1V), using a low- pass filter order = 4 and a cut-off frequency = 2Hz.

Note that when the frequency of the noise is close to the modulation frequency, the output DC signal starts to become noisy even if the SNR is not too low. In the extreme case, when the noise frequency includes the modulation frequency, it is not possible to obtain a steady output even with longer integration times. Figure 4.16 shows the output of the lock-in amplifier when noise has amplitude 10 times greater than the original signal and its range of frequencies includes the modulation frequency (the band-stop of the noise generator was disabled for this illustration).

104

Figure 4.16 The graphs shows that the output starts becoming less steady if the noise starts becoming close to the modulation frequency. Amplitude square signal = 0.1V, amplitude noise = 1V, settling time = 0.8 seconds.

There were favourable results performing the simulations. For example, the lock-in amplifier algorithm was able to recover the amplitude of the input signal, which was mixed with white noise of 10 times greater (including noise at the modulation frequency). The output signal was within 0.14% of the known signal using the LabVIEW function to estimate DC, and within 0.76% applying a total average operation over a range of the output array. However, the variations dropped near 0.017% and 0.037% (method 1 and method 2 respectively) with noise up to 1000 times the input signal if there was not noise within ±15Hz from modulation frequency. Further work will test different modulation frequencies in order to determine the most suitable option for this project.

Thus far, after the lock-in amplifier has been designed, the principles of its operation illustrated and its performance simulated, the author considers that this lock-in amplifier algorithm is a feasible option to be used in the proposed spectrometer design.

4.5. Performance of the lock-in amplifier algorithm with real data coming from the APD module APD110A2/M

Once the simulations provided satisfactory results, the LabVIEW program was adapted to acquire measurements using the ADC instrument NI-USB4431(National Instruments, 2009) (see

105 Figure 4.17). This instrument possess a resolution of 24 bits, sampling rates from 1kS/s (samples per second) to 102.4kS/s, and input voltages from ˗10Vpk to +10Vpk. peak to peak. Additionally, it is completely compatible with LabVIEW, since it is manufactured by NI.

Figure 4.17 ADC instrument NI-USB4431 (National Instruments, 2009)

The modulation signal is independently generated by an electronic device. Its output frequency was fixed avoiding 50Hz, 100Hz or any of their multiples, because lamps in the lab are modulated to those frequencies. The ability of the electronic circuit to accurately generate square wave signals is limited to the frequencies listed in Table 4.1, because of the internal oscillator of the circuit. Preliminary tests showed similar results using all frequencies listed in Table 4.1, the modulation frequency chosen for this experimental stage was 80Hz. Note that the modulation frequency is not controlled by the lock-in amplifier interface yet, however, this will be considered in further work.

Table 4.1 List of modulation frequencies generated by the electronic circuit.

Modulation Time period frequency 10 100 ms 16 62.5 ms 20 50 ms 25 40 ms 32 31.25 ms 40 25 ms 80 12.5 ms 125 8 ms 160 6.25 ms

Because the output of the APD sensor is current instead of voltage, the module APD011A2/M (THORLABS, 2011) was initially used since it has an integrated current to voltage convertor. The module includes a UV-enhanced Silicon APD with 1mm of diameter (operating from 200nm to 1000nm) and an ultra-low noise amplifier (see Figure 4.18).

106

Figure 4.18 Avalanche photodetector module APD11A02/M (THORLABS, 2011), as well as its quantum efficiency (Hamamatsu, 2004).

The tested incident beam was generated by the UV LED M365L2 (THORLABS, 2015) with a nominal wavelength near 365nm. Its spectral response was measured using the TE-cooled CCD array spectrometer model BTC112E with subtracted dark-current measurements (see Figure 4.19). Since, the beam from the UV LED is not collimated, a set of fore-optics was used to

collimate the beam and allow variation in the collimated beam width.

Relative Intensity (Logarithmic) Intensity Relative

Wavelength (nm)

Figure 4.19 Measured spectrum of a UV LED with substracted dark-current measurements, using a TE cooled CCD array spectrometer (model BTC112E) (central wavelength is near 365nm).

The AOTF used was the model QZAF-0.25-0.40, which operated in the range of wavelengths from 250 to 400nm with resolutions from 0.4 to 0.8nm depending of the wavelength. Figure 4.20 displays the provisional water cooling system used by the AOTF.

107

Figure 4.20 AOTF with the water cooling system.

The polarizers were place as close as possible to the AOTF to avoid gaps between the optics, diminishing the ratio of stray light coming from sources other than the incident beam with a extinction ratio for the output beam near to 100 000:1 (THORLABS, 2015). The polarizer located between the AOTF and the APD was fixed into a horizontal position, while the other polarizer, mounted in a rotational base, was located between the entrance optics and the AOTF and positioned perpendicular to the first one. Fine rotation adjustments were performed by a graphic interface provided by THORLABS to set the two polarizers perfectly perpendicular to each other.

Another important issue related to the polarizers is their optical alignment along the beam, due to their field of view is wavelength dependent (see Figure 4.21). Further research will fully characterize the optical system, determining if the scattered, unpolarised light that escapes from the polarizer may affect the measurements.

Figure 4.21 Field of view of polarizers (THORLABS, 2015)

Figure 4.22 displays the measurements made by the spectrometer BTC112E, after the incident beam was filtered by the AOTF and the polarizers at 365nm (subtracted dark-current measurements).

108

Relative Intensity (Logarithmic) Intensity Relative

Wavelength (nm)

Figure 4.22 The plot displays the measured spectrum of a UV LED with substracted dark-current measurements, after being filtered by the AOTF. The spectrometer used was a TE cooled CCD array model BTC112E.

