STUDIES OF SATURN S ULTRAVIOLET AURORAS USING THE HUBBLE SPACE TELESCOPE

Transcription

1 STUDIES OF SATURN S ULTRAVIOLET AURORAS USING THE HUBBLE SPACE TELESCOPE Thesis submitted for the degree of Doctor of Philosophy at the University of Leicester by Calum James Meredith MPhys Radio and Space Plasma Physics Group Department of Physics and Astronomy University of Leicester March 2015

2 Abstract Studies of Saturn's Ultraviolet Auroras using the Hubble Space Telescope Calum James Meredith In this thesis we study Saturn s ultraviolet dayside auroras mainly using images from the Hubble Space Telescope. We investigate a variety of different types of auroral emission observed in the database of images compiled between 1997 and In equinoctial data from 2009 two different types of features are investigated. In the dawn sector patches of emission are observed that are found to lack direct conjugacy between the two hemispheres and instead are displaced in local time. A production mechanism related to ULF waves is suggested, plausibly driven by drift-bounce resonance. Dusk transient auroral emission is also observed and found to be strictly non-conjugate. A suggested explanation is that the transient patches are related to newly-opened flux tubes where the hemispheric symmetry is broken via the interplanetary magnetic field Y component. A further study uses data from the Cassini spacecraft during passages through the solar wind in conjunction with Hubble Space Telescope images from 2011 and 2012 in order to investigate how changes in the Interplanetary Magnetic Field strength affect the auroral emission at Saturn. It is found that during intervals of positive interplanetary magnetic field the auroral emission in the dusk sector is enhanced as opposed to cases of negative field strength where there is little emission observed in the dusk sector. This supports earlier interpretations that this emission is due to low latitude dayside reconnection and open flux production. The final study in this thesis uses the entire dataset of Hubble Space Telescope images to investigate auroral storm events. There are 12 such events found in the database with statistics showing that storms are present ~12% of the time. We deduce from the statistics and observations that storms are likely to last ~16 hours or ~1.5 Saturn rotations. 2

3 Declarations The research undertaken during the course of this doctoral programme has led to the submission and acceptance for publication of the following three scientific papers: Meredith, C. J., S.W.H. Cowley, K.C. Hansen, J.D. Nichols, and T.K. Yeoman (2013), Simultaneous conjugate observations of small-scale structures in Saturn s dayside ultraviolet auroras - implications for physical origins, J. Geophys. Res., 118, , doi: /jgra Meredith, C. J., I. I. Alexeev, S. V. Badman, E. S. Belenkaya, S. W. H. Cowley, M. K. Dougherty, V. V. Kalegaev, G. R. Lewis, and J. D. Nichols (2014), Saturn s dayside UV auroras: Evidence for morphological dependence on the direction of the upstream interplanetary magnetic field, J. Geophys. Res., 119, , doi: /2013JA Meredith, C. J., S.W.H. Cowley, and J.D. Nichols (2015), Survey of Saturn auroral storms observed by the Hubble Space Telescope: Implications for storm time scales, J. Geophys. Res., 119, , doi: /2014JA

4 Acknowledgements I thank Stan Cowley for being a truly inspiring supervisor, for his indefatigable attention to detail and for always making sure I kept on going. I also thank the Radio and Space Plasma Physics group for being so friendly, welcoming and for being such a comfortable place to work. I particularly thank Jon Nichols for helping with access to the HST data. I thank all of my family, particularly my mum, dad and sister for all of their influence and care growing up and still today. I thank my wife Melony for all of her support, for putting up with me working late at night and for always knowing how to make me smile. The Catholic Chaplaincy to the University of Leicester has been an enormous source of support during my studies and I thank all at Holy Cross Priory Church, both friars and co-conspirators. I acknowledge the STFC who funded this body of work. Ad Majorem Dei Gloriam My soul glorifies the Lord, my spirit rejoices in God, my Saviour. He looks on his servant in her lowliness; henceforth all ages will call me blessed. The Almighty works marvels for me. Holy his name! His mercy is from age to age, on those who fear him. He puts forth his arm in strength and scatters the proud-hearted. He casts the mighty from their thrones and raises the lowly. He fills the starving with good things, sends the rich away empty. He protects Israel, his servant, remembering his mercy, the mercy promised to our fathers, to Abraham and his sons for ever. 4

5 Table of Contents Abstract... 2 Acknowledgements... 4 Table of Contents... 5 List of Tables... 7 List of Figures... 8 List of Abbreviations Chapter 1 Introduction The Solar Wind and the Interplanetary Magnetic Field Convection of Plasma in a Magnetosphere Particle Bounce and Drift Current Systems The Production of Auroral Emission The Contribution of Wave Activity Chapter 2 Auroral Emissions Observed at Earth, Jupiter and Saturn Auroral Emission at Earth Auroral Emission at Jupiter Auroral Emission at Saturn Chapter 3 Instrumentation Hubble Space Telescope Advanced Camera for Surveys Advanced Camera for Surveys Instrument Cassini Spacecraft Cassini Spacecraft MAG instrument CAPs Chapter 4 Observations of Small-Scale Features in the Dawn sector of Saturn s Aurora Introduction

6 4.2 Data Employed in this Study Observations of Saturn s aurora at Dawn Physical Origin of Dawn Sector Patches Conclusions Chapter 5 - Observations of Small-Scale Features in the Dusk sector of Saturn s Aurora Introduction Dusk Sector Transients Physical Origin of Dusk Sector Patches Summary and Discussion Chapter 6 - Saturn s dayside ultraviolet auroras: Evidence for morphological dependence on the direction of the upstream interplanetary magnetic field Introduction Cassini IMF Data Morphology and IMF Dependence of the 2011 and 2012 HST Campaign Auroras Summary and Discussion Chapter 7 - Survey of Saturn auroral storms observed by the Hubble Space Telescope: Implications for storm time scales Introduction Dataset HST Image Presentation and Analysis Discussion and Summary Chapter 8 Summary and Future Work Summary Future Work Final Conclusions Appendix 1 - Formulae for Alfvén Resonance Box Model

7 Appendix 2 HST Image Identifier Formats References List of Tables Table 3.1 Tabulated SBC filters together with their spectral window and a brief description 60 Table 6.1 Averaged Lagged IMF Components and Clock Angle for Each HST Visit 122 Table 7.1 Saturn auroral storm events observed by HST observed

8 List of Figures Figure 1.1 Heliosphere and Parker Spiral 15 Figure 1.2 Heliospheric Current Sheet 19 Figure 1.3 Corotating Interaction Regions 19 Figure 1.4 Chapman-Ferraro Magnetosphere 20 Figure 1.5 Magnetic Reconnection 23 Figure 1.6 Dungey Cycle Magnetospheric Configuration 25 Figure 1.7 Corotation 27 Figure 1.8 Stagnation Point 28 Figure 1.9 Vasyliunas Cycle 30 Figure 1.10 Magnetospheric Current Systems 32 Figure 1.11 Dungey and Vasyliunas Cycle Currents 35 Figure 1.12 ULF Waves 39 Figure 2.1 Earth s Aurora 42 Figure 2.2 Jupiter s Aurora 45 Figure 2.3 Saturn s Aurora Quiet 50 Figure 2.4 Saturn s Aurora Compressed 53 Figure 3.1 Hubble Space Telescope 58 Figure 3.2 Advanced Camera for Surveys Optical Design 59 Figure 3.3 Cassini Spacecraft 61 Figure 3.4 Cassini Magnetometer Instrument 63 8

9 Figure 3.5 CAPS Instrument 65 Figure 4.1 Modelled Solar Wind and Magnetopause Parameters 68 Figure 4.2 Observations of Small Scale Features at Dawn 70 Figure 4.3 Propagation of Dawn Patches 72 Figure 4.4 Local Time Extent 74 Figure 4.5 Statistical Properties of Dawn Patches 76 Figure 4.6 Dawn Patch Angular Momentum 78 Figure 4.7 Drift-Bounce Resonance 87 Figure 4.8 Properties of the Magnetosphere in the Equatorial Plane 90 Figure 4.9 Wright and Allen Box Model 92 Figure 5.1 Observations of Auroral Features in the Dusk Sector 99 Figure 5.2 Propagation of Dusk Patches 100 Figure 5.3 Local Time Extent 101 Figure 5.4 Statistical Properties of Dusk Patches 103 Figure 5.5 Dusk Patch Angular Frequencies 104 Figure 5.6 Mechanism for the Production of Dusk Patches 108 Figure 6.1 Trajectory of the Cassini Spacecraft 116 Figure 6.2 Cassini Data for Figure 6.3 Cassini Data for Figure 6.4 HST Campaign Images 124 9

10 Figure 7.1 Auroral Storm Observation Statistics 139 Figure 7.2 Images of Storm Figure 7.3 Images of Storm Figure 7.4 Images of Storm Figure 7.5 Images of Storm 11 and Figure 7.6 Images of Storm Figure 7.7 Images of Storms 2 and Figure 7.8 Images of Storm Figure 7.9 Images of Storm

11 List of Abbreviations ACS Advanced Camera for Surveys CAPS Cassini Plasma Spectrometer CDA Cosmic Dust Analyser CIRS Composite Infrared Spectrometer COS Cosmic Origins Spectrograph ELS Electron Spectrometer ESA European Space Agency FGM Flux Gate Magnetometer FGS Fine Guidance Sensor FOC Faint Object Camera HRC High Resolution Channel IBS Ion Beam Spectrometer IMF Interplanetary Magnetic Field IMS Ion Mass Spectrometer INMS Ion and Neutral Mass Spectrometer ISS Imaging Science Subsystem KH Kelvin Helmholtz kr KiloRayleigh LISM Local Interstellar Medium LT Local Time MAG Magnetometer Instrument MAMA Multi-Anode Microchannel Array 11

12 MIMI Magnetospheric Imaging Instrument NASA National Aeronautics and Space Administration NICMOS Near Infrared and Multi-Object Spectrometer OCFLB Open-Closed Field Line Boundary OTA Optical Telescope Assembly RPWS Radio and Plasma Wave Science instrument RSS Radio Science Subsystem SBC Solar Blind Channel STIS Space Telescope Imaging Spectrograph ULF Ultra Low Frequency Wave UV Ultraviolet UVIS Ultraviolet Imaging Spectrograph VIMS Visible and Infrared Mapping Spectrometer V/SHM Vector / Scalar Helium Magnetometer WFC Wide Field Channel WFC3 Wide Field Camera 3 WFPC2 Wide Field Planetary Camera 2 12

13 Chapter 1 Introduction 1.1 The Solar Wind and the Interplanetary Magnetic Field Material from the Sun is constantly flowing radially away due to the Sun s inability to balance the gravitational force of its mass with the pressure exerted outwards at hydrostatic equilibrium. This material originates in the corona and consists of hot hydrogen and helium ions (protons and α particles) together with electrons, overall electrically neutral, i.e. a plasma, which flows outward with velocities ~400 km s -1, although this can vary between ~ km s -1. It is possible to see this by considering the momentum equation ρ dv dd = ρg P + ρ qe + j B, (1.1) where ρ is the mass density, V is the velocity, g is the acceleration due to gravity, P is the gas pressure, ρ q is the charge density, E is the electric field vector, j is the current density and B is the magnetic field vector. The magnitude of the electric force term is generally negligible compared to the magnetic term, and is neglected, while the j B term can also be rewritten by using Ampère s law j = 1 μ 0 B where μ 0 is the vacuum permeability to yield j B = 1 μ 0 ( B) B (1.2) 13 = 1 μ 0 (B )B B2 2μ 0 using a vector identity [Kivelson and Russell, 1995, section 2]. The second term of this equation represents the magnetic pressure force while the first term represents a magnetic tension force that acts to straighten bent magnetic field lines reducing the tension. For the case of the Sun the magnetic forces can be neglected beyond a few solar radii to yield ρ dv dd = ρg P. (1.3) For a solution where V is dependent only on the radial component, r, the continuity equation can be expressed as r 2 ρρ = const in spherical polar coordinates. For the radial component the momentum equation can be written as

14 ρ dv dd = GM ρ r 2 dd dd, (1.4) where G is the gravitational constant, M ʘ is the mass of the Sun. This can be used with the continuity equation and the hydrostatic solution dd dd = m gg k B T to yield V dd dd = GM r 2 + k BT m 2 r 1 dd V dd, (1.5) where m is the averaged mass of the particles, k B is Boltzman s constant, and T is the temperature. The equation can be solved through separation of variables yielding a velocity of V 2C s ln r r c, (1.6) where C S is the speed of sound in the gas, assumed isothermal, r is the radial distance and r c is a critical radius defined by r c = GM m 2k B T. (1.7) The Sun also has its own internal magnetic field generated by dynamo action. It is thought this field is wound around the outer layers of the Sun by differential rotation giving rise to sunspots and other solar activity. The field is seen to reverse polarity every ~11 years at solar maximum when the winding of the field is maximum. This 11 year cycle is known as the solar cycle with a full cycle returning to the original magnetic configuration every ~22 years. The magnetic field extends far beyond the Sun itself out in to the heliosphere, where it meets interstellar space at the heliopause [Kivelson and Russell, 1995, section 3] and a diagram of this magnetic field configuration is shown in Figure

15 Heliosphere and Parker Spiral Figure 1.1 The Sun is indicated in the centre of the diagram with the Parker spiral magnetic field configuration (solid black) shown out through the heliosphere to the heliopause (dot-dash line) prior to its interaction with the local interstellar medium (LISM) ( Magnetic field indicated by solid grey line). Flow of plasma in the solar wind and in the LISM is indicated by solid grey lines with large arrows [Milan, 2011]. 15

16 In plasmas where the magnetic field and electric field are relatively constant over space and time the frozen-in flow approximation can be used. To understand why this is the case it is important to understand the motion of charged particles in magnetic and electric fields. A charged particle in a uniform magnetic field B gyrates around the field line with a gyrofrequency Ω = ee m and with uniform v, i.e. a helix. If the particle is subject to another force F it also drifts perpendicular to B with a velocity v f = 1 F B q B 2, (1.8) where q is the charge on the particle. This can be used with a variety of different force contributions to determine particle drift motion. One of the simplest cases is when an electric field is perpendicular to a magnetic field, the plasma will move with a motion known as E B drift. In this case the force is F = qe and the velocity is therefore v E = E B B 2 (1.9) independent of q. Here the particles drift in a direction perpendicular to both the electric field, E, and the magnetic field, B. Transforming to the frame of this flow, the electric field disappears and only the gyration about B remains. The electric field, but not the magnetic field, depends on the frame of reference. Due to E B drift and the motion allowed along a field line, the gyrational guiding centres of particles in a magnetic field remain on the same field line in the plasma. This follows from Faraday s law B = E, (1.10) together with the E B drift equation rewritten in the form E = v E B, which together yield B = (v B), (1.11) relating the change in the magnetic field over time to the velocity of the particles. From this equation it can be shown that the particles move with the field, and vice versa, such that the field is said to be frozen in [Kivelson and Russell, 1995, section 2]. The magnetic Reynolds number R m can be used to describe the importance of either the 16

17 frozen-in condition or the diffusion of the plasma. This is the ratio of the frozen-in term (equation 1.11) as compared to the diffusion term and is written as (v B), R m = (1 μ 0 σ 0 ) 2 B (1.12) where σ 0 is the conductivity of medium. Where R m is greater than unity the frozen-in condition dominates whereas when it is less than unity diffusion is dominant. This frozen in approximation in the solar wind is well satisfied. The solar wind moves outwards from the Sun radially. However, because the feet of the field lines are frozen into the rotating Sun it is wound into a spiral pattern within the solar system. This spiral pattern was first theorised by Eugene Parker and is known as the Parker spiral [Parker, 1958]. The heliosphere and the Parker spiral are shown in Figure 1.1, where this spiral magnetic field pattern is shown extending out from the Sun into the heliosphere accompanied by the radially flowing solar wind. For this reason the magnetic field configuration that makes up the Parker spiral pattern is often called the Interplanetary Magnetic Field (IMF). At Earth the angle of the Parker spiral is ~45 as measured by the Ulysses spacecraft [Forsyth et al., 1996]. At Jupiter the same dataset showed this angle to be ~80 and data from the Cassini spacecraft at Saturn showed this angle to be ~83 [Jackman et al., 2008]. The Parker spiral magnetic field and radial solar wind flow are idealised pictures of the magnetic environment and flow in the heliosphere, and in reality there are a variety of processes that result in less predictable behaviour due to localised variable solar wind source velocities. Co-rotating interaction regions are one such phenomenon that can cause dramatic behaviour when interacting with planetary bodies. The solar wind flowing the Sun can be categorised as either slow solar wind travelling at ~400 km s -1 or the fast solar wind that propagates at ~750 km s -1 [Feldman et al., 2005]. The fast solar wind originates for the most part from open coronal hole in the polar regions of the Sun while the slow solar wind is from close field arcades near the dipole equator. Due to the tilt of the Sun s magnetic dipole and the solar wind outflow that creates the Parker spiral, a heliospheric current sheet is created that moves up and down as observed from a fixed position as the Sun rotates. A diagram of this heliospheric current sheet is shown in Figure 1.2. This configuration can cause regions of fast and slow solar wind to interact and create CIRs. As the fast solar wind catches up to a region of slow 17

18 solar wind a shock wave is formed that propagates ahead of the interaction region, this is shown in Figure 1.3. The shock wave is manifested as a jump in solar wind speed across the boundary together with an associated increase in density, temperature and magnetic field magnitude [Balogh et al., 1999]. As the occurrence of CIRs is related to the heliospheric current sheet they tend to be coincident with changes in the sector structure of this current sheet. During normal conditions a two sector structure of the heliospheric current sheet exists over the course of a solar rotation (~25 days) such that a CIR may be expected every half a solar rotation period, yet at the height of the solar cycle there can instead be four or more sectors with a corresponding increase in the expected periodicity of CIR occurrence. The directional components of the IMF also vary together with the heliospheric current sheet structure with the z component reversing from north to south across this boundary. Coronal mass ejections are another phenomenon where material is ejected from the Sun at high velocity resulting in a travelling shock front ahead of a magnetic loop with associated out flowing plasma. This complicated system can result in a variety of magnetic field orientations and particle populations when it comes to interact with a solar system body. 18

19 Heliospheric Current Sheet Figure 1.2 A representation of the heliospheric current sheet within the heliospheric looking down from above depicted out to a distance of ~15 AU [Jokipii and Thomas, 1981]. Corotating Interaction Regions Figure 1.3 A diagram of the formation of corotating interaction regions where fast solar wind interacts with slow solar wind producing shock fronts and regions of rarefaction [Hundhausen, 1972]. 19

20 Chapman-Ferraro Magnetosphere Figure 1.4 The magnetopause and tail current systems for a planet with a magnetic dipole directed south (thin black lines). The Chapman-Ferraro currents are indicated by thick lines directed from dawn to dusk around the magnetopause and annotated as j mp. Also shown in the diagram is the tail current in the equatorial plane indicated by j tail and by the large arrow across the tail [Lapenta, 2011]. 20

21 Many planets in the solar system also generate magnetic fields through dynamo action in their interiors, past or present, and the plasma in their environments is also frozen in to the planet s magnetic field to a first approximation. When the magnetised solar wind flows past a magnetised body, the two regimes therefore cannot mix to a first approximation, thus creating a cavity in the solar wind flow with a current layer separating the two. This magnetic cavity is known as a magnetosphere, and is shown in Figure 1.4 for the case in which the magnetic field of the planet is northwardly orientated in the equatorial region with a current layer directed in the duskward direction as indicated. The magnetic field of planets that have a dynamo generated field is to first approximation dipolar. Due to the dynamic pressure of the solar wind, the front of magnetosphere on the dayside becomes compressed until the pressure of the compressed magnetic field strength of the planet balances the dynamic pressure of the solar wind. Conversely on the nightside the magnetosphere is elongated forming a tail as magnetic field lines are carried tailward due to the interaction with the surrounding solar wind plasma flow. The solar wind plasma flows around the outside of the magnetosphere along a current layer boundary called the magnetopause. As the solar wind flow is supersonic, due to the obstacle in the flow a bow shock is also formed ahead of the magnetopause. At the bow shock the solar wind is slowed down, heated, and compressed in a region between the bow shock and the magnetopause known as the magnetosheath. The current sheet at the magnetopause is known as the Chapman- Ferraro current with its direction dependent on the orientation of the dipole and magnetosheath field [Chapman and Ferraro, 1931]. In Figure 1.4 for a northward directed equatorial planetary field the Chapman-Ferraro current is shown directed eastward at the dayside magnetopause. At Saturn the dipole is oppositely orientated with the Chapman-Ferraro current being directed westward at the dayside magnetopause. It is possible to estimate the position of the dayside magnetopause for any such body in the solar wind by the simple balance of forces of plasma dynamic pressure and magnetic pressure. In such a case we can neglect the solar wind s thermal and magnetic pressure as the wind s dynamic pressure is usually considerably higher, we can also neglect the plasma pressure of the planet s magnetosphere to estimate the relationship between the internal and external pressures as 2 P SS B MM 2μ 0, (1.13) 21

22 where P SW is the dynamic pressure of the solar wind and B MP is the strength of the magnetic field just at the magnetopause. simple dipolar field where This magnetic field can be estimated as a B MM 2B ee R 3 MM, (1.14) where B eq is the equatorial magnetic field strength and R MP is the radial distance to the magnetopause. The factor of two is a result of the Chapman-Ferraro current that acts to strengthen the magnetic field at the magnetopause boundary by approximately a factor of two. Using the two equations 1.10 and 1.11 and rearranging for R MP yields a magnetopause standoff distance of R MM 2B ee μ 0 P SS (1.15) Although this simple approximation works well for Earth, at Jupiter and Saturn the underlying assumption that the magnetospheric plasma pressure can be neglected becomes less valid [Kivelson and Russell, 1995, section 6], resulting in the boundary being formed somewhat further out. At Saturn the plasma pressure is much greater due to the ejection of matter from the moon Enceladus. This and other factors have been considered by Kanani et al. [2010] to produce a model of the magnetopause standoff distance at Saturn using data from the Cassini spacecraft. This model balances the solar wind dynamic pressure with the magnetic and thermal pressures found inside the boundary. Although frozen-in flow is valid as a first approximation, as described above, under the right conditions magnetic field lines whose directional vectors are oppositely orientated can reconnect with one another. This is shown in Figure 1.5, where a magnetic field line directed down the page on the left interacts with a magnetic field line directed up the page on the right. Under these conditions the frozen-in approximation is no longer valid due to the thin nature of the current sheet between the two field configurations allowing other motions to come into play, such that the condition in equation 1.11 no longer holds true. The magnetic field can then diffuse through the plasma within the current layer and as shown in Figure 1.5 in this region the magnetic field lines then reconnect so that each end of the downwardly orientated field is connected separately to the upwardly directed field and are no longer themselves connected. The two newly 22

23 reconnected field lines then experience a force due to the tension of the bent field line that moves them out of the reconnection region due to the j B force as shown in Equation 1.2. Magnetic Reconnection Figure 1.5 A sequence of diagrams showing the stages of magnetic reconnection. Oppositely directed magnetic field lines (black lines) are shown to reconnect at a current sheet boundary (shown by the dashed line) resulting in plasma flows indicated by the red lines [Tapia, 2005]. 1.2 Convection of Plasma in a Magnetosphere This reconnection process can be very important at a magnetopause where two different magnetic fields encounter each other with a thin current sheet separating the two regimes as discussed previously. The magnetic dipole field of the magnetized planets is of fixed orientation (on these time scales), whereas due to the solar wind s variability the interplanetary magnetic field experienced at their magnetopause can also be variably orientated. When the IMF is oppositely orientated to the planetary magnetic field reconnection can occur at the dayside magnetopause [Kivelson and Russell, 1995, section 9]. This scenario is shown in Figure 1.6 where a magnetosphere is shown surrounded by the solar wind with the magnetic field lines of each indicated, and with this dayside reconnection site shown at point 1 in the diagram. The reconnected magnetic field lines have one end connected open to the solar wind with the other end 23

24 of the field line connected to the polar region of the planet. The field line tension and the continued flow of the solar wind connected to the open end of the field line drags the newly opened field lines down the tail of the magnetosphere over the poles. This also represents a plasma flow due to the plasma being frozen to the magnetic field after it has emerged from the current sheet region. In the tail there then exists a tail current sheet between the oppositely directed field lines connected to the two polar regions and as at the dayside magnetopause the magnetic field lines reconnect. This reconnection occurs at point 4 of Figure 1.6, where the orientation of the magnetic field lines in the tail is also shown. This reconnection creates a new closed magnetic field line connected to the planet at both ends and also a new interplanetary magnetic field line that is connected to the solar wind at both ends. This new interplanetary magnetic field line is dragged further out through the solar wind carrying with it a mix of plasma. The newly closed field lines move back towards the planet though the force of the field line tension, accelerating trapped plasma. The particles are accelerated into the auroral zone close to midnight, where particularly rapid connection events create substorms [Akasofu, 1964]. The field lines themselves flow back to the dayside around the sides of the planet and will then become reconnected again at the dayside magnetopause. This cyclical process is known as the Dungey cycle after Jim Dungey who first described it in 1961 [Dungey, 1961, 1963]. When IMF conditions do not favour Dungey cycle convection there can still be reconnection at the dayside magnetopause. At times when the IMF z component is directed in the same direction as the internal field reconnection at the subsolar point is inhibited. However, reconnection is possible with antiparallel open field lines with footprints at high latitudes in the magnetospheric cusp region. This reconnection results in a magnetic field line still with one end attached to the ionosphere and one end in the IMF, however the rearranged field line geometry results in a field line tension allowing the injection of energised particles at this location into the ionosphere [Cowley and Lockwood, 1992, Milan et al., 2003]. At the foot of these field lines the energetic particles collide with the neutral atmosphere close to noon resulting a spot of auroral emission just poleward of the main auroral oval. 24

25 Dungey Cycle Magnetospheric Configuration Figure 1.6 A schematic detailing the Dungey cycle. Solid lines indicate magnetic field lines, dashed lines indicate plasma flow lines and dots indicate the electric field. The numbering represents the stages of the Dungey cycle starting at point 1 on the dayside, reaching point 4 on the nightside before returning to point 7 [Milan, 2011]. In addition to this Dungey cycle there is also the internal rotation of the planet to consider. Without any other influences the magnetic field of the planet and its internal plasma will rotate around with planetary rotation. The reason for this corotation enforcement is due to collisions in the planet s ionosphere between neutral atmospheric particles and ions attached to magnetic field lines. The neutral atmosphere exerts a torque that acts to force the ions to rotate at the same rate as the planet. As the ions in the ionosphere are connected to the magnetic field lines, these are also forced to corotate with the planet. This situation is depicted in Figure 1.7 where the rotating planet s ionosphere is connected to the equatorial region by the magnetic field. Due to the frozen-in flux condition the rotating magnetic field lines also carry with them the magnetospheric plasma in their environment. The rotating planet transfers angular momentum from its neutral atmosphere to the magnetospheric plasma [Kivelson and Russell, 1995, section 10]. If a magnetosphere is large enough, such as that of Jupiter and Saturn, at a certain point this co-rotation enforcement process can no longer keep the magnetospheric plasma rotating with the same period. This can happen when there is an internal source of plasma present inside the magnetosphere, for example a moon 25

26 with active volcanism. This additional plasma is transported radially outwards, and in order to conserve angular momentum, its angular velocity falls. This slowing of the equatorial plasma is an additional torque that the ionosphere must exert to enforce corotation, at a certain point this becomes impossible and co-rotation breaks down. At planets where this rotation is particularly strong this plasma flow can have important consequences. The Dungey cycle convection and internal corotation are taken as separate phenomena yet in reality they can co-exist and it is necessary to assess the contributions from each so as to understand the different impact they have at different planets. Figure 1.8 shows the equatorial plasma streamlines when the Dungey cycle is combined with simple corotation flows. For the Dungey cycle the flow inside the magnetopause is directed from the nightside to the dayside so that the flow is directed towards the Sun. This is due to closed reconnected magnetic field lines in the tail returning to the dayside. In this simple approximation all the flow is directed in the x direction so the velocity of the equatorial flow can be written as V cccc = E 0 B x = E 0 ( r ) 3 x, B ee R p (1.16) where V conv is the plasma velocity and r is the radial distance at any point. For the corotation flow it is easy to specify the plasma flow with reference to the planetary angular velocity, ω p, so that V ccccc = rω p φ. (1.17) On the dawn side of the planet the flows are orientated in the same direction, whereas at dusk the flows are oppositely directed. At dusk where V ccccc = V cccc there exists a stagnation point at a certain radius which can be determined by combining equation 1.16 and 1.17 and solving for r yielding, 26

