The Dynamics of Saturn's Main Aurorae

Saturn's main aurorae are thought to be generated by plasma flow shears associated with a gradient in angular plasma velocity in the outer magnetosphere. Dungey cycle convection across the polar cap, in combination with rotational flow, may maximize (minimize) this flow shear at dawn (dusk) under strong solar wind driving. Using imagery from Cassini's Ultraviolet Imaging Spectrograph, we surprisingly find no related asymmetry in auroral power but demonstrate that the previously observed “dawn arc” is a signature of quasiperiodic auroral plasma injections commencing near dawn, which seem to be transient signatures of magnetotail reconnection and not part of the static main aurorae. We conclude that direct Dungey cycle driving in Saturn's magnetosphere is small compared to internal driving under usual conditions. Saturn's large‐scale auroral dynamics hence seem predominantly controlled by internal plasma loading, with plasma release in the magnetotail being triggered both internally through planetary period oscillation effects and externally through solar wind compressions.

Plain Language Summary Saturn's main aurorae are thought to be generated as a result of sheared plasma flows near the boundary between the rapidly rotating magnetosphere of Saturn and interplanetary space. It is often assumed that the steady flow of the solar wind away from the Sun has an impact on this flow shear; due to the direction of Saturn's rotation the aurorae would then have to be brighter at the planet's dawnside than on its duskside, which was observed in previous studies. Here we analyze a large set of auroral images taken by Cassini's ultraviolet camera, but we cannot find any sign of such an asymmetry. This indicates that the impact of the solar wind on Saturn's aurorae must be smaller than previously thought and that they must instead mainly be controlled from within the system. This assumption is supported by our observations of bright auroral patches at dawn, which are likely a signature of plasma being released from Saturn's magnetosphere and appear at quite regular periods corresponding to Saturn's rotation period.
Under strong solar wind driving (increased solar wind velocity and density), active Dungey cycle reconnection between the interplanetary magnetic field and Saturn's magnetic field at the dayside magnetopause may prompt an antisunward flow in the slowly subcorotating polar open field region just like at Earth (Dungey, 1961). At dawn, this Dungey cycle convection across the polar cap-here oppositely directed to the subcorotating magnetospheric plasma flow-would act to enhance the (rotational) plasma flow shear associated with the generation of Saturn's main aurorae and hence also the auroral brightness. Conversely, strong solar wind driving should lead to a reduction of this plasma flow shear and the auroral brightness at dusk (e.g., Cowley, Bunce & Prangé 2004;Jackman & Cowley, 2006). Adding to this local time (LT) asymmetry, the Dungey and Vasyliunas cycle return flows are expected to pass from the magnetotail toward the dayside via dawn due to the rapid rotation of the magnetosphere (e.g., Cowley, Bunce & Prangé 2004;Vasyliūnas, 1983). However, the importance of Dungey cycle convection at Saturn is disputed as magnetopause reconnection may be inhibited across parts of the magnetopause (e.g., Desroche et al., 2013;Masters et al., 2012Masters et al., , 2014 and viscous interactions mediated by Kelvin-Helmholtz instabilities may instead be the main coupling mechanism between the solar wind and Saturn's magnetosphere (e.g., Delamere & Bagenal, 2010;Delamere et al., 2013).
Previous studies using auroral imagery obtained by the Hubble Space Telescope in the ultraviolet (UV) wavelength band (e.g., Kinrade et al., 2018;Lamy et al., 2009Lamy et al., , 2018Nichols et al., 2016) and by the Cassini spacecraft at infrared (IR) and UV wavelengths (e.g., Bader et al., 2018;Badman et al., 2011;Carbary, 2012) have statistically identified such a brightness asymmetry, seemingly confirming that Saturn's main aurorae are indeed significantly solar wind driven. However, most of these studies used rather small sets of single exposures lacking context and/or short observation series without good time resolution to obtain statistical averages, hence not taking into account the complicated dynamics of Saturn's aurora which had already been observed by the Voyager spacecraft (Sandel & Broadfoot, 1981;Sandel et al., 1982).