Note that the spectrometer model BTC112E was used only to obtain a visual representation of the entire spectrum of the light before and after being filtered by the AOTF. Then, it was replaced by the APD module to perform measurements applying the lock-in amplifier algorithm previously described.

Figure 4.23 displays a general diagram of the optical system used to test the lock-in amplifier algorithm developed in section 4.4.2, and Figure 4.24 is a picture of the real optical system.

Polarizers

UV LED Entrance APD ADC LabVIEW AOTF 365nm optics module NI˗USB4431 interface (PC)

RF controller

Square wave generator (at modulation frequency)

Figure 4.23 Block diagram of the optical system used to test the lock-in amplifier.

109

Figure 4.24 Picture of the optical system used in the experimentation stage.

In order to extract a weak signal from the APD module, there are two main issues to address. The first one is the estimation of the settling time required by the filter to stabilize the output signal, which increases if the cut-off frequency is reduced or if the order of the filter is increased. And the second one is the integration time that the APD detector requires to sense a certain rate of photons, which is strongly related to the modulation frequency.

Further experiments as well as improvements in the LabVIEW program will allow automatic adjustments of the modulation frequency, the settling time and the order and cut-off frequency of the filter. However, in order to determine the performance of the lock-in amplifier in this early stage, the modulation frequency was fixed to 80Hz, the sampling rate to 100kS/s and the settling time was changed manually according to the criteria set by the author.

Figure 4.25 shows the graphic interface developed using LabVIEW, where the left panel is in charge of controlling the parameters of the AOTF (see section 4.2), the upper-right panels displays the output of the acquisition instrument in time and frequency domain, and the bottom- middle panel represent the DC output of the lock-in amplifier algorithm. The AOTF is configured to pass the central wavelength of the UV LED (365nm).

The results obtained in Figure 4.25 show a steady output amplitude of 1.17mV and a background noise with amplitude near 0.5mV. While Figure 4.26 displays a noisy DC output near 158μV and noise near 0.5mV. In both cases the settling time was fixed to 0.8ms, with a 4th low- pass filter order with cut-off frequency equal to 4Hz.

110

Figure 4.25 Graphic interface designed in LabVIEW to manupulate the RF control of the AOTF and visualize the input and output signal of the lock-in amplifier (modulation frequency equal to 80Hz).

Figure 4.26 Left panel of the graphic interface designed in LabVIEW, which displays the input and output signal of the lock-in amplifier when the intensity of the light was reduced using the UV LED driver.

Due to the fact that the APD operates on a broad range of wavelengths (from 200 to 1000nm), the background noise generated by other light sources in the laboratory was also measured by the photodetector. However, the lock-in amplifier algorithm was able to remove most of the noise in order to recover the amplitude of the irradiance from the incident beam at 365nm. Although, the results match with the previous simulations, the LABVIEW program needs some improvements in the low-pass filter stage, which needs to be modified to reduce the settling time of the DC output. Since, a general Butterworth low-pass filter was the only filter tested, further work will analyse others and customize their parameters to improve the rejection of noise.

111 4.6. Outline feasibility study

The AOTF and the APD technology are currently used by experimental spectrometers on- board satellites to estimate atmospheric variables through measuring the infrared spectrum. However, there is no evidence about a spectrometer using them together in the UV range with the same purposes.

During the one-year MPhil in the Faculty of Engineering and Physical Sciences, the AOTF technology and the lock-in amplifier algorithm were tested in order to assess the feasibility of developing a new spectrometer based on such technologies. The results obtained suggest that such an instrument should be capable of performing reliable measurements of total column of ozone, and able to achieve similar accuracy to current Brewer spectrometers with advantages of using fewer moving parts, lower temperature dependence, reduced weight and size, lower price and lower power consumption. However, in order to confirm these assumptions, it will be necessary to develop a first prototype and compare it with the standard Brewer spectrophotometer to corroborate the feasibility of the future widespread use of this solid-state technology for measuring ozone.

Once the first spectrometer prototype is completed, an intercomparison, with more than one spectrophotometer, should be performed preferably at the RBCCE Intercomparison campaign in Spain, where the new prototype will perform measurements from the direct solar beam in ideal clear conditions. Further analysis of its field performance will be undertaken to aid with future improvements and provide additional characterization of the instrument.

112 Chapter 5. Conclusion and further work

5.1. Conclusion

The work described in this thesis has provided a general description of the technology to be implemented in a spectrometer to measure the total column ozone. Meeting the minimum requirements discussed in Chapter 3, the new spectrophotometer aims to achieve an accuracy comparable to actual Brewer double spectrophotometer, plus improvements in the measuring time, the temperature dependence, the number of moving parts, the power consumption, as well as reductions in size, weight and production costs.

Currently, there is no a documented spectrometer whose design includes an AOTF, an APD and a lock-in amplifier to measure UV spectrum. Thus, a cost-effective spectrometer with all the advantages of solid-state technology will be well received as long as the accuracy and reliability of its measurements would be comparable to those obtained with the Brewer spectrometer.

The benefits of replacing the diffraction grating with the AOTF are the absence of concerns about the alignment of the equipment due to the contraction or expansion of all the mechanical parts, as well as a faster response of the device selecting a desired wavelength from a wider spectrum (250nm to 400nm). A PMT is certainly one of the best photon detectors, nonetheless, some of its weaknesses such as having mechanical parts, using high voltage (typically over one thousand volts), its size, its elevated cost, having low quantum efficiency and being easily damaged were the reasons to choose a replacement. Therefore, the APD was the most suitable device to use in the spectrometer, because of its simplicity, high quantum efficiency, internal gain, faster response (1MHz up to 100MHz), a lower power consumption and its ability to work in linear mode or photon count mode. However, further analysis will compare the performance of the PMT with the APD in conjunction with the digital data processing, in order to obtain a detailed analysis.