27 Corotation Figure 1.7 The diagram depicts how a source of additional plasma inside a rotationaldominant magnetosphere can produce field-aligned currents [Cowley and Bunce, 2001]. The solid lines starting from the polar regions of the planet show the magnetic field of the planet. The large dot indicates an outgassing moon (in this case indicated as Io) with the smaller dots indicating the region where this plasma resides. The corotation enforcement currents are shown by the dashed lines with arrows. R SS R P = ω PB ee R p E 0 1 (1.18) 2, where R SP is the radius of the stagnation point on the dusk side. This stagnation point is identified in Figure 1.8a on the dusk side and shows where the two regimes dominate. We may then estimate the radius at which the effects of the two flow cycles become important for different planets. Where this radius is small, the Dungey cycle would play a larger role. Conversely where the stagnation radius is large, or beyond the magnetopause, the system will be dominated from rotational flows [Kivelson and Russell, 1995, section 10]. This situation is shown in Figure 1.8b for Saturn where the rotationally dominated region extends much further out with a much smaller Dungey cycle region. 27

28 Stagnation Point Figure 1.8 (a) A plot of modelled flow streamlines surrounding a planet in the equatorial X-Y plane. The solid lines show equipotentials and flow streamlines with a combination of corotation close to the planet and convection further away. The stagnation point between the rotational dominant inner magnetosphere and the Dungey cycle is indicated [Milan, 2011]. (b) The same streamlines for Saturn show the dominance of rotation within this system [Cowley et al., 2004]. 28

29 Where such rotational flows are large it is expected that another type of plasma circulation can occur known as the Vasyliunas cycle [Vasyliunas, 1983]. As magnetic field lines rotate with the frozen-in magnetospheric plasma the field lines are distended outwards as shown in Figure 1.7 due to the presence of an azimuthal current whose j B balances the centrifugal force of the plasma. In addition to this there is a further distension of the field lines on the nightside where the rotation is strong enough so that they distend further outwards and are then swept into the tail as plasma flows outwards. Such motion is inhibited on the dayside by the solar wind dynamic pressure. Once these magnetic field lines extend far down the tail, the magnetic field connected to both polar regions forms a tail-like current sheet at which reconnection and pinch-off will occur. The tailward portion is released down tail as a plasmoid and the planetward portion moves back towards the planet on the dawn side returning to near-rigid co-rotation. This reconnected magnetic field is then swept around through dawn back towards noon as a return flow. This cycle is shown in Figure 1.9 incorporating the flows for corotating and the Vasyliunas cycle. In the right hand side of this diagram is shown the process whereby a plasmoid is formed from internal magnetic field lines that is released downtail. These flows are also shown in Figure 1.8b incorporating the Vasyliunas cycle flows at dusk together with the Dungey cycle flows at dawn. This and similar diagrams allow the relative importance of co-rotating, the Dungey cycle and Vasyliunas cycle to be understood. 29

30 Vasyliunas Cycle Figure 1.9 A diagram of the plasma flows for the Vasyliunas cycle and co-rotation alone (left). Here noon is to the left in the sunward direction with dawn to the top of the diagram The Magnetic X-Line indicates the reconnection site where plasmoids are formed and flow becomes tailward. This process is shown in the right panel where internal magnetic field lines reconnect to produce a plasmoid with the remainder returning to the planet [Vasyliunas, 1983]. 1.3 Particle Bounce and Drift Previous discussion is mainly restricted to the behaviour of low temperature plasma following the frozen-in flux condition involving E B drift and motion along the field lines. In addition to this motion, particles moving along magnetic field lines in a dipole field experience the magnetic mirror effect [Kivelson and Russell, 1995, section 10]. For a particle gyrating about a field line with velocity v, in a mirror geometry field it is found that it would conserve the first adiabatic invariant, v 2 B = ccccc, (1.19) where v is the perpendicular velocity of the particle and B is the magnetic field strength. The result of this conservation means that as the magnetic field gets stronger along the field line towards the planet the velocity of the particle perpendicular to magnetic field also increases. As the particle s kinetic energy is also constant as the 30

31 particle moves along the field, v 2 = v 2 + v 2 = ccccc, the parallel velocity falls as the perpendicular increases due to the increased magnetic field strength. At a certain point the parallel velocity becomes zero and the particle is reflected in the opposite direction. This is referred to as magnetic mirroring. As well as this bounce motion and E B drift, the particles can also experience drifts related to gradients in the magnetic field. The effect of these can be analysed by recourse to Equation 1.19 that allows the velocity of a particle subjected to a force, F, to be determined. One such drift motion is the gradient-b drift that occurs due to a change in the strength of the magnetic field, where F = μ B, and where the velocity can be expressed as v B = ε B B qq B 2, (1.20) where ε is simply the perpendicular kinetic energy of the particle ε = mv 2. As well as a change in the magnetic field strength resulting in a particle drift, a curvature of a field line can also cause a drift. In this case the velocity of the drift is v R = 2ε R c B qqr2 c B 2, (1.21) 2 where ε is the parallel kinetic energy of the particle ε = mv and R c is the radius of curvature of the field lines. These two drifts are often considered together as the gradient-curvature drift resulting from the inhomogeneity of the magnetic field. One important consequence of these drifts arises due to charge dependence, which results in electrons and ions drifting in opposite directions thus producing current transverse to B. Trapped particles will thus produce an azimuthal current that can distend field lines similar to that shown in Figure 1.7, which together with the magnetization current, can from a ring current within a magnetosphere. 31

32 Magnetospheric Current Systems Figure 1.10 A diagram of the principal current systems surrounding Earth. Solid lines indicate currents present general with arrows indicating their directions, while dashed lines indicate currents that are present only under certain conditions. The different types of currents are labelled [Russell and Huddleston, 1997]. 1.4 Current Systems Dungey cycle convection and the co-rotation related Vasyliunas cycle can both produce auroral emission. To understand how this can occur, it is important to understand the electric currents generated by these processes [Kivelson and Russell, 1995, section 14]. For the Dungey cycle at Earth the currents associated with this plasma convection are shown in Figure Of particular importance for auroral studies are the current systems that connect to the ionosphere thereby allowing particles to be accelerated to 32

33 these regions. It is these magnetosphere-ionosphere coupling current systems that are involved in the transfer of stress between these two regions. At the magnetopause there exists a current that flows from the dusk side to the dawn side, part of the Chapman- Ferraro current as is also shown in Figure 1.4. At the dawn side in the equatorial region close to the magnetopause in a region known as the low latitude boundary layer, a magnetic field-aligned current flows to the ionosphere. The location that this current maps to in the ionosphere is roughly at the boundary between the feet of magnetic field lines opened to the solar wind by the Dungey cycle and those that remain closed, known as the open closed field line boundary (OCFLB). A similar current returns from the ionosphere close to the OCFLB to the dusk side low latitude boundary layer and the magnetopause current. Taken together these two currents are known as Region 1 currents as indicated in Figure This current loop is partially completed by Pedersen current flow across the polar cap from dawn to dusk. However in addition to this flow from dawn to dusk there also exists a second set of currents known as the Region 2 currents. These currents cause the Region 1 currents to be enhanced by allowing part of the current flow to circulate through a nightside ring current produced by gradient-curvature drift of hot trapped plasma. As shown in Figure 1.10, the Region 2 currents flow out of the ionosphere on the dawn side at lower latitude than the Region 1 currents. The Region 1 currents are connected to these Region 2 currents by a further Pedersen current flowing between the two regions. The current then flows around through the nightside equatorial plane before returning to the ionosphere on the dusk side at a similar location. Here too the Region 2 current circuit is completed by a Pederson current flowing from Region 2 to Region 1. There is in addition a Hall current that flows around the plasma streamlines in the ionosphere across the auroral zone. In general the height integrated ionospheric current, i, is the sum of these Pedersen and Hall currents and can be expressed as i = Σ P E + Σ H b E, (1.22) where Σ P is the height-integrated Pedersen conductivity, Σ H is the height-integrated Hall conductivity, E is the electric field in the neutral air rest frame, and b is the unit vector along B (i.e. b = B/B). Where upward currents are found, so too are downward flowing electrons, since fieldaligned currents are carried mainly by mobile electrons in the planetary rest frame. It is 33

34 these electrons accelerated by field-aligned voltages associated with field aligned currents that cause the brightest discrete auroral emission. That field aligned currents might be associated with auroral emission was suggested by Kristian Birkeland in 1903 [Potemra et al., 1988]. This pattern of Region 1 and Region 2 currents therefore creates auroral emission at the feet of these field lines being itself dependent on the behaviour of the magnetosphere in response to the solar wind. Diffuse auroral emission also occurs from precipitation of trapped flux, due to scattering of particles that sufficiently lowers their mirror height in the atmosphere. Another crucial aspect of the Dungey cycle flow is the return of magnetic flux that is stored in the magnetotail and the currents associated with this. In the magnetotail a current sheet exists that flows across the tail in the equatorial plane dependent on the sense of the planetary field. In a planetary field directed south to north in the equatorial plane, as reconnection in the tail occurs a substorm current wedge can form where the tail current is redirected down to the ionosphere towards the dawn-night side flowing through midnight as an auroral electrojet and then returning to the tail towards dusk. This current system is also capable of generating auroral emission as electrons are accelerated in regions of upward current into the ionosphere. This type of auroral emission is associated with substorms and can last a number of hours and can reoccur on an hour timescale while southward IMF conditions prevail. At Saturn the build up of magnetic flux in the magnetotail can produce a similar auroral storm that manifests itself as intense aurora emission offset to the dawn side of the planet by Vasyliunas outflow. Similar tail currents are generated across the tail that connect to the ionosphere on the dawn side. As part of the Vasyliunas cycle where plasmoids in the tail are detached from the rotating plasma through tail reconnection a field-aligned current is also formed in the ionosphere. The current is directed downwards (upward flowing electrons) in a narrow region of the dawn potentially leading to a region that that is aurorally dark. Upwards currents (downward electrons) flow on the dusk side of the planet possibly leading to discrete auroral emission in this region [Cowley et al., 2003]. 34

35 Dungey and Vasyliunas cycle currents Figure 1.11 A sketch of the flows and currents relatated to the Dungey cycle, the Vasyliunas cycle and co-rotation enforcement [Cowley et al., 2003]. Upward and downward currents are indicated by circles with a dot or a cross respectively. The open flux (polar cap) region is indicated by the hatched area together with the sites of reconnection (X-lines) mapped into the ionosphere for the Dungey cycle and Vasyliunas cycle. Field-aligned currents are also generated by the enforcement of plasma rotation inside a rotation-dominated magnetosphere. As material is accelerated azimuthally due to the sub-rotation of the plasma with the planetary period, as described above and depicted in Figure 1.7, a current system is generated along the magnetic field lines that communicates the angular momentum from the ionosphere to the equatorial plasma. For Earth with a magnetic dipole field orientated north to south in the equatorial plane, this 35

36 current flows equatorwards in the ionosphere associated with a j B force directed oppositely to the planet s rotation that balances the frictional force due to ion-neutral collisions in the Pedersen layer of the ionosphere. In the equatorial plane there exists a Pedersen current that flows outwards associated with another j B force in the direction of the rotation of the planet serving to accelerate this equatorial plasma towards co-rotation. These two currents are then connected by two field-aligned currents that are directed from the equatorial plasma sheet in its outermost regions in to the ionosphere and then out of the ionosphere to the equatorial current sheet at the point where rigid co-rotation begins to break down. This current system is shown in Figure 1.7 by dashed lines directed into and out of the ionosphere.. At Saturn where the magnetic field dipole is oppositely orientated these currents would flow in opposite directions. The upward currents may require magnetospheric electrons to be accelerated down to the ionosphere to produce auroral emission. The Vasyliunas cycle also has associated currents that are shown in Figure 1.11 together with the flows, together with those of the Dungey cycle. Downward currents are generated in a narrow section of the dawnside ionosphere. Just as the plasma is convected in different ways simultaneously due to differing processes, so too must the current systems associated with these flows co-exist, dependent on the sense of momentum flow to or from the ionosphere. Where a rotationally dominant system exists the Dungey current system is likely to be found towards the dawn side as the Vasyliunas cycle and co-rotation enforcement plays a larger role at dusk. This sort of current system is depicted in Figure 1.11 where the two current systems mapped to the ionosphere are identified in a rotationally dominant system. 1.5 The Production of Auroral Emission The above discussion has limited itself as to how field aligned currents can be produced without explicitly discussing how this produces auroral emission. As has been mentioned, the occurrence of upwardly directed field-aligned currents can lead to the acceleration of electrons down into the ionosphere. A simple model as first discussed by Knight [1973] where the field-aligned voltage, Φ, required by a field-aligned current, j i to produce a current in excess of that carried by precipitating magnetospheric electrons alone was calculated to be 36

37 Φ (j i ) = W th e j i j ii 1, (1.23) where W th is the thermal energy of the electron source population, e is the electric charge, and j ii = ee W th 2πm e where N is the density of the source population and m e is the electron mass [Cowley and Bunce, 2001]. The auroral power output per square metre is proportional to a first approximation to the energy flux E f = E f0 2 j 2 i + 1, j ii (1.24) As energetic charged particles collide with the neutral atmosphere they electronically excite the atmospheric atoms or molecules. This newly acquired excess energy is then emitted as photons at the appropriate wavelength for a particular energy state transition [Kivelson and Russell, 1995, section 14]. Due to the variety of different interactions permitted and different atmospheric compositions auroral emission can be found at a variety of different wavelengths at different planets. At Earth the atmosphere predominantly consists of O 2 and N 2, with mainly O in the upper atmosphere. The predominant colours observed in the visible are greens and reds due to the presence of oxygen, yet a variety of different wavelengths are observed, in the visible, ultraviolet (UV), X-ray, and infrared. At Jupiter and Saturn the atmosphere is made mostly of H 2, with H present at high altitudes and a small amount of He. Direct excitation of H produces Lyman series emissions in the UV, and Balmer in the visible, while excitation of H 2 produces banded UV emissions termed the Lyman-Werner bands. In addition ionization of H 2 in the upper atmosphere rapidly forms the H + 3 ion through H 2 + H + 2 H H, (1.25) which radiates strong in the infrared due to ro-vibration transitions. 1.6 The Contribution of Wave Activity Field-aligned acceleration of electrons in quasi-static current systems explains the production of much discrete auroral emission, however there has also been substantial work on the role of time-dependent acceleration by wave activity [Kivelson and Russell, 1995, section 11]. A variety of waves can be found in the magnetosphere at Earth but 37

38 particularly important for the work discussed in this thesis are ultra low frequency (ULF) waves. The properties of ULF waves depend on the wave modes and also the boundary conditions. For a ULF wave in the magnetosphere, the ionospheres in the north and south act as near-perfectly conducting boundaries of the flux tube. At these boundaries the ionosphere can reflect the ULF Alfvénic wave signals. In addition to this the most equatorial ionosphere acts as an inner boundary with the magnetopause as the outer boundary. If straightened to form a box with the ionosphere boundaries at the top and the bottom it is easy to see this situation in a similar way to a wave on a string. This box model is depicted in Figure 1.12 for two different modes [Yeoman and Wright, 2001]. Only Alfvén waves with certain wavelengths can therefore satisfy the reflection condition, giving allowed wavelengths along the field of λ = 2l n where l is the length of the field line and n is any integer. The allowed angular frequencies are then ω = υ A k = υ A 2π λ where υ A is the Alfvén velocity (υ A = B μ 0 ρ) and k is the parallel wave number. B is the magnetic field strength and ρ is the plasma mass density. Substituting in the wavelength it can be shown that ω = nn υ 2l A and thus that the allowed frequencies depend on the length of the field line, the strength of the magnetic field, and the density of the plasma. Figure 1.12 shows how odd and even modes result in different magnetic field and electric field perturbations with nodes and antinodes at the equator or ionosphere. For the odd mode fundamental condition the electric field is symmetric about the equator while the magnetic field is anti-symmetric. For the even mode at the second harmonic the perturbations result in an anti-symmetric electric field with a node in the ionosphere and a symmetric magnetic field. Field aligned currents are then associated with the divergence of the ionospheric currents required by the B perturbations resulting in a small Pedersen current in the ionosphere that closes along the magnetic field lines. These currents are often said to be generated by field line resonances due to the way in which an Alfvén wave produces a resonance on the magnetic field line. 38

39 ULF waves Figure 1.12 Trajectories of two ions (solid and dashed lines) in the wave rest frame. Shaded and unshaded regions represent positive and negative electric field respectively. The ions indicated with a solid line are in resonance with the N = 0 drift mode, and the ions represented by the dashed line with the N= 1 drift-bounce resonance condition Panel (a) shows a wave in the fundamental mode, while Panel (b) shows the second harmonic[yeoman and Wright, 2001]. Alfven wave can be driven in a number of different ways that generate different wave properties. At Earth many of the drivers are thought to be related to disturbances at the magnetopause boundary such as when the dynamic pressure changes abruptly or due to Kelvin-Helmholtz instabilities. Another source of wave energy at Earth comes from the ring current. When particles are injected into the ring current, at Earth often associated with substorm and storm activity, these energetic particles can drift and bounce as described above in phase with ULF waves (see Figure 1.12). This can cause a resonance and result in the wave power to grow in favourable circumstances. Wave related phenomena have also been theorised and in some cases observed at Jupiter and Saturn. At Jupiter ULF waves have been observed in-situ with the Voyager 2 spacecraft that detected pulsations of the middle magnetosphere observed as perturbations in the magnetic and particle pressures [Khurana and Kivelson, 1989]. Most notably Kelvin-Helmholtz instabilities have been suggested as an important mechanism of production at boundary layers. At Jupiter such Kelvin-Helmholtz instabilities have been theorised as a possible mechanism of reconnection related to 39

40 solar wind interaction [Delamere and Bagenal, 2010]. The evidence for ULF waves at Saturn has also been studied by Cramm et al. [1998] who concluded that such waves were possible at Saturn and tentatively identified one such ULF wave in Voyager 1 insitu data. At Saturn too Kelvin Helmholtz instabilities have been theorised, their characteristic features observed with in-situ data from the Cassini spacecraft including their magnetic signature [Galopeau et al.,1995; Masters et al., 2009; 2010; Delamere et al,, 2013]. 40

41 Chapter 2 Auroral Emissions Observed at Earth, Jupiter and Saturn 2.1 Auroral Emission at Earth At Earth the magnetopause standoff distance, R MP, can be estimated using Equation 1.15 to be ~11 R E when R E is the Earth s radius (6384 km). The magnetopause standoff distance is the shortest distance from the planet to the magnetopause. Along each flank close to dawn and dusk the distance to the magnetopause is ~15 R E. Similarly the stagnation point, R SP, between the Dungey cycle flow and the co-rotation flow can be estimated using Equation 1.18 using values of ω p = rad s -1, B eq 31,000 nt and E V m -1 (resulting in a cross-magnetosphere potential of ~16 kv [Boyle et al., 1997]) to give a value of ~8 R E. Thus we see that at dusk where co-rotation can dominate further away from the planet this stagnation point is still well within the magnetopause allowing the Dungey cycle to dominate overall. The result of this simple calculation is that Earth s magnetosphere responds strongly to solar wind driving and therefore the Dungey cycle is dominant throughout the magnetosphere except for a co-rotating core known as the plasmasphere. This is observed in the aurora where particles with medium energies are accelerated as part of the Region 1 and Region 2 field-aligned currents producing auroral emission in a near circular pattern slightly tipped back from the Sun. The oval itself lies close to ~15 colatitude at noon and ~20 co-latitude at midnight where it is more intense [Akasofu, 2012]. This auroral oval structure can be seen in Figure 2.1 where more complex variation is also evident. Discrete auroras are often brighter in the Region 1 current at dawn. These boundaries move in response to solar wind activity as the Region 1 and Region 2 current locations in the ionosphere are related to the OCFLB. The Earth s magnetic field is directed south to north in the equatorial plane so that when the interplanetary magnetic field is directed southwards reconnection is possible. Where this occurs over an extended period this allows the OCFLB to move equatorward and so too the auroral oval. 41

42 Earth s Aurora Figure 2.1 Four ultraviolet images of Earth s auroral oval taken using the WIC instrument on the IMAGE spacecraft. The images are separated by ~120 s showing the development of a substorm. Midnight is located toward the bottom of these images with dusk to the left and dawn to the right. The bright arc region is the auroral oval with growing red region showing the growing substorm-related emission [Gérard et al., 2004]. As a result of the Dungey cycle reconnected magnetic field lines in the tail return to closer to the planet due to the magnetic tension force. This dipolarisation accelerates the particles along field-aligned currents discussed in section 1.4 as part of the substorm current wedge and results in bright auroral emission close to midnight. The development of a substorm typically undergoes a period of growth where magnetic flux is stored in the tail through dayside reconnection, a period of expansion where tail reconnection produces auroral emission, and then followed by a period of recovery as the amount of open flux in the tail is reduced. Figure 2.1 shows the development of a substorm over a number of minutes. The substorm formation process can last for multiple hours but as shown in Figure 2.1 the expansion phase can occur over a short period of time. The emission is broad around the midnight sector reaching a maximal extent at expansion before the recovery phase returns the system to a narrow main auroral oval [Akasofu, 1964]. Also observed at Earth is cusp emission related to lobe reconnection when the IMF is directed northward at Earth [Milan et al., 2000b]. This type of auroral emission is found close to noon and often has the appearance of a spot. Cusp auroral emission may be related to transpolar arcs, features that bisect the auroral oval [Fear and Milan, 2012]. This cusp emission can also occur for southward IMF poleward of the OCFLB at noon. 42

43 There is also diffuse auroral emission at Earth that is generally less intense than the main auroral oval [Lui et al., 1973]. This results from precipitation of high energy particles escaping through the loss cone from being trapped by magnetic mirroring bounce motion around the planet. 2.1 Auroral Emission at Jupiter Auroral emission at Jupiter is very different to that of Earth. The reason for this becomes apparent when the stagnation point between Dungey cycle and co-rotation convection is calculated. For the case of Jupiter ω p rad s -1, R p 71,400 km, B eq 500,000 nt and E V m -1 so that R SP 390 R p. The standoff distance to the dayside magnetopause at noon varies widely between ~50 and 100 R J [Joy et al., 2002], but even at its most extended, the stagnation point lies far outside of the magnetopause. This means that the impact of the Dungey cycle should be limited to the dawn side magnetosphere, while co-rotation and the Vasyliunas cycle should be dominant through most of the magnetosphere. In addition to the dominance of co-rotation, the moon Io plays an important role by producing a large amount of SO 2 from volcanic activity that is ejected into Jupiter s magnetosphere at ~6R J. The atoms are photodissociated and then ionised by the Sun and by warm electron impact producing 10 3 kg s -1 of plasma contributing to a near-equatorial plasma sheet [Shemansky et al., 1988; Barbosa, 1994, Lichtenberg and Thomas, 2001; Bagenel and Delamere, 2011]. As described above in section 1.2, the enforcement of co-rotation can only be maintained out to ~20 R J in the equatorial plane after which point it becomes increasingly ineffective. Beyond this, angular momentum is approximately conserved in the outflow resulting in the angular velocity decreasing as ω ~ 1 r 2. It is at this transition between these behaviours where the main field aligned currents that communicate the momentum transfer are located as outlined above. Cowley and Bunce [2001] showed that this mechanism would result in a circumpolar ring of upward fieldaligned current located at ~16 co-latitude which is the location of Jupiter s auroral oval. They also showed that it is capable of producing auroral emission through acceleration of magnetospheric electrons down the field lines to energies of ~ kev through the mechanism described in section 1.5, resulting in auroral emission on the order of a few ~100 kr which has been observed in Jupiter s aurora previously 43

44 [Prangé et al., 1998]. This auroral emission forms the main part of Jupiter s total aurora emission and supports the idea that Jupiter is dominated by rotational flows. The auroral emission at Jupiter is shown in Figure 2.2, where the main auroral oval is clearly indicated. Panel (a) shows the auroral emission in the north (left) and the south (right) as viewed by the Hubble Space Telescope (HST, see section 3.1) where noon is to the bottom, dawn to the left and dusk to the right. Panel (b) shows the northern emission with its constituent parts labelled as discussed below. When a CIR arrives at Jupiter the magnetosphere is compressed, this causes the interior sub-corotating plasma to speed up resulting in a reduction in the required field-aligned currents generated to maintain corotation. These reduced field-aligned currents ought to produce less intense auroral emission. Counter-intuitively when a CIR interacts with Jupiter s magnetosphere the auroral emission should be reduced [Southwood and Kivelson, 2001]. However other studies have shown increases in auroral activity to be correlated with solar wind intensity [Gurnett et al., 2002]. Nichols et al [2007] used Jupiter fly-by data from the Cassini spacecraft in combination with HST images of Jupiter s aurora to investigate this behaviour and found particularly intense auroras during the passage of compression regions in the solar wind, however this was not present throughout the compression region and it was suggested to be at a time of modest magnetospheric expansion. Clarke et al. [2009] used HST data together with solar wind conditions propagated from solar wind monitors close to Earth to show that at Jupiter the total auroral power is not particularly strongly correlated by the solar wind in either sense. 44

45 Jupiter s Aurora Figure 2.2 Jupiter s aurora as observed in the ultraviolet by the Hubble Space Telescope. Panel (a) shows projections of these images in a latitude-longitude grid for the northern (left) and southern (right) hemispheres. The intense arc region is the auroral emission at Jupiter. Panel (b) allows the different features of Jupiter s aurora to be identified. The main oval is indicated, together with the polar emission and satellite footprints of Io, Ganymede and Europa [Bagenal et al., 2004]. Due to the rotational dominance at Jupiter it is expected that the Vasyliunas cycle may play a large part in dynamic auroral emission when tail reconnection occurs. As described above, when this occurs as part of the Vasyliunas cycle, a plasmoid is released down tail into the solar wind while the reconnected planetary magnetic field lines return close to the planet under magnetic tension. This provides an accelerating 45

46 mechanism for particles trapped on these field lines. These particles precipitate down to the ionosphere on the dusk side and there produce discrete auroral emission [Badman and Cowley, 2007]. While from simple considerations of the position of the stagnation point it is thought that the Dungey cycle ought to not to play a major role at Jupiter, it is possible that emission poleward of the main auroral oval is related to dayside reconnection with the solar wind. At Jupiter the open-closed field line boundary as mapped to the ionosphere is thought to lie poleward of both the co-rotation enforcement currents and the Vasyliunas cycle currents. This polar swirl of aurora observed at Jupiter is thought to be related to Dungey cycle convection and also reconnection in the lobe resulting in cusp emission as discussed for Earth above [Grodent et al., 2003]. This polar swirl region is indicated in Figure 2.2 panel b as Polar emission. However this polar region is still a topic of current research. At Jupiter a key feature of its auroral emission is the presence of the footprints of its satellites Io, Ganymede and Europa. Io is the brightest of these footprints due to the large amount of plasma it generates [Clarke et al., 1996; 2002]. The footprint positions in the ionosphere can be mapped back to the equatorial plane through magnetic field models. Having done this calculation these locations are shown to be coincident with the positions of Io, Ganymede and Europa [Bonford, 2012]. These moon auroral footprint signatures are also indicated in Figure 2.2 panel (b) and shows the important role they play at Jupiter. 2.3 Auroral Emission at Saturn It was assumed from physical characteristics that Saturn s auroral emission could be explained as an intermediate example between Jupiter and Earth. At Saturn ω p rad s -1, R p 58,232 km, B eq 21,000 nt and E V m -1 so that R SP 35 R p, a value intermediate between Earth and Jupiter. While this characterisation is partially true, it is a complex system in its own right. As at Jupiter, at Saturn rotation plays an important role. The co-rotation enforcement currents in the middle magnetosphere exist with plasma provided by Saturn s moon Enceladus. Enceladus provides up to 200 kg s -1 that is ionized, followed by pick-up and radially transported requiring co-rotation enforcement [Hansen et al., 2006] However, Cowley and Bunce [2003] showed that the currents produced by the co-rotation enforcement 46

47 mechanism map to ~20 co-latitude rather than co-latitude observed for the auroral emission as observed by Voyager and the HST. In addition Cowley and Bunce [2003] showed that the auroral emission produced by the process would be ~1 kr as opposed to up to ~75 kr as observed [Gerard et al., 2004]. While this auroral emission may exist it is clearly not the main auroral emission at Saturn as observed. Physical characteristics that differ from both Earth and Jupiter also affects the importance of the Vasyliunas cycle that causes pinch-off of plasmoids in the tail, although it is likely that this mechanism produces auroral emission in a similar way to at Jupiter [Cowley et al., 2004, McAndrews et al., 2009]. With the main auroral oval of Saturn lying poleward of these processes, the likely candidate is solar wind-magnetosphere coupling as described for the Dungey cycle above. However, unlike at Earth where the Dungey cycle is dominant, at Saturn it is important to consider the effect that the stronger rotation has on this system, as was discussed for Jupiter. Cowley et al. [2004] described this combination of plasma flows at Saturn and the ionospheric currents they create. As at Earth it is expected that upward field-aligned currents will be located close to the OCFLB resulting in auroral emission, due to the shear in the rotational flow across the boundary towards higher angular velocities on closed field lines. At Saturn this was determined to be at roughly 15 co-latitude in agreement with the location of the dayside auroral emission. In a similar way to Jupiter, the plasma near-rigidly co-rotates with the planet in the innermost region within radial distances of ~3 R S [Wahlund et al., 2005; Wilson et al., 2009]. Outside this distance the plasma increasingly subcorotates, being confined on the dayside by the solar wind flow. The plasma is then lost via the dusk flank magnetotail through the Vasyliunas cycle forming plasmoids [Jackman et al., 2011]. This process is confined more to dusk and pre-midnight than at Jupiter with fast flowing return in the post-midnight and dawn sector. The final component of this system is the Dungey cycle. Saturn s magnetic dipole field is orientated opposite to that of Earth so that the field is directed north to south in the equatorial plane away from the planet. This field orientation has implications for the solar wind interaction in that dayside reconnection is possible under northward interplanetary magnetic field conditions. This leads to the presence of open flux at higher latitudes where the plasma angular velocity is expected to be low. It is the shear between the flow in this region and the higher flows expected on closed field lines that is expected to produce the auroral upward currents in the 47