In this study we use extensive sets of auroral imagery obtained by the Cassini spacecraft to investigate the dynamics of Saturn's main aurorae and shed more light on its generation mechanisms. We present the data set and describe our analysis methods in section 2. In section 3 we analyze observations consistent with quiet auroral conditions to reveal the structure of subcorotationally driven main aurorae and their modulation by planetary period oscillations (PPOs), while in section 4 we describe the added complexity brought into the system by magnetotail dynamics, causing transient large-scale brightenings. We summarize our findings and propose an updated model of Saturn's main aurorae in section 5.

Data and Methods
NASA's Cassini spacecraft orbited Saturn for over 13 years, providing a rich set of auroral observations in the UV spectrum with its Ultraviolet Imaging Spectrograph (UVIS; Esposito et al., 2004). Here we investigate Saturn's auroral dynamics and therefore select observation windows where many images were taken in quick succession (exposure time <20 min) for several hours. This corresponds to auroral observations from high apoapsis where Cassini moved relatively slowly, preserving the same viewing geometry for long periods; and where the large distance from Saturn allowed UVIS to cover the entire auroral oval with a single slit scan, allowing for low exposure times. Nearly all available observations of this kind fall into 2014/2016/2017, and all are from Saturn's northern hemisphere.

Cassini-UVIS Imagery
The Cassini-UVIS instrument includes two telescope-spectrographs observing in the 56-to 118-nm (extreme ultraviolet) and 110-to 190-nm (far ultraviolet, or FUV) wavelength ranges; most of Saturn's auroral UV emissions are observed in the FUV band. The UVIS FUV slit has a field of view of 1.5 × 64 mrad, with 64 spatial pixels of size 1.5 × 1 mrad each arranged along a single line. Pseudo-images of the aurora are obtained by scanning this slit across the auroral region. Several successive scans may be necessary to cover the entire region of interest depending on Cassini's distance from Saturn, increasing the exposure time of auroral images. The total exposure time for a pseudo-image of the entire auroral oval can vary between 6 and 180 min.
Each image is polar projected onto a planetocentric polar grid with resolution 0.5 • × 0.25 • (lon × lat) at an altitude of 1,100 km above Saturn's 1-bar pressure surface (oblate spheroid with R SEQ = 60,268 km and R SPO = 54,364 km as equatorial and polar radii), the approximate altitude of Saturn's auroral emissions ). Cassini SPICE pointing information is used to perform the projection. The spectrum recorded by each pixel of the UVIS FUV sensor, observed in 1,024 spectral bins, is reduced to total unabsorbed H 2 emission intensity (70-170 nm) by multiplying the intensity measured in the 155-to 162-nm range by the factor 8.1 (Gustin et al., 2016(Gustin et al., , 2017. Using this method, dayglow emission and hydrocarbon absorption affect the estimated total unabsorbed H 2 intensity as little as possible. Even so, some dayglow is still apparent in most UVIS images; it is removed as previously described in  in order to obtain accurate auroral brightnesses and emission powers.
Many of the images in this study have quite low spatial resolutions, with single pixels extending over up to 5 • in colatitude or 1 hr in LT. However, this issue is circumvented by integrating over the auroral brightness to obtain the emitted radiant flux, or "auroral power," as laid out in the supporting information of this paper. A large instrument pixel covering a small bright auroral feature and its surroundings is dimmer than the actual brightness maximum of the observed emission-however, the pixel brightness corresponds to the average brightness of the area it subtends during the time of the exposure. Integrating over this area therefore gives a quite exact measure of the auroral power nevertheless. We reduce each image by integrating its auroral brightness between 8 • and 22 • colatitude in 36 LT bins and thereby obtain a distribution of auroral power per hour of LT. This latitudinal range fully includes the statistical position of the main aurorae and associated uncertainties . Arranging these integrated powers of all images along the horizontal axis-taking into account the start and stop times of each exposure-we obtain a keogram.