The core of the signal processing is the lock-in amplifier technique, which extracts the amplitude of the signal of interest from the noisy one obtained at the output of the APD detector. After the digital filtering, performed by the lock-in amplifier algorithm, the output DC signal is interpreted as the ratio of UV radiation at a selected wavelength. A complete simulation of the lock-in amplifier to evaluate its performance showed encouraging results that were applied to real signals from the APD detector (see Chapter 4). However, due to the fore-optics were not fixed, the irradiance of the incident beam (UV-LED) reaching the first polarizer and the irradiance of the beam at the APD detector were not measured and compare by an external photo-detector. Further measurements in the lab will also include a complete characterization of the light source.

The stray light rejection of the spectrometer is directly proportional to the efficiency of the polarizers. Thus, the importance of choosing the correct ones is crucial to achieve high stray-light rejection. On the other hand, the features and configuration of the APD will determine the noise

113 rejection of the new spectrometer. Thus, a complete analysis of those components is planned to be performed on the first year of the PhD, in order to accomplish a successful and complete comparison with the Brewer spectrophotometer, taking into account the temperature dependence and seasonality of the measurements.

The promising results of this thesis, obtained during this early phase of experimentation (Chapter 4), led to suggest the feasibility to proceed to the next stage of this project, which is building a prototype and perform field measurements. This new stage is planned to be continued by the author during the PhD, where more accurate specifications of the spectrometer will be detailed, being aware the expectations, limitations and improvements of the new device compared with actual instruments.

5.2. Further work

5.2.1. Optical system of the spectrometer

The experiments showed in this thesis only uses a UV LED as a light source. However, it will be necessary to modify the system to allow performing measurements with different light sources, such as lasers, mercury lamps, UV lamps, as well as test the optical system using a solar simulator, to characterize the optical parts of the spectrometer and the spectrometer as a whole. By adapting the system to use optic fibre, interchanging light sources will be simpler in further experiments. Nonetheless, other options must be evaluated to solve this issue.

The system also needs to be optimized against stray light. For this reason, a bandpass filter will be used within the fore-optics to reject wavelengths higher than 400nm (maximum wavelength operated by the AOTF). The AOTF housing will be modified to mount the polarizers as close as possible to the AOTF. Thus, all the optical system will be compacted, avoiding gaps between each optical element. Currently, it was used a water-cooling system to stabilize the temperature in the AOTF. However, it is necessary to analyse other options such as thermo-gel, a Peltier device or a compact water-cooling system. Additionally, the new spectrometer prototype will be cover by a customized enclosure to minimize stray light coming from external sources others than the incident beam.

Once the system has been optimized, it will be necessary to perform a characterization the prototype. First, the extinction coefficient of the polarizers has to be evaluated at different wavelengths by removing the AOTF, perpendicularly mounting each polarizer without any gap, and measuring the amount of irradiance that reaches the photodetector. The light that was not blocked by the polarizers has to be considered apart the dark current of the photodetector, which is temperature dependence. The use of a rotating third polarizer to attenuate the incident beam will be also analysed because in conjunction to the RF controller, which modifies the attenuation of the AOTF by changes in the amplitude of the RF signal, they can attenuate the incident beam, avoiding the use of neutral density filters.

114 The characterization of the RF controller of the AOTF will ensure that the frequency and amplitude displayed in the interface corresponds to the generated RF signal. Then, by using lasers at different wavelengths as well as a mercury lamp, the RF controller of the AOTF will be calibrated to have a relation between wavelengths and radio frequencies.

The output of the APD module has to be calibrated by comparing this output voltage with the irradiance measured by a reference spectrometer. However, it will be also analysed the option of use the APD as a photon counting device (Williams and Huntington, 2011). The temperature dependence of the detector must be careful characterized, because as many other detectors, such as CCDs, PDAs, or photomultiplier tubes, the accuracy in measurements relies on stabilizing the temperature of the detector.

Once the designed instrument is temperature stabilized, characterized and tested in the laboratory with artificial light, obtaining the FWHM of the optical system at the wavelengths of interest, it will be necessary to calibrate the new instrument against a calibrated and well maintained reference spectrophotometer during field tests. It is planned to use the measurements of the Brewer #172, which is located at University of Manchester, as a reference to calibrating the new instrument and performing long-term stability tests.

Figure 5.1 shows a proposed spectrometer prototype to be developed during the PhD, which includes a case and a solar tracker. The first prototype should separate the fore optics and the solar tracker from the AOTF and the APD, in order to achieve better temperature stabilization over such devices. This first prototype will include the selection of the most suitable type of cable to connect the ADC and the PC as well as choosing an appropriate protocol of communications. The diagram of Figure 5.1 is not definitive, due to it must be tested first. Thus, further improvements will be made to this initial design in hardware and software.

Fixed polarizers

AOTF Incident Fore optics APD

(over solar (temperature ADC

beam module tracker) stabilized) Optical Fiber

Rotating polarizer

LabVIEW RF controller interface (PC) Serial cable Square wave generator (lock-in amplifier frequency)

Figure 5.1 Block diagram of the proposed design of the spectrometer prototype.

5.2.2. Improvements to the LabVIEW program

Some improvements to the graphic interface of the spectrometer are pending, such as using customized functions, simplifying the processes in the program, estimating the offset and the dark

115 current of the spectrometer and obtaining a calibration file to change from frequency to wavelength domain. Also, it is necessary to test different digital filters and select the one with the best response, as well as determining the optimum modulation frequency of the AOTF, which will be least affected by environmental noise.