48 Cowley et al. [2004] model, with magnetospheric electrons being accelerated down the field lines as described in section 1.5. While in theory this would produce a ring of auroral emission around the open-closed field line boundary, the currents are expected to be stronger at dawn than at dusk due to increased shear from the Dungey cycle flow at dawn than at dusk. As a result of these asymmetric currents, asymmetric auroral emission is also generated with the strongest auroral emission generated on the dawnto-noon side. This is consistent with observations. Theoretical analysis yields auroral emission on the order of ~5 to ~50 kr on the dawnside and 0.5 to 5 kr emission on the dusk side, also consistent with observations [Gérard et al., 2004; Bunce et al., 2008]. Observations of Saturn s aurora support this picture in outlines but also suggest a much more complicated system. Over the last ~15 years Saturn s aurora has been observed in the ultraviolet by the HST and more recently by the ultraviolet imaging spectrograph (UVIS) on the Cassini spacecraft. A study by Carbary [2012] using UVIS data provided an averaged determination of Saturn s auroral morphology in agreement with previous work by Badman et al. [2006] and others using the HST. These observations have shown that the aurora on Saturn is near-circular with a width of ~4 at dawn and ~8 at dusk for both the north and the south. The peak emission occurs close to 5 h LT in both hemispheres with ~40 kr for the north and ~60 kr for the south intensity. This dawn-dusk asymmetry was predicted by Cowley et al. [2004] and first observed by Clarke et al. [2005] The auroral circle is offset from the pole towards a midnight-dawn region by ~1-2 in the north and ~2-3 in the south. The co-latitude at which the auroral emission peaks varies over local time with an average of 15.1 in the north and 15.9 in the south. This difference results from the north-south asymmetry in Saturn s internally-generated magnetic field, associated with the quadrupole moment [Dougherty et al., 2005]. At dawn and around the nightside it is more equatorward, while at dusk it located more poleward [Gérard et al., 2004; Badman et al., 2006; Nichols et al, 2009; Carbary, 2012]. An image of Saturn s northern aurora can be seen in Figure 2.3 for an inferred relatively expanded magnetosphere with the absence of any auroral storm effects due to solar wind compression of the magnetosphere. This image was obtained in March 2012 using the HST and has been projected in a latitudelongitude grid where noon is to the bottom, dawn to the left and dusk to right. The properties discussed above can be identified, perhaps most obviously its local time dependence about noon. However, as can be observed, the width of the aurora is 48

49 narrower than the averaged oval suggested by the work of Carbary [2012]. This auroral picture largely supports the Cowley et al [2004] explanation with brighter emission at dawn, roughly at the location of the open-closed field line boundary. The work of Bunce et al. [2008] provided further evidence for this interpretation through the use of the HST together in-situ data from the Cassini spacecraft. However Belenkaya et al. [2008, 2010, 2014] used a modelled magnetic field to calculate an OCFLB that often lies several degrees of latitude poleward of the auroral emission particularly at dawn. 49

50 Saturn s Aurora Quiet Figure 2.3 Saturn s aurora as observed in the ultraviolet by the Hubble Space Telescope while the magnetosphere is inferred to be relatively expanded. The image has been projected and plotted onto a latitude-longitude grid. The bright dawn arc can be seen together with the less intense noon-to-dusk region. Noon is at the bottom of the image, dawn to the left and dusk to the right. This is an image from the dataset of images used in this thesis. 50

51 In addition to this circular emission there are other types of auroral emissions observed. One type of recurring feature is dawn brightenings associated with an expansion of the auroral emission at dawn to higher latitudes, associated with a sudden compression of the magnetosphere by the solar wind. This was observed by Clarke et al. [2005] who noted that Saturn s auroral emission was different to either Earth s or Jupiter s in this regard, although compression induced contractions in the auroral oval are occasionally seen on the nightside at Earth [Boudouridis et al., 2003, 2004; Hubert et al. 2006]. The compression of the magnetosphere by the solar wind may cause tail reconnection as magnetic field lines are forced closer together. This causes a poleward expansion of the auroral emission as flux from the tail is closed shrinking the OCFLB where upward currents are located. This emission is centred in the dawn sector as a result of planetary rotation [Prangé et al., 2004; Clarke et al., 2005; Cowley et al., 2005; Nichols et al., 2014]. An example of this behaviour is shown in Figure 2.4 in a HST image of Saturn s southern aurora from February The strikingly different appearance of the auroral emission to that in the Figure 2.3 is particularly noted. Dayside reconnection is thought to behave differently at the outer planets due to the effect of magnetospheric plasma rotation at the magnetopause, dayside reconnection is thought to be favoured at dusk due to the reduced shear flow as compared to the case at dawn where the planet and the magnetosheath flow are flowing in opposite directions [Desroche et al., 2013]. The importance of the plasma beta parameter is also stressed by Masters et al [2012], this parameter is the ratio of the plasma pressure to the magnetic pressure. For this reason at Saturn it may be the case that to produce reconnection it is necessary that the two fields are closely anti-parallel restricting reconnection to particular locations on the magnetopause where the plasma beta parameter is low. The importance of the magnetosonic Mach number is also highlighted by Scurry and Russell [1991] that suggest that reconnection may be inhibited under high magnetosonic Mach number conditions (>~7) that are common in the outer solar system [Achilleos et al., 2006]. Huddleston et al. [1997] identified some signatures of reconnection at the outer planets with spacecraft data but suggested that reconnection may be limited. Signatures of reconnection have been observed in Cassini data by McAndrews et al. [2008], Badman et al. [2013] and Fuselier et al. [2014]. In addition to these indicators, Radioti et al. [2011] have presented evidence in UVIS data for reconnection-related auroral features in Saturn s dayside ultraviolet auroras, taking the form of sequential extended 51

52 bifurcations in the auroral oval in the noon to dusk sector. It was discussed above that the dusk sector is favoured for reconnection at Saturn due to the reduced flow shear across the boundary, between sub-corotating magnetospheric plasma on the inside and anti-sunward flowing magnetosheath plasma on the outside. The auroral bifurcations are found to recur on ~1-2 h time scales, and endure as discrete features for comparable or longer intervals, so that more than one such feature is often present simultaneously, moving slowly poleward and eastward at ~15% of rigid co-rotation [Radioti et al., 2013]. In the events studied by Radioti et al. [2011], the largest arcs were found to contain ~2 GWb of magnetic flux, corresponding to ~10% of the pre-existing flux lying poleward of the auroral oval, with the oval boundary at other local times expanding equatorward accordingly as the arcs moved poleward. Similar auroral structures have also been reported by Badman et al. [2012] in infrared emissions observed by the Cassini Visual and Infrared Mapping Spectrometer (VIMS), but in this case centred closer to noon. This dayside reconnection would be expected to occur during intervals of northward interplanetary magnetic field although this could not be observed easily in the work of Badman et al. [2012]. Small scale auroral features have also been observed that are suspected to be cusp emission associated with lobe reconnection during intervals of southward IMF. They have been observed as poleward patches of auroral emission close to noon as reported by Badman et al. [2012] and Gérard et al. [2005], and as seen at Earth by Milan et al. [2000]. Sub-structures have also been observed within the dawn arc, that appear to subcorotate with the planet. Grodent et al. [2011] reported the occurrence in highresolution UVIS pseudo-images of groups of small-scale auroral spots in the dawn and noon-sector oval. These were ~ km in size (~1000 km corresponds to ~1º co-latitude in the north-south direction), persisted for more than a few tens of minutes, and were found to rotate eastward at ~70% of rigid corotation. It was suggested that these structures might be formed by fluctuations in flow associated with Kelvin-Helmholtz (KH) waves at the magnetopause and/or boundary layer. Radioti et al. [2009] using HST images also reported the occurrence of isolated auroral patches in the dusk sector which were shown to be transient in nature, brightening and decaying on ~10-30 min time scales. It was suggested on the basis of comparisons with nearcontemporaneous in situ Cassini data that these could be formed by hot plasma injections within the magnetosphere, resulting in enhanced particle precipitation into 52

53 the atmosphere. Plasma injections are where particles are accelerated by enhanced fieldaligned currents to the ionosphere. They have been found to be present at all local times [Hill et al., 2005] and have been associated with rapid tail reconnection [Bunce et al., 2005b]. They are identified in in-situ Cassini data by the identification of localised hot plasma a reduction in magnetic field strength and field fluctuations. Often accompanied with direct particle and magnetic field indications are brief increases in the Saturn Kilometric Radiation. Saturn s Aurora - Compressed Figure 2.4 Saturn s aurora as observed in the ultraviolet by the Hubble Space while the magnetosphere is inferred to be relatively contracted. The same format is used as in Figure 2.3. The poleward expansion of the dawn arc is shown. This is an image from the dataset of images used in this thesis 53

54 Saturn s auroras are complex with other separate behaviours observed. There appears to be a separate outer ring of emission as observed by Grodent et al. [2010] at ~23 colatitude possibly associated with scattering of trapped hot magnetospheric particles into the loss cone. Also observed is the auroral footprint of Enceladus in a similar way to the auroral footprint of Io can be observed at Jupiter [Pryor et al., 2011]. However for Enceladus this emission ought to be an order of magnitude dimmer than that of Io at Jupiter [Pontius and Hill, 2006] at a few tenths of kilorayleigh. Pryor et al. [2011] found variable emission of ~10kR possibly due to variations in Enceladus plume activity. An important behaviour at Saturn that can direct the location and intensity of the auroral emission are the planetary period oscillations. Oscillations in many parameters related to Saturn s magnetosphere have been noted with the first such being the Saturn Kilometric Radiation by Desch and Kaiser [1981]. Since then oscillations of the same planetary period have been noticed in the charged particle properties, magnetic field data and have been found to be ubiquitous in the Saturnian system [Andrews et al., 2008]. Andrews et al. [2008, 2010, 2012] and work by Provan et al. [2009, 2013] have shown the existence of rotating current systems in the northern and southern hemispheres that have separate rotational periods of the order of ~10.6 and ~10.8 h respectively. Importantly however these rotational periods have drifted throughout the Cassini era almost converging at equinox before slowly diverging once again [Provan et al., 2013]. This rotating current system is still a subject of much study but one suggested mechanism is that of Southwood and Cowley [2014] that connects the current systems to the polar cap and open field lines. Importantly for this study the planetary period oscillations have been shown to influence the location of the auroral oval, produce a dawn-dusk oscillation and also vary the auroral intensity as the rotating current system moves around with the planetary period [Nichols et al., 2008, 2010a, 2010b]. Thus we see at Saturn a system that is strongly affected by rotational flows connected to planetary rotation but are also importantly influenced by the solar wind both through dynamic pressure changes and by dayside reconnection. This thesis explores different small-scale features observed in Saturn s aurora as previously observed by Gérard et al. [2011] and by Radioti et al. [2009], investigates how different interplanetary magnetic 54

55 field conditions can result in different auroral observations and provides a survey of auroral storm observations. 55

56 Chapter 3 Instrumentation Throughout the studies presented in this thesis instrumentation has been used principally from two sources, the Hubble Space Telescope (HST) and the Cassini spacecraft presently in orbit around Saturn. The Advanced Camera for Surveys (ACS) instrument was mainly used to image Saturn s aurora in the ultraviolet. The Cassini orbiter s flux gate magnetometer (MAG) was used to acquire in situ magnetic field data and the electron spectrometer (ELS) component of the Cassini Plasma Spectrometer (CAPS) was used to measure plasma electrons energies. These instruments are described in detail below. 3.1 Hubble Space Telescope Advanced Camera for Surveys The HST is currently in low Earth orbit, launched in 1990 by the Space Shuttle and subsequently serviced during Space Shuttle service missions in 1993, 1997, 1999, 2002, and 2009 where various instruments were swapped and repairs carried out. A basic schematic of the HST is shown in Figure 3.1 showing that the largest component is the optical assembly. There are presently six scientific instruments on board HST that have replaced six former instruments on the telescope. The Advanced Camera for Surveys (ACS) is used in the work presented in this thesis and is described in detail below. It is used as the primary imaging instrument covering the ultraviolet to near-infrared spectrum. The Space Telescope Imaging Spectrograph (STIS) instrument has been used in previous work for imaging Saturn with a main function of providing a spectrograph with imaging seen as subsidiary. Images from STIS are used Chapter 7 alongside the newer ACS images. The Cosmic Origin Spectrograph (COS) instrument is designed to study faint point sources in the ultraviolet. The Wide Field Camera 3 (WFC3) is designed for imaging principally in the visible but can image in the near infra-red with a wide field of view. The near infrared camera and multi-object spectrometer (NICMOS) is an instrument that can image in the infrared and is a spectrometer but has largely been superseded by WFC3 due to its greater versatility in a similar spectral window. The final instrument is the Fine Guidance Sensor (FGS) which acts as a guiding system using interferometry of target stars as it is important for a telescope to be able to target and track accurately. However the instrument has been used as a scientific instrument to detect binary star systems. Before being replaced the Wide Field Planetary Camera 2 (WFPC2) was also used during a small number of auroral imaging campaigns, notably 56

57 for a period when the ACS malfunctioned in 2007 and utilised in Chapter 7 despite the greatly reduced imaging quality as compared to the ACS. Hubble Space Telescope Figure 3.1 Diagram of the HST s spacecraft components and assembly. The instruments are contained within the Axial Science Instruments identified section at the rear of the spacecraft [Lallo et al., 2012]. 3.2 Advanced Camera for Surveys Instrument As briefly described above, the Advanced Camera for Surveys instrument is the HST s primary imaging instrument in the ultraviolet to infrared range of the spectrum and is fully described by Ubeda et al. [2014]. It replaced the Faint Object Camera (FOC) on the 2002 servicing mission. The ACS instrument features three separate channels that serve different purposes. The Wide Field Channel (WFC) takes its name from its large field of view of 202 x 202 arcsec 2 with two 2048 x 4096 charge-coupled devices as detectors. It has a spectral range of nm. The High-Resolution Channel (HRC) has a 1024 x 1024 pixel charge-coupled device with smaller field of view than the WFC but with much greater spatial resolution. It is particularly efficient in the near-ultraviolet spectral window. The Solar Blind Channel (SBC) is a photon counting device 57

58 particularly efficient in the ultraviolet spectrum between 115 and 170 nm and is the channel used for the observations used in this work. It is solar blind as it blocks visible light to allow fainter ultraviolet light to be detected. It consists of a microchannel plate that is solar-blind and made of Caesium Iodide (CsI) with a multianode microchannel array (MAMA) readout. The MAMA has 1024 x 1024 pixels with each of these pixels being 25 x 25 μm with an average resolution of ~0.032 arcsec pixel-1. It is capable of two-dimensional imaging in the ultraviolet and some spectroscopic capability with a field of view of 34.6 x 30.8 arcsec2. Advanced Camera for Surveys Optical Design Figure 3.2 Diagram of the ACS s instruments Solar Blind Channel and High Resolution Channel shared optical assembly together with the filter wheels [Ubeda et al., 2014]. 58

59 The SBC and the HRC partially share optical channels which is shown in Figure 3.2. Light entering the ACS instrument from the Optical Telescope Assembly (OTA) is first reflected by two mirrors, indicated as M1 and M2 in the diagram. To select the HRC detector, a third mirror, M3, is moved into the light path. To select the SBC this mirror is moved out of the light path. It then enters a filter wheel capable of selecting a variety of different wavelength bands before interacting with the MAMA of the SBC channel. The filter wheel for the SBC of the ACS instrument contains eight different filters selectable for different observing purposes. Considering all filters, the SBC is capable of imaging from nm. The table below outlines the specific window of each filter.. Table 3.1 Tabulated SBC filters together with their spectral window and a brief description [Ubeda et al., 2014]. Filter name Spectral Window (nm) Description F115LP MgF 2 long pass F125LP CaF 2 long pass F140LP BaF 2 long pass F150LP Crystal quartz long pass F165LP Fused silica long pass F122M Ly-α PR110L LiF Prism PR130L CaF 2 Prism In this project the filters used have mainly been the F115LP filter and the F125LP filter. The F125LP filter excludes the contribution from Lyman-α which can be necessary due to the geocoronal Lyman-α emission in the vicinity of the HST. The F115LP filter was also used in Earth s shadow where this geocoronal emission is not problematic. The raw images obtained by the instrument are then corrected for geometric distortion, flatfielded and dark-count calibrated using calibration files from the Space Telescope Science Institute. 3.2 Cassini Spacecraft 59

60 The Cassini spacecraft was launched in 1997, and through gravity assists at Venus, Earth and Jupiter arrived at Saturn in July It is a collaborative project between NASA and ESA with the broad scientific goal of exploring the Saturnian system. It is a large spacecraft with a variety of scientific instruments. It also carried the Huygens lander to Saturn before it was released for its successful landing on Titan. The spacecraft is three axis stabilised such that it has control over its attitude using thrusters and reaction wheels while the instruments are fixed to its body. An illustration of the spacecraft is shown in Figure 3.3 identifying the main subsystems found on the spacecraft. Cassini Spacecraft Figure 3.3 Depiction of the Cassini spacecraft with a number of its systems identified. The magnetometer boom is shown, while the Field and Particle pallet is also identified that carries the CAPS instrument [NASA, 1996]. The spacecraft has a suite of twelve instruments used for a variety of different scientific work. The Cosmic Dust Analyser (CDA) can measure the characteristics of dust grains 60

61 in Saturn s environment. The Composite Infrared Spectrometer (CIRS) can determine temperatures and compositions of atmospheric gases. The Ion and Neutral Mass Spectrometer (INMS) directly analyses particles to learn about Saturn s environment. The Imaging Science subsystem (ISS) images in the visible and infrared range. The magnetospheric imaging instrument (MIMI) can image energetic particles in Saturn s magnetosphere with its ion and neutral camera. The Radar instrument can produce maps of moon surfaces. The Radio and Plasma Wave Science Instrument (RPWS) measures radio signals from Saturn and its environment and can detect electric and magnetic waves present in Saturn s magnetosphere. The Radio Science subsystem (RSS) can characterise objects by using small changes in radio waves it transmits from Cassini back to Earth. The Ultraviolet Imaging Spectrograph (UVIS) captures ultraviolet spectra of Saturn and can determine compositions. It can produce ultraviolet images through a rastering process. The Visible and Infrared Mapping Spectrometer (VIMS) can capture images in the infrared to determine composition and structures. Of particular use in this study are the magnetometer instrument (MAG) and the Cassini Plasma Spectrometer (CAPS). MAG is a dual technique magnetometer that can measure the direction and strength of the magnetic field around Saturn while CAPS can measure the energy and electrical charge of particles in Saturn s environment. These two instruments are explained a more detail below. 3.3 Cassini Spacecraft MAG instrument The magnetometer instrument on Cassini is known as a dual technique magnetometer and is fully described by Dougherty et al. [2004]. It consists of a fluxgate magnetometer (FGM) together with a Vector/Scalar Helium Magnetometer (V/SHM). An 11m long non-metallic boom is attached to the spacecraft in order to isolate the two parts of the magnetometer from the magnetic field generated by the spacecraft. The FGM part of the instrument is located halfway along the boom, with the V/SHM located at the end of the boom. This dual technique can have its advantages by allowing the magnetic field of the spacecraft to be better characterised and background subtracted. The two instruments can also be used to calibrate each other and allow for redundancy. The V/SHM instrument failed in November 2005 leaving the FGM as the only working magnetometer. This remaining instrument is still capable of characterising the system without its companion magnetometer. The FGM system and its electronics are shown in Figure

62 Cassini Magnetometer Instrument Figure 3.4 Picture of the the Cassini Magnetometer instrument and its associated electronics board [Dougherty et al., 2004] The FGM measures the strength of the surrounding magnetic field in three mutual orthogonal directions. In order to accomplish this it uses three individual sensors, one for each spatial dimension. Each of these sensors produces a voltage output that relates to the magnetic field detected along its axis. In order to prevent changes in alignment between the three sensors that would introduce error, the sensors are housed in a ceramic block that minimizes changes through thermal expansion by having a low thermal expansion coefficient. Each individual sensor from the three axis instrument works in the same way. For any fluxgate magnetometer the method used is the same. A highly magnetically permeable alloy is used in a ring wrapped in a drive coil winding which is itself wrapped by a second coil known as the sense winding. When a square wave form is driven through the coil the central ring goes through a cycle of magnetisation. The two halves of the 62

63 ring core generate an oppositely orientated field. The square wave waveform produces a cycle of magnetized, unmagnetised, inversely magnetised, unmagnetised and a return to magnetised in each half of the ring with a π phase difference between the cycle in each half. When an external magnetic field is applied (such as the one to be measured) the half-ring that is at that time oppositely orientated to the field will be less magnetised than were the external field not present and the length of time for which the ring is magnetised is reduced. The opposite is true of the half ring magnetized in the same direction as the field. For this ring the time spent magnetised is increased. When there is no external magnetic field the two ring fields cancel each other out completely, whereas when an external field is applied this is no longer the case. This change in flux due to the external field induces a voltage in the sense winding. This induced voltage has a frequency that is twice that of the input square wave signal and an amplitude that is proportional to the external magnetic field strength. The magnetometer is capable of recording 32 samples s -1 although this level of precision is not made use of in this work where instead one minute averages are used. At Saturn the instrument must also be capable of detecting magnetic fields that change over several orders of magnitude. For this reason the magnetometer can change between four different range modes depending on the concurrent magnitude of the magnetic field. 3.4 CAPs The CAPs instrument is itself made up of three separate sensors with related functions and is fully described by Young et al. [2004]. The entire instrument schematic is shown in Figure 3.5 with its three constituent parts. The ion mass spectrometer (IMS) and the ion beam spectrometer (IBS) focus on determining the flux and energies per charge of positive ions. The IBS can also determine energy to charge ratios in the range of 1 ev 50 kev in order to determine ion species and the angle of arrival at the instrument. The ion mass spectrometer s energy also ranges from 1 ev 50 kev. The third instrument is the electron spectrometer (ELS) that is used in this study that allows the flux of electrons with energy per charge and angle of arrival at the instrument aperture and can measure energies in the range of ,000 ev. The CAPS instrument is situated on a rotating actuator that moves the instruments through 208. This movement is required 63

64 by the three axis stabilisation of the spacecraft in order for the instruments to be used in a variety of different fields of view. CAPS Instrument Figure 3.5 The CAPS instrument on Cassini. The ELS instrument is shown towards the top of the image with its characteristic hemispherical analyser [Young et al., 2004]. The ELS instrument is a hemispherical top hat electrostatic analyser with a baffled collimator. The aperture allows electrons to enter through a field of view of 5 by 160. This field of view consists of eight different pixel elements of 20 width each. An electric potential applied to a set of plates at the aperture allows through only electrons within a certain energy range. Those electrons that fail to meet this criterion instead hit the walls of the instrument. Having passed through these hemispherical plates the electrons hit micro-channel plates and are thus registered by the electronics of the 64

65 instrument. By stepping through a series of 96 different values of electric potential of the analyser plates it is therefore possible to produce an energy spectrum. 65

66 Chapter 4 Observations of Small-Scale Features in the Dawn sector of Saturn s Aurora 4.1 Introduction Grodent et al. [2011] reported the occurrence in high-resolution UVIS pseudo-images of groups of small-scale auroral spots in the dawn and noon-sector oval as discussed in Chapter 2. It was suggested that these structures might be formed by fluctuations in flow associated with Kelvin-Helmholtz (KH) waves at the magnetopause and/or boundary layer. It is evident that simultaneous conjugate observations of such features could provide key information about their physical origin. Equatorially-generated KH waves, for example, would produce emissions that are spatially conjugate in the two hemispheres. The simultaneous conjugate images of Saturn s dayside auroras obtained during the 2009 HST campaign can be used to investigate the conjugacy of these auroral features. Initial results were discussed by Nichols et al. [2009], who noted that while large-scale auroral characteristics generally varied in concert north and south, numerous examples of non-conjugate features were also present. In this chapter, smallscale auroral structures found in the dawn-noon sector are examined in more detail using this unique equinoctial data set, in particular the issue of their conjugacy, or otherwise, in the two hemispheres, and discuss implications for the physical origins of these features. 4.2 Data Employed in this Study The Saturn equinoctial campaign used the Advanced Camera for Surveys (ACS) instrument on the HST, specifically employing the Solar Blind Channel (SBC) [Nichols et al., 2009]. At Saturn s distance from Earth (8.39 AU during this campaign), this field of view is sufficient to encompass the whole of the planet plus a portion of the rings, thus allowing both northern and southern auroras to be observed simultaneously. The spatial resolution of the images, determined by the ACS/SBS point spread function (corresponding to ~2 detector pixels), is ~400 km at normal incidence. The 2009 campaign consisted of 32 one-orbit visits that took place between 23 January and 7 March 2009, with opposition between Saturn and Earth occurring on 8 March The visits were organized into seven groups termed A-G, each consisting of either six (A and B) or four (C-G) individual visits, thus termed A1 to A6, 66

67 B1 to B6, and so on. However, for operational reasons planned visit A2 was delayed and undertaken out of sequence between visits A5 and A6, such that for clarity we have here re-labelled visits A3-A5 and A2 as Aʹ2-Aʹ5 in correct time sequence. The vertical dashed lines in Figure 4.1, which spans the interval January-March 2009 (day of year (DoY) 1-90), show the times of each visit. The interval between each visit varies over a wide range of time scales, from successive HST orbits separated by ~1.5 h within groups, to intervals up to ~11 days between groups. These times are superposed on modelled solar wind parameters at Saturn projected from near-earth observations using the MHD code of Zieger and Hansen [2008]. The accuracy of these modelled solar wind parameters depends strongly upon Saturn being close to opposition. This fidelity decreases dramatically beyond 60 days either side of opposition. The visits in this campaign were all within 45 days of opposition (8 March 2009, DoY 67). From top to bottom we show the solar wind number density, velocity, dynamic pressure, and field strength, together with the radial distance of Saturn s subsolar magnetopause obtained from the dynamic pressure using the Kanani et al. [2010] model. It can be seen that the modelled solar wind velocity is relatively low throughout the interval, ~ km s -1, while the number density is much more variable particularly toward lower values, thus leading to modest dynamic pressures typically below a few 10-2 npa, sometimes dropping as low as 10-3 npa. Modelled subsolar magnetopause distances then lie typically in the range ~20-30 R S, corresponding to HST groups A, B, E, and F, while sometimes expanding as far as ~40 R S, as during groups C, D, and G. (R S is Saturn s equatorial 1 bar radius, equal to 60,268 km.) It may be noted that this interval corresponds to the bottom of the recent deep and extended minimum in the solar cycle, with averaged sunspot numbers near to zero. During each ~40 min visit nineteen individual images were obtained each with 100 s exposure, except for visit G4 when seventeen images were obtained. The overall data set thus consists of 606 images. Two filters were also employed, F115LP and F125LP, covering the FUV wavelength ranges nm and nm, respectively. The former thus admits both H Lyman-α and H 2 Lyman/Werner band emission, while the latter excludes H Lyman-α, which is important when contamination from Earth s geocorona is likely. In groups A and B the first seven images in each visit were obtained using the F115LP filter, while the remaining twelve employed the F125LP filter. In groups C-G the F115LP filter was used throughout. The images were then 67

68 subject to a pipeline reduction process involving calibration using relevant files obtained from the Space Science Telescope Institute, location of the planetary disc and Modelled Solar Wind and Magnetopause Parameters Figure 4.1 Plot showing modelled solar wind and magnetopause parameters at Saturn for the interval January-March 2009 (DoY 1 90). From top to bottom we show the solar wind number density (m -3 ), velocity (km s -1 ), dynamic pressure (npa), and field strength (nt), together with the radial distance to the subsolar magnetopause (R S ). The solar wind parameters were propagated from Earth using the one-dimensional MHD model of Zieger and Hansen [2008], while the magnetopause distance was obtained from the dynamic pressure using the model of Kanani et al. [2010]. The superposed vertical dashed lines show the times of the 32 HST visits whose images are studied in this study. scaling to a standard distance of 8.2 AU, and removal of reflected sunlight (see Nichols et al. [2009] for further details). Here we have also co-added the data from each pair of adjacent images to improve the counting statistics, though not for adjacent images in groups A and B that use different filters. This results in a final data set of 562 co-added images, 132 using the F125LP filter (groups A and B only), and 430 using the F115LP filter.. 68