PPO Systems
Each of Saturn's hemispheres is associated with one PPO system, a complex array of FACs spanning the entire magnetosphere of Saturn (e.g., Hunt et al., 2014;Provan et al., 2011;Southwood & Kivelson, 2007) likely associated with vortical flow structures in Saturn's polar ionospheres (e.g., Hunt et al., 2014;Southwood & Cowley, 2014). Their rotation at roughly the planetary period generates periodic signatures in all plasma properties and processes in Saturn's environment, the two systems exhibiting close but distinct periods that vary with time (e.g., Provan et al., 2013Provan et al., , 2016. Each PPO system is usually dominant in one hemisphere, but its associated system of FACs partly closes in the opposite hemisphere such that each hemisphere experiences a double modulation of, for example, auroral FACs by both the northern and southern PPO systems (e.g., Bader et al., 2018;Hunt et al., 2015;Provan et al., 2018).
A sketch of the northern PPO system is shown in Figure S1 in the supporting information, with Figure  S1a showing the magnetic field and electric currents in the equatorial plane and Figure S1b showing the electric currents and atmospheric/ionospheric flows in the northern polar ionosphere. The southern PPO system effects the same pattern of upward/downward FACs in the northern hemisphere as shown here for the northern system. Depending on the relative orientation between the two systems, their associated FACs can combine to intensify or negate one another. The orientation of the two PPO systems is described by the PPO phase angles N,S , the counterclockwise azimuthal angle between the PPO magnetic perturbation dipoles in the equatorial plane and local noon. In this study we use the phase angles determined by Provan et al. (2016Provan et al. ( , 2018. PPO-fixed reference frames are defined using the phase values N,S , giving the clockwise angle from the PPO dipole direction. In the northern hemisphere, the PPO-associated upward FACs maximize at N,S = 90 • , with the downward FACs maximizing at N,S = 270 • (e.g., Hunt et al., 2014). The modulation effect is hence largest when the two PPO systems are in phase, their perturbation dipoles parallel. In the keograms shown through this study and in the supporting information, N,S = 90 • is marked with yellow lines.
The PPO-induced modulation of the equatorial current sheet thickness shows a different phasing; the current sheet being thinnest at N = 0 • and S = 180 • Cowley & Provan, 2017;Jackman et al., 2016). This modulation is therefore emphasized when the two PPO systems are in antiphase. In Figures 4 and S4, the two systems were within 45 • of antiphase-orange dotted lines hence indicate the approximate location at which the PPO-related thinning of the current sheet is expected to be most pronounced.

Saturn's Quiet Main Aurora: Subcorotational and PPO Systems
In quiet and steady auroral conditions, the main aurorae should form a quasi-static ring of emission around both poles corresponding to the region of peak flow shear between the rapidly rotating magnetospheric (a-c) Three UVIS images within this sequence, each about 5-6 hr apart. The view is from above the planet down onto the north pole, with noon/the Sun toward the bottom. White numbers around each image mark local time (LT), and gray concentric circles mark the northern colatitude in steps of 10 • . The gray shaded and hatched regions (colatitudes >22 • and <8 • ) were ignored for the integration of UV powers. The start and exposure time of each observation are given on top. Shown is the background-subtracted auroral brightness in kilo-Rayleigh; note the logarithmic scale. (d) UV power keogram of all images in this sequence; logarithmic power scale. The UV power between 8 • and 22 • colatitude was integrated in 36 LT bins for each image and is arranged by the image collection time such that UT increases to the right. Diagonal lines mark planetary period oscillation (PPO) upward field-aligned current regions propagating around the planet at their respective PPO rotation rate. Dashed horizontal lines limit the "dawn" (blue) and "dusk" (red) LT bins whose UV powers were added for the line plots shown in the bottom panel. Black arrows on top of the panel mark the collection times of the example images shown in (a-c). (e) Line plots of the total, dawn and dusk UV powers. plasma and the slowly rotating plasma in the polar open field region (e.g., Cowley, Bunce & O'Rourke 2004;Cowley, Bunce & Prangé 2004;Cowley et al., 2005;Stallard et al., 2007;Vasyliūnas, 2016). Lacking continuous upstream solar wind monitoring, we cannot know for sure the solar wind conditions during most of Cassini's observation sequences. We therefore identify "quiet conditions" as imaging sequences where no large-scale transient brightenings (total power > 20 GW for >5 hr) were observed, indicating low magnetic reconnection activity at both dayside and nightside as such events would manifest as bifurcations at noon-dusk LTs (e.g., Badman et al., 2013;Radioti et al., 2011; or as bright transient features at midnight-dawn LTs (e.g., Jackman et al., 2013;Lamy et al., 2013). Figure 1 shows an auroral keogram of one such period without transient events, covering more than two full Saturn rotations (∼25 hr) with near-continuous imagery.
