We present observations of the oscillation of the dawn-dusk locations of Saturn's northern and southern UV auroral ovals obtained using the Hubble Space Telescope during the 2009 equinoctial campaign. We determine the dawn-dusk locations of the centers of the southern and northern auroral ovals from the mean of the dawnward and duskward extents of the emission, and order the computed locations by the phases of the respective SKR oscillations for each hemisphere. We show that statistically significant ∼1–2° oscillations of the dawn-dusk location of the auroral ovals are evident, with the duskward displacement being in lagging quadrature with the SKR power. These results, the first to indicate that the location of the northern auroral oval oscillates, show that the cause of the oscillation is an external magnetospheric current system, and the sense of the oscillations is consistent with the expected displacement caused by magnetic perturbations observed throughout Saturn's magnetosphere.
 Despite the high degree of axisymmetry of Saturn's internal magnetic field [Dougherty et al., 2004], an oscillation in the location of Saturn's southern ultraviolet (UV) auroral oval location at a period near to that of planetary rotation has been reported by Nichols et al. . These authors used Hubble Space Telescope (HST) images obtained in 2007 and 2008 to show that for both intervals examined the southern oval oscillated around eccentric ellipses with ∼1–2° semi-major axes oriented toward pre-noon/pre-midnight, with period 10.76 h ± 0.15 h, i.e., close to the period of the southern oscillations in Saturn Kilometric Radiation (SKR) intensity [e.g., Kurth et al., 2008]. However, as they noted, Nichols et al.  were unable to conclusively determine whether the oscillation was caused by a small, previously unobserved dipole tilt of ∼1°, or an external magnetospheric current system, such as those envisaged to be produced by, e.g., non-axisymmetric convection of internally produced plasma [Gurnett et al., 2007; Goldreich and Farmer, 2007] or seasonal asymmetries [Southwood and Kivelson, 2007; Khurana et al., 2009; Provan et al., 2009a]. They concluded that this ambiguity in the result would remain until simultaneous observations of the oscillations of both north and south conjugate ovals were obtained, to determine their relative phase. Since then, Gurnett et al.  have reported that the periods of the oscillations of the northern and southern SKR intensities differ, and Provan et al. [2009b] have shown that the sense of the oscillations in the southern oval observed by Nichols et al.  is consistent with the expected offset due to a quasi-uniform equatorial field of a few nT in amplitude rotating with the southern SKR period that has been observed in the quasi-dipolar ‘core’ region of the magnetosphere. Recently, Andrews et al.  showed that magnetic perturbations observed by Cassini at high latitudes take the form of planet-centered transverse dipoles rotating at the respective SKR periods, while Nichols et al.  have shown that both the southern and northern observed auroral powers oscillate at the varying periods of the SKR. These results suggest that the auroral emission is influenced by an external magnetospheric current system, but observations of the oscillation of the conjugate ovals, and thus their respective phases, has yet to be presented. In mid-2009, Saturn passed equinox, providing a unique opportunity to observe both northern and southern auroral ovals simultaneously using HST [Nichols et al., 2009], albeit from an oblique viewpoint. In this paper we thus present observations of the oscillation in the dawn-dusk displacement of Saturn's conjugate auroral emission in order to determine their respective phases.
 We employ images of Saturn's conjugate UV auroral emission obtained by the Solar Blind Channel (SBC) of the Advanced Camera for Surveys (ACS) onboard HST over days 23–66 (i.e., January–March) 2009 [Nichols et al., 2009]. The processing of these images has been extensively discussed previously [see, e.g., Clarke et al., 2009; Nichols et al., 2009], such that here we provide only a brief overview. The ACS/SBC detector is a 1024 × 1024 Multi-Anode Microchannel Array, with a bandpass of 115–170 nm and an average resolution of ∼0.032 arcsec pixel−1, such that the field of view is 35 × 31 arcsec2. Nineteen 100 s images were obtained during each ∼1 h observing interval, which executed in groups of 1–5 over the observing interval. The images were rotated such that the north pole was oriented toward the top and scaled to a standard distance of 8.2 AU. The center pixel of the planet was found by fitting the planet's major rings and limb to a simulated profile that takes into account the planet's obliquity and terminator. In this respect the major rings are extremely useful since they provide a very sharp boundary to which to fit, and the accuracy of the center pixel location is estimated to be 2 pixels in both the horizontal and vertical directions. Movies of the images showing the full disk of the planet and individual image start times can be found in the auxiliary material presented by Nichols et al. . The reflected solar UV disk was subtracted by fitting modified Minnaert functions separately to the dawn and dusk halves of the disk, and applying a latitudinal intensity pattern obtained by summing the images. Finally, the images were summed over groups of 5 images in order to increase the signal-to-noise at the cost of ∼5° blurring at the central meridian longitude (CML) for corotating features. Four example processed images, focusing on the auroral regions, are shown in Figure 1.