69 The polar auroral data from each co-added image were then mapped onto latitudelongitude grids in each hemisphere, simulating the view looking down onto the planet from the north in both cases, with the images being clipped near the limb to avoid overstretching as the view becomes increasingly oblique. The mapping assumes an auroral height of 1100 km above the 1 bar reference spheroid, corresponding to the peak in the emission height profile [Gérard et al., 2009]. These mapped images represent the primary resource employed in this study, as discussed in the following sections. 4.3 Observations of Saturn s aurora at Dawn Auroral morphology and dynamics Figure 4.2 shows a representative set of images from visit D3 on 18 February 2009, where the four sets of plots display the northern and southern auroral emissions from co-added images 5, 9, 13, and 17, separated from each other by intervals of ~9 min. The visit and image number are shown at the top of each set, together with the centre time of the data in the plot. The top two panels in each plot show polar strips containing the auroral emissions in the original reduced images for the northern (left) and southern (right) hemispheres, where the white dashed line shows the planetary limb. The log intensity scale is shown at the top of the figure, saturated red at 100 kr. The middle panels show the corresponding projected images, plotted for ease of comparison with noon at the bottom and dawn on the left in both hemispheres, such that the view is looking down from the north in each case (thus through the planet in the south). White dotted lines show co-latitude circles at 10º intervals and LT meridians at 2 h intervals. The blue and red dotted circles show magnetically conjugate latitude bands in the north and south, respectively, which encompass the bulk of the auroral emission in both hemispheres. The emission intensity (in kr) has then been averaged in latitude over each of these bands in LT strips 0.05 h wide, and is plotted versus local time (LT) in the panel beneath the images for both the northern (blue) and southern (red) hemispheres. These profiles have also been subject to a simple correction for limb brightening through multiplication by the cosine of the viewing angle to the local vertical [e.g., Grodent et al., 2005]. 69

70 Observations of Small Scale Features at Dawn 70

71 Figure 4.2 Saturn auroral observations from HST visit D3 on 18 February 2009, displaying data from co-added images 5, 9, 13, and 17, as indicated at the top of each plot set. The centre times of the co-added images are also given, separated from each other by intervals of ~9 min. The top two panels in each set show polar strips containing the auroral emissions in the original reduced images for the northern (left) and southern (right) hemispheres, where the white dashed line shows the planetary limb. The intensity scale is shown at the top of the figure, saturated red at 100 kr. The middle panels in each set show the corresponding projected images, both plotted with noon at the bottom and dawn on the left. White dotted lines show co-latitude circles at 10º intervals and LT meridians at 2 h intervals, while the blue and red dotted circular segments in the north and south, respectively, show magnetically conjugate latitude bands that span the main emissions in both hemispheres. The lower panels in each set show the emission intensity in kr averaged in latitude over these bands in the north (blue) and south (red), to which a simple correction for limb brightening has been applied, and plotted versus LT. The lettered arrows in this panel mark the principal peaks in emission (other than the standing dawn arc), the positions of which are determined as indicated in section 4.2. The corresponding features in the images are also indicated by similarly lettered arrows. The overall auroral morphology in these images follows the usual pattern, with bright arc-like emissions at dawn and weaker forms at dusk in both hemispheres. However, it can also be seen that in addition to the quasi-steady dawn arc, patch-like emissions are observed between dawn and noon in both hemispheres, in general labelled a, b, c, and so on, in the north, and a, b, c in the south, that yield similarly-labelled peaks in the intensity-lt line plots beneath the images. (The near-fixed peaks due to the quasisteady dawn arcs are not so labelled, however.) Examination of the location of these patches and corresponding peaks from plot-to-plot in the figure shows consistent eastward motion in both hemispheres, as quantified below. As also shown below (section 4.3), such dawnside patches represent a very common feature in our data set, whose location in the oval, spatial scale, and motion are all similar to the small-scale spot auroral structures previously identified in Cassini UVIS pseudo-images (albeit at greater spatial resolution in the best images) by Grodent et al. [2011]. The similarity is even more striking in the UVIS images presented in Figure 4 of the latter paper, which have a spatial resolution ~ km, more comparable to that of the HST images employed here. If we newly examine the relation between features observed in the north and south, however, we see that while patches are indeed simultaneously present in both hemispheres, they are not generally closely conjugate, but instead are significantly displaced from each other in LT. In Figure 4.2, in particular, the maximum associated with patch a in the north clearly falls near-midway between patches a and b in the 71

72 south throughout the interval, with peak emissions in opposite hemispheres being separated by ~1 h LT. The weaker near-noon maxima (labelled b in the north and c in the south), however, are more nearly coincident. Propagation of Dawn Patches Figure 4.3 Plots showing the LT of principal maxima in the intensity-lt line plots (see Figure 4.2) plotted versus image time for the sequence of images obtained from a given visit. Blue triangles and red inverted triangles correspond to maxima in the north and south, respectively, and the image time is relative to the center time of the first coadded image in the visit. Least-squares linear fits to image-to-image sequences of maxima are also shown, whose gradients and corresponding rotation periods (with formal uncertainty estimates) are indicated on the right, identified by corresponding letters a, b, c, and a, b, c, and so on. The gradients are given as fractions of h rigid corotation, corresponding to angular velocity Ω S. Results are shown for (a) visit D3 (as in Figure 4.2), and (b) visit Aʹ4 where the one-point data gap at ~0.23 h is due to the change in ACS/SBC filter, such that a co-added image is not formed at this time. 72

73 In order to quantify the relative locations of these patches, and their longitudinal motions, we have determined the LT position of the principal intensity maxima, and have followed them from image to image within each visit. The position of a maximum has been determined by considering the LT range in the intensity-lt line plot where the intensity lies continuously within 80% of a peak value, and calculating an intensityweighted averaged LT within this range. This procedure yields the positions of the maxima shown by the labeled vertical lines in the intensity-lt in plots Figure 4.2. We then plot the LT of these maxima versus image time for each of the images in a given HST visit, with results for visit D3 (as in Figure 4.2) being shown in panel (a) of Figure 4.3. Blue symbols (triangles) correspond to maxima in the north, red (inverted triangles) to maxima in the south, and the time scale is relative to the centre time of the first co-added image in each visit. It can be seen that the sequence of maxima from image-to-image corresponding to a given moving patch is very clear, displaying sequentially displaced peaks in the north and the south in the post-dawn sector, with more nearly coincident features near to noon. All these features drift rather steadily eastward with time over the interval, with the near-noon features losing their identity before the end of the visit as they propagate into the early afternoon sector. Least squares linear fits to these data then yield the overall angular velocities Ω of these motions, indicated (together with the formal uncertainty estimate) in the boxes to the right of the plot, labeled by the patch identifier letter. These angular velocities are expressed as fractions of rigid corotation with Saturn, where the rotation period of the latter is taken for definiteness to correspond to exactly h (such that S rad s ). Internal planetary periods of h were inferred by Ω Anderson and Schubert [2007] based on planetary oblateness, and h by Read et al. [2009] based on considerations of Saturn s zonal winds, with the value taken here lying between these two. It can be seen that the rotation rates in this case correspond typically to ( ΩΩS) , not dissimilar to the results of Grodent et al. [2011], though feature a is significantly slower. The corresponding rotation periods of the auroral features, typically ~12-15 h, are also given in the figure. A second more complicated example is shown in panel (b) of Figure 4.3, taken from visit Aʹ4, where the one-point data gap at ~0.23 h results from the change in ACS/SBC filter as mentioned in section 4.2. In this case interleaved northern and southern maxima are again evident in the pre-noon sector, as seen in the sequence a, a, b with

74 increasing LT in the first half of the visit. However, the picture is then complicated by a patch bifurcation event in the northern hemisphere that occurs around ~0.35 h, in which northern patch a splits into patches b and c. Patch b remains located between southern patches a and b, moving eastward more slowly than before (all modestly sub-corotating as in panel (a)), while non-conjugate patch c moves with significantly super-corotating speed past b into the post-noon sector. Patches c and d are initially nearly co-located at ~13 h LT, similar to c and b in D3 (panel (a)), but then significantly diverge due to the sub-corotational motion of c, and the modestly supercorotational motion of d. Local Time Extent Figure 4.4 Plot showing the LT ranges on the vertical axis over which either single or multiple dawn auroral patches were observed during each HST visit, the latter being indicated sequentially in time along the horizontal axis. Blue bars correspond to the northern hemisphere and red to the southern. Diamonds indicate where patches were formed in the field of view during a given visit, and stars where they lost their identity. Where no such symbols are shown, the patches were either already present at the start of the visit, or had not disappeared at its end. For further discussion see section 4.2. The central dashed line indicates the noon meridian. 74

75 Statistical Analysis We now provide an overview of the dawn patch phenomenon determined from all of the equinox campaign visits, analysed in the same manner as in section In Figure 4.4 we first show the LT ranges over which either single or multiple patches were observed, if any, during each visit, where the blue bar corresponds to the northern hemisphere and the red to the southern. It can be seen that patches are identified in 23 out of 32 visits, corresponding to ~72% of cases, such that it is a common phenomenon as indicated in section Furthermore, in 22 out of these 23 cases, patches are identified in both hemispheres, such that when they occur, they usually occur both north and south. Of the nine visits in which patches were not identified, four correspond to the closely-spaced visits of group G extending over a ~5 h interval, and a further four to two closely-spaced pairs in B3-B4 and C2-C3. Comparison with Figure 4.2 does not suggest any consistent relation between the occurrence of these patches, or lack thereof, and the modeled state of expansion of the magnetosphere. Figure 4.4 also shows that the patches are primarily a dawnside phenomenon as indicated above, only rarely being observed beyond ~13 h LT. The diamonds and stars plotted on the vertical bars show where patches formed or lost their identity within a given visit, respectively. Where no such symbols are shown, patches were either already present at the start of the visit or had not disappeared at its end. Patch formation within the field of view centres near to ~9 h LT, though with many being present at earlier LTs at the start of the visits. Disappearances are centred near to ~12 h LT, the dashed horizontal line in the figure indicating noon. Diamonds drawn a short way along the bar (visit B6 north and south) indicate the centre of a newly-formed patch that extended somewhat westward in LT, but, as in all cases, subsequently propagated eastward. Stars drawn part-way along the bars (visits Aʹ2 south, Aʹ5 north and south) indicate the presence of multiple patches, one of which loses its identity at an earlier LT than one of the others. In Figure 4.5 we examine some typical properties of the patches. In panel (a) we consider the LT extent of individual patches, defined here to be the width within which the intensity remains above 80% of the peak value in the intensity-lt plots, representing the width of the brightest central region of a patch, generally lying within a more extended but less well-defined region of emission. Values have been averaged 75

76 Statistical Properties of Dawn Patches Figure 4.5 Histograms showing overall properties of the dawn auroral patches determined from the ensemble of equinox campaign visits. Panel (a) shows the LT extent of individual patches, defined as the width within which the intensity remains above 80% of the peak value in the intensity-lt line plots, with values being averaged over all the images within a visit over the lifetime of the patch. Blue (downward hatching left to right) corresponds to northern patches, red (upward hatching left to right) to southern, and black to the overall data set. Mean values and standard deviations of these distributions are given numerically in the panel, as well as being indicated for the overall distribution by the vertical arrow and horizontal bar. Panel (b) shows in the same format a corresponding histogram of peak limb-brightening corrected emission intensities (kr) similarly averaged over the lifetime of a patch on a given visit. Panel (c) shows the LT displacement between adjacent pairs of patches in the northern and southern hemispheres, determined from the mean displacements of the linear fits (e.g., Figure 4.3) over the intervals of time when both patches were present. Positive values correspond to the northern patch being earlier in LT than the southern, and negative values vice versa, with all successive patch pairs being represented. 76

77 over the lifetime of an individual patch within a given visit to produce the histograms shown, blue corresponding to northern patches, red to southern, and black to the overall data set. No significant difference is observed between the north and south (see the statistical data within the figure panel), the overall mean and standard deviation (SD) being 0.22±0.11 h LT. Overall, the distribution of widths spans from ~0.08 h LT, essentially the instrument point spread function (PSF), to ~0.37 h LT, equivalent to ~4.5 times the PSF. The PSF describes how well an imaging system can resolve a particular point source and the extent to which is spread out (blurred) in the image. This distribution also corresponds to a range of east-west distances from the ~400 km resolution limit up to ~2000 km, with the mean value corresponding to ~1000 km. These scales are thus comparable to the ~ km scales of the auroral spots previously studied by Grodent et al. [2011]. Panel (b) shows a corresponding histogram of peak limb-brightening corrected emission intensities similarly averaged over the patch lifetime, typical values extending from ~2 to ~7 kr. These values are somewhat smaller than the ~10-30 kr range for individual spots quoted by Grodent et al. [2011], likely as a result of the latitudinal averaging undertaken here. Panel (c) of Figure 4.5 newly addresses the issue of patch conjugacy. Here we show the LT displacement between adjacent pairs of patches in the northern and southern hemispheres, determined from the mean displacements of the linear fits (such as those shown in Figure 4.3) over the intervals of time when both patches were present. A positive value corresponds to the northern patch being at an earlier LT than the southern, and a negative value vice versa, with all successive patch pairs being represented. If the patches were an essentially conjugate phenomenon, the distribution would be strongly peaked near to zero displacement, with corresponding patches lying within ~±0.2 h LT of each other according to the above width determinations. The histogram in panel (c) shows that very few patch pairs are conjugate within this limit, however, with observed separations typically being more than double this, extending up to ~±1 h LT. Rather, the distribution of separations is distinctly bi-modal between positive and negative values, indicating a preference for longitudinally displaced peaks north and south, such as those shown in Figures 4.2 and 4.3. The means and SDs of the positive and negative values taken separately are 0.56±0.33 and -0.62±0.51 h LT, respectively. However, roughly equal numbers of positive and negative values are represented in the plot (29 positive and 21 negative), indicating no significant 77

78 preference for patches in the north to lead patches in the south or vice versa. The mean and SD of the distribution overall is 0.05±0.72 h LT, consistent with zero. Dawn Patch Angular Frequencies Figure 4.6 Histograms of (a) dawn patch normalized angular velocities, and (b) corresponding rotation periods, determined from linear fits to the image-to-image patch position determinations such as those shown in Figure 4.3. The angular velocities are normalized to the planetary angular velocity, taken to correspond to a period of exactly h. Blue corresponds to northern patches, red to southern, and black to the overall distribution, as in Figure 4.5. The overall mean value and standard deviation is again shown by the vertical black arrow and horizontal bar. The purple arrow indicates the planetary rotation frequency/period, while the orange arrow indicates the frequency/period of the rotating planetary period oscillations (PPO), which is essentially the same on this scale. The green horizontal bars indicate the range of 78

79 plasma rotation angular velocities/periods inferred from Cassini ion velocity measurements in differing equatorial radial ranges. Bar A corresponds to the results of Wilson et al. [2009] for 6-10 R S, B to Thomsen et al. [2010] for R S, and C to Arridge et al. [2011] for R S. Panel (a) of Figure 4.6 shows histograms of the dawn patch rotation angular frequencies Ω normalized to Ω S as above, determined from linear fits such as those shown in Figure 4.3, while for ease of comprehension panel (b) shows the same data in terms of rotation periods. Blue again corresponds to northern patches, red to southern, and black to the overall distribution. The distributions are seen to be very broad in both hemispheres, spanning ( ΩΩS) 0.3 2, corresponding to periods to ~5-30 h. The purple arrow indicates near-rigid corotation with the planet at the period of h as discussed above, while the orange arrow emphasizes that the periods of the rotating planetary period oscillations (PPOs) in Saturn s magnetosphere, ~ h [e.g., Andrews et al., 2012], are also almost the same as rigid corotation on this scale. The histograms show that while a small fraction of the patches (~20%) are observed to super-corotate, most of these corresponding to bifurcation events such as that shown in panel (b) of Figure 4.3, the majority significantly sub-corotate. The overall mean and SD of the patch normalized angular velocity is ( ΩΩS) 0.78 ± 0.32 (the standard error of the mean being ±0.04), while the mean and SD of the rotation period is 15.8±5.9 h. Again, no significant differences within the uncertainties are observed between the separate northern and southern values, as given in the figure. It is of central physical interest to compare the patch angular frequencies with those of the thermal plasma measured in situ within the magnetosphere. Magnetic models indicate that the dayside auroras map to the outer part of the magnetosphere between the outer ring current and the magnetopause [Belenkaya et al., 2011], typically, say, to radial distances ~20 R S when the magnetosphere is reasonably expanded [Bunce et al. 2008a], as seems appropriate to most of the interval examined here according to Figure 4.1. The green horizontal bars in the upper parts of the panels in Figure 4.6 show the plasma rotation angular frequencies/periods derived from Cassini equatorial ion velocity measurements in three recent studies spanning radial ranges at increasing distance. Bar A from Wilson et al. [2009] corresponds to the radial range 6-10 R S in the dayside magnetosphere, mapping to lower latitudes than the main auroral oval, where 79

80 ( p S) Ω Ω with periods ~13-15 h. These values are seen to be comparable to the mean patch values, though spanning a much narrower range. Bar B corresponds to the results of Thomsen et al. [2010] at larger equatorial distances of R S, encompassing a wide range of LTs including the dayside, that cover a broader and lower range of values ( ) Ω Ω, corresponding to periods ~15-26 h. These p S values are seen to span the lower range of auroral patch angular velocities, with a mean value at the larger distances of ( ) 0.5 Ω Ω, rather lower than the mean angular p S velocity of the patches. Bar C then corresponds to the results of Arridge et al. [2011] at R S near Titan s orbit, which indicates a similarly broad range of values ( p S) Ω Ω, corresponding to periods ~18-35 h, centred on an even smaller median value of ( ) 0.4 velocities ( ) p Ω Ω. Overall, these results suggest typical plasma angular S p S Ω Ω in the region to which the patches likely map in the outer magnetosphere, compared with mean angular velocities of ( ΩΩS) 0.8 for the patches themselves. We thus infer that the patches typically propagate eastward relative to the plasma at modest angular velocities, as must certainly be the case for the fastest near-rigidly rotating and super-corotating patches. 4.4 Physical Origin of Dawn Sector Patches Relevance of the Kelvin-Helmholtz Instability We now consider the physical origin of the auroral phenomena discussed above, beginning with the eastward-propagating dawn patches. Grodent et al. [2011] suggested that such small-scale auroral structures might be caused by Kelvin-Helmholtz (KH) waves occurring at the outer boundary of the magnetosphere, either at the magnetopause or in the adjacent boundary layer, specifically through the currents that couple the resulting perturbed plasma flow with the ionosphere. Alternatively, ultralow frequency (ULF) field line resonances (FLRs) can be excited within the magnetosphere by such boundary wave sources, which similarly couple to the ionosphere [e.g., Southwood, 1974; Southwood and Hughes, 1983]. It is evident, however, that a KH source is inconsistent with the patch properties determined in this 80

81 study, both with regard to their inter-hemispheric symmetry, and to their azimuthal propagation. We briefly discuss each in turn. The KH instability operates principally in the magnetospheric equatorial region (see, e.g., Desroche et al. [2012] in the case of Jupiter), the effects of which are transmitted along the field to the ionosphere by Alfvén waves carrying field-aligned currents. The field perturbations produced by the equatorial vortical flows associated with the instability are perforce in opposite directions on either side of the equator, a configuration we refer to as being of odd symmetry, and hence so are the field-aligned currents. A relevant discussion is given, e.g., by Masters et al. [2010]. Enhanced auroral emissions are expected to be associated with downward electron acceleration in regions of field-aligned current directed upward with respect to the ionosphere, which with the above odd field symmetry implies auroral patches that are conjugate in the two hemispheres. This is directly contrary to the observations presented in section 4.3, illustrated in Figures 4.2 and 4.3, which show that patches in the two hemispheres are characteristically displaced in longitude, with maxima in one hemisphere being located near-centrally between maxima in the other. This requires instead that the magnetic perturbations and field-aligned currents associated with the patches have the same direction on either side of the equator, i.e. have even symmetry, and are thus oppositely-directed with respect to the ionosphere in conjugate hemispheres. Exactly the same issue concerns the FLRs that might be excited inside the magnetosphere by coupling to KH waves on the boundary. Under near-equinoctial conditions of northsouth symmetry in the magnetosphere, a source of odd symmetry on the boundary such as KH waves can excite odd mode FLRs inside the magnetosphere, principally the fundamental mode, but cannot excite even mode FLRs such as is required by our observations. We note that in both of the two terrestrial cases in which a magnetospheric ULF wave has been directly linked through a FLR with a KH wave at the boundary, the ULF wave has indeed been inferred to be in the fundamental mode [Rae et al., 2005; Agapitov et al., 2009]. Turning now to the azimuthal propagation characteristics of the patches, we note that unstable KH waves driven by flow shear at a boundary propagate at a speed which is intermediate between the speeds on either side of the boundary. In the context of Saturn s dayside magnetopause in which the magnetospheric plasma sub-corotates at ~100 km s -1 adjacent to the boundary [Thomsen et al., 2010; Arridge et al., 2011; 81

82 Wilson et al., 2012], while the magnetosheath plasma flows tailward from the sub-solar region at generally comparable speeds, the implication is that KH waves in the pre-noon sector perforce propagate at all points at an eastward speed that is less than or equal to the plasma speed inside the boundary. A corresponding angular velocity would also be imposed on any FLRs driven inside the magnetosphere by KH waves. This is again contrary to the properties of the dawn patches, which as shown in section and Figure 4.6, rotate eastward at a mean angular velocity which is considerably in excess of the plasma angular velocity in the outer magnetosphere, ~0.8 plasma angular velocity implied by the above flow speed of ~0.4 Ω S compared with a Ω S. As indicated in section 4.3.2, the field perturbations and currents associated with the patches thus propagate eastward relative to the plasma, rather than westward as required by a KH source. More specifically, growing KH modes are stationary in the frame in which the mass fluxes of the plasma (mass density times flow speed) are equal and opposite on either side of the boundary [e.g., Southwood, 1978]. At the dayside boundary at Saturn the mass densities are comparable on either side, though dominated by protons in the magnetosheath and water ions in the magnetosphere [Masters et al., 2010], so that the speed of the growing modes will be approximately the mean of the two flow speeds. If so, KH waves in the near-noon region where the magnetosheath flow is small will propagate eastward at half the speed of the magnetospheric plasma inside the boundary, i.e. at ~0.2 Ω S according to the above discussion. The eastward propagation then slows at earlier LTs around the boundary, to near zero where the tailward magnetosheath speed reaches ~100 km s -1. According to simple magnetosheath models this would occur near ~10 h LT [e.g., Stahara et al., 1989], in agreement with the results of Wilson et al. [2012]. At still earlier LTs the propagation then reverses to tailward. We note in this context that in a study of magnetopause boundary normal deflections, Masters et al. [2012a] found that these were consistent with tailward propagation in ~80% of cases, both at dawn and dusk. By contrast, the dawn auroral patches studied here move consistently eastward at all LTs later than ~8 h LT covered by the HST images, with relatively constant angular velocities over their observed lifetimes (e.g., Figures 4.3) averaging ~0.8 Ω S Relevance of Drift-Bounce Resonance Instability 82

83 These considerations thus show conclusively that the dawn patch phenomenon is not driven by KH waves at the boundary, nor by FLRs excited by them within the magnetosphere. A second possibility, however, is that ULF FLRs can be driven within the magnetosphere by resonant interactions with trapped magnetospheric particles. Such waves are known to commonly occur within the Earth s magnetosphere, with modes of even magnetic and current symmetry as indicated by our observations (principally second harmonic waves) being driven by drift-bounce resonance with trapped ring-current ions [e.g., Hughes et al., 1978; Yeoman and Wright, 2001; Wright et al., 2001; Baddeley et al., 2002, 2005]. Here we therefore make an initial assessment of the relevance of this mechanism to the dawn patch phenomenon at Saturn, starting with the conditions for resonance. The condition for local drift-bounce resonance derived by Southwood et al. [1969] is given by 83 ω mω = Nω, (4.1) d b where ω is the angular frequency of the wave in the inertial frame in which we work, m is the azimuthal wave number such that the wave varies with azimuthal angle φ as e -imφ, ω d is the particle azimuthal drift frequency due both to the bulk flow of the plasma in the inertial frame (the E B drift) and to grad-b and curvature drifts (eastward for ions at Saturn), ω b is the particle bounce frequency along the field lines, and N is any integer, positive, negative, or zero. However, net interchange of energy between waves and particles during the wave cycle requires that N be zero or even for waves whose magnetic fields are anti-symmetric about the equator ( odd modes), such that the wave electric fields associated with the field line motion are symmetric, and that N be odd for waves whose magnetic fields are symmetric about the equator ( even modes), such that the wave electric fields are anti-symmetric [e.g., Southwood and Kivelson, 1982]. As indicated above, our observations suggest the latter condition pertains to the dawn patches, such that we focus here upon N = ± 1 drift-bounce resonance with symmetric (e.g. second harmonic) even magnetic modes, rather than N = 0 drift resonance with anti-symmetric (e.g. fundamental) odd magnetic modes. We start by considering the wave parameters in equation (4.1) suggested by our observations. First, the azimuthal wave number m, equal to the total number of wavelengths in azimuth around the planet, can be found from the LT separation of

84 successive auroral patches in the ionosphere, corresponding to one wavelength, or equivalently from the separation of patches in opposite hemispheres, corresponding to half a wavelength. From the histogram in Figure 4.5c it can be seen that the typical value of the latter is ~0.6 h LT, thus implying typically that m 20. Corresponding ULF waves in the Earth s magnetosphere are also found to be of high m number, typically m [e.g., Wright et al., 2001; Baddeley et al., 2002, 2005]. Second, the angular frequency of the waves in the inertial frame ω can be found from the eastward angular velocity Ω of the auroral patches, the latter corresponding to the angular velocity of the wave phase fronts in the inertial frame. The angular frequency is then given by ω = m Ω. (4.2) From Figure 4.6 we find a typical value of -4-1 Ω rad s corresponding to a patch rotation period of ~14 h, which with m 20 gives ω rad s, corresponding to an oscillation period in the inertial frame of ~40 min. We note that Kleindienst et al. [2009] have shown that packets of Alfvénic field fluctuations at such frequencies (the above corresponding to ~0.4 mhz) are indeed commonly observed with significant amplitudes in Saturn s equatorial outer magnetosphere (see e.g. their Figures 1 and 2). However, they have not been extensively studied to date. Turning now to the particle frequencies, the drift frequency ω d is the sum of those due to plasma rotation at angular velocity p Ω (due to E B drift) and to the bounceaveraged sum of the grad-b and curvature drifts associated with magnetic field inhomogeneity at angular frequency ω B. We note that the resonance condition given by equation (4.1) can then be written as ( m ) B = N, where m b ( p ) ω ω ω ω = Ω Ω is the angular frequency of the waves in the plasma rest frame. From the discussion of Figure 4.5 in section 4.1.2, the results of Thomsen et al. [2010] and Arridge et -4-1 al. [2011] suggest typical plasma angular velocities of Ω rad s in the outer magnetosphere, corresponding to plasma rotation periods of ~25 h. The oscillation frequency in the plasma rest frame is then typically p ω -3-1, rad s corresponding to a period of ~80 min. We also assume a dipole field approximation as a first estimate of the bounce-averaged magnetic inhomogeneity drift, such that 84

85 85 ( αeq ) f B LW ω B qb R, (4.3) eq 2 S where L is the equatorial radius of the field lines in planetary radii ( R = 60,268 km as indicated above), W is the particle kinetic energy, q is the particle charge (where we assume singly-charged ions with q= e, the elementary charge), B = 21,136 nt is the dipole field strength at the planet s equator [Burton et al., 2010], and f B ( αeq ) eq S is a slowly-varying function of equatorial pitch angle α with values between 2 and 3 for eq α between 0º and 90º. We note that N = ± 1 drift-bounce resonance in a symmetric eq magnetic wave field favours smaller pitch-angle particles that spend significant time away from the node in the anti-symmetric wave electric field at the equator. The particle bounce frequency in the dipole field approximation is similarly ( α ) 12 π W ωb, (4.4) f LR 2m b eq S where m is the particle mass, and b ( eq ) f α is another slowly-varying function of equatorial pitch angle with values between 1.38 and 0.74 for α between 0º and 90º. eq We note that functions f B ( αeq ) and b ( eq ) f α are defined by integrals over the particle bounce motion that must in general be evaluated numerically [Hamlin et al., 1961], and such values have been employed in the results shown here (values are given in the caption to Figure 4.8). For most purposes, however, the simple approximate forms B ( eq ) sin eq and ( ) sin f α α et al., 2005]. f α α are sufficient [e.g., Baddeley b eq eq Results are shown in Figure 4.7, where the solid lines show the left side of equation (4.1) plotted versus ion energy W on a log scale, where the purple, red, green, and blue lines correspond to equatorial pitch angles α eq of 0º, 30º, 60º and 90º, respectively. The similarly color-coded dashed lines correspond to the right side of equation (4.1), for N = + 1 in the upper half of the plot and N = 1 in the lower half as indicated. Drift-bounce resonance for given N then occurs where the solid and dashed lines for a given pitch angle cross each other. For the case of protons in panel (a), driftbounce resonance with N = + 1 occurs at energies of a few kev, e.g. at ~5 kev for 30º