In addition, the intensity of the input beam at the entrance of the APD detector must be regulated either by attenuating the signal in the AOTF, or by controlling a third polarizer.

Finally, before starting the field measurements, the calibration file must include the slit function of the instrument, the ozone coefficients, offset and dark current corrections, as well as the adjustments to the photon counting. Noticing that adjustments to the file will be necessary during all the testing process to obtain a stable and accurate spectrometer.

116 References

ACE 2007. BS ISO 21348:2007 - Space environment (natural and artificial). Process for determining solar irradiances. AKIBA, M., TSUJINO, K. & SASAKI, M. 2010. Ultrahigh-sensitivity single-photon detection with linear-mode silicon avalanche photodiode. Optics Letters, 35, 2621-2623. ÅNGSTRÖM, A. 1964. The parameters of atmospheric turbidity. Tellus, 16, 64-75. BAIS, A. F., ZEREFOS, C. S. & MCELROY, C. T. 1996. Solar UVB measurements with the double- and single-monochromator Brewer ozone spectrophotometers. Geophysical Research Letters, 23, 833-836. BASS, A. M. & PAUR, R. J. 1985. The ultraviolet cross-sections of ozone, Springer Netherlands. BERNHARD, G., EVANS, R. D., LABOW, G. J. & OLTMANS, S. J. 2005. Bias in Dobson total ozone measurements at high latitudes due to approximations in calculations of ozone absorption coefficients and air mass. Journal of Geophysical Research-Atmospheres, 110. BERTAUX, J.-L., KORABLEV, O., PERRIER, S., QUÉMERAIS, E., MONTMESSIN, F., LEBLANC, F., LEBONNOIS, S., RANNOU, P., LEFÈVRE, F., FORGET, F., FEDOROVA, A., DIMARELLIS, E., REBERAC, A., FONTEYN, D., CHAUFRAY, J. Y. & GUIBERT, S. 2006. SPICAM on Mars Express: Observing modes and overview of UV spectrometer data and scientific results. Journal of Geophysical Research: Planets, 111, n/a-n/a. BHARTIA, P. K., MCPETERS, R. D., FLYNN, L. E., TAYLOR, S., KRAMAROVA, N. A., FRITH, S., FISHER, B. & DELAND, M. 2013. Solar Backscatter UV (SBUV) total ozone and profile algorithm. Atmospheric Measurement Techniques, 6, 2533-2548. BHATTACHARYA, A. B., JOARDAR, S. & BHATTACHARYA, R. 2008. Astronomy and astrophysics. BOVENSMANN, H., BURROWS, J. P., BUCHWITZ, M., FRERICK, J., NOEL, S., ROZANOV, V. V., CHANCE, K. V. & GOEDE, A. P. H. 1999. SCIAMACHY: Mission objectives and measurement modes. Journal of the Atmospheric Sciences, 56, 127-150. BREWER, A. W. 1973. A replacement for the Dobson spectrophotometer? Pure and Applied Geophysics, 106-108, 919-927. BREWER, A. W., MCELROY, C. T. & KERR, J. B. 1973. Nitrogen Dioxide Concentrations in the Atmosphere. Nature, 246, 129-133. BRIMROSE 2014. Manual AOTF. BRION, J., CHAKIR, A., DAUMONT, D., MALICET, J. & PARISSE, C. 1993. High-resolution laboratory absorption cross-section of O3 - Temperature effect. Chemical Physics Letters, 213, 610- 612. BRÖNNIMANN, S., CAIN, J. C., STAEHELIN, J. & FARMER, S. F. G. 2003. Total ozone observations prior to the IGY. II: Data and quality. Quarterly Journal of the Royal Meteorological Society, 129, 2819-2843. BUCHOLTZ, A. 1995. Rayleigh-scattering calculations for the terrestrial atmosphere. Applied Optics, 34, 2765-2773. BURROWS, J. P., HÖLZLE, E., GOEDE, A. P. H., VISSER, H. & FRICKE, W. 1995. SCIAMACHY— scanning imaging absorption spectrometer for atmospheric chartography. Acta Astronautica, 35, 445-451. BURROWS, J. P., RICHTER, A., DEHN, A., DETERS, B., HIMMELMANN, S., VOIGT, S. & ORPHAL, J. 1999a. Atmospheric remote-sensing reference data from GOME—2. Temperature- dependent absorption cross-sections of O3 in the 231–794nm range. Journal of Quantitative Spectroscopy and Radiative Transfer, 61, 509-517. BURROWS, J. P., WEBER, M., BUCHWITZ, M., ROZANOV, V., LADSTATTER-WEISSENMAYER, A., RICHTER, A., DEBEEK, R., HOOGEN, R., BRAMSTEDT, K., EICHMANN, K. U. & EISINGER, M. 1999b. The global ozone monitoring experiment (GOME): Mission concept and first scientific results. Journal of the Atmospheric Sciences, 56, 151-175.