We notice a periodic modulation of the emitted UV auroral power, which is well explained with rotating patterns of upward and downward FACs associated with Saturn's PPO systems. In this case, the two PPO systems are aligned nearly parallel and rotating in phase-their upward and downward FAC regions overlap and enhance the associated modulations of the static main aurorae. The dawn UV power is largest roughly when the expected PPO upward FAC maxima pass and weakest during opposite PPO orientations and varies by nearly a factor of 10. Consequently, the main oval seemingly disappears near dawn as the combined PPO downward FAC regions sweep over and negate the subcorotational system's upward currents (see Figure 1b). While this modulation should theoretically be of comparable strength at all LTs , it is here barely discernible at dusk. This difference in modulation amplitude agrees with statistical findings  and might be related to a seemingly larger spread of the PPO currents at dusk than at dawn .
Neither the keogram (Figure 1d) nor the summed dawn and dusk UV powers (Figure 1e) show an asymmetry as expected during periods of significant solar wind driving-this is not surprising, as the time period considered here shows rather quiet auroral conditions, probably indicating quiet solar wind conditions and low Dungey cycle activity. Surprisingly though, the duskside is noticeably brighter than the dawnside during most of the observation sequence. This can partly be explained with quasiperiodic flashes, possibly a sign of small-scale magnetodisc reconnection observed preferentially at dusk . These have been shown to occur near-constantly and manifest as spikes in the dusk power (Figure 1e), but they do not fully account for the underlying steady asymmetry between dawn and dusk which we observe here. At Jupiter, a similar asymmetry was observed and suggested to be related to a partial ring current in the nightside magnetosphere (Bonfond et al., 2015), but it is unclear whether a similar process could be important in Saturn's magnetosphere. The case study presented in Figure 1 is not the only quiet sequence observed. Considering only sequences with quasi-continuous coverage of at least one Saturn rotation, we find additional quiet sequences at 2014 DOY 130/147/158-159/311 ( Figure S2)-including overall 476 images with 67 hr of total exposure time, corresponding to just over six Saturn rotations. A UV power-LT histogram for these images is shown in Figure 2a, with the mean and median power per LT added as line plots; the dawn and dusk slices of this histogram are compared in Figure 2b. We observe similar UV powers through all LTs, disagreeing with previously discussed UV and IR auroral intensity distributions (e.g., Badman et al., 2011;Bader et al., 2018;Carbary, 2012;Kinrade et al., 2018;Lamy et al., 2009Lamy et al., , 2018Nichols et al., 2016) with a brightness peak at dawn probably due to our choice of quiet periods. Centered on roughly 0.5 GW per 40 min LT bin, the powers are more variable and feature a more prominent tail toward lower powers at dawn/noon than at dusk/midnight. The occurrence of UV powers below the lower histogram limit (see Figure 2b) is much larger at dawn, indicating longer intervals with a complete absence of auroral emissions.