86 smaller pitch angle particles, corresponding to the high-energy tail of the low-energy proton population in the outer magnetosphere with typical energies ~100 ev [McAndrews et al., 2009], while drift-bounce resonance with N = 1 occurs at energies of a few MeV (off the scale of the plot), corresponding to the high-energy tail of the hot proton population with typical energies ~10 kev [Dialynas et al., 2009]. Neither condition thus corresponds to the central energy of a major proton population in Saturn s magnetosphere. For the case of water group ions in panel (b) (mass 17 assumed), drift-bounce resonance with N = + 1 occurs at energies of a few tens of kev, e.g. at ~30 kev for smaller 30º pitch angle particles, corresponding to typical energies of the hot water ion population in the outer magnetosphere, while drift-bounce resonance with N = 1 occurs at energies of a few hundred kev, corresponding to the high-energy tail of this population [Dialynas et al., 2009]. The former condition is thus well placed to tap the energy of the main hot water ion population in the outer magnetosphere if the particle distribution function contains free energy in the form of non-monotonic features. With regard to the disappearance of the auroral patches near to noon, we note that modelling of FLRs in the terrestrial context shows that they are strongly damped in the presence of auroral electron acceleration in the upward field-aligned current regions associated with their ionospheric coupling [Damiano and Johnson, 2012], such as we infer to be present at Saturn. This suggests that the disappearance of the auroral patches near to noon may be related to the diminution of the free energy source within the driving hot ion population, followed by rapid damping of the waves. The hot ions are likely injected near to midnight and drift eastward via dawn, in conformity with the LT properties of the patches. 86

87 87 Drift-Bounce Resonance

88 Figure 4.7 Plots showing conditions for drift-bounce resonance with ULF waves corresponding to the eastward-drifting dawn auroral patches, for (a) protons, and (b) water group ions (mass 17 assumed). The waves have azimuthal wave number m = 20, -3-1 an oscillation frequency in the plasma rest frame of ω = rad s, and are taken to have symmetric (e.g., second harmonic) magnetic perturbations relative to the equator so that principal resonances are those with N = ± 1. In each panel the solid lines show the left side of equation (4.1) plotted versus ion energy W on a log scale, where purple, red, green, and blue correspond to equatorial pitch angles α eq of 0º, 30º, 60º, and 90º, respectively (these curves being independent of ion mass). The similarly color-coded dashed lines correspond to the right side of equation (4.1), for N = + 1 in the upper half of the plot and N = 1 in the lower half as indicated. Numericallyintegrated values of the pitch-angle functions employed for α eq = 0, 30º, 60º, and 90º are f B ( αeq ) = 2.000, 2.552, 2.871, and 3.000, and fb ( α eq ) = 1.380, 1.000, 0.806, and 0.741, to three decimal places. Drift-bounce resonance occurs at ion energies where the solid and dashed lines of a given color cross. Drift resonance with N = 0 also occurs where the solid lines pass through zero near ~100 kev independent of ion mass, but the wave-particle energy exchange is zero for such particles for waves having symmetric magnetic and anti-symmetric electric perturbations Properties of Second Harmonic Field Line Resonances Having shown that the hot water ion population forms a suitable drift-bounce resonant energy source in the outer magnetosphere, we now outline the nature of the ULF waves proposed to be excited, namely an even mode FLR, specifically the second harmonic. To illustrate the properties of such waves we employ the simple box model of Wright and Allan [1996], following Southwood and Hughes [1983], in which Cartesian x, y, and z coordinates represent the radial, azimuthal, and northward directions in the magnetosphere, respectively. In the unperturbed system the plasma is at rest (i.e. we work in the frame of the large-scale flow), the magnetic field is uniform in the z direction, taken here to be given by B= B z ˆ to represent the north-to-south field of Saturn, and the plasma density ρ varies in the x direction, taken here to fall with x so that the Alfvén speed rises. The system is then limited in the z direction by two plane boundaries of large but finite Pedersen conductivity Σ P, representing the two ionospheres, located at z = ± l, such that z = 0 represents the equator. The waves in this system are characterized by three scale lengths or wave numbers, corresponding to the three spatial coordinates, namely δ x, k, and k y z. The wave numbers are k y in the y (azimuthal) direction, such that the wave varies as 88

89 exp j k y ( y ωt), and k z in the z (northward) direction, to lowest order quantized by the ionospheric boundary conditions according to k z nπ =, (4.5) 2l for n any positive integer. When n is odd, as studied by Wright and Allan [1996], the transverse magnetic and electric perturbations vary as sin ( kz ) and cos( ) z kz, respectively, thus being anti-symmetric and symmetric about the equator, as indicated above. However, when n is even, as studied here, the transverse magnetic and electric perturbations vary as cos( kz ) and sin ( ) z kz, respectively, thus being symmetric and z anti-symmetric about the equator. In both cases, to lowest order, the ionospheres form nodes in the wave electric field (and plasma flow), and anti-nodes in the transverse magnetic field. The corresponding angular frequency ω of the wave for a mode n field line resonance at position where V ( x) B µρ( x) A x = x is given by R nπ ω = kzva( xr) = VA( xr), (4.6) 2l = o is the Alfvén speed. The spatial scale x region in the x direction about x R is then given for any n by δ of the resonant z 2 1 d x = nπ µ Σ ( dv dx) = o P A x xr. (4.7) This equation may be obtained from equation (18) of Wright and Allan [1996], with the use of their equations (15) and (16) and the above expression for the Alfvén speed VA ( ) x (see also the Appendix). We now apply this model to the dawn patch observations presented in section 4.3, assuming that this phenomenon corresponds to a resonance of the required north-south symmetry, n = 2, propagating eastward in the outer magnetosphere. First, the wave number k in the y direction should correspond to azimuthal wave number m 20 at a y radial distance of ~20 R S, thus implying an east-west wavelength 2π 6R such -1 that k y 1R S. Second, equation (4.5) gives kl z π with n = 2, where l is the effective length of the field lines from the equator to the Pedersen-conducting ionospheric boundary. Unlike the box model, however, the Alfvén speed V A varies λy S 89

90 Properties of the Magnetosphere in the Equatorial Plane Figure 4.8 Plots showing radial profiles in Saturn s equatorial plane of (a) the magnetic field strength, (b) the plasma mass density, and (c) the Alfvén speed. The field and plasma parameters were derived by Kellett et al. [2011] from eleven orbits of closely equatorial Cassini data, these data being divided into four color-coded LT quadrants as indicated in the figure. 90

91 strongly along field lines in Saturn s magnetosphere, with equation (4.6) being generalized to ω L L nπ ds V A ( s), (4.8) where s is distance along field lines of total length 2L. Consideration of equation (4.8) shows that length l corresponds essentially to the half-width of the central plasma layer where the density remains high due to equatorial confinement of the rotating plasma and the field strength remains low, such that the Alfvén speed is also low, thus making the dominant contribution to the propagation time integral. Plasma observations indicate that l 5RS is reasonable in the outer magnetosphere [Richardson, 1995; -1 Thomsen et al., 2010], such that λz 10 RS for n = 2 and kz π R S. Third, we check the consistency of equation (4.6) with our above estimate of the wave angular -3-1 frequency in the plasma rest frame, ω rad s, such that we need to know the value of the equatorial Alfvén speed. This is shown in Figure 4.8, where in panels (a) and (b) we show radial profiles spanning ~3 to ~20 R S of the equatorial magnetic field strength and plasma mass density, respectively, determined by Kellett et al. [2011] from eleven orbits of the Cassini spacecraft, together with the consequent Alfvén speed in panel (c). The data have been divided into four color-coded LT sectors as indicated in the figure, plotted over the radial ranges for which data exists, of which the 6-12 h LT data (yellow lines) is most germane here. It can be seen that the Alfvén speed falls rapidly with distance in the inner magnetosphere, before flattening and then rising again -1 with much variability in the outer dayside region, where V 100 km s. Substitution of this value into equation (4.6) with n = 2 and l 5RS then yields a frequency A ω -3-1, in excellent consistency with the value rad s ω rad s -3-1 deduced from observations. Fourth, the radial scale of the resonance δ x is not obviously determinable from equation (4.7), since although we may estimate Σ 8 mho in the morning auroral zone from the work of Galand et al. [2011], P Figure 4.7 suggests that a wide range of spatial scales may be present in the outer region. Here we simply assume sufficiently steep spatial gradients in the Alfvén speed, comparable to the largest suggested in the figure, such that the resonant region is 91

92 sharply defined. Specifically, we take for definiteness δ x 0.3 RS, with the main current region then extending radially over ~1 R S in the equatorial plane (see below). This range maps to approximately half a degree of latitude in the ionosphere (~500 km north-south), rather less than is indicated by the projected images in Figures 4.2 and 4.8. However, as indicated in section 4.3, the latter are significantly broadened by the projection of obliquely viewed emission, which also has a finite vertical extent in the atmosphere. Wright and Allen Box Model 92

93 Figure 4.9 Montage of plots illustrating the properties of a second harmonic ( n = 2 ) Alfvén resonance obtained from the Wright and Allan [1996] box model (see Appendix), specifically for model parameters k yδ x = 0.3, and kl= y 5. The top three rows of plots show contours of normalized parameters plotted in the x z plane, while the bottom row similarly shows contours in the x y plane at the two ionospheric boundaries ( zl ) = ± 1. The x coordinate represents the radial direction in the equatorial region and the poleward direction in both hemispheres, y represents the azimuthal direction positive eastward, and z is anti-parallel to the background field, northward in the equatorial region. The x coordinate is normalized to δ x and shown relative to the position of the resonant field line at x = xr, y to π k y, and z to the effective half-length of the field lines l. Zero contours are indicated by black dotted lines, while red and blue contours show positive and negative normalized values, respectively, differing by factors of 2 from each other, starting with lowest values of ±0.5 nearest the zero lines. In the x z plots, the dashed and solid contours correspond to ky y π = 0, 2, 4... and ky y π = 0.5, 2.5, , respectively. Values for the other half wave cycle may be obtained simply by interchanging red and blue. The top row shows the normalized field line displacements in the x and y directions x xy, and velocities u xy,, the second row the normalized magnetic and electric fields b xy, and e xy,, and the third row the normalized current densities j xyz,,. The x y plots in the bottom row show the height-integrated Pedersen currents i xy, in the northern ionosphere, the sign of which reverses in the southern hemisphere, together with the ionospheric field-aligned current density j z applicable to both hemispheres by virtue of the symmetry. The plots correspond specifically to time t = 0 and arbitrary phase φ = 0 in the formulas in Appendix 1, with the perturbations then propagating in the y direction with phase speed ω k y, with ω given by equation (4.6). In Figure 4.9 we show a montage of wave parameter plots obtained from the Wright and Allan [1996] box model, the formulas for which are given in the Appendix, illustrating the nature of the magnetosphere-ionosphere perturbations considered here. Following the above discussion, the model parameters employed are n = 2 (second harmonic), k δ = 0.3, and kl= 5. The top three rows of plots show contours of y x y normalized parameters plotted in the x z plane, recalling that x represents the radial direction, and z northward anti-parallel to the background field. The x coordinate is normalized to δ x given by equation (4.7) and shown relative to the resonant field line for angular frequency ω at x = x over the range ±5 (representing a physical length at R the equator of ~3 R S ), while z is normalized to the effective half-length of the field lines 93

94 l (a physical length of ~5 R S ), such that the highly conducting model boundaries are located at the top and bottom of the plot at ±1. The y coordinate representing the azimuthal direction, positive eastward, is then directed into the plane of these diagrams, normalized to π k (a physical length of ~3 R y S ). Zero values are indicated by black dotted lines, while red and blue contours show positive and negative normalized values, respectively, which differ by factors of 2 from each other, starting with lowest values of ±0.5 nearest the zero lines. Dashed and solid contour lines correspond to k y π = 0, 2, 4... and k y π = 0.5, 2.5, , respectively, with values for the other y y half wave cycle being obtained simply by interchanging red and blue. The plots correspond specifically to time t = 0 in the formulas in the Appendix (and arbitrary phase φ = 0 ), with the perturbation then propagating in the y direction with time at phase speed ω k. y The top row of plots in Figure 4.9 show the normalized field line displacements in the x and y directions x and velocities xy, u, the second row the normalized magnetic and xy, electric fields b and xy, e, and the third row the normalized current densities xy, j. As xyz,, discussed above, the magnetic perturbations are symmetric about the equator, with nodes at ( zl ) = ± 0.5, and anti-nodes at the equator and at the boundaries ( zl ) = ± 1. Conversely, the electric field, together with the plasma velocities, field displacements, and transverse currents, are anti-symmetric about the equator, with antinodes at ( zl ) = ± 0.5, and nodes at the equator and at the boundaries. The field-aligned current j, however, is again symmetrical about the equator, with nodes at ( zl ) = ± 0.5 z, and anti-nodes at the equator and at the boundaries. This symmetry means, however, that the field-aligned current is oppositely directed relative to the ionosphere at conjugate points in the two hemispheres, such that if the current is into the ionosphere in one hemisphere, it is out of the ionosphere in the other, and vice versa. The ionospheric currents themselves are shown in the bottom row of Figure 4.9, where we plot contours of normalized parameters in the x y plane at the two ionospheric boundaries ( zl ) = ± 1, where in the ionosphere positive x represents the poleward direction in both hemispheres, and y is again positive eastward. Specifically we show contours of the field-aligned current j z applicable to both hemispheres by virtue of the 94

95 symmetry, together with the height-integrated Pedersen currents i which close this xy, current specifically for the northern ionosphere. The sign of the latter currents is reversed in the southern hemisphere. It can be seen that strong oscillatory peaks in the upward and downward field-aligned current occur within a few δ x of the resonance, as indicated above. Upward currents, potentially associated with downward electron acceleration and auroras, correspond to blue regions in the northern hemisphere and red regions in the southern, thus displaying the symmetry properties observed in the dawn auroral patches. For the parameters chosen, the main upward current regions have a north-south physical dimension of ~500 km in the ionosphere, with successive regions of given sign being displaced ~5000 km east-west in a given hemisphere Conclusions The equinox campaign images show that eastward-propagating patches are commonly present in the dawn-to-noon sector, being observed on ~70% of all HST visits. When they are observed, they are nearly always observed in both northern and southern hemispheres during a visit. Unexpectedly, however, the emission maxima are found generally not to be conjugate north and south, but are instead typically displaced in LT by 0.5~1 h. When multiple patches are present, maxima in one hemisphere typically fall near minima in the other. Angular velocities of rotation, while broadly spread, are found to average ~80% of rigid corotation, similar to typical ~70% values reported for auroral spots by Grodent et al. [2011]. These values are somewhat larger than the typical plasma angular velocities of ~40-50% of rigid corotation reported by Thomsen et al. [2010] and Arridge et al. [2011] for the outer magnetosphere to which these emissions are likely conjugate. Our results thus suggest that the patches propagate eastward through the plasma at modest relative angular speeds. These properties are consistent with those expected for a second harmonic ULF FLR wave propagating eastward through the plasma, for which the magnetic field and fieldaligned current perturbations are symmetric about the equator, thus being consistent with the above conjugacy findings. In this interpretation the waves have azimuthal wave number m 20, and propagate eastward with a wave period in the plasma rest frame of ~80 min, compared with ~40 min in the inertial frame. This period is consistent with expectations for a second harmonic resonance in the outer magnetosphere for an observed equatorial Alfvén speed of ~100 km s -1 and a plausible

96 effective length of the field lines of ~10 R S. With regard to the origin of the ULF wave, both symmetry and propagation properties have been shown to be inconsistent with a Kelvin-Helmholtz source excited in the magnetospheric boundary region. However, drift-bounce resonance with water ions at energies of a few tens of kev is a plausible mechanism for wave generation, such ions forming a major component of the hot plasma population in Saturn s outer magnetosphere. 96

97 Chapter 5 - Observations of Small-Scale Features in the Dusk sector of Saturn s Aurora 5.1 Introduction This study follows Chapter 4 directly as it makes use of the same HST campaign of 2009 where both the northern and southern auroral emission of Saturn is visible. It also uses a parallel approach to data selection, processing and analysis which is expounded in section 4.2 with a broader explanation of instrumentation found in Chapter 3. Figure 4.1 of Chapter 4 is a useful overview of the entire 2009 campaign together with propagated solar wind data from the period. However in this study rather than focusing on auroral emission at dawn, it is transient behaviour observed at dusk that is studied in this Chapter. As well as the small-scale features observed in Saturn s aurora by Grodent et al. [2011] as discussed in Chapter 4, Radioti et al. [2009] using HST images also reported the occurrence of isolated auroral patches in the dusk sector which were shown to be transient in nature, brightening and decaying on ~10-30 min time scales. It was suggested on the basis of comparisons with near-contemporaneous in situ Cassini data that these could be formed by hot plasma injections within the magnetosphere, resulting in enhanced particle precipitation into the atmosphere. Similar transient auroral patches in dusk sector were identified in the 2009 campaign and can in this case be studied by observing the two hemispheres to allow us to gain a better understanding of the mechanism of their formation. 5.2 Dusk Sector Transients Examples In Figure 5.1 we show a set of images from visit C4 in essentially the same format as that shown in Chapter 4 as Figure 4.2, although the optimized emission colour-scale has now been saturated red at 60 kr as indicated at the top of the figure. Eastward-moving non-conjugate auroral patches are again present in the dawn sector, while here we instead focus on emissions in the dusk sector. Examining the northern hemisphere data, it can be seen that the emission brightens substantially at latitudes modestly higher than that of the dawn patches over a broad post-noon sector centred near ~15 h LT between images 5 and 9, producing peak b in the intensity-lt line plot for the latter image. The 97

98 patch remains bright in image 13, though of diminished intensity, and then fades significantly in image 17. Examination of all the images in the visit shows that the lifetime of the transient emission is ~20 min, typical of the similarly-located transient events examined by Radioti et al. [2009]. Newly examining the conjugate emissions in these images, however, it can be seen that no related enhancement occurred at all in the southern hemisphere, an unexpected finding that will be shown below to be a general property of these events. In Figure 5.2a we show the LT of the auroral intensity maxima versus time for visit C4 in the same format as Figure 4.3, together with linear fits to the clear long-lived tracks. The dawn-side tracks again show the presence of non-conjugate patches in both northern and southern hemispheres propagating eastward. The near-stationary peak of dusk transient b in the northern hemisphere is also evident for ~20 min near ~15 LT, with no corresponding southern feature. The linear fit to these data indicate an overall eastward drift at ( ΩΩS) 0.45 ± 0.51, consistent with zero. A second example is shown in Figure 5.2b for visit G2. Unusually in this case, no dawn patches are present in either hemisphere in conformity with Figure 4.4, while two patches are present in the southern hemisphere only, patch a which is near-stationary close to noon with ( ΩΩS) 0.34 ± 0.12, and patch b which is present for ~10 min centred near ~15 LT which travels eastward at a super-corotational rate ( ΩΩS) 1.72 ± 0.36 corresponding to a rotation period of ~6 h Statistical Analysis We now provide an overview of the dusk transient phenomenon determined from all of the equinox campaign data, and begin in Figure 5.3 by showing the occurrence and LT extent of the phenomenon for all of the visits in the same format as Figure 4.4. Dusk patches were observed in 13 out of 32 visits, corresponding to ~41% of cases, showing that it is a common phenomenon, but does not occur as frequently as the dawn sector patches. The events are also seen to be well-distributed throughout the dataset, with at least one event occurring in each group of visits, except for group F where no dusk patches were observed over a ~5 h interval. Comparison with Figure 4.1 again does not suggest a clear relation between the occurrence of these transients and the modelled 98

99 Observations of Auroral Features in the Dusk Sector Figure 5.1 Auroral observations from HST visit C4 on 17 February 2009, in the same format as Figure 4.2, except that the emission intensity color scale is now saturated red at 60 kr as indicated at the top of the figure. 99

100 Propagation of Dusk Patches Figure 5.2 Plots showing the LT of principal maxima in the intensity-lt line plots, plotted versus image time for (a) visit C4 (as in Figure 7), and (b) visit G2, in the same format as Figure 4.3. Linear fits are shown to the data for the clear long-lived tracks. state of the magnetosphere. Most importantly, however, it is seen that although these events can be observed in both northern and southern hemispheres with roughly equal frequency within the small overall numbers involved, in no cases was a corresponding event observed in the conjugate hemisphere. The increased prevalence of diamonds and stars in Figure 5.3 compared with Figure 4.4 results from the short-lived nature of the dusk transient events compared with the dawn patches studied in Chapter 4, such that we generally see both the formation (diamonds) 100

101 Local Time Extent Figure 5.3 Plot showing the LT ranges on the vertical axis over which dusk transient auroral patches were observed during each HST visit, the latter being indicated sequentially in time along the horizontal axis. The format is the same as for Figure 4.4. and decay (stars) of the dusk events within a given visit. The mean location of formation is found to be ~14.5 h LT, while the mean location of decay is at ~15 h LT. Two single-hemisphere (southern) transient events are also observed close to noon on consecutive visits G2 and G3, marked by the dashed lines. Although their properties are similar to the more typical mid-afternoon events, it is not completely evident whether these form part of the same phenomenon, or whether they represent an additional rarer noon transient phenomenon. It may be noted from Figure 4.1 that these visits occurred during an extended interval of unusually low modelled dynamic pressure and consequent magnetospheric expansion. In Figure 5.4 we examine some typical properties of the dusk transient events, determined in a similar way and shown in a similar format to the Chapter 4 dawn patch 101

102 properties in Figure 4.5. The parameters of the two noon transients from visits G2 and G3 have been excluded in this figure. In panel (a) we show a histogram of the width in LT of the transient patches averaged over the patch lifetime, whose mean and standard deviation are found to be 0.35±0.19 h LT, respectively, with no significant difference between transients observed in the northern and southern hemispheres. These widths are thus somewhat larger than those of the dawn patches, which according to Figure 4.5 extend on average 0.22±0.11 h in LT. Panel (b) similarly shows a histogram of peak limb-corrected emission intensities averaged over the patch lifetime, with a mean value and standard deviation of 2.9±1.0 kr, similar to but slightly smaller than the mean value of 4.2±3.0 kr for the dawn patches. Panel (c) then shows a histogram of the lifetime of these events, spanning a range from ~6 to ~30 min, but with an overall mean and standard deviation of 13.6±7.4 min. We note that these values are very similar to the range of ~7 to 30 min quoted by Radioti et al. [2009], though the intensities given above are somewhat less than the ~10 kr quoted by these authors. In Figure 5.5 we show the angular frequency and rotation periods of the dusk transients determined from the linear fits to the image-to-image peak emission positions, as illustrated in Figure 5.2. Despite the smaller number of examples and the shorter lifetimes compared with the dawn patches, the distribution of values is even broader, spanning events that are essentially stationary in LT over their lifetime as for visit C4 shown in Figures 5.1 and 5.2a, and those that appear to super-corotate up to ~4 times the planetary angular velocity. Notwithstanding the breadth of the distribution, however, none of the dusk transients is found to move significantly westward against planetary rotation. 102

103 Statistical Properties of Dusk Patches Figure 5.4 Histograms showing overall properties of dusk auroral transients determined from the ensemble of equinox campaign visits, in a similar format to Figure 4.5. Panel (a) shows the LT extent of individual transients, defined as the width within which the intensity remains above 80% of the peak value in the intensity-lt line plots, with values being averaged over all the images within a visit over the lifetime of an event. Panel (b) shows a corresponding histogram of peak limb-brightening corrected emission intensities (kr) similarly averaged over the lifetime of the event. Panel (c) shows a histogram of the lifetime of the dusk transients, determined as the duration over which the event can be identified by a clear peak in the intensity-lt line plots. 103

104 Dusk Patch Angular Frequencies Figure 5.5 Histograms of (a) dusk transient normalized angular velocities, and (b) corresponding rotation periods, determined from linear fits to the image-to-image patch position determinations such as those shown in Figure 5.1. The format is similar to Figure

105 5.3 Physical Origin of Dusk Sector Patches The exclusively non-conjugate nature of the dusk transient events demonstrated in section 5.2 shows that they cannot be produced by processes on closed field lines such as magnetospheric hot plasma injections or ULF waves that would inevitably result in simultaneous emission from both hemispheres. Instead, our results suggest that open field lines may be involved for which conjugacy is not mandatory, which we therefore consider here. Specifically, we discuss emissions associated with bursts of magnetic reconnection at the dayside magnetopause, and consider their location, conjugacy, and lifetime. With regard to the post-noon location of the events, we first note that magnetic reconnection is expected to be suppressed by strong velocity shear across the boundary, when the shear becomes comparable with the Alfvén speed [Owen and Cowley, 1987; La Belle-Hamer et al., 1995; Cassak and Otto, 2011]. As discussed previously for the case of Jupiter by Desroche et al. [2012], the subsolar and dusk side of the magnetopause at Saturn is thus favoured for large-scale reconnection compared with dawn due to the asymmetry imposed by the interior rotation flow. The plasma angular velocity in the outer magnetosphere envisaged in section 5.2 corresponds to an eastward flow of ~100 km s -1 adjacent to the boundary, which is likely to be matched in the postnoon magnetosheath near ~14 h LT for the low solar wind flow conditions pertaining here (Figure 4.1) [e.g., Stahara et al., 1989], thus reducing the flow shear across the boundary to minimum values in this sector. A link between reconnection events and transient emissions with a mean location of formation at ~14.5 h LT as shown in Figure 5.4 (section 5.2), thus seems plausible. We then consider the effects associated with a burst of reconnection at the post-noon magnetopause, such that patches of new open flux are formed lying initially just equatorward of the pre-existing open-closed field boundary. Although magnetosheath plasma will then precipitate near-equally into the atmosphere in these patches in both hemispheres, containing equal amounts of newly-opened magnetic flux, the associated energy flux is entirely insufficient to produce bright detectable UV auroras [e.g., Cowley et al., 2004]. Instead, we need to consider the field-aligned currents associated with velocity perturbations on the open flux tubes, the north-south symmetry of which is broken by 105

106 the presence of the east-west (Y) component of the interplanetary magnetic field (IMF). This is sketched in Figure 5.6, where panels (a) and (b) show the situation for positive and negative IMF B y (and positive B z ), respectively. The sketches on the left of these panels show newly-opened field lines in the post-noon sector in each case, in views looking from the direction of the Sun. It can be seen for B y positive that the effect of the field tension on the newly-opened flux tubes aids the duskward flow of the field lines in the north, while retarding it in the south, and vice versa for B y negative. The effects on the plasma flow and currents in the northern and southern ionospheres are then sketched in the upper and lower diagrams on the right of each panel, respectively, where the arrowed long-dashed lines show plasma streamlines, and the patch of new open flux in each hemisphere is contained between the equatorward-displaced openclosed boundary on the equatorward side (solid line), and the perturbed former openclosed boundary on the poleward side (short dashed lines). We have assumed for simplicity similar significantly sub-corotating flows on both open and closed field lines in the unperturbed state, such that an approximately uniform ionospheric Pedersen current flows equatorward across the boundary in both hemispheres, this simple initial state allowing us to focus on the essential asymmetry associated with the newly-opened flux tubes. Considering first the effects for IMF B y positive in panel (a), it can be seen that the enhancement of the duskward plasma flow due to the field tension effect on newly-opened flux tubes in the northern hemisphere reduces the equatorward current within the patch, hence requiring upward field-aligned currents to flow on its poleward side, indicated by circled dots, and equal and opposite downward currents to flow on its equatorward side, indicated by circled crosses. Simultaneously, reduction of the duskward flow on the newly-opened flux tubes in the southern hemisphere enhances the equatorward current within the patch, requiring downward currents to flow on its poleward boundary and equal upward currents to flow on its equatorward boundary. Consideration of auroras then focuses on the regions of upward current, which for IMF B y positive in panel (a) will be carried by cool dense magnetosheath plasma at the poleward boundary of the patch in the northern hemisphere, generally not requiring strong field-aligned acceleration of the electrons, but at least partially by hot tenuous magnetospheric plasma at the equatorward boundary of the patch in the southern hemisphere, the latter then requiring strong field-aligned electron acceleration leading to bright auroras [e.g., Cowley et al., 2004]. For IMF B y positive, therefore, we may expect transient bright auroras to be associated with patches of newly-opened flux tubes 106