117 BURROWS, J. P. E. 2011. The Remote Sensing of Tropospheric Composition from Space / edited by John P. Burrows, Peter Borrell, Ulrich Platt, Berlin, Heidelberg : Springer Berlin Heidelberg : Imprint: Springer. CALLIES, J., CORPACCIOLI, E., EISINGER, M., HAHNE, A. & LEFEBVRE, A. 2000. GOME-2 - Metop's second-generation sensor for operational ozone monitoring. Esa Bulletin-European Space Agency, 28-36. CAVALCANTI, M. D. S., MENDONÇA, F. A. & RAMOS, R. V. 2011. Spectral method for characterization of avalanche photodiode working as single-photon detector. Optics Letters, 36, 3446-3448. COOPER, P. I. 1969. The absorption of radiation in solar stills. Solar Energy, 12, 333-346. CORNU, A. 1881. Sur l'absorption atmosphérique des radiations ultra-violettes. J. Phys. Theor. Appl., 10, 5-17. CZERNY, M. & TURNER, A. F. 1930. Über den Astigmatismus bei Spiegelspektrometern. Zeitschrift für Physik, 61, 792-797. CHANG, I. C. 1977. Tunable acousto-optic filters: An overview. Optical Engineering, 16, 455-460. CHAPMAN, S. 1930. On ozone and atomic oxygen in the upper atmosphere. Philosophical Magazine, 10, 369-383. DAUMONT, D., BRION, J., CHARBONNIER, J. & MALICET, J. 1992. Ozone UV spectroscopy .1. Absorption cross-sections at room-temperature. Journal of Atmospheric Chemistry, 15, 145-155. DOBSON, G. M. B. 1931. A photoelectric spectrophotometer for measuring the amount of atmospheric ozone. Proceedings of the Physical Society, 43, 324-339. DOBSON, G. M. B. 1968. 40 years research on atmospheric ozone at oxford - A history. Applied Optics, 7, 387-&. DOUGLASS, A. R., NEWMAN, P. A. & SOLOMON, S. 2014. The Antarctic ozone hole: An update. Physics Today, 67, 42-48. DUFFIE, J. A. 1980. Solar engineering of thermal processes / John A. Duffie, William A. Beckman, New York, Chichester, Wiley. DUNAI, D., ZOLETNIK, S., SÁRKÖZI, J. & FIELD, A. R. 2010. Avalanche photodiode based detector for beam emission spectroscopy. Review of Scientific Instruments, 81, 103503. EBERT, H. 1889. H. Ebert: Wied. Ann. 38, 489. ENVIRONMENT CANADA 2013. User Guide to the WMO/GAW World Ozone Data Centre. Canada. ESA. 2012. Envisat [Online]. Available: http://m.esa.int/Our_Activities/Observing_the_Earth/Envisat/ESA_declares_end_of_ mission_for_Envisat [Accessed February 26 2016]. ET ENTERPRISES LIMITED 2011. Undestanding Photomultipliers. EUROPEAN COOPERATION IN SCIENCE AND TECHNOLOGY. 2005. A European Brewer Network (EUBREWNET) [Online]. Available: http://www.cost.eu/COST_Actions/essem/ES1207 [Accessed February 16,2016. FABRY, C. & BUISSON, H. 1913. L'absorption de l'ultra-violet par l'ozone et la limite du spectre solaire. J. Phys. Theor. Appl., 3, 196-206. FABRY, C. & BUISSON, H. 1921. A study of the ultra-violet end of the solar spectrum. Astrophysical Journal, 54, 297-322. FARMAN, J. C., GARDINER, B. G. & SHANKLIN, J. D. 1985. Large losses of total ozone in Antarctica reveal seasonal ClOX/NOX interaction. Nature, 315, 207-210. FASTIE, W. G. 1952. A Small Plane Grating Monochromator. Journal of the Optical Society of America, 42, 641-647. FASTIE, W. G. 1991. Ebert spectrometer reflections. Phys. Today, 44, 37-43. FIOLETOV, V. E., KERR, J. B., HARE, E. W., LABOW, G. J. & MCPETERS, R. D. 1999. An assessment of the world ground-based total ozone network performance from the comparison with satellite data. Journal of Geophysical Research: Atmospheres, 104, 1737-1747.