Nichols et al.  determined the location of the equatorward boundary of the auroral emission by fitting circles to latitude-longitude projections of the auroral emission. However, the viewing geometry of the equinoctial images obtained in 2009 is such that only the dayside aurora is visible, and from highly oblique angles such that the auroral curtain is significantly stretched in such projections. It is thus not possible to obtain detailed information regarding the noon-midnight motion of the auroral ovals in 2009, although the dawn-dusk motion is still clearly observable. In this study the dawn-dusk boundaries of the emission in each image are thus determined by vertically averaging the intensity over horizontal strips 15 pixels wide over the auroral region, as indicated by the white horizontal dotted lines in Figure 1, smoothing the averaged intensity profiles over a modest boxcar width of 5 pixels to reduce noise, and obtaining the locations of the dawnward and duskward extremities at which the integrated brightness is greater than 3 kR. The threshold of 3 kR was chosen since it is 3 standard deviations above the noise floor of ∼1.2 ± 0.6 kR as measured just outside the auroral region, and is significantly below the typical emission peak at ∼10 kR. The smoothed vertically-averaged intensity profiles obtained from each of the four examples displayed in Figure 1 are shown above each image, and the dawnward and duskward boundaries are shown by the left- and right-hand vertical red lines, respectively. As is apparent from Figure 1, the threshold of 3 kR reasonably represents the boundary of the fainter edges of the auroral emission and is well away from the more poleward regions of peak emission. The dawn-dusk location of the center of the auroral oval Y (so-termed in conformity with the coordinate system used by Nichols et al. , i.e., Y positive toward dusk and X positive toward the Sun) is then given by the mean of the dawnward and duskward boundaries, as shown by the middle vertical red line is each panel. This procedure allows a simple combination of results with ovals of differing sizes.
 With the results of Provan et al. [2009b], Andrews et al.  and Nichols et al.  in mind, we order the dawn-dusk locations of the auroral ovals by the phases ΦSKR N and ΦSKR S of the northern and southern SKR oscillations, respectively. The polynomial fits to the southern SKR phase obtained by Kurth et al. [2007, 2008] are not valid for 2009, and we thus employ L. Lamy's (manuscript in preparation, 2010) numerical data set of phases determined empirically from the Cassini Radio and Plasma Wave Science (RPWS) instrument [Gurnett et al., 2004]. These empirical phase values were used previously by Nichols et al.  to successfully order the auroral powers and Andrews et al.  to order the magnetic field pertubations, and the data set and full details of its derivation are available online at the LESIA Observatoire de Paris website (http://www.lesia.obspm.fr/kronos/guest.php). During the equinoctial HST interval in 2009, the periods of the northern and southern SKR oscillations were ∼10.6 h and ∼10.8 h, respectively.
 Results are shown in Figure 2, in which the oval center values are shown versus SKR phase for both the north (Figure 2a) and south (Figure 2b). Results from individual images are shown by the crosses, and the means of ten 36° bins are shown by the joined pluses. It is apparent that, notwithstanding the scatter in the data arising from intrinsic morphological variability, overall the dawn-dusk locations of the ovals in both hemispheres exhibit oscillations with positive Y in lagging quadrature with the respective SKR oscillations. Thus, both the northern and southern ovals are typically displaced duskward of the mean locations (shown by the horizontal dotted lines) for phases less than ∼180° and dawnward for phases greater than ∼180°.
 The statistics of the oscillations are shown in Table 1. Specifically, we show the peak-to-peak amplitudes δ of the solid lines in Figure 2, the standard error between the individual locations in each 36°-wide bin and the bin means indicating the spread around the solid lines, the linear correlation coefficients r between Y and sin ΦSKR, and the false-alarm probabilities p for the correlation coefficients, i.e., the probability that Y and sin ΦSKR are uncorrelated, given by p = erfc(|r|) where N is the number of data points [Press et al., 2007]. The variation of oval location with phase is slightly more apparent for the southern data than the northern, but in both cases the peak-to-peak variation of the binned data is significantly larger than the spread around the bin means, and the correlation coefficients of 0.5–0.6 are both highly significant and similar to the values obtained by Nichols et al.  for the auroral power. In addition, the standard errors between the data and bin means of ∼0.6° is consistent with the standard errors of ∼0.5–0.6° obtained by Nichols et al.  for the dawn-dusk component of the southern oval motion, despite the fact that these authors used only images exhibiting quasi-circular ovals, a selection criterion we have not applied here. In order to verify that each oval does oscillate independently, the correlation coefficients of the locations of each oval with the opposite hemisphere phases were also computed. The values of ∼0.03 and ∼0.23 for the southern and northern ovals, respectively, are essentially insignificant compared to the results obtained with the same hemisphere phases. Further, in order to determine whether the oscillation of the center location is due to one boundary moving with respect to the other fixed boundary, representing, e.g., a periodic equatorward expansion on one flank rather than the motion of the oval as a whole, the correlation coefficient between the widths of the auroral regions and sin ΦSKR was determined for both hemispheres. The values of the correlation coefficients r are ∼0.25 and ∼0.00 (to 2 decimal places) for the southern and northern hemispheres, respectively. Thus, the width of the ovals does not exhibit variation with SKR phase, such that the oscillation of the center is due to motion of the oval rather than periodic equatorward expansion. This result is also borne out in the examples shown in Figure 1, in which it is apparent that, despite the significantly asymmetric morphologies in Figures 1b and 1d in which significant poleward expansions are evident on the dawn side, both dawnward and duskward boundaries exhibit motion and the oval widths (for each hemisphere) remain similar. We have also computed the mean locations of the centroids of the emission on the dawn and dusk halves of the images to determine if only the equatorward edges of the emission were oscillating, but effects due to noisy auroral morphology and limb brightening rendered the results inconclusive. However, the 4 images from the present data set shown in Animation 3 of Nichols et al.  suggest that the ovals as a whole are moving and do not exhibit, e.g., rotating equatorward expansions. Although not a major focus of this study, it is also worth noting that the mean half-width of the northern auroral region is ∼18.7°, i.e., ∼1.5° less than the southern value of ∼20.2°, a result confirming the difference reported by Nichols et al. , who obtained this value using the 4 images in this data set to which circles can be reasonably fitted. This confirms that the boundary locating algorithm used here returns locations consistent with the circle-fitting routine used previously to discover the oscillation of the southern oval.