There appears to be a dip in the average power at noon, somewhat reminiscent of the noon discontinuity in the Jovian main emission (e.g., Radioti et al., 2008;Ray et al., 2014). The currents associated with Jupiter's main emission are thought to be internally driven by the breakdown of corotation in the magnetodisc, which is less significant at the solar wind-compressed dayside (e.g., Chané et al., 2017). images, corresponding to ∼122 hr of exposure within the ∼411-hr observation window-a data set quite representative of Saturn's typical auroral dynamics, likely capturing a variety of different solar wind conditions. As each observation block covers roughly one full Saturn rotation (or PPO phase cycle) or more, we assume no significant bias in PPO phases. A keogram of the entire set is shown in Figure S3, including solar wind properties propagated from OMNI which indicate initially typical solar wind conditions, likely with average Dungey cycle activity, followed by rather quiet conditions. Note that two of the observation blocks (2014 DOY 147/158-159) were considered to show quiet auroral conditions and included in the corresponding analysis above as well as here. Figure 2c differs from the histogram of the quiet aurora (Figure 2a) significantly only at dawn to postnoon LTs. We see a much wider spread in UV power at dawn than in quiet conditions, but do not observe a significant statistical dawn brightening (see Figure 2d). On the contrary, again the median UV power is larger at dusk than at dawn. The mean and median UV power distributions (Figure 2c) are in close agreement between noon and midnight but clearly differ near dawn-the mean maximizing here, while the median minimizes. The mean auroral power agrees very well with intensity averages of previous observations which all showed a distinct peak between 6 and 9 LT (e.g., Bader et al., 2018;Badman et al., 2011;Carbary, 2012;Kinrade et al., 2018;Lamy et al., 2009Lamy et al., , 2018Nichols et al., 2016)-but, as seen here, the mean UV intensity/power is obviously not a good representation of the typical state of the aurora. The median directly shows that in more cases than not, the dawn aurora is dimmer than the dusk aurora and not brighter; it is the few transient high-power events subcorotating through dawn which skew the mean power to unrepresentative high values at these LTs. Figure 3 compares the mean and median brightness of the actual images in this data set.

Typical Auroral Conditions and Periodic Magnetotail Dynamics
A detailed view of the 2014 DOY 156-162 keograms is shown in Figure 4 ( Figure S4 shows 2014 DOY 144-149). Figures 4a-4d show an example UVIS image from each observation block-note that the observation geometry worsens toward the end, with the last images lacking coverage beyond ∼20 • colatitude from the pole between 18-24 LT. The integrated UV powers at these LTs are hence more uncertain as empty pixels have been filled with longitudinally averaged values of each latitudinal bin before integration.
The quiet auroral oval is overlaid with repeated powerful auroral plasma injection events (Mitchell et al., 2015) at Saturn's dawnside, which almost never rotate past noon as the perturbed source population's free energy is gradually deposited in Saturn's atmosphere, generating aurorae. The related rotating injected hot plasma populations seen in energetic neutral atom images do not stall at noon but continue rotating near rigidly with diminishing intensity back into the nightside sector where they appear to be reenergized with every pass (Carbary & Mitchell, 2017;Mitchell et al., 2009). All injections commence near dawn, indicating nightside reconnection and the consequent magnetic dipolarization (Yao et al., 2017) as a likely cause (Radioti et al., 2016)-considering the significant bendback of the magnetic field at dawn, this LT region maps well into Saturn's nightside. An auroral signature of this process may be the result of particle acceleration and precipitation during the dipolarization (Mitchell et al., 2015).
The injection events vary strongly in power, but show a regularity indicating a trigger mechanism internal to Saturn's magnetosphere. One known instigator of magnetotail reconnection is the PPO-induced modulation of the current sheet thickness Cowley & Provan, 2017;Jackman et al., 2016), which is most pronounced when the two PPO systems rotate in antiphase. This is the case in Figure 4e; the approximate location at which the current sheet is expected to be thinnest and reconnection is more likely to occur is indicated with orange dotted lines. Most of the injections observed are triggered within some 3-hr LT of these highlighted locations, suggesting the PPO current sheet thinning effect to indeed be a main influence on the occurrence of the observed large-scale disturbances.

Discussion and Conclusions
It is clear that Saturn's main aurorae are more dynamic than previous statistical studies may suggest. We conclude that the presently called "main aurorae" are associated with three different magnetospheric processes: the subcorotational FAC system, the two PPO FAC systems, and the occurrence of large-scale magnetotail reconnection events.