118 FIOLETOV, V. E., KERR, J. B., MCELROY, C. T., WARDLE, D. I., SAVASTIOUK, V. & GRAJNAR, T. S. 2005. The Brewer reference triad. Geophysical Research Letters, 32, n/a-n/a. FIOLETOV, V. E., KERR, J. B., WARDLE, D. I., KROTKOV, N. & HERMAN, J. R. 2002. Comparison of Brewer ultraviolet irradiance measurements with total ozone mapping spectrometer satellite retrievals. Optical Engineering, 41, 3051-3061. GREEN, A. C., BARTLETT, L. M. & VAUGHAN, G. 2001. SAOZ measurements of total ozone at Aberystwyth. Journal of Quantitative Spectroscopy and Radiative Transfer, 69, 231-243. GRÖBNER, J., KOUREMETI, N., SODER, R., WASSER, D., COULON, E. D., GYO, M., DÜRIG, F. & COULON, E. D. 2014. A precision solar spectroradiometer for spectral aerosol optical depth and solar irradiance measurements. UVnet Workshop 2014, Davos, Switzerland, PMOD/WRC. GRÖBNER, J., WARDLE, D. I., MCELROY, C. T. & KERR, J. B. 1998. Investigation of the wavelength accuracy of Brewer spectrophotometers. Applied Optics, 37, 8352-8360. GUSHCHIN, G. P., SOKOLENKO, S. A. & KOVALYOV, V. A. 1985. Total-Ozone Measuring Instruments Used at the USSR Station Network. In: ZEREFOS, C. S. & GHAZI, A. (eds.) Atmospheric Ozone: Proceedings of the Quadrennial Ozone Symposium held in Halkidiki, Greece 3–7 September 1984. Dordrecht: Springer Netherlands. HAMAMATSU 2001. Photon counting. Using Photomultiplier Tubes. HAMAMATSU 2004. Si APD S5343 to S5345, S907 to S975. HAMAMATSU PHOTONICS 2006. Photomultipliers tubes - Basic applications. Third Edition ed. HARTLEY, W. N. 1881. XXI.—On the absorption of solar rays by atmospheric ozone. Journal of the Chemical Society, Transactions, 39, 111-128. HAUTEFEUILLE, M. P. & CHAPPUIS, J. 1880. On the liquefaction of ozone, and on its color in the gaseous state. Journal of the Franklin Institute, 110, 345-347. HENDRICK, F., POMMEREAU, J. P., GOUTAIL, F., EVANS, R. D., IONOV, D., PAZMINO, A., KYRO, E., HELD, G., ERIKSEN, P., DOROKHOV, V., GIL, M. & VAN ROOZENDAEL, M. 2011. NDACC/SAOZ UV-visible total ozone measurements: improved retrieval and comparison with correlative ground-based and satellite observations. Atmospheric Chemistry and Physics, 11, 5975-5995. HERMAN, J., ABUHASSAN, R., EVANS, A., PETROPAVLOVSKIKH, N., MCCONVILLE, I. & CEDE, G. 2015. Comparison of ozone retrievals from the Pandora spectrometer system and Dobson spectrophotometer in Boulder, Colorado. Atmospheric Measurement Techniques, 8, 3407-3418. HERMAN, J., CEDE, A., SPINEI, E., MOUNT, G., TZORTZIOU, M. & ABUHASSAN, N. 2009. NO2 column amounts from ground-based Pandora and MFDOAS spectrometers using the direct-sun DOAS technique: Intercomparisons and application to OMI validation. Journal of Geophysical Research-Atmospheres, 114. HUGGINS, W. 1890. On a New Group of Lines in the Photographic Spectrum of Sirius. Proceedings of the Royal Society of London, 48, 216-217. INN, E. C. Y. & TANAKA, Y. 1953. Absorption Coefficient of Ozone in the Ultraviolet and Visible Regions. Journal of the Optical Society of America, 43, 870-873. JOSEFSSON, W. A. P. 1986. Solar ultraviolet radiation in Sweden. Sweden: SMHI. JOSEFSSON, W. A. P. 1992. Focused sun observations using a Brewer ozone spectrophotometer. Journal of Geophysical Research: Atmospheres, 97, 15813-15817. KAUR, P. & KAUR, S. 2015. Acousto Optical Tunable Filters. IJECT - International Journal of Electronics and Communication Technology, 6. KAYSER, H. 1910. Handbuch der Spectroscopie / von H. Kayser. 5.Band : mit 2 Tafeln und 3 Figuren, Leipzig, Leipzig : S. Hirzel. KERR, J. B. 2002. New methodology for deriving total ozone and other atmospheric variables from Brewer spectrophotometer direct sun spectra. Journal of Geophysical Research: Atmospheres, 107, ACH 22-1-ACH 22-17.

119 KERR, J. B. 2010. The Brewer Spectrophotometer. In: GAO, W., SLUSSER, J. R. & SCHMOLDT, D. L. (eds.) UV Radiation in Global Climate Change: Measurements, Modeling and Effects on Ecosystems. Berlin, Heidelberg: Springer Berlin Heidelberg. KERR, J. B., EVANS, W. F. J. & ASBRIDGE, I. A. 1985. Recalibration of Dobson field spectrophotometers with a traveling Brewer spectrophotometer standard. KERR, J. B., MCELROY, C. T., WARDLE, D. I., OLAFSON, R. A. & EVANS, W. F. J. 1981. The Automated Brewer Spectrophotometer. In: ZEREFOS, C. S. & GHAZI, A. (eds.) Atmospheric Ozone: Proceedings of the Quadrennial Ozone Symposium held in Halkidiki, Greece 3–7 September 1984. Dordrecht: Springer Netherlands. KIPP & ZONEN 2014. Brewer MkIII photometer - Operator's Manual for Single Board. Revision E ed. KOMHYR, W. D. 1980. Operations handbook - Ozone observations with a Dobson spectrophotometer. In: 1469, W. T.-N. (ed.). KOMHYR, W. D. & GRASS, R. D. 1972. Dobson Ozone Spectrometer Modification. Journal of Applied Meteorology, 11, 858-863. KOMHYR, W. D., GRASS, R. D. & LEONARD, R. K. 1989. Dobson spectrophotometer-83 - A standard for total ozone measurements, 1962-1987. Journal of Geophysical Research- Atmospheres, 94, 9847-9861. KOMHYR, W. D., MATEER, C. L. & HUDSON, R. D. 1993. Effective Bass-Paur 1985 ozone absorption coefficients for use with Dobson ozone spectrophotometers. Journal of Geophysical Research: Atmospheres, 98, 20451-20465. KONDRATYEV, K. Y. 1969. Radiation in the atmosphere. KORABLEV, O., BERTAUX, J.-L., FEDOROVA, A., FONTEYN, D., STEPANOV, A., KALINNIKOV, Y., KISELEV, A., GRIGORIEV, A., JEGOULEV, V., PERRIER, S., DIMARELLIS, E., DUBOIS, J. P., REBERAC, A., VAN RANSBEECK, E., GONDET, B., MONTMESSIN, F. & RODIN, A. 2006. SPICAM IR acousto-optic spectrometer experiment on Mars Express. Journal of Geophysical Research: Planets, 111, n/a-n/a. KORABLEV, O., BERTAUX, J. L., GRIGORIEV, A., DIMARELLIS, E., KALINNIKOV, Y., RODIN, A., MULLER, C. & FONTEYN, D. 2002. An AOTF-based spectrometer for the studies of Mars atmosphere for Mars Express ESA mission. Advances in Space Research, 29, 143-150. KOUREMETI, N., BAIS, A. F., BALIS, D. & ZYRICHIDOU, I. 2013. Phaethon: A System for the Validation of Satellite Derived Atmospheric Columns of Trace Gases. In: HELMIS, G. C. & NASTOS, T. P. (eds.) Advances in Meteorology, Climatology and Atmospheric Physics. Berlin, Heidelberg: Springer Berlin Heidelberg. KYRÖ, E. 1993. Intercomparison of total ozone data from Nimbus 7 TOMS, the Brewer UV Spectrophotometer and SAOZ UV-Visible Spectrophotometer at High Latitudes Observatory, Sodankylä. Geophysical Research Letters, 20, 571-574. LAENG, A., HUBERT, D., VERHOELST, T., VON CLARMANN, T., DINELLI, B. M., DUDHIA, A., RASPOLLINI, P., STILLER, G., GRABOWSKI, U., KEPPENS, A., KIEFER, M., SOFIEVA, V., FROIDEVAUX, L., WALKER, K. A., LAMBERT, J. C. & ZEHNER, C. 2015. The ozone climate change initiative: Comparison of four Level-2 processors for the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS). Remote Sensing of Environment, 162, 316-343. LAKKALA, K., KYR, J., AROLA, G., HEIKKIL, A., KAUROLA, A., KOSKELA, J., LINDFORS, T., MEINANDER, E., TANSKANEN, A., GR?BNER, O. & H?LSEN, A. 2008. Quality assurance of the Brewer spectral UV measurements in Finland. Atmospheric Chemistry and Physics, 8, 3369-3383. LAMB, K. 2015. Calibration Report on Brewers #172, #126 and #075 in Spain, May 27 - June 4, 2015. LAMM, L. O. 1981. A new analytic expression for the equation of time. Solar Energy, 26, 465. LEVELT, P. F., VAN DEN OORD, G. H. J., DOBBER, M. R., MALKKI, A., HUIB, V., DE VRIES, J., STAMMES, P., LUNDELL, J. O. V. & SAARI, H. 2006. The ozone monitoring instrument. Geoscience and Remote Sensing, IEEE Transactions on, 44, 1093-1101.