107 in the south, but not in the north. For IMF B y negative shown in panel (b), however, the sense of the current asymmetry is reversed, leading to the expectation of bright auroras at the equatorward boundary of the patch in the north but not at the poleward boundary in the south by the same argument. Although simplistic, this discussion nevertheless illustrates how magnetopause reconnection events can lead to non-conjugate transient emissions that occur in only one hemisphere, but with roughly equal numbers of such events occurring in the northern and southern hemispheres. With regard to the lifetime of the transient events, we note that in the above scenario this relates to the time scale for relaxation of the field tension effect leading to the perturbed flows, which should correspond to some fraction of the time scale for evolution of the open flux tubes over the dayside magnetopause into the magnetospheric tail. The latter time scale is of order several tens of minutes, so that transient auroral lifetimes of ~10-30 min (Figure 5.5) seem plausible. This discussion clearly prompts consideration of the hemispheric occurrence of these dusk transient events in relation to the IMF vector upstream of the magnetosphere, particularly with regard to the Y component that governs the hemispheric asymmetries as outlined above, as well as the Z component that regulates the overall reconnection rate. We note, however, that the correlation study presented by Zieger and Hansen [2008] between observed and modelled interplanetary parameters shows that the IMF Y component is only moderately well predicted, while the Z component cannot be predicted at all. It is thus perhaps unsurprising that examination revealed no clear correspondence between the northern and southern transient events found here and the modelled Y component corresponding to the propagated parameters in Figure 4.1. Chapter 6 presents a study to better understand the IMF as encountered at Saturn where coincident HST images are available, however data was insufficient to establish the connection between IMF B y and dusk transient events. Further study of the Cassini dataset alone may establish typical IMF B y behaviour and its timescales for change. 107

108 Mechanism for the Production of Dusk Patches Figure 5.6 Sketches illustrating the north-south asymmetries on newly-opened flux tubes due to the presence of IMF B y, where panels (a) and (b) show the cases of positive and negative IMF B y (and positive B z ), respectively. The sketches on the left in each panel show newly-opened field lines in the post-noon sector in each case, in views looking from the direction of the Sun, showing the senses of the field tension force associated with the east-west field. For near-anti-parallel reconnection, the reconnection site is displaced into the southern hemisphere for positive B y and into the northern hemisphere for negative B y, as shown. The effects on the plasma flow and currents in the northern and southern ionospheres are sketched in the upper and lower diagrams on the right in each panel, respectively. The solid lines show the open-closed boundary in each hemisphere, displaced equatorward in the central region by the reconnection event forming the equatorward limit of the patches of new open flux, while the short-dashed lines show the former boundary marking the poleward limit of the patches, perturbed poleward as they transfer into the polar cap. The arrowed longdashed lines show plasma streamlines, while the circled dots and crosses show the regions of upward and downward field-aligned currents, respectively, on the boundary of the open patch. 108

109 5.4 Summary and Discussion The equinox data set shows that transient dusk emissions are also a common phenomenon, but not as common as the dawn patches studied in Chapter 4, being observed on ~40% of the HST visits of ~40 min total duration. These emissions are also found to be non-conjugate, but now in the more strict sense that transient enhancements in one hemisphere are entirely unaccompanied by enhancements in the other. However, examples of such enhancements were observed in both the northern and southern hemispheres, those reported in pre-equinox HST data by Radioti et al. [2009] perforce being located exclusively in the south. Such extreme non-conjugate behaviour suggests an association with open flux tubes for which approximate conjugacy is not mandatory, and we have discussed one scenario in which north-south symmetry is broken on newly-opened dayside flux tubes through the agency of the eastwest (Y) component of the IMF, generally the major IMF component in the outer solar system. This picture has been shown to be plausibly consistent with preferential event occurrence in the post-noon sector, equal overall numbers of events north and south, and time scales of a few tens of minutes, though the expected relationship with the sign of the IMF Y component remains to be established observationally. The occurrence of lobe reconnection, one hemisphere at a time, during intervals of southward IMF is an alternative if possibly less attractive scenario (since the origin of the preference for dusk may then be less obvious). 109

110 Chapter 6 - Saturn s dayside ultraviolet auroras: Evidence for morphological dependence on the direction of the upstream interplanetary magnetic field 6.1 Introduction In Chapter 2 the debate as to whether external forcing by the solar wind or internal dynamics is responsible for auroral activity at Earth, Jupiter and Saturn was discussed at length. It is thought that large-scale Earth-like dayside reconnection at Saturn is unlikely but it was also discussed how at dusk where flow shear is reduced reconnection may be more favourable than suggested by scaling models [Desroche et al., 2013]. In addition it was discussed that reconnection may be restricted to loci where magnetospheric and magnetosheath fields are closely anti-parallel [Masters et al., 2012]. Further in-situ evidence for reconnection has been found by McAndrews et al. [2008], Badman et al. [2013] and Radioti et al. [2013] who found reconnection related plasma heating, magnetic flux connection across the boundary and plasma injection signatures in the cusp. Auroral emission in the dusk sector observed by Cassini s Ultraviolet Imaging Spectrograph (UVIS) is thought to be related to reconnection [Radioti et al., 2011]. These UVIS observations found sequential arc-like bifurcations in the noon-to-dusk sector reoccurring on ~1-2 h time scales. In Chapter 5 evidence has been presented that suggested that dusk transient patches in Saturn s auroral emission could be a result of newly opened flux with inter-hemispheric symmetry broken by the IMF B y component. The bifurcated features observed by Radioti et al. [2011] have strong morphological similarities with the dayside auroras associated with FTEs at Earth [Milan et al., 2000a], but in the latter case having recurrence times of ~5-10 min and lasting for ~10-20 min, so that more than one event is also typically present at any given time. It seems likely that the difference in time scales between Earth and Saturn lies in the different spatial scales of these systems combined with similar plasma and field line propagation speeds. In particular, a time scale relevant to the lifetime of dayside auroral features is the time required for open flux tubes (specifically their magnetopause intersection point) to propagate from the dayside reconnection site into the lobe of the tail. Newly-opened flux tubes generate ionospheric flow, field-aligned current, and precipitation that result in dayside auroras, but these die away as the flux tubes are 110

111 transported down-tail, are assimilated into the tail lobe, and become aurorally dark [e.g., Cowley and Lockwood, 1992; Milan et al., 2000a]. Taking for simplicity the down-tail path length on the nightside to be comparable to that on the dayside, appropriate to the size of the system, the overall flux tube propagation path will typically be ~35 Earth radii (R E ) at Earth and ~65 Saturn radii (R S ) at Saturn, for typical subsolar magnetopause radial distances of ~11 R E and ~21 R S, respectively. With similar flux tube propagation speeds comparable to a typical solar wind speed of ~450 km s -1, the relevant time scales are thus ~10 min at Earth and ~2.5 h at Saturn. If the Radioti et al. [2011, 2013] events correspond to FTEs at Saturn, as seems likely, it is then unsurprising that the study of magnetopause phenomena by Lai et al. [2012], looking for features occurring on terrestrial time-scales, failed to detect any. In support of this suggestion, Radioti et al. [2013] have shown that when Cassini was located inside the magnetosphere and near-conjugate to one such arc, it was immersed in magnetosheathlike cusp plasma near to its poleward exit into the polar plasma void region beyond. Similarly, in a related auroral case study when Cassini was located nearer to the magnetopause, Badman et al. [2013] have shown that the simultaneous near-noon magnetosheath field was directed northward, favouring lower-latitude reconnection and open flux production at Saturn, with signatures present of escaping magnetospheric electrons. We note in this regard that Saturn s magnetic dipole is directed parallel to the planet s spin axis, opposite to the case of the Earth, such that open flux production is favoured for northward-directed interplanetary magnetic field (IMF) at Saturn, rather than for southward IMF at Earth. In this case, we should expect the morphology of Saturn s dayside auroras to depend on the sense of the upstream IMF, in a similar manner to that previously discussed by Bunce et al. [2005] and Gérard et al. [2005]. Specifically, for northward IMF favouring lower-latitude reconnection, we should expect auroral dynamics to occur on newly-opened field lines as discussed above, consisting principally of multiple oval bifurcations in the noon to dusk sector produced by pulsed reconnection if the results of Radioti et al. [2011, 2013] and Badman et al. [2013] represent a guide. For southward IMF we would then expect such features to disappear, while auroral patches at high latitudes poleward of the dayside oval may then occur associated with reconnection between the IMF and the open tail lobe field [e.g., Milan et al., 2000b], i.e. lobe 111

112 reconnection, possibly corresponding to the emissions reported by Badman et al. [2012]. Opportunities to test these expectations are very limited, however, due to a lack of auroral images combined with concurrent suitably lagged upstream IMF values. Examination has firstly shown that available Cassini UVIS and VIMS images provide little detailed information under these circumstances, due to limited spatial resolution during intervals when Cassini was located beyond the bow shock, combined with nearequatorial viewing geometries. A significant catalogue of UV images is, however, available from Hubble Space Telescope (HST) observations. Belenkaya et al. [2010] studied southern hemisphere UV auroras observed during the 2008 HST Saturn campaign when Cassini was located near apoapsis in the solar wind. However, this interval appears to have been somewhat disturbed, with bright auroras extending to high latitudes in the dawn sector as mentioned above, while here we wish to focus attention on more usual auroral oval morphologies under quieter conditions. No simultaneous IMF and image data were obtained during the 2009 Saturn equinox HST campaign [e.g., Nichols et al., 2009], and none either in the most recent 2013 HST campaign [e.g., Nichols et al., 2013]. However, during the post-equinox 2011 and 2012 HST Saturn campaigns, thus observing northern auroras, we have found seven imaging intervals exhibiting usual oval emissions when Cassini simultaneously lay in the solar wind measuring the upstream IMF over intervals of at least several hours. An additional imaging interval in the 2011 campaign which exhibited dawn emissions extending towards the pole indicative of magnetospheric compression effects was excluded from the study. Of the seven cases exhibiting usual ovals, four have northward concurrent IMF and three southward. Here we examine these images to determine to what extent the above morphological expectations are fulfilled. In section 6.4 we also compare our findings with the results of a parallel study of these images by Belenkaya et al. [2014], where the bright auroral features have been mapped along magnetic field lines using the paraboloid model of Saturn s magnetospheric magnetic field which employs the concurrent Cassini IMF values as input. 6.2 Cassini IMF Data Propagation Delay Between Cassini IMF Measurements and Dayside Auroral Response 112

113 In this section we describe the Cassini IMF measurements used in this study, and begin by discussing the propagation delay that associates a HST auroral image obtained at a particular time (corrected for light travel between Saturn and Earth) with a particular interval of IMF data measured by Cassini. Evidently, the IMF which is influencing specifically the dayside auroral morphology at a particular time is that measured at an earlier time, due to the frozen-in propagation of the IMF from the spacecraft to the dayside magnetopause, together with the one-way Alfvénic communication time along outer magnetospheric field lines from potential reconnection sites to the ionosphere. The frozen-in propagation time consists of the time from the spacecraft to the bow shock in the upstream solar wind, plus the transit time through the magnetosheath. We now estimate these times, using a similar procedure to that employed by Nichols et al. [2007] in a study of HST images and concurrent solar wind data during the Cassini Jupiter fly-by in 2000/01. For the solar wind segment we reasonably assume that the phase fronts of IMF variations in the solar wind are near-perpendicular to the planet-sun line, the average field spiral angle at these distances also being within a few degrees of perpendicularity. This time is therefore given by SW ( Cass BS ) SW τ = X R V, where X Cass is the X-position of the spacecraft in KSM coordinates, R BS is the sub-solar radial distance of Saturn s bow shock, and V SW is the anti-sunward speed of the solar wind. (In KSM coordinates X points from the centre of the planet towards the Sun, the X-Z plane contains the planet s magnetic/spin axis, and Y completes the orthogonal right-handed set pointing towards dusk. We note that for Saturn the magnetic and spin axes are co-aligned to within measurement uncertainties of ~0.1º [Burton et al., 2010].) Unfortunately, solar wind plasma parameters cannot routinely be determined from Cassini ion data due to instrument pointing limitations [Young et al., 2004]. Noting, however, that the auroral images examined do not include disturbed morphologies in the sense discussed in section 1, here we make estimates based on typical values. Specifically, we use a typical solar wind speed and dynamic pressure of ~450 km s -1 and ~0.03 npa [e.g., Arridge et al., 2006], respectively, to determine typical positions of the bow shock and magnetopause (these values implying a typical solar wind number density of ~0.08 cm -3 at Saturn). The typical subsolar radial distance of the bow shock, given in Saturn radii by Masters et al. [2008] as P, where P dyn SW is the dynamic pressure of the dyn SW 113

114 solar wind, is then ~27.8 R S (Saturn s 1 bar equatorial radius R S is km), such that the solar wind delay in hours is given by τ ( X ( R S ) 27.8) SW Cass. This expression can take negative values if Cassini is located in the solar wind off the planet- Sun line at a smaller X than the subsolar shock, as was the case for both campaign intervals examined here. To determine the magnetosheath propagation time we also need the position of the subsolar magnetopause, given in Saturn radii by Kanani et al. [2010] as dyn SW P, thus equal to ~20.8 R S for the above dynamic pressure. The propagation time in the magnetosheath is taken to be given by the formulation of Khan and Cowley [1999], in which the speed of the magnetosheath plasma is taken to decrease linearly from a shocked speed of one quarter of the solar wind speed just down-stream of the sub-solar shock (appropriate to the high Mach number (M~10) solar wind at Saturn, see, e.g., section of Kivelson and Russell [1995]) to a speed V MP at the magnetopause, the latter representing the inflow speed normal to the magnetopause at near-subsolar reconnection sites. The typical value of V MP is unknown in detail, but may reasonably be taken to be a few tens of km s -1, consistent with reconnection rates of a several tens of kv [e.g., Jackman et al., 2004; Badman et al., 2005] imposed along a magnetopause reconnection line of length, say, ~10 R S. Here for definiteness we have taken a value of 30 km s -1, with a resulting propagation delay of τ 1.9 h. However, the result is not very sensitive to this choice unless V MP is taken to become very small. Finally, rough estimates of the one-way Alfvénic propagation time from the reconnection sites to the ionosphere at Saturn have been made by Meredith et al. [2013] and are found in Chapter 5. These estimates recognize that such sites may generally be located off-equator due to the requirement of near-antiparallel magnetic fields across the magnetopause, the sense of the displacement then depending on the sense of the IMF sector, toward or away. The estimated propagation times will then be less on the short field branch, ~15 min, than on the long field branch, ~1 h. Here for simplicity we have taken a fixed compromise propagation time of Sh τ Alf 0.5 h. Adding this to the fixed magnetosheath propagation time of ~1.9 h and the solar wind propagation time to the shock, the overall time delay (lag) from Cassini to the subsolar ionosphere is thus taken to be given by 114

115 ( X ( R ) 27.8) 2.4 hours τ τ + τ + τ S +. (6.1) Lag = SW Sh Alf Cass Values are ~1.7 h and ~2.2 h for the 2011 and 2012 campaign images, respectively, with estimated uncertainties of ~±0.4 h, obtained by considering the usual range of variation of solar wind parameters at Saturn [e.g., Arridge et al., 2006], combined in quadrature with the comparable uncertainty in the Alfvénic propagation delay to the ionosphere. The dayside auroral morphology is not just determined by the instantaneous value of the IMF at this earlier time, however, but by a history of the IMF over a preceding interval comparable to the lifetime of auroral features produced by magnetopause dynamics. Following the discussion of the time scales of post-noon arc bifurcations and Saturn FTEs in section 1, this time scale should clearly be taken to be a few hours, and for definiteness is taken here to be 2.5 h. As indicated in section 1, this corresponds to the typical propagation time of open flux tubes over the magnetopause from reconnection sites on the dayside to comparable distances down-tail from the planet on the nightside, such that it accommodates additional possible delays resulting from displacements of the reconnection sites away from the subsolar region towards dusk, and also largely subsumes the ~±0.4 h uncertainties in the lag time estimates. We thus employ the Cassini IMF data for each HST image centred at Saturn time t over a 2.5 h t t Lag to t t Lag interval from ( h) (during which X Cass hardly changes), noting that the results are insensitive to the exact interval chosen over a reasonable range. The IMF value averaged over this interval is then taken to be representative of the concurrent IMF for a given image. We note, however, that such intervals relate specifically to the expected time scale for dayside auroral morphology response. As discussed in section 1, the time scale for other responses, such as the large-scale size of the auroral oval determined by the amount of open flux present, will usually be significantly longer, days not hours in this case [Jackman et al., 2004; Badman et al., 2005], though intervals of somewhat more rapid change may also occur [Radioti et al., 2011]. 115

116 Trajectory of the Cassini Spacecraft Figure 6.1 Plot showing the Cassini trajectory projected onto the KSM X-Y plane (units of R S ) on two orbits during which the 2011 and 2012 HST Saturn UV auroral campaigns took place. The 2011 campaign took place during the orbit shown by the blue trajectory (Revs 146 outbound and 147 inbound), while the 2012 campaign took place during the orbit shown by the red trajectory (Revs 163 outbound and 164 inbound). Filled circles show apoapsis and periapsis positions on these orbits, while the arrows indicate the direction of spacecraft travel. The solid segments of these trajectories indicate 6-day intervals near apoapsis for which Cassini magnetic field and electron data are shown in Figures 6.2 and 6.3. The dashed black lines indicate typical magnetopause (inner) and bow shock (outer) locations for a solar wind dynamic pressure of 0.03 npa according to the empirical models of Kanani et al. [2010] and Masters et al. [2008], respectively, showing that the spacecraft is expected to have been located in the magnetosheath or solar wind near the bow shock during these intervals. The unfilled circles indicate the positions of the spacecraft at the centre times of seven ~44 min HST imaging sequences examined in this study, for which the spacecraft was located in the solar wind continuously over several hours. 116

117 6.2.2 Cassini IMF Data During the 2011 and 2012 HST Saturn Campaigns The data employed in this study were obtained during the 2011 and 2012 HST Saturn campaigns, undertaken during intervals when Saturn was in near-opposition at Earth. During these intervals the Cassini trajectory was near-equatorial with apoapsis in the pre-dusk sector as shown in Figure 6.1, where the relevant orbits have been projected into the X-Y KSM plane (units of R S ). Here the blue and red dashed lines show the orbits relevant to the 2011 and 2012 campaigns, respectively, with apoapses and periapses indicated by filled circles. Given that Cassini orbit revolution or Rev numbers are defined from apoapsis to apoapsis, the orbits correspond to Revs 146 (outbound) and 147 (inbound) for the 2011 campaign, and 163 (outbound) and 164 (inbound) for the 2012 campaign. The black dashed lines show the Masters et al. [2008] and Kanani et al. [2010] model bow shock and magnetopause positions for the typical solar wind dynamic pressure of 0.03 npa as employed in section 6.2.1, from which it can be seen that the spacecraft is expected to have been located in the solar wind just upstream from the bow shock for an interval spanning apoapsis in both cases. The solid line segments on each trajectory indicate a 6-day interval in each case for which Cassini data are shown below in Figures 6.2 and 6.3, generally containing both magnetosheath and solar wind intervals as may be expected, but where HST obtained auroral images while Cassini was continuously located in the solar wind over intervals of at least several hours. The position of the spacecraft at the light travel corrected centre times of these ~44 min HST imaging intervals are shown in Figure 6.1 by the open circles, three during the 2011 campaign on Revs 146/147, and four during the 2012 campaign on Rev 163. Figure 6.2 shows relevant Cassini data for Revs 146/147, spanning days 90-95, inclusive, of Panel (a) shows a color-coded electron count rate spectrogram spanning energies between 0.6 ev and kev, obtained by the ELS electron spectrometer of the Cassini CAPS plasma package [Young et al. 2004], where we note that the intense component at lowest energies ~1 ev is dominated by spacecraft photoelectrons. Panels (b)-(e) show magnetic field data obtained from the fluxgate magnetometer [Dougherty et al. 2004], where panel (b) shows the field magnitude (nt) on a logarithmic scale, while panels (c)-(e) show the x, y, and z KSM field components 117

118 Cassini Data for 2011 Figure 6.2 Plot showing Cassini data for the six-day interval during the 2011 HST Saturn auroral campaign shown by the blue solid line segment of the trajectory in Figure 6.1, spanning days 90 95, inclusive. From top to bottom we show (a) a colorcoded count rate spectrogram of plasma electrons from 0.6 ev kev, (b) the magnetic field strength shown on a log scale (nt), (c)-(e) the KSM B x, B y, and B z magnetic field components (nt), and (f) the clock angle of the magnetic field about the X axis (the planet-sun direction) given by tan -1 (B y /B z ) defined between 0º and 180º. Spacecraft position data are given at the bottom of the figure at the start of each day, specifically its position in Cartesian KSM coordinates, together with the total radial distance from the planet (R S ). Vertical dashed lines indicate the light travel time corrected centre times of ~44 min HST observing visits during the campaign, with visit identifiers indicated at the top of the figure. Red stripes indicate the lagged 2.5 h intervals used to determine the mean concurrent IMF vectors associated with the imaging intervals, as discussed in section respectively. Panel (f) shows the clock angle of the field about the planet-sun line, given by = arctan( ) θ B y B z defined between 0º and 180º. A clock angle in the range 0º-90º thus indicates a northward field potentially favourable for low-latitude reconnection and open flux production, while a clock angle in the range 90º-180º indicates a southward field unfavourable for low-latitude reconnection, but possibly favourable for lobe reconnection. Spacecraft KSM coordinates and radial distance (R S ) 118

119 are given at the day of year (DoY) boundaries at the bottom of the plot. Intermittent magnetosheath intervals are readily identified by the presence of warm electrons extending to several tens of ev energy with simultaneously elevated field strengths, these intervals having sharp boundaries as the spacecraft crosses the bow shock. Otherwise the spacecraft is located in the solar wind upstream from the shock. The light travel corrected centre times of the three ~44 min HST imaging sequences, termed visits, that occurred in the interval shown in Figure 6.2 are marked by the vertical dashed lines, with visit identifiers H1-H3 being shown at the top of the plot (the excluded image mentioned above in section is H4 occurring after the interval shown). It can be seen that Cassini had been located in the solar wind for a substantial prior interval, many hours, in each case. The red vertical bars over-plotted in the magnetic field panels show the lagged 2.5 h intervals of IMF data relevant to the dayside auroral morphology, derived using equation (6.1) and the centre time of each visit. In section 6.3 we employ slightly refined intervals corresponding to subsets of the image data obtained during each visit, but since these differ in centre time by only ~16 min in either direction from the overall centre time employed here, this does not significantly affect the averaged IMF values derived. It can be seen that the IMF was directed near-continuously northward in each of the intervals indicated by the red bands, thus potentially favourable for low-latitude reconnection and open flux production, with mean and standard deviation (SD) B z values of 0.16±0.02, 0.13±0.06, and 0.45±0.27 nt, respectively. It can thus be seen that the largest B z field, by a factor of ~3, occurred for visit H3. The corresponding averaged clock angles are 50º±6º, 52º±20º, and 67º±14º. These averaged IMF component values and clock angles are collected together in Table 6.1 for convenience and comparison, together with the computed lag times employed. Figure 6.3 then shows Cassini data for Rev 163, spanning days 91-96, inclusive, of 2012, in the same format as Figure 6.2. The spacecraft is located in the magnetosheath at the start of the interval, unsurprisingly given the trajectory of the spacecraft shown in Figure 6.1, but passes across the bow shock into the solar wind early on day 92, with intermittent magnetosheath intervals occurring thereafter presumably during transient decreases in solar wind dynamic pressure. The light-travel corrected centre times of four HST visits are again shown by the vertical dashed lines marked with identifiers I5-I8 at the top of the plot, together with four lagged intervals relevant to the dayside 119

120 Table 6.1 Averaged Lagged IMF Components and Clock Angle for Each HST Visit HST visit HST visit center time h:min:s DoY/Year Saturn visit center time h:min:s Lag time/h a IMF B x /nt b IMF B y /nt b IMF B z /nt b IMF clock angle θ/deg c DoY/Year H1 05:03:51 91/2011 H2 03:23:19 93/2011 H3 06:30:29 95/2011 I5 10:47:15 92/ :52:13 91/ :11:41 93/ :18:51 95/ :34:29 92/ ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±26 I6 10:48:56 09:36: ± ± ±5 93/ / ±0.03 I7 10:45:49 09:33: ± ± ±7 94/ / ±0.06 I8 13:54:13 12:41: ± ± ±14 95/ / ±0.06 a Uncertainties in lag time values are ~±0.4 h, see section 2.1. b Values given are the mean and standard deviation over 2.5 h lagged intervals based on the visit center time, see section 2.1. c Values given are the mean and standard deviation obtained using the directional statistics approach of Mardia and Jupp [2000] appropriate to circular measure, determined over the same 2.5 h lagged intervals as the field values. 120

121 Cassini Data for 2012 Figure 6.3 Plot showing Cassini data for the six-day interval during the 2012 HST Saturn auroral campaign shown by the red solid line segment of the trajectory in Figure 6.1, corresponding to DoY 91 96, inclusive. The format is the same as Figure 6.2. auroras shown by the red bars, computed using equation (6.1). As in Figure 6.2, all these intervals were located exclusively in the solar wind, though the first of these, for visit I5, appears to have been disturbed by field fluctuations associated with the bow shock. For this case the IMF was again near-continuously northward, with mean and SD B z values of 0.15±0.08 nt over the interval, similar to visits H1 and H2, and a highly variable clock angle of 33º±26º (Table 6.1). For the subsequent three visits, however, the IMF was directed consistently southward, unfavourable for low-latitude dayside reconnection, and had generally been so directed for a number of hours previously, though a brief interval of northward-directed field had occurred shortly before the lagged interval for visit I8. The mean and SD B z values for I6-I8 are -0.54±0.03, -0.41±0.06, and -0.18±0.06 nt (Table 1), such that the southward fields for I6 and I7 are of comparable strength to the strong northward field for H3, while the lesser southward field for I8 has a similar strength to the weaker northward fields of H1, 121

122 H2, and I5. The corresponding clock angles for I6-I8 are 148º±5º, 134º±7º, and 124º±14º, respectively (Table 6.1). 6.3 Morphology and IMF Dependence of the 2011 and 2012 HST Campaign Auroras HST Image Data and Display The HST UV auroral images during the 2011 and 2012 Saturn campaigns were obtained using the Solar Blind Channel of the Advanced Camera for Surveys, the detector of which is a pixel Multi-Anode Microchannel Array with highest throughput in the far-uv waveband nm. The average resolution of the detector is ~0.032 arcsec pixel -1, with an instrument point spread function (PSF) of two pixels. At Saturn s distance from Earth during these campaigns of ~8.6 AU, this PSF translates to a spatial resolution in the images in the noon-sector auroral region of ~400 km east-west (~0.1 h LT), and ~1000 km north-south (~1º latitude) due to the oblique view. We note that the latter value is of comparable order to the ~ km north-south spatial scales of the UVIS dusk-side arcs studied by Radioti et al. [2011, 2013], such that individual arc structures are at the limit of resolution in these HST images. Nevertheless, such emissions should certainly be observed by HST, forming a broad structured band of high-latitude UV aurora located in the noon to dusk sector. During each campaign visit, 19 individual images were obtained over an interval of ~44 min, each with an exposure time of 100 s. The first and last five images employed the F125LP filter that has a band-pass of nm, thus observing the Lyman and Werner band emissions of H 2, but excluding H Lyman-α at nm (to avoid image contamination by geocoronal emission when HST is not located in Earth s shadow). The nine central images then usually employed the F115LP filter with a band-pass of nm, which thus includes H Lyman-α emission. All of these images show essentially similar features. However, here we choose to concentrate on the initial and final sets of five images using the F125LP filter, since these can then be compared with each other (but not directly with the central nine images) to provide an indication of temporal variability and motion of auroral features, or the lack thereof, during a given 122

123 visit. The data from each such set of five images have then been co-added to improve the signal-to-noise, such that here we examine two co-added images from each visit corresponding to the initial and final 5-image sets, each spanning an interval of ~11 min, with ~33 min between the two centre times (i.e., ~±16 min about the overall visit centre time). For ease of reference, the co-added image from the initial interval of each visit will be termed image A for that visit, while that from the final interval will be termed image B. We note that in the ~11 min interval contributing to a single coadded image a near-rigidly corotating feature would rotate through only 0.4 h of LT, while in the ~33 min between them, such a feature would rotate through a well-resolved 1.2 h of LT. The co-added images for the two campaigns are shown in Figures 6.4a and 6.4b, respectively, projected onto a latitude-longitude grid at a height of 1100 km above the 1 bar reference spheroid, the latter corresponding to the typical peak in the emission height profile [Gérard et al., 2009]. Noon is at the bottom of each image, dawn to the left, and dusk to the right, with white dotted circles and radial lines showing 10º intervals of latitude and 30º intervals of longitude (2 h LT), respectively. Each image has been truncated past the dawn-dusk meridian to avoid over-stretching the pixels as the HST view approaches the planetary limb, with a somewhat more expanded view being available for the 2012 campaign compared with 2011 due to the developing northern spring season at Saturn. The same color-coded intensity (kr) scale is used for all the images, shown upper right in the figure, saturated red at 40 kr. The co-added images obtained from the initial (A) and final (B) intervals of each visit of the 2011 campaign are shown in the top and bottom rows of Figure 6.4a, respectively, with the images from visits H1-H3 being shown in the columns from left to right. The images from visits I5-I8 of the 2012 campaign are shown in the same format in the top and bottom rows of Figure 6.4b. The header above each image gives the HST visit identifier, the date (year-month-day) and UT centre time (hr:min:sec) of the image, together with the corresponding averaged lagged IMF components in KSM coordinates, and the averaged clock angle. These values have been determined over 2.5 h intervals lagged relative to the centre time of the individual co-added images shown, but since, as indicated above, these centre times are displaced by only ~16 min either side of the overall centre time marked in Figures 6.2 and 6.3, the values generally differ only marginally from those discussed in section and shown in Table