The subcorotational system is a largely or completely LT-invariant system of FACs which are likely generated by flow shears between plasma populations subcorotating at different speeds in the middle and outer magnetosphere . This agrees with field line mapping of the main aurorae which places the main upward FAC sheet at an equatorial distance beyond 10 R S , outward from the middle ring current (e.g., Belenkaya et al., 2014;Talboys et al., 2011). The flow of the solar wind and the associated Dungey cycle activity (e.g., Cowley, Bunce & Prangé 2004;Jackman & Cowley, 2006) seem to have little to no impact on this system, since no significant LT asymmetries in auroral FACs  and auroral brightness are observed, contrary to previous findings (e.g., Bader et al., 2018;Badman et al., 2011;Carbary, 2012;Kinrade et al., 2018;Lamy et al., 2009Lamy et al., , 2018Nichols et al., 2016), where observed asymmetries were likely an artifact of small data sets and averaging procedures unsuitable for determining the full variability of Saturn's auroral dynamics. This is supported by earlier studies estimating the Dungey cycle contribution to magnetic flux transport to be roughly an order of magnitude lower than the contribution arising from rotational flows in quiet solar wind conditions such that no asymmetry in auroral brightness is expected (e.g., Badman & Cowley, 2007;Badman et al., 2005). During solar wind compressions, significant asymmetries should theoretically arise (e.g., Badman & Cowley, 2007;Jackman et al., 2007) but will in reality be subsumed into the major auroral dynamics, that is, poleward extending auroral storms which occur simultaneously. The subcorotational system alone would cause a rather steady ring of upward FACs and associated auroral emissions around Saturn's poles corresponding to the region of highest flow shear, possibly with secondary emissions associated with corotation breakdown currents like Jupiter's main aurorae Stallard et al., 2007Stallard et al., , 2008).
This subcorotational system is enhanced and reduced by the asymmetric PPO-related FACs flowing at the same latitudes (e.g., Hunt et al., 2014Hunt et al., , 2015. The slightly differing periods of the two PPO systems result in a double-sinusoidal modulation of the main oval's auroral brightness through LT, as the PPO and subcorotational FACs add up on one side of the planet but nearly negate each other on the opposite side )-we found this modulation to be significantly stronger at dawn than at dusk. These two current systems combine to generate what should be considered the "main emission". Unintuitively though, the main (quasi-static and continuous) emission is often not dominant in Saturn's aurora, as it is quite dim (up to ∼10 kR). It is overpowered significantly by large and bright patches which are likely a consequence of magnetic dipolarization events (e.g., Jackman et al., 2013;Lamy et al., 2013;Radioti et al., 2016) and which usually emerge between midnight and dawn LTs. They subcorotate and usually disperse before reaching dusk. Their occurrence seems to be partly governed by the PPO-induced thinning of the current sheet Cowley & Provan, 2017;Jackman et al., 2016); this was already observed in modeling studies Zieger et al., 2010) and is likely related to similarly periodic plasma heating and ring current intensifications observed in energetic neutral atom measurements (Mitchell et al., 2009;Nichols et al., 2014). We observe such auroral plasma injection events about once per Saturn rotation, in rough agreement with direct plasmoid observations (Jackman et al., 2011 and Saturn's estimated magnetospheric refresh rate (Rymer et al., 2013).
Previous studies have further observed a clear dependence of magnetotail reconnection on solar wind conditions, as, for example, solar wind compression regions are known to trigger magnetotail reconnection and auroral storms (e.g., Badman et al., 2016;Clarke et al., 2005Clarke et al., , 2009Cowley et al., 2005;Crary et al., 2005;Kidder et al., 2012;Palmaerts et al., 2018), roughly about once per week . Quiet solar wind conditions can lead to an expansion of the magnetotail and an accumulation of open flux as magnetotail reconnection is impeded Jackman et al., 2010), and fewer or no auroral injections are observed (Gérard et al., 2006). Moreover, higher magnetopause reconnection rates cause higher flux loading, thereby indirectly promoting magnetotail reconnection events Badman et al., 2014;Jackman, 2004).
These results are an important step toward a better understanding of the global dynamics of Saturn's magnetosphere and the internal and external factors at play, providing a crucial framework for future studies. Analyzing in situ data from past Saturn missions as well as modeling the system theoretically in the light of these new findings will help investigate Saturn's global plasma circulation more thoroughly, helping unravel the physics of rotating magnetospheres in general.