120 MALICET, J., DAUMONT, D., CHARBONNIER, J., PARISSE, C., CHAKIR, A. & BRION, J. 1995. Ozone UV spectroscopy. II. Absorption cross-sections and temperature dependence. Journal of Atmospheric Chemistry, 21, 263-273. MARENCO, F., DI SARRA, A. & DE LUISI, J. 2002. Methodology for determining aerosol optical depth from Brewer 300-320-nm ozone measurements. Applied Optics, 41, 1805-1814. MATSUMURA, Y. & ANANTHASWAMY, H. N. 2004. Toxic effects of ultraviolet radiation on the skin. Toxicology and Applied Pharmacology, 195, 298-308. MCARTHUR, L. J. B., FIOLETOV, V. E., KERR, J. B., MCELROY, C. T. & WARDLE, D. I. 1999. Derivation of UV-A irradiance from pyranometer measurements. Journal of Geophysical Research- Atmospheres, 104, 30139-30151. MIELENZ, K. D. 1964. Theory of mirror spectographs .3. Focal surfaces + slit curvature of Ebert + Ebert-Fastie spectographs. Journal of Research of the National Bureau of Standards Section C-Engineering and Instrumentation, C 68, 205-+. MIYAZAKI, K., IWASAKI, T., SHIBATA, K. & DEUSHI, M. 2005. Roles of transport in the seasonal variation of the total ozone amount. Journal of Geophysical Research-Atmospheres, 110. MOLINA, L. & MOLINA, M. 1986. Absolute absorption cross-sections of ozone in the 185- to 350- nm wavelength range. Journal of Geophysical Research, 91, 14501-14508. MOLINA, M. J. & ROWLAND, F. S. 1974. Stratospheric sink for chlorofluoromethanes: chlorine atomic-catalysed destruction of ozone. Nature, 249, 810-812. MORISON, W. L. 1989. Effects of ultraviolet-radiation on the immune-system in humans. Photochemistry and Photobiology, 50, 515-524. MURTY, M. V. 1974. Theory and Principles of Monochromators, Spectrometers and Spectrographs. Optical Engineering, 13, 130123-130123-. NASA. Ozone Hole Watch [Online]. Available: http://ozonewatch.gsfc.nasa.gov/. NASA 1996a. Meteor–3 Total Ozone Mapping Spectrometer (TOMS) Data Products User's Guide. NASA 1996b. Nimbus–7 Total Ozone Mapping Spectrometer (TOMS) Data Products User's Guide. NASA 1998a. ADEOS Total Ozone Mapping Spectrometer (TOMS) Data Products User's Guide. NASA 1998b. Earth Probe Total Ozone Mapping Spectrometer (TOMS) Data Products User's Guide. NATIONAL INSTRUMENTS 2009. Sound and Vibration Data Acquisition NI-USB4431. PALMER, C. A. & LOEWEN, E. G. 2005. Diffraction grating handbook, NEWPORT CORPORATION. PAULESCU, M. 2013. Weather Modeling and Forecasting of PV Systems Operation / by Marius Paulescu, Eugenia Paulescu, Paul Gravila, Viorel Badescu, London : Springer London : Imprint: Springer. PERKINELMER INSTRUMENTS 2000. Low level optical detection using lock-in amplifier techniques. PHOTONIS 2002. PHOTOMULTIPLIER TUBES principles & applications. Brive, France. PHYSIKALISCH-METEOROLOGISCHES OBSERVATORIUM UND WELTSTRAHLUNGSZENTRUM DAVOS 2014. Annual Report. Davos Dorfo: World Radiation Center. PLATT, U. 2008. Differential Optical Absorption Spectroscopy Principles and Applications / by Ulrich Platt, Jochen Stutz, Berlin, Heidelberg : Springer Berlin Heidelberg. R. & J. BECK LTD Instructions for use of dr. Dobson's spectrophotometer. London. RAVANAT, J. L., DOUKI, T. & CADET, J. 2001. Direct and indirect effects of UV radiation on DNA and its components. Journal of Photochemistry and Photobiology B-Biology, 63, 88-102. REDONDAS, A., EVANS, R., STUEBI, R., HLER, U. K. & WEBER, M. 2014. Evaluation of the use of five laboratory-determined ozone absorption cross-sections in Brewer and Dobson retrieval algorithms. Atmospheric Chemistry and Physics, 14, 1635-1648. REDONDAS, A. & RODRIGUEZ-FRANCO, J. 2014. Ninth Intercomparison Campaign of the Regional Brewer Calibration Center Europe (RBCC-E). Izaña Atmospheric Research Center, AEMET, Tenerife, Canary Islands, Spain: WMO/GAW. RENKER, D. 2006. Geiger-mode avalanche photodiodes, history, properties and problems. Nuclear Inst. and Methods in Physics Research, A, 567, 48-56.