Table 1. Oscillation of Saturn's Auroras
3.12 × 10−12
4.58 × 10−7
5. Summary and Discussion
 In this paper we have determined the dawn-dusk extent of Saturn's equinoctial UV auroral emission as imaged by HST in 2009, from which we then derived the dawn-dusk location of the centers of the southern and northern auroral ovals. We have ordered the computed oval center locations by the phase of the respective SKR oscillation phases in each hemisphere, and shown that statistically significant ∼1–2° oscillations in the dawn-dusk location of the auroral ovals are evident, with the duskward motion in lagging quadrature with respective SKR phases. We have also determined that the ovals do not oscillate at a single phase, and the oscillation of the oval centers is caused by motion of the ovals as a whole, rather than periodic asymmetric expansion.
 These results are the first to indicate that the location of the northern auroral oval oscillates, with an amplitude that is consistent with that of the southern oval observed by Nichols et al. . We are also able to definitively rule out the dipole tilt hypothesis suggested by Nichols et al.  as a possible cause for the oscillation of the location of the southern oval observed in images obtained in 2007 and 2008 since the ovals do not oscillate in anti-phase with identical periods. The cause of the oscillation is thus an external magnetospheric current system, and we show schematically in Figure 3 that the sense of the oscillations is consistent with the magnetic perturbations described by Andrews et al. . These authors have shown that the high-latitude perturbations take the form of planet-centered transverse dipoles rotating at the respective periods of the northern and southern SKR intensity oscillations, each dominating the respective high latitude regions, while the equatorial plane was dominated until equinox by the southern perturbation. For details of the morphology of current system, see Andrews et al. , but in Figure 3 we show simply how the addition of such dipolar perturbations to the planetary field affects the location of the auroral ovals. Figure 3a shows the southern perturbation, while Figure 3b shows the northern perturbation, each viewed from the dawn side (thus with the Sun to the right) at ΦSKR = 0, i.e., at respective SKR maxima. The dipole moment of the planet is shown, along with those of the southern and northern perturbations Δ and Δ, the former being oriented roughly toward midnight and the latter directed essentially toward noon, respectively, i.e., the senses determined by Andrews et al. . The ‘auroral axes’ and resulting from the superposition of the planetary and perturbation dipoles, which cone about the planetary axis at the angular velocities ΩSKR S and ΩSKR N respectively are shown, along with indications of the location of each auroral oval (note that in this schematic the magnitude of the perturbation dipole moment vector and the tilt of the auroral axes from the planetary axis are greatly exaggerated for clarity). It is apparent that, since at respective SKR maxima the directions of the perturbation dipole moments are opposite, both auroral ovals are oriented toward noon. A quarter of a rotation later, both ovals are directed toward dusk, and after another half a rotation they are tilted toward dawn. Although the noon-midnight motion is not observable in these data, the dawn-dusk oscillation shown in Figure 2 is consistent with this picture. We finally note that while the field perturbations at high latitudes are consistent with “dipole tilts” of ∼5–10°, the effect observed in the auroral oval is much smaller, indicating that the perturbations are associated with an external current system, and not a true planet-centered dipole. The smaller amplitude observed in the auroral motion is consistent with computations of the effect of an external current system as shown by Nichols et al. .
 This work is based on observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by AURA, Inc. for NASA. JDN and SWHC were supported by STFC grant ST/H002480/1. LL was supported by the CNES agency. The authors acknowledge the support of ISSI, as this paper was discussed by ISSI International Team 178.