124 HST Campaign Images 124

125 Figure 6.4 Set of projected HST images of Saturn s northern UV aurora from (a) the 2011 campaign (visits H1-H3), and (b) the 2012 campaign (visits I5-I8), during which Cassini was located in the solar wind upstream of Saturn s bow shock over intervals of at least several hours. The upper rows of panels (a) and (b) correspond to the five coadded F125LP filtered images at the start of each visit, while the lower rows similarly correspond to the five co-added F125LP filtered images at the end of each visit. These are termed the A and B images, respectively, as marked in the figure. HST visit identifiers are given above each image, together with the date (year-month-day) and UT center time (hr:min:sec). Beneath this are shown the mean and SD values of the three KSM components of the IMF determined from the lagged 2.5 h intervals of Cassini data associated with the center time of each individual image, together with the corresponding mean and SD values of the IMF clock angle. The images are projected onto a polar grid assuming an auroral height of 1100 km above the 1 bar reference spheroid, the grid being shown by white dotted latitude circles at 10º intervals and white dotted longitude lines at 30º intervals (2 h LT), with noon shown at the bottom and dawn to the left. The images have been truncated somewhat past the dawn-dusk meridian to avoid pixels becoming over-stretched on approaching the planetary limb. The same color-coded intensity scale, shown upper right, is employed for all images, saturated red at 40 kr Case by Case Dayside Auroral Morphologies In common with previous studies of Saturn s auroral morphology, Figure 6.4 shows that Saturn s dayside auroras usually consist of a relatively narrow arc spanning from the pre-dawn sector towards noon, giving way to intermittent patchy forms at higher latitudes in the post-noon sector [e.g., Gérard et al., 2004, 2005; Grodent et al., 2005]. We now examine and discuss these emissions on a case-by-case basis before drawing overall conclusions in subsequent sub-sections. In the images from visit H1 shown in Figure 6.4a, the dawn arc is centred near ~17º colatitude, and is truncated pre-noon near ~10 h LT. At later LTs the emissions then step poleward, forming a broader, weaker, spatially structured band spanning ~10º colatitude that extends beyond the dusk meridian. The detailed form of the latter emissions is seen to change somewhat in the ~33 min between images A and B, but the overall position appears to be fixed to a first approximation, with notable patches near to noon as well as in the pre-dusk sector. Comparing this morphology with those presented, e.g., by Radioti et al. [2011, 2013] and Badman et al. [2012, 2013], it seems reasonable to suggest that the noon to dusk emissions correspond to the similarly located high-latitude structured auroras observed in the UVIS and VIMS data, which lie at the limit of resolution of individual arc features in these HST images as discussed 125

126 in section The association of these emissions with dayside reconnection-related phenomena is supported in the present case by the positive IMF B z known to have been prevailing over an extended prior interval, having a mean concurrent value of ~0.16 nt with a mean clock angle of ~50º (Table 6.1). The auroral morphology during visit H2, for which the mean concurrent IMF B z was slightly weaker at ~0.13 nt with a clock angle of ~52º, is similar to H1 (Table 6.1). The brighter wider dawn arc in this case, centred near ~14º co-latitude, is again truncated near ~10 LT, stepping poleward to weaker variable patchy emissions at around ~10º co-latitude near noon (particularly image B) and at later LTs extending towards dusk (particularly image A). The noon to dusk emissions are somewhat less extensive than those observed during visit H1, however, possibly resulting from the weaker IMF B z prevailing, and the lesser time for which the IMF had been northwardpointing prior to the visit (Figure 6.2). Visit H3 then occurred in association with the largest positive IMF B z in this data set by a factor of ~3, ~0.45 nt with a clock angle of ~67º (Table 6.1), for which it can be seen that auroras are continuously present across the full range of LTs in both images. Structured dawn arc emissions centred at ~14º co-latitude near dawn move poleward to ~12º in the pre-noon sector, and then give way post-noon to a broad band of highlatitude emissions poleward of ~10º that extend past the dusk meridian. Similar features are observed in both H3 images, although in image B the emission near dusk appears to have contracted slightly poleward compared with image A. Again, the broad band of post-noon emissions may well correspond to a multiply-bifurcated oval, such as those shown in Figures 1 and 2 of Radioti et al. [2011, 2013], the detailed structure of which remains unresolved here. Continuing with the images from the 2012 HST campaign shown in Figure 6.4b, visit I5 also occurred during an interval of northward-directed IMF, with a mean lagged B z of ~0.15 nt and a clock angle of ~33º, this IMF B z value being similar to the weaker northward field cases H1 and H2 (Table 6.1). In this case a narrow dawn arc centred at ~16º co-latitude extends continuously from beyond the dawn meridian to ~9-10 LT, giving way at later LTs to a high-latitude patch of emission poleward of ~10º in the noon sector. In image A the patch is bright and extends in a ~3 h LT band centred near noon, while in image B it is less intense with a centre moved toward the pre-noon 126

127 sector. No significant emission is present at later local times in the pre-dusk sector. However, structured forms are observed in both images A and B at lower latitudes, ~13º co-latitude, in the post-dusk sector, a region that was inaccessible to HST during the 2011 campaign. This morphology appears similar to that for visit H1 in Figure 6.4a occurring under similar northward IMF B z conditions, in which the dawn arc is also truncated near ~10 LT, giving way at later LTs to a similar patch of high-latitude auroras near to noon which is somewhat separated from the pre-dusk emission, particularly in image B. Unlike the four cases discussed above, the three remaining images from the 2012 campaign all occurred under continuously southward IMF conditions as outlined in section 6.2.2, with stronger concurrent negative IMF B z values during visits I6 and I7, and a weaker negative B z during I8. For visit I6, for which the mean lagged B z value and clock angle were ~-0.54 nt and ~148º (Table 6.1), respectively, it can be seen that a narrow dawn arc is centred at ~17º co-latitude near the dawn meridian. The emission then dims and becomes more scattered in the mid-morning sector between ~8 and ~10 h LT, before brightening again and moving poleward to ~12º near noon, finally terminating post-noon near ~13 h. The auroral enhancement near to noon has some similarity to that observed for northward IMF in visits H1, H2, and I5, but in this case it is centred at lower latitudes, and appears to be joined by weaker structured emission to the arc spanning dawn, particularly in image B. We thus infer that the noon emission in this case forms a continuation of the dawn arc into the noon sector, an interpretation also supported by examination of the emissions for visit I7 described below, observed under similar IMF conditions. We note that significant spatial structuring is typical of dawn arc emissions in the mid-morning hours [Meredith et al., 2013; Chapter 5]. There are then no significant emissions for visit I6 in the post noon to dusk sector, though emissions are again present past the dusk meridian, similar to those observed (for northward IMF) in visit I5. The IMF conditions for visit I7 were similar to those for I6, as just noted, with a mean lagged B z value and clock angle of ~-0.41 nt and ~134º (Table 6.1), respectively. In this case a wider brighter band of variable dawn arc emission stretches continuously from pre-dawn, and terminates post-noon at ~13 h LT. The emission is centred near ~15º co-latitude at dawn, and moves poleward to ~12º near noon. The LT extent and latitudinal variation of this emission thus has much in common with that observed for 127

128 visit I6, but is now clearly continuous in LT from pre-dawn to post-noon, thus supporting the view that the dawn-to-noon emissions for I6 also represent extended but variable dawn arc emissions under similar IMF conditions, as indicated above. Also as for visit I6, the post-noon region is then devoid of significant emissions, and remains so in this case beyond the dusk meridian throughout the field of view. For visit I8, with a weaker mean southward field of ~-0.18 nt and a clock angle of ~124º (Table 6.1), we again observe a narrow bright dawn arc centred near ~15º colatitude near dawn, which appears to extend into weaker structured forms in the midmorning sector that vary somewhat between images A and B, and approach ~10º colatitude pre-noon. As for the other visits with negative IMF B z, there are again no patchy emissions present between noon and dusk extending around the high-latitude oval. However, a well-defined patch of polar emission now lies poleward of the structured dawn arc, near-stationary in images A and B principally on the dawn side of the noon-midnight meridian, extending to the magnetic/spin pole itself. We suggest that this patch is likely located on open field lines, and is associated with high-latitude lobe reconnection between the southward-directed magnetosheath field and open field lines in the northern tail lobe, as previously discussed for Saturn by Bunce et al. [2005] and Gérard et al. [2005]. This suggestion will be discussed further in relation to the noon sector auroral morphologies for positive IMF B z in section Dawn Arc In the following sub-sections we now compare the auroral features observed during the seven HST imaging visits, focusing on possible IMF dependencies, and begin by considering the dawn arc. Figure 6.4 shows that the dawn arc is a persistent feature in this data set, being present in all of the images examined, though we note it is very occasionally absent in the wider Saturn auroral archive. It can also be seen, however, that it exhibits somewhat variable properties from visit to visit. While the arc is typically centred near ~15º co-latitude in the dawn sector, its position at a given LT clearly varies by a few degrees over the 1-2 day intervals from visit to visit, as noted in section above, though not discernibly over the ~33 min intervals between the two images from a given visit. In addition, the latitudinal thickness of the arc is somewhat variable over the range ~1 to ~4, while the intensity varies from small values up to ~40 kr. As mentioned in section 6.3.2, it also often exhibits spatially structured sub- 128

129 corotating patches of emission in the mid-morning hours, previously analysed by Meredith et al. [2013], as discussed in Chapter 4, using the 2009 equinoctial campaign images, that are particularly evident here in the image pairs for visits H2, H3, and I7. Examination shows, however, that none of these features are simply related to the mean lagged concurrent IMF vectors determined here, for any KSM component. For example, if we consider the sub-set of images with positive IMF B z, H1-H3 and I5, these contain examples of slightly expanded (H1) and contracted (H3) latitudinal positions, as well as wide and bright (H2) and narrow and dimmer (I5) arcs. The subset with negative IMF B z, I6-I8 shows similar variability, such that these features clearly do not depend on the concurrent IMF B z. The same conclusion can be drawn by considering the image sub-sets, e.g., with positive (H1, H2, and I8) and negative (H3 and I5-I7) concurrent IMF B y (Table 6.1). One possible exception, however, concerns the LT extent of the dawn arc in the near-noon sector, which is often truncated pre-noon near ~10 h LT for northward IMF (visits H1, H2, and I5), but can extend post-noon to ~13 h LT for southward IMF (visits I6 and I7). Overall, however, it thus appears that the main properties of the quiet time dawn arc do not depend on the value of the concurrent IMF. It is known on a statistical basis that the dawn emissions are modulated by the planetary period oscillations observed ubiquitously in Saturn s magnetosphere [Nichols et al., 2010]. However, examination shows that the variations found in the individual examples considered here are not clearly related to the northern oscillation phase either. The origin of these dawn arc variations thus requires further study Noon to Dusk Emission While the dawn arc is thus a persistent if somewhat variable feature in these images, the discussion in section shows that the emissions in the noon to dusk sector are considerably more intermittent. The images from visits H1-H3 and I5, associated with northward IMF, all show the presence of high-latitude emissions in this sector, beyond the LT extent of the dawn arc. However, the form of these emissions varies significantly from visit to visit, and also somewhat over the ~33 min intervals between images A and B of a given visit, thus indicating the presence of a dynamic phenomenon varying on time scales down to a few tens of minutes. In addition, the most extensive of these noon to dusk emissions has been found to be associated with the strongest 129

130 positive IMF B z. However, such emissions are conspicuously absent for visits I6-I8, i.e., for all three cases for negative IMF B z. Based on this IMF B z dependency, and on the variable nature of these emissions observed both within visit and from visit-to-visit, it seems reasonable to suggest that these emissions are associated dayside reconnection, related to the structured flows, field-aligned coupling currents, and precipitation associated with dayside events occurring at various locations on the magnetopause, observed at various stages of their few-hour auroral development. Indeed, it seems likely that these emissions are manifestations of the same high-latitude noon to dusk auroral phenomenon reported by Radioti et al. [2011, 2013] and Badman et al. [2013], though observed here with lesser spatial resolution. These authors suggested that the structured auroras observed in their images are associated with dayside reconnection, a proposal that is strongly supported here by the observed IMF B z dependency Polar Emission In the discussion of individual visits in section we suggested that the polar patch of emission observed on visit I8 is associated with lobe reconnection occurring under the southward IMF conditions then prevailing. We note, however, a superficial similarity with the auroral morphology observed for visit I5, and also for H1, in which a well-defined patch of emission also appears at relatively high latitudes, ~10 co-latitude and poleward thereof, centred near to noon. Without additional information, it might seem reasonable to suggest that these patches too could be associated with lobe reconnection. However, the concurrent northward direction of the IMF, with a clock angle of only ~33 for visit I5, the most northward of all the IMFs in this data set (Table 6.1), shows that this cannot be the case. Rather, as in section 6.3.2, these emissions are suggested to form part of the continuum of variable high-latitude patchy auroras that form the noon to dusk oval for northward IMF, related to similarly variable lower-latitude reconnection events occurring at the magnetopause. We note that for visits H1 and I5 for northward IMF, the dawn arc is truncated in the pre-noon sector before the emission steps poleward near noon and at later LTs. This suggests an auroral oval that is located mainly on closed field lines in the dawn sector (noting the lack of IMF-dependence of the dawn arc), but moves poleward onto newly-opened field lines in the noon to dusk sector associated with lower-latitude reconnection. For visit I8 for 130

131 southward IMF, however, the polar patch of emission lies near-stationary principally in the dawn-to-noon sector poleward of the mid-morning extension of the dawn arc, reaching to much higher latitudes than for the northward IMF cases, encompassing the magnetic/spin pole itself. These results therefore show that care must be taken when considering the physical origins of dayside emissions at Saturn. 6.4 Summary and Discussion In this study we have examined a unique set of HST observations of Saturn s dayside auroras during which the Cassini spacecraft was located in the solar wind just upstream from Saturn s bow shock, measuring the IMF over concurrent intervals of at least several hours. Confining the study to intervals exhibiting usual oval auroral morphologies, as opposed to rarer disturbed conditions when the dawn auroras expand poleward to high latitudes, we have found seven such simultaneous intervals, all of which occurred during the post-equinox 2011 and 2012 HST auroral campaigns, thus involving the northern UV auroras. As in previous studies, the dayside auroras generally exhibit a relatively narrow arc extending continuously from pre-dawn towards noon, together with intermittent broader patchy forms at higher latitudes between noon and dusk. In the images examined here the dawn arc is a persistent feature, but centred variably in the range ~12º-17º colatitude, with variable latitudinal width and intensity. None of these variations has been found to be related to suitably lagged and averaged KSM components of the concurrent IMF (nor indeed to the concurrent phase of the northern planetary period oscillations). Future study of the origins of this variability is thus warranted. However, auroral features at LTs beyond the dawn arc in the noon to dusk sector show a consistent dependency on IMF B z. Specifically, all four cases with positive IMF B z show the presence of patchy auroras in the noon to dusk sector, more broadly distributed in latitude than the dawn arc and located at higher latitudes, ~10º co-latitude and poleward thereof, while none of the three cases with negative IMF B z show similar features. The three cases with weaker positive IMF B z exhibit variable patchy high-latitude forms, a noon patch and/or a patchy arc in the pre-dusk sector, while the case with a significantly stronger IMF B z, by a factor of ~3, shows continuous emissions from noon to beyond the dusk meridian. All these cases also show some variation in the post-noon emissions over the ~33 min interval between the initial and final images obtained on a given HST 131

132 visit, thus indicating evolution of the related physical process on such time scales. While the images for negative IMF B z show no such auroras, one (for visit I8) does show the presence of a prominent polar patch centred in the dawn-to-noon sector poleward of a structured dawn arc, extending to the magnetic/spin pole itself. Comparison with the auroral morphologies discussed using UVIS data by Radioti et al. [2011, 2013] and Badman et al. [2013] suggests that the patchy high-latitude auroras observed between noon and dusk correspond to the structured forms in this sector that these authors propose to be associated with dayside reconnection, though here observed with lesser spatial resolution. The exclusive association found here of such emissions with northward IMF, favourable for low-latitude dayside reconnection and open flux production at Saturn, provides support for this view. In this scenario, the variability of the forms observed under this condition, both during a visit and from visit-to-visit, can be ascribed to observations of reconnection-related phenomena occurring at differing locations on the magnetopause, and in varying stages of their few-hour evolution from the dayside magnetopause into the tail. The confinement of such features to the noon to dusk sector is likely due to the suppression of magnetopause reconnection at dawn by the strong flow shear across the boundary at such LTs, an effect that is reduced in the post-noon sector [Desroche et al. 2013]. We emphasize that if the reconnection events inferred from the auroral observations by Radioti et al. [2011, 2013] and Badman et al. [2013] have similar character to FTEs at Earth, as seems likely, the time scales of these events will be few-hours and not few-minutes, as a result principally of the much larger spatial scale of the Saturn system. The failure of Lai et al. [2012] to detect fewminute Earth-like FTEs at Saturn is then not a major surprise. Along related lines, the lack of correspondence noted in section 3 between the latitude of the auroras, either the dawn arc or the post-noon emissions, and the concurrent few-hour mean value of IMF B z is also not surprising, since the time scale for open flux accumulation in the system that influences the size of the open field region and hence that of the auroral oval, is usually days and not hours [Jackman et al., 2004; Badman et al., 2005]. Modest, fewdegree, changes in latitude are observed in this data set on the 1-2 day intervals between visits, but are not discernible on the ~30 min intervals during visits. The polar patch observed in one case (visit I8) for relatively weak southward IMF is then suggested to be associated with intermittent lobe reconnection on open field lines, following the earlier discussions of Bunce et al. [2005] and Gérard et al. [2005], and 132

133 the additional possible examples shown by Badman et al. [2012, 2013]. The pre-noon preference of this patch in the present case may be associated with the positive IMF B y conditions prevailing (Table 6.1), which via the anti-parallel field condition would favour lobe reconnection pre-noon in the northern hemisphere and post-noon in the southern hemisphere at Saturn, similar to but opposite in sense to the effect observed at Earth due to the opposite polarity of the planetary dipole field [e.g., Milan et al., 2000b]. We would not expect plasma rotation effects to play such a significant role in the lobe reconnection process at polar latitudes as they appear to do in open flux production nearer the equator. We conclude by noting that the above discussion of the physical origins of these emissions is in excellent accord with the results of a parallel study of these images by Belenkaya et al. [2014]. In order to examine the auroral source regions, these authors employed the paraboloid model of Saturn s magnetosphere to magnetically map the regions of bright auroral emission from the northern ionosphere into the magnetosphere. The magnetic model includes the internally-generated field of the planet, typical ring and tail current fields, the shielding effect of the current flowing on a parabolic magnetopause, and a partially penetrating IMF derived from the concurrent Cassini field data essentially as in section They found that the dawn arcs map typically from near the inner edge of the dawn-sector ring current at ~7 R S radial distance, outwards to either the centre of the ring current at ~10 R S for narrow arcs, or to its outer edge near ~15 R S for the broader dawn emissions, or to the outer magnetosphere close to the magnetopause for the broadest dawn emissions such as those that occur for visit H3. These auroras thus lie essentially wholly on closed field lines in the outer part of the dawn magnetosphere, and are thus likely to be associated with flow shears and pressure gradients within the hot ring current plasma injected from the nightside region. The lack of dependence on the concurrent IMF found here is thus unsurprising. The reason for the strong dawn-dusk asymmetry in these emissions, with no equivalent emission being observed in the post-noon sector for either northward or southward IMF, at least within the sensitivity of the present HST image data, is unclear. However, it could possibly relate to the persistent dawn-dusk asymmetry associated with the Vasyliunas flow cycle, with slow outflows of mass-loaded flux tubes down the tail at dusk, and rapid returns of pinched-off flux tubes via dawn [Vasyliunas, 1983; Cowley et al., 2004a, 2004b]. More recent observational studies have used plasma flow data to 133

134 map Vasyliunas flow cycles supporting this interpretation [McAndrews et al., 2009; Kane et al., 2013; Thomsen et al., 2014]. The higher-latitude patchy forms observed for northward IMF in visits H1-H3 and I5 in the noon to dusk sector, however, are all found to straddle the model open-closed field line boundary, consistent with our discussion of processes connected with lowerlatitude reconnection and open flux production at the dayside magnetopause. We also note that the near-noon emissions for southward IMF, which can sometimes adopt a patchy form, particularly for visits I6 and I8, were found to be located principally on closed model field lines near the outer boundary of the ring current and in the outer magnetosphere. This supports our conclusion in section that these emissions are associated with the dawn arc and its termination in the noon (I8) to post-noon (I6 and I7) sector under these conditions. The post-dusk arcs seen for northward and southward IMF during visits I5 and I6 are also located principally on closed model field lines. Finally, the polar emissions observed on visit I8 are found to map wholly to open field lines in the model, consistent with our suggested lobe reconnection origin. The results of Belenkaya et al. [2014] are thus found to be fully consistent with the discussion given here. Overall, the HST auroral images examined here provide new evidence of significant IMF-dependence of Saturn s dayside auroras, though using only the limited data set of seven simultaneous observation intervals with usual auroral morphologies presently available. Clearly, this study should be followed by further investigation if more such data sets become available in future. 134

135 Chapter 7 - Survey of Saturn auroral storms observed by the Hubble Space Telescope: Implications for storm time scales 7.1 Introduction The occurrence of auroral storms at Saturn was discussed in Chapter 2 and such storms are notable as being some of the most dramatic auroral behaviour observed at Saturn. For this reason individual events have been observed and reported on relatively frequently from the time that HST was capable of observing such activity [Prangé et al., 2004; Clarke et al., 2005; Crary et al., 2005; Grodent et al., 2005; Bunce et al., 2006]. During storm events, the auroras brighten significantly and expand poleward on the dawn side of the planet, sometimes reaching Saturn s spin/magnetic pole itself [Clarke et al., 2005], while the total UV auroral power per hemisphere increases from a typical range ~5-10 GW to values ~20-30 GW [Badman et al., 2005; Clarke et al., 2009]. Image sequences obtained during extended observation campaigns using the Hubble Space Telescope (HST) suggest that these events occur with a typical separation of 5-10 days, and have durations in excess of a few hours [Clarke et al., 2005; Grodent et al., 2005; Bunce et al., 2006; Clarke et al., 2009]. Due to the lack of high-cadence imaging over more extended intervals, however, the latter lifetime estimate remains uncertain and forms a central topic of this Chapter. The association of storms with solar wind compressions of the magnetosphere has been inferred from tracking heliospheric disturbances from the Sun to Saturn via Earth and Jupiter [Prangé et al., 2004], or more directly from observations of the interplanetary medium upstream from Saturn during the Cassini planetary approach phase in early 2004 [Crary et al., 2005; Clarke et al., 2005], or more recently by propagating solar wind conditions observed near Earth to Saturn under conditions of favourable planetary alignment using a computational model [Clarke et al., 2009]. With regard to their physical origin, it has been suggested that storms result from compression-induced bursts of rapid reconnection in Saturn s magnetic tail which over a few hours closes a significant fraction of the open flux present [Cowley et al., 2005; Badman et al., 2005]. A similar phenomenon is also occasionally observed to occur at Earth [e.g., Zesta et al., 2000; Boudouridis et al., 2003, 2004]. Open flux closure 135

136 injects heated and compressed tail plasma into the outer nightside magnetosphere, leading to poleward expansion of the auroras in this sector, followed by sub-corotation of the injected hot plasma around the planet via dawn [e.g., Cowley et al., 2005; Mitchell et al., 2009]. Nichols et al. [2014] have recently studied small-scale auroral structures within the Saturn storm bulge, showing in one case a brightened arc at the poleward border that expanded eastward at ~3 times the corotation speed (Saturn s rotation period being ~10.7 h), taken to be the signature of on-going tail reconnection. It is no doubt the case that tail reconnection events at Saturn occur on a range of spatial and temporal scales, leading to a corresponding range of auroral events. These include the small-scale transient tail events and nightside emissions examined by Jackman et al. [2013] as mentioned above, as well as medium-scale events such as those discussed by Mitchell et al. [2009] (their Figure 7) and Badman et al. [2014] (their Figure 4). Storms can also be identified by the increase in the intensity of the Saturn Kilometric Radiation (SKR) variations [Mitchell et al., 2005]. In this Chapter we examine the data base of HST Saturn UV auroral images acquired over the past ~17 years, now encompassing ~2000 individual exposures, in order to better characterize large-scale storm events. 7.2 Dataset The overall UV image data set examined here was acquired on 143 individual orbits of the HST that took place between 1997 and 2013, yielding a total of 74.4 h of accumulated exposure time. Figure 7.1 provides an overview of these data, where Figure 7.1a first shows the planetocentric latitude of the Sun at Saturn over this interval, illustrating the developing seasons from southern summer to northern spring across equinox (zero sub-solar latitude) in mid-2009, which strongly influences the view of the aurora from Earth as indicated in section 7.3. Figure 7.1b then shows a histogram of the distribution of HST Saturn exposure time by calendar year, together with the number of individual HST orbits involved. Most of these data were acquired during the interval of Cassini approach to the planet in 2004 and during subsequent in-orbit operations, generally obtained in few-week observing campaigns when Saturn was in near opposition at Earth. Since Saturn is visible to the HST for approximately half of each ~96 min (low Earth) orbit, individual observations generally consist of ~45 min data sequences termed visits, variously subdivided into sequential images possibly 136

137 employing different instrument filter settings. The earlier data were obtained using the Space Telescope Imaging Spectrograph (STIS), with, e.g., 4 images per visit being obtained during the Cassini approach campaign in early Following the subsequent failure of STIS in mid-2004, however, the Solar Blind Channel of the Advanced Camera for Surveys (ACS) has been employed to obtain a larger number of UV images (typically 19 most recently) during each visit. The only exceptions have been a group of three visits at the end of the 2007 campaign obtained using the Wide Field Planetary Camera 2 (WFPC2), following an ACS safemode event that occurred within the campaign interval. Each of the latter ACS and WFPC2 visits has a two-symbol identifier, where the initial letter identifies the observing campaign, or sub-section thereof, while the following number or letter identifies the usually (but not invariably) sequential individual HST orbits within that campaign, or campaign subsection. Due to operational constraints, however, we note that for one visit the observing sequence was split between two consecutive HST orbits, this being visit W9 during the 2008 Saturn campaign, the data for which will be discussed in section 7.3 (Figure 7.4 introduced below). With regard to the cadence of the one-orbit Saturn image data sequences (thus usually visits ), it is evident that the minimum interval between such data sets corresponds to the HST orbital period, with a minimum interval of ~50 min thus elapsing between the end of one data sequence and the beginning of the next. However, while a number of such high-cadence image sets extending, e.g., over ~3-6 HST orbits (~4-8 h) have been acquired during observing campaigns, the visits within typical ~1-2-week HST campaigns, or campaign sub-sections, are more usually separated by ~1-2 days [e.g., Grodent et al., 2005; Clarke et al., 2009; Meredith et al., 2013]. Auroral storm intervals within these images are generally identified by the presence of bright (tens of kr) structured auroral forms which are distributed broadly in local time (LT) via dawn, and which extend significantly poleward of usual locations, often forming a broad band of bright emissions in the dawn sector. To date, similar features have not been observed in the noon to dusk sector. Under usual conditions, the UV aurora at dawn lie in the co-latitude range ~ in the northern hemisphere, and in the southern [Carbary, 2012], the difference between the hemispheres relating to the north-south axial quadrupole asymmetry of the planetary magnetic field [Burton et al., 2010]. The above statement on storm morphology is thus taken to refer 137