121 SAVASTIOUK, V. & MCELROY, C. T. 2005. Brewer spectrophotometer total ozone measurements made during the 1998 Middle Atmosphere Nitrogen Trend Assessment (MANTRA) campaign. Atmosphere-Ocean, 43, 315-324. SCI-TEC INSTRUMENTS INC 1999. Brewer MkIV spectrophometer - Operator's Manual SCHARMER, K. & GREIF, J. 2000. The European Radiation Atlas. Vol. 1: Fundamentals and maps. SERDYUCHENKO, A., GORSHELEV, V., WEBER, M., CHEHADE, W. & BURROWS, J. P. 2014. High spectral resolution ozone absorption cross-sections - Part 2: Temperature dependence. Atmospheric Measurement Techniques, 7, 625-636. SHARPE, M. R. 1984. Stray light in UV-VIS spectrophotometers. Analytical Chemistry, 56, A339- &. SMEDLEY, A. R. D., KIFT, R. C. & WEBB, A. R. 2015. Assessment of a Dual-Channel Array Spectrometer for Ground-Based Ozone Retrievals. Journal of Atmospheric and Oceanic Technology, 32, 1464-1477. SOLOMON, K. R. 2008. Effects of ozone depletion and UV-B radiation on humans and the environment. Atmosphere-Ocean, 46, 185-202. STAEHELIN, J., KERR, J., EVANS, R. & VANICEK, K. 2003. Comparison of total ozone measurements of Dobson and Brewer Spectrophotometers and recommended transfer functions. World Meteorological Organization. Gobal Atmosphere Watch, 149. STINE, W. B. & GEYER, M. 2001. Power from the sun. THOMASON, L. W., HERMAN, B. M. & REAGAN, J. A. 1983. The effect of atmospheric attenuators with structured vertical distributions on air-mass determinations and Langley plot analyses. Journal of the Atmospheric Sciences, 40, 1851-1854. THOMPSON, A., EARLY, E. A., DELUISI, J., DISTERHOFT, P., WARDLE, D., KERR, J., RIVES, J., SUN, Y. C., LUCAS, T., MESTECHKINA, T. & NEALE, P. 1997. The 1994 North American interagency intercomparison of ultraviolet monitoring spectroradiometers. Journal of Research of the National Institute of Standards and Technology, 102, 279-322. THORLABS 2011. APD110x Series - Avalanche Photodetectors: APD110A, APD110A2, APD110C. THORLABS. 2015. Catalogue of products [Online]. Available: http://www.thorlabs.de/navigation.cfm [Accessed 06 of March 2016]. TOMASI, C., VITAKE, V. & DE SANTIS, L. V. 1998. Relative optical mass functions for air, water vapour, ozone and nitrogen dioxide in atmospheric models presenting different latitudinal and seasonal conditions. Meteorology and Atmospheric Physics, 65, 11-30. TRAGER, F. N. 2007. Springer Handbook of Lasers and Optics, Dordrecht, Dordrecht : Springer. TZORTZIOU, M., HERMAN, JR., CEDE, A. & ABUHASSAN, N. 2012. High precision, absolute total column ozone measurements from the Pandora spectrometer system: Comparisons with data from a Brewer double monochromator and Aura OMI. J. Geophys. Res.- Atmos., 117. UNEP 2012. Handbook for the Montreal Protocol on Substances that Deplete the Ozone Layer. the Montreal Protocol on Substances that Deplete the Ozone Layer. WILLIAMS, G. M. & HUNTINGTON, A. S. Single photon counting linear mode avalanche photodiode technologies. 2011. 81551M-81551M-10. WMO. Observing Systems Capability Analysis and Review Tool (OSCAR) [Online]. Available: http://www.wmo-sat.info/oscar/instruments/view/167 [Accessed 24 February 2016]. WMO 2001. Global Atmosphere Watch Measurements Guide (WMO TD No. 1073). GAW Report No. #171. Geneva, Switzerland: World Meteorological Organization. WMO 2010a. Scientific assessment of ozone depletion : 2010, World Meteorological Organization. WMO 2015. Seventh Intercomparison Campaign of the Regional Brewer Calibration Center Europe (RBCC-E) Lichtklimatisches Observatorium, Arosa, Switzerland, 16-27 July 2012, WORLD METEOROLOGICAL ORGANIZATION / GLOBAL ATMOSPHERE WATCH WMO, W. M. O. 2010b. Guide to meteorological instruments and methods of observation. YOUNG, J. E. H. & SHI-KAY, Y. 1981. Design considerations for acousto-optic devices. Proceedings of the IEEE, 69, 54-64.

122

123