138 to bright auroras at dawn extending poleward to ~10 co-latitude and beyond in the northern hemisphere, and to ~12 co-latitude and beyond in the southern, these positions thus lying at least ~4 poleward of the usual poleward limits in this LT sector. Twelve individual events approaching or exceeding these criteria have been identified, most of which have previously been discussed individually in the papers cited above. Here, however, we consider what may be learned by newly examining the ensemble of such events. Storm overview information is given in Figure 7.1c, where we show the fraction of the yearly exposure time in Figure 1b for which storm features were present, together with the number of individual events identified. Overall, storm images so identified correspond to ~12% of the HST Saturn exposure time. Further details of the events are given in Table 7.1. Specifically the columns of the table indicate the storm event identifier employed in this chapter numbered in time sequence together with the relevant figure number, the corresponding HST instrument employed and number of orbits of storm images observed (usually equal to the number of visits), the time and date of the storm and the interval over which the storm was observed by the HST, the time between the last quiet image and the first storm image if any prior images exist within a 3 day interval, the corresponding time between the last storm image and the first quiet image if any exist within a 3 day interval, and finally selected references to prior discussions of these events in the literature. 7.3 HST Image Presentation and Analysis Storm Images and Figure Format In Figure 7.2 we make an initial presentation of image data that exemplifies the figure format, as well as the varying view of auroral storms over the ~17 year data interval. The data reduction process employed is similar for STIS and ACS exposures as described in a number of previous works [e.g., Grodent et al., 2005; Clarke et al., 2009; Nichols et al., 2008; Meredith et al., 2014], with background-subtracted images being projected onto a latitude-longitude grid 1100 km above the IAU 1 bar reference spheroid, this generally corresponding to the peak height of UV auroral emission [Gérard et al., 2009]. The WFPC2 images, however, are of insufficient quality to warrant such processing. Figures 7.2a and 7.2b show two 480 s southern hemisphere images obtained by STIS in December 2000 encompassing storm 1, while Figures 7.2c 138

139 Auroral Storm Observation Statistics Figure 7.1 Overview of HST Saturn observations employed in this study. Panel (a) shows the planetocentric latitude of the Sun at Saturn (deg) over the interval from the start of 1997 to the end of 2013, indicating the developing planetary seasons that strongly influence the view of the auroras from Earth orbit. Panel (b) shows a histogram by calendar year of HST Saturn exposure time (hours), together with the number of HST orbits involved, generally equal to the number of visit imaging data sequences. Panel (c) shows a corresponding histogram by calendar year of the fraction of the exposure time in panel (b) for which an auroral storm was present in the image data, together with the number of individual storm events that were present. 139

140 Table 7.1 Saturn auroral storm events observed by HST observed Storm identifier HST instrument HST storm interval (hour:min UT & date) Interval after Interval before Selected references Figure number HST storm orbits HST storm observation duration a /h before image b /h after image c /h 1 Figure 2 STIS 1 11:26-11:34 7 Dec None 22.4 Prangé et al. [2004] 2 Figure 7 STIS 1 18:39-19:23 26 Jan Clarke et al. [2005] Grodent et al. [2005] 3 Figure 7 STIS 1 01:04-01:48 28 Jan Clarke et al. [2005] Grodent et al. [2005] 4 Figure 8 ACS 2 d (?) 02:50 20 Jan :33 21 Jan Clarke et al. [2009] d (?) 5 - WFPC2 1 15:16-15:52 11 Feb None Clarke et al. [2009] Mitchell et al. [2009] 6 Figure 3 ACS 3 03:53-07:49 8 Feb Clarke et al. [2009] 7 Figure 4 ACS 3 21:34 12 Feb :19 13 Feb Clarke et al. [2009] Belenkaya et al. [2011] Figure 6 ACS 2 09:22-14:54 7 Feb None 18.5 Nichols et al. [2009] 9 Figure 9 ACS 1 e (?) 06:04-06:48 7 April e (?) Belenkaya et al. [2014] 10 Figure 2 ACS 1 18:49-19:33 5 April None None Nichols et al. [2014] 140

141 11 Figure 5 ACS 3 18:41-22:36 20 May Nichols et al. [2014] 12 Figure 5 ACS 1 21:52-22:36 20 May Nichols et al. [2014] a Interval from start of first storm image to end of last b Interval from the end of the before image to the start of the first storm image (limited to 72 h) c Interval from the end of the last storm image to the start of the after image (limited to 72 h) d Unclear whether this interval represents one storm as assumed here or successive activations e Unclear whether weak emissions in the after image represent decayed storm emissions of long duration or are not necessarily storm connected as assumed here and 7.2d similarly show the first and last of s northern hemisphere ACS images of storm 10 obtained on a single visit in April Since storm auroras are generally very bright, except during their late decay phase, only individual (not co-added) image frames are shown throughout. The intensity colour-scale employed is shown on the right of the figure, with the same scale also used consistently throughout the chapter. At the top of each image we indicate the HST instrument employed and image identification information on the left (see appendix for details), together with the date and start and stop time (at Earth) of the image on the right. Saturn times are ~70 min earlier. The images for both northern and southern hemispheres are shown with noon at the bottom, dawn to the left and dusk to the right, such that the view in southern hemisphere cases is through the planet from the north. White dotted circles show colatitude from the corresponding pole at 10 intervals, while white dotted radial lines show longitude at 2 h intervals of LT. The images have been truncated on the nightside to avoid over-stretching of the image pixels as the view approaches the planetary limb. For the December 2000 data under near-southern solstice conditions (Figure 7.1a), almost the whole of the southern auroral oval is visible, while for the April 2013 data obtained during northern spring (Figure 7.1a), the northern images are truncated beyond a co-latitude of ~11 on the midnight meridian. Only during the HST equinoctial campaign from late January to early March 2009 could simultaneous images be obtained of both northern and southern emissions, though both with a very oblique view 141

142 of the dayside auroras in this case (Figure 7.6 introduced below) [e.g., Nichols et al., 2009; Meredith et al., 2013 (see Chapter 4 and 5)]. Examining the images in Figure 7.2, we note that Figure 7.2a shows the single STIS southern hemisphere image obtained on 7 Dec 2000 which displays the first storm feature as identified by Prangé et al. [2004]. This shows a moderately bright bifurcated oval which extends to moderately high southern latitudes in the mid-morning sector. No relevant earlier images were obtained in this case that constrain the prior evolution of the event, while the single subsequent image in Figure 7.2b shows that this feature had disappeared ~22.4 h later. The latter image instead shows a more typical quiet narrow arc extending from post-midnight to post-noon, discussed by Cowley et al. [2004], similar to the weaker lower-latitude emission seen in Figure 2a. Figures 7.2c and 7.2d then show the first and last of 19 northern hemisphere images of storm 10 obtained during a single visit in April 2013, for which there are no other relevant images either before or after (Table 7.1). The data on this event are thus confined to the ~44 min interval of the single visit itself. These images show the presence of a broad bright auroral bulge that extends well poleward of ~10 co-latitude throughout the post-midnight to post-dawn sector, containing highly structured arc-like forms that vary significantly over the interval of the visit. In particular, as mentioned in section 7.1 above, an arc at the poleward edge of the bulge which is just visible in the post-midnight sector in Figure 7.2c is seen to expand eastward at ~3.3 times the corotation rate to span the entire bulge edge into the post-dawn sector by Figure 7.2d [Nichols et al., 2014], this taken to be the signature of a tail reconnection activation event within the storm interval. The overall envelope of the emission, however, containing the overall newly closed flux, is seen to vary only modestly during the visit, the leading poleward edge propagating from dawn towards noon at ~60% of rigid corotation. Such a speed suggests an evolution of the structure over a number of hours, corresponding, e.g., to an interval of ~5 h (half a Saturn rotation) to propagate across a significant ~6 h interval of LT in the midnight-to-dawn sector. As we will see below, the overall auroral morphology of these events suggests that storm 10 in Figures 7.2c and 7.2d corresponds to a relatively early stage of development, perhaps a few hours after initial bulge formation in the post-midnight sector, while storm 1 likely corresponds to a stage of fairly late decay. A central goal of 142

143 Images of Storm 1 Figure 7.2 Images exemplifying the figure format employed in this chapter, together with the change in view of Saturn s UV auroras over the interval , from near southern summer solstice to northern spring. Each image represents a single HST exposure (not co-added), the date of which is given in year-month-day format to the upper right of the image, together with the (Earth) start and stop times in hour: minute: second format. At the top left of each image we indicate the HST instrument employed, either STIS or ACS, together with the image identifier (see appendix 2 for details). The arrow between the images indicates the time sequence, labelled with the time interval between the end of the last image shown and the beginning of the next. Panel (a) shows the southern hemisphere storm features identified by Prangé et al. [2004] in single STIS image V1/1 in December 2000, here termed storm 1, while panel (b) shows the quiet oval auroras observed ~1 day later in single image V2/1 discussed by Cowley et al. [2004]. Panels (c) and (d) represent the first and last ACS images acquired of northern hemisphere storm 10 on visit J1 in April 2013, discussed previously by Nichols et al. [2014]. No relevant images were obtained either before or after in this case. The emission intensity scale (kr) for all these images is shown on the right of the figure, the same scale being employed throughout the chapter. 143

144 this study is to determine what limits can be set on the time scale involved between growth to decay, from examination of the overall storm data set Primary Storm Image Data Sets In Figures we show the three best-observed storm sequences, for storms 6 and 7 in April 2008 [Clarke et al., 2009] and storm 11 in May 2013 [Nichols et al., 2014], respectively. Each involves a set of storm-time images obtained on three consecutive HST orbits encompassing a ~4 h interval from beginning to end, exemplified in the central block of six images in each figure. These show the first and last images obtained on each storm orbit. We note, however, that these involve only a single visit-worth of 19 images split across two orbits in the case of visit W9 during storm 7 in Figure 7.4, as mentioned in section 7.2 above. In each case we also have relevant image sets obtained both before and after these observations, though with significantly greater displacement in time than for the central block of images. At the top of each figure we thus show the last of the individual exposures obtained prior to the storm interval, while at the right of the figure we show the first of the individual exposures obtained after the storm interval. These will be referred to in the discussion below as the before and after images. The sequence of images in these figures is again indicated by the blue arrows, each labelled with the time interval from the end of the previous image to the beginning of the next. The interval over which storm images were obtained by the HST, from the start of the first image to the end of the last, is given in the third column of Table 7.1, while the intervals between the storm images and the before and after images are given in the fourth and fifth columns, respectively. In Figures 7.3 and 7.4 for storms 6 and 7, storm-related features are continuously present in the southern hemisphere throughout the central block of images in each case (Figures 7.3b-7.3g and 7.4b-7.4g). For storm 6 in Figures 7.3b-7.3g these initially show a band of bright auroras extending just poleward of ~10 co-latitude at dawn, brightest near the poleward border, which extend to ~10 h LT in the pre-noon sector. With increasing time over the ~3.9 h interval these are seen to contract somewhat in LT away from noon, but to broaden in latitude in the post-dawn sector to cover much of the range from ~10-20 co-latitude in bright variable structured forms, similar to storm 10 in Figures 7.2c and 7.2d in the observed LT range. These observations show that such forms can persist in the dawn sector for intervals of at least ~4 h, comparable to the 144

145 Images of Storm 6 Figure 7.3 Image sequence of southern hemisphere storm 6 in February 2008, where the central interval of storm observations on three consecutive HST orbits (in order, visits W5, W3, and W4) are shown in panels (b)-(g), with before and after images of the immediately preceding (W2) and succeeding (W6) visits being shown in panels (a) and (h), respectively. As in Figure 7.2, the arrows indicate the time sequence, labelled with the elapsed time between the end of one image shown and the beginning of the next. The HST observation duration of the storm (3.9 h) is given in Table

146 expansion time scale estimated in section 3.1. For storm 7 in Figures 7.4b-7.4g, observed at an interval of ~4.6 days following storm 6, the auroral distribution is initially similar to storm 6, though extending with weaker emissions into the post-noon sector in Figures 7.4b-7.4e. The emissions extend further poleward in the post-dawn sector in Figures 7.4d and 7.4e compared with Figures 7.4b and 7.4c, suggestive of an eastward expansion toward higher latitudes similar to that observed for storm 10 in Figures 7.2c and 7.2d, before bifurcating into two bright regions in Figures 7.4f and 7.4g, one again encompassing a substantial fraction of the range ~10-20 co-latitude in the post-dawn sector, similar to storm 6 in Figures 7.3f and 7.3g, and the other extending to very high latitudes well poleward of ~10 in the immediate pre-noon sector. Again, we find bright spatially extended storm forms persisting over the full ~3.7 h interval of these observations. Importantly, however, neither the before images in Figures 7.3a and 7.4a, observed at prior intervals of ~17 and ~35 h, respectively, nor the after images in Figures 7.3h and 7.4h, at following intervals of ~24 and ~30 h, respectively, show any evidence of storm-related emissions. Instead, these show much weaker emissions, with a narrow arc centred between ~15 and ~19 from dawn to pre-noon, giving way to patchier higher-latitude forms in the noon to dusk sector, corresponding to more usual conditions [e.g., Gérard et al., 2004; Grodent et al., 2005; Meredith et al., 2014]. Figure 7.5 shows similar coverage of storm 11, in which the central block of images in Figures 7.5b to 7.5g appears to encompass the late phase of a storm event [Nichols et al., 2014]. In the first set of storm images in Figure 7.5b and 7.5c we see a highlatitude band of bright auroras in the dawn-to-noon sector, which extends poleward of ~10 co-latitude, with the brightest forms at the poleward border. In the second set in Figures 7.5d and 7.5e these auroras dim and again appear to bifurcate, with a weak form extending in LT from pre-dawn to noon at more usual northern dawn arc locations near ~15, and a continuing brighter form lying poleward of ~10 between noon and postdawn. The similarity of this bifurcated morphology to that of storm 1 in Figure 7.2a is notable. With increasing time in Figure 7.5f the pre-noon forms continue to weaken, and have almost disappeared at the end of the sequence in Figure 7.5g. Higher latitude post-noon to dusk forms are also intermittently present in these images (e.g., Figures 7.5c and 7.5f), likely relating to intermittent magnetopause reconnection processes [Meredith et al., 2014; Chapter 6]. 146

147 Images of Storm 7 Figure 7.4 Image sequence of southern hemisphere storm 7 in February 2008, where the central interval of storm observations on three consecutive HST orbits (W9 split between two orbits plus X3) are shown in panels (b)-(g), with before and after images of the immediately preceding (W8) and succeeding (X4) visits being shown in panels (a) and (h), respectively. The format is the same as Figure

148 Images of Storms 11 and 12 Figure 7.5 Image sequence of northern hemisphere storm 11 in May 2013, where the central interval of storm observations on three consecutive HST orbits (visits JB, JC, and JD) are shown in panels (b)-(g), with before and after images of the immediately preceding (JA) and succeeding (JE) visits being shown in panels (a) and (h), respectively. Storm 12 then emerges in the pre-dawn sector in panels (f) and (g). The format is the same as Figure

149 Despite the fact that these images appear to document the decay phase of storm 11 over an interval of ~3.9 h, it is notable that no storm features are observed in the before image in Figure 7.5a at an interval of ~22 h compared with the start of the first storm image, which shows only a relatively bright narrow arc that extends across dawn at usual co-latitudes. This image, together with the central block of images (Figures 7.5b-7.5g) thus demonstrates that the duration of this storm was less than ~25.5 h. Nichols et al. [2014] have established that the bright emissions observed in this late phase were associated with the trailing region of a hot plasma injection in the mid-to-outer magnetosphere observed in a concurrent Cassini/INCA energetic neutral atom (ENA) image, and inferred that the emissions were associated with the upwarddirected field-aligned current expected to be associated with such an azimuthal gradient in plasma pressure. Transforming the injection backward in time using the observed rotation rate of ~50% of rigid corotation, we further infer that the leading edge of the injection formed in the midnight sector at roughly ~5 UT on the storm day (20 May 2013), consistent with the lack of observation of the storm in the before image ~9 h earlier. Given that the time of decay of the storm was at ~22:30 UT on that day, taken to be essentially the time of Figure 7.5g, the implication is that the lifetime of this storm was ~17.5 h. Certainly no storm features were observed in the after image in Figure 7.5h, obtained following an interval of ~18 h. We further note with Nichols et al. [2014] that a second storm began within the third of the three-orbit sequence shown in Figure 7.5, termed here storm 12. This is observed as a bright high-latitude arc that expanded rapidly at ~1.9 times rigid corotation from the post-midnight sector towards dawn in Figures 7.5f and 7.5g, that was not present ~50 min earlier in Figure 7.5e. This speed is thus comparable with the ~3.3 times rigid corotation found by Nichols et al. [2014] for the eastward expansion of the poleward arc in storm 10 shown in Figures 7.2c and 7.2d, and is thus similarly taken to represent the signature of a tail reconnection activation event observed even earlier in the storm interval. Despite this, no storm signatures are present in the after image in Figure 7.5h, implying an overall lifetime of this event of less than ~20 h.. In Figure 7.6 we show images of storm 8 in February 2009 [Nichols et al., 2009], observed prior to Saturn equinox in August that year. Temporal coverage is less complete than for storms 6, 7, and 11 in Figures , but in this case data for both northern and southern hemispheres are simultaneously available (Figures 7.6a-7.6e and 149

150 Images of Storm 8 Figure 7.6 Image sequence of storm 7 in February 2009, close to Saturn equinox in August 2009, where panels (a)-(e) show the auroras in the northern hemisphere, while panels (f)-(j) show the simultaneous auroras on the same HST visits in the southern hemisphere. Storm-related auroras are observed in panels (a)-(d) and (f)-(i) on the central two visits on non-consecutive HST orbits separated by 4.1 h (visits B1 and B2), with bright auroras of more usual morphology being observed in the after images in panels (e) and (j) (visit B3). No relevant before image was obtained in this case. The overall format is similar to Figure

151 7.6f-7.6j, respectively), albeit with a highly oblique view, allowing approximate conjugacy to be examined. Here the storm images were obtained on two visits separated by 4.1 h, together with an after image obtained following an interval of ~18 h. No relevant before images are available in this case. The images from the first visit (Figures 7.6a and 7.6b in the northern hemisphere, and 7.6f and 7.6g in the southern) show the presence of broad bright auroral forms extending from dawn into the pre-noon sector in both hemispheres, similar to those observed in Figures , but not extending strongly poleward of usual locations at this time. However, the images from the second visit (Figures 7.6c and 7.6d in the northern hemisphere, and 7.6h and 7.6i in the southern), show the presence of bright patchy forms extending to high latitudes in the pre-noon sector together with a weaker narrow dawn arc at lower latitudes, that have features in common with those observed in the later images for both storms 7 (Figures 7.4f and 7.4g) and 11 (Figures 7.5d-7.5g). Overall, these features show approximate conjugacy in the two hemispheres, consistent with a phenomenon located on closed dayside flux tubes, though with some possibly projection-related distortions and differences in intensity between the two hemispheres. At the same time, a general lack of conjugacy is observed in the dusk side emissions, possibly indicative of reconnection-related processes at the magnetopause thus involving open flux tubes [e.g., Meredith et al., 2013, 2014, and earlier chapters]. The after image observed ~18 h later, however, shown for the north and south in Figures 7.6e and 7.6j, respectively, again exhibits more usual morphologies in both hemispheres Implications of Results for Storm Time Scales We now consider what these results imply about the time scale of Saturn auroral storms, given the evident fact that no one set of storm observations captures the whole event from onset to decay. First, we note that in all four observations in which storm features were observed during intervals of imaging covering ~4-5 h (Figures ), such features were observed over the whole interval, though possibly only just so for storm 11 in Figure 7.5, which had decayed to the presence of weak high-latitude forms in the pre-noon sector at the end of the main imaging interval. This implies that the time scale of such events is generally a significant factor longer than ~4-5 h, otherwise such imaging intervals would be dominated by examples in which the storm either appeared or disappeared, or both, within the ~4-5 h observation interval. We already noted in regard to storm 10 in Figures 7.2c and 7.2d that only small storm evolutions 151

152 are observed over the ~40 min intervals of individual visits, a finding that is consistent throughout the examples in Figures However, from the motion towards noon observed in the dawn-sector auroral bulge in storm 10 we also inferred an evolution time for expansion over a ~6 h LT interval into the pre-noon sector of ~5 h. Correspondingly, significant storm evolution is indeed found on the ~4-5 h observation time scales for storms 7, 8, and 11 in Figures , each involving the formation of bifurcated emissions in the pre-noon sector extending to high polar latitudes. Storm 1 in Figure 7.2a appears to be of similar morphology. No such feature was observed in storm 6 in Figure 7.3, however, in which bright high-latitude dawn forms extended equatorward, but not significantly towards noon, over the 4 h interval of observations, indicating observation of an interval following the initial expansion but prior to the onset of decay. From ancillary Cassini ENA image data presented by Nichols et al. [2014], a lifetime of ~17.5 h was also inferred for storm 11, whose decay was observed over a ~4 h interval by HST (Figures 7.5b-7.5g). Second, however, we note that in no case were related storm features observed in the images acquired before or after the central ~4-5 h of observations, being displaced in time typically by intervals of ~20 h. In the three cases for which we have such information, the ~4 h sequence of storm images were located near-centrally within an interval bounded by non-storm images which was ~46 h long for storm 6 in Figure 7.3, ~57 h long for storm 7 in Figure 7.4, and ~44 h long for storm 11 in Figure 7.5, thus showing that these storms must certainly have endured for shorter intervals than these. The overall implication therefore is that the time scale of storms is typically significantly shorter than such intervals, otherwise it is likely that events would have been observed in which storm features bridged the ~20 h gaps between the ~4 h blocks of observations and the before or after image sets. In addition, consistent with this conclusion, we estimated a lifetime of ~17.5 h for storm 11, while the lifetime of storm 12 was certainly less than ~20 h. Our results thus directly imply that storm intervals must generally last for intervals significantly longer than ~4-5 h, but significantly shorter than, say, ~50 h. If roughly the same factor applies in both cases, the overall implication is that the duration of typical storms is likely around ~16 h, i.e., roughly ~3 times longer than ~5 h, and ~3 times shorter than ~50 h. This estimate, essentially ~1.5 Saturn rotations, is then entirely consistent with the lifetime of ~17.5 h inferred for storm 11, and the upper limit 152

153 of ~20 h for storm 12. A reasonable likely range lies between ~1 and ~2 Saturn rotations, i.e., ~11 to ~21 h, where ~11 h still qualifies as larger than ~4-5 h by a factor of ~2, and ~21 h as smaller than ~50 h by a similar factor. Assuming these lifetime values, we may then consider the probability that one will observe either the beginning or the end of a storm within a given ~4-5 h observing sequence, noting that the results in Figures suggest a value no larger than ~25%, given the essential disappearance of storm 11 within such an interval in Figure 7.5. We assume that in order to confidently identify either the beginning or the ending of a storm in an observation of length τ, that the centre of the observing interval must lie within ~ τ 4 on either side of that occurrence, so that the start or end of the storm is observed for at least a quarter of the interval. In that case a storm will be detected if the randomly-placed centre time of the observation lies within a total interval of length T S + ( τ 2), where T S is the duration of the storm, while the beginning or ending of the storm will be observed if the centre time of the observation lies within an interval of length τ within that overall time. The probability of observing a beginning or ending within a set of storm images of length τ is thus P τ ( T + ( τ 2) ) S, such that if we put τ 5 h and T 16 h, then we find P 0. 27, commensurate with the results discussed S above. Lifetimes between ~11 and ~21 h yield probabilities spanning between 0.37 and 0.21, respectively, the latter value appearing to be more likely than the former. Overall, it thus seems safe to conclude that Saturn auroral storms generally last for between one and two Saturn rotations, ~11-21 h, typically around one and a half rotations or ~16 h. A storm can persist for multiple rotations of the planet because the magnetic field lines associated with the region of field-aligned currents sub-corotate with respect to the planet. This sub-corotation occurs as the planet fails to maintain rigid co-rotation of equatorial plasma beyond a certain radial distance. Thus a lifetime of between one and two Saturn rotations is expected from consideration of subcorotating plasma and the typically observed latitude of emission. The further implication of the overall ~12% occurrence frequency of storm-related exposure time in our data set is that the time between individual storms is typically ~5.5 days, within the likely range ~4-7 days, which seems reasonable in terms of known behaviours [e.g., Clarke et al., 2005, 2009]. Alternatively, if we group the HST data into observing units in which a set of HST Saturn visits lying within one Saturn 153

154 rotation period (comparable to a storm lifetime) is counted as one unit, separated from the next by at least one Saturn rotation period (generally longer than a storm interval), we find that storm signatures are present for 11 out of 96 such units (one unit contains two storms as seen in Figure 7.5). This corresponds to a very similar storm percentage occurrence of ~11.5%, from which essentially the same conclusions follow. Further discussion of these findings is provided in section

155 Images of Storms 2 and 3 Figure 7.7 Image sequence of southern hemisphere storms 2 and 3 observed by STIS in January Storm auroras are observed in the images in both visits V11 and V12 shown in panels (b)-(e) separated by ~30 h, suggested here to represent separate activations and hence given separate storm identification numbers. Before and after images from visits V10 and V13 are shown in panels (a) and (f), respectively. The overall format is similar to Figure Further Storm Examples In Figures we show most of the remaining storm examples in a similar format to Figures , and provide brief discussion in relation to the above conclusions. 155

156 Figure 7.7 shows southern hemisphere STIS images obtained during the Cassini approach campaign in January 2004, that are known to relate to the arrival at Saturn of a ~4-day solar wind corotating interaction region (CIR) compression [Clarke et al., 2005; Crary et al., 2005; Grodent et al., 2005; Bunce et al., 2006]. In this case, the before image shown in Figure 7.7a, the last of four STIS images from visit V10 (see appendix) shows a typical narrow dawn arc, while the images obtained ~42 h later in visit V11 in Figures 7.7b and 7.7c show storm 2, in which a broad region of bright highly-structured emission centred in the dawn sector extends almost to the pole. To date these images represent the best-observed highly disturbed storm emissions, though clearly similar in nature to the emissions observed in other storm intervals, particularly for storm 10 in Figure 7.2d. The subsequent images obtained ~30 h later in visit V12, shown in Figures 7.7d and 7.7e, also show bright high-latitude emissions in the dawn sector, but having completely altered morphology compared with the earlier storm emissions in Figures 7.7b and 7.7c. In view of the above discussion of storm lifetimes, typically ~16 h in the range ~11-21 h, it thus seems likely that this represents a new disturbance observed early in its evolution, with a high-latitude auroral bulge forming in the post-midnight sector and extending perceptibly eastward over the interval of the visit. For this reason we have termed this storm 3 within the present study, though given the known nature of the event, disturbed auroras may have been present over much of the ~4 day CIR-related compression interval, of which storms 2 and 3 represent separate activations. The final after image in Figure 7.7f, obtained ~65 h later in V13, then again shows a more usual oval morphology, brighter at dawn than at dusk, but with an unusual number of bright patchy structures still being evident. Indeed, given the fact that this image was obtained towards the end of the on-going CIR event, it may represent the late decay phase of yet further activation, not dissimilar in post-dawn bifurcated morphology to storm 1 in Figure 7.2a, or storm 11 in Figure 7.5d and 7.5e. However, since V13 was the final visit in the 2004 HST campaign, no further information relevant to this issue is available. Figure 7.8 shows southern hemisphere images for storm 4, observed in January 2007 [Clarke et al., 2009], for which the features are much less well marked. The before image in Figure 8a shows a relatively bright oval of usual morphology, while the storm images observed after an interval of ~16.4 h in Figures 7.8b and 7.8c show similar but broader bright auroral forms that extend to somewhat higher latitudes in the dawn 156

157 Images of Storm 4 Figure 7.8 Image sequence of southern hemisphere storm 4 observed in January Storm auroras are observed in panels (b) and (c) (visit S9), and (possibly) panels (d) and (e) (visit T0) separated by ~23.2 h, with before and after images being shown in panels (a) and (f) (visits S8 and T1). The overall format is similar to Figure 7.3. sector. Similar but dimmer and patchier emissions persist after an interval of ~23.2 h in Figures 7.8d and 7.8e, with bright high-latitude forms now present in the pre-noon sector, similar to storms 7, 8, and 11 in Figures , which have disappeared in the 157

158 after image in Figure 7.8f, observed after a further ~20.0 h. Due to the lack of continued imaging between the two central visits exhibiting disturbed high latitudes forms (Figures 7.8b-7.8e), the overall development remains unclear in this case. However, due to the continued similarity of the emissions, this may possibly represent a case of unusual longevity, at least ~24.7 h between the start of Figure 7.8b and the end of Figure 7.8e, somewhat beyond the longer limit of the ~11-21 h lifetime range discussed above. Images of Storm 9 Figure 7.9 Image sequence of northern hemisphere storm 9 observed in April Storm auroras are observed in panels (b) and (c) (visit H4). A before image is shown in panel (a) (visit H3). In the after image in panel (e) (visit H5), weak high-latitude emissions remain present, but with a temporal separation of ~63 h from the storm images, this is not taken necessarily to be physically connected with the storm. The overall format is similar to Figure 7.3. Figure 7.9 shows images for storm 9 observed in the northern hemisphere in April The before image in Figure 7.9a shows relatively weak emissions of usual morphology, while the storm images observed in a single visit in Figures 7.9b and 158