Icarus 190 (2007) 127–144 www.elsevier.com/locate/icarus Self-gravity wakes and radial structure of Saturn’s B ring J.E. Colwell a,∗ , L.W. Esposito b , M. Sremčević b , G.R. Stewart b , W.E. McClintock b a Department of Physics, University of Central Florida, Orlando, FL 32816-2385, USA b LASP, University of Colorado, 392 UCB, Boulder, CO 80309-0392, USA Received 19 July 2006; revised 2 March 2007 Available online 13 April 2007 Abstract We analyze stellar occultations by Saturn’s rings observed with the Cassini Ultraviolet Imaging Spectrograph and find large variations in the apparent normal optical depth of the B ring with viewing angle. The line-of-sight optical depth is roughly independent of the viewing angle out of the ring plane so that optical depth is independent of the path length of the line-of-sight. This suggests the ring is composed of virtually opaque clumps separated by nearly transparent gaps, with the relative abundance of clumps and gaps controlling the observed optical depth. The observations can be explained with a model of self-gravity wakes like those observed in the A ring. These trailing spiral density enhancements are due to the competing processes of self-gravitational accretion of ring particles and Kepler shear. The B ring wakes are flatter and more closely packed than their neighbors in the A ring, with height-to-width ratios <0.1 for most of the ring. The self-gravity wakes are seen in all regions of the B ring that are not opaque. The observed variation in total B ring optical depth is explained by the amount of relatively empty space between the self-gravity wakes. Wakes are more tightly packed in regions where the apparent normal optical depth is high, and the wakes are more widely spaced in lower optical depth regions. The normal optical depth of the gaps between the wakes is typically less than 0.5 and shows no correlation with position or overall optical depth in the ring. The wake height-to-width ratio varies with the overall optical depth, with flatter, more tightly packed wakes as the overall optical depth increases. The highly flattened profile of the wakes suggests that the self-gravity wakes in Saturn’s B ring correspond to a monolayer of the largest particles in the ring. The wakes are canted to the orbital direction in the trailing sense, with a trend of decreasing cant angle with increasing orbital radius in the B ring. We present self-gravity wake properties across the B ring that can be used in radiative transfer modeling of the ring. A high radial resolution (∼10 m) scan of one part of the B ring during a grazing occultation shows a dominant wavelength of 160 m due to structures that have zero cant angle. These structures are seen at the same radial wavelength on both ingress and egress, but the individual peaks and troughs in optical depth do not match between ingress and egress. The structures are therefore not continuous ringlets and may be a manifestation of viscous overstability. © 2007 Elsevier Inc. All rights reserved. Keywords: Saturn, rings; Planetary rings 1. Introduction Saturn’s B ring is the broadest and most massive planetary ring in the Solar System, extending from 92,000 to 117,500 km from Saturn’s center and with an estimated mass of 3 × 1019 kg, or ∼80% of the entire ring system (Cuzzi et al., 1984). Measurements of the optical depth from various occultation techniques can still place only a lower limit on the optical depth in the core of the B ring, which lies between 99,000 and * Corresponding author E-mail address: jcolwell@physics.ucf.edu (J.E. Colwell). 0019-1035/$ – see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.icarus.2007.03.018 108,000 km. New measurements by the Cassini spacecraft show the structure of the B ring in unprecedented detail. A strong azimuthal brightness asymmetry in the A ring, exterior to the B ring, is due to the presence of aligned self-gravity wakes in that ring (Camichel, 1958; Colombo et al., 1976; Lumme and Irvine, 1976; Reitsema et al., 1976; Lumme et al., 1977; Gehrels and Esposito, 1981; Thompson et al., 1981; Dones and Porco, 1989; Dones et al., 1993; Dunn et al., 2004; Nicholson et al., 2005). The geometric properties of the selfgravity wakes in the A ring have been determined from occultation measurements of the dependence of the ring transparency on viewing angle (Colwell et al., 2006; Hedman et al., 2007). Here we examine the B ring where similar variations 128 J.E. Colwell et al. / Icarus 190 (2007) 127–144 in transparency are observed in Cassini occultation measurements. Asymmetries in the B ring have been seen in imaging from Hubble Space Telescope (French et al., 2007) and radar observations (Nicholson et al., 2005). Unlike the A ring, which is characterized by broad regions of slowly varying background optical depth, punctuated by density and bending waves, the B ring is filled with poorly understood structure. The B ring is more opaque than the A ring, making the signals of occulted stars that much weaker and therefore noisier. The outer few 100 km of the B ring is not circular due to forcing from the moon Mimas (e.g., Cuzzi et al., 1984) so we restrict our analysis to ring plane radii R < 117,000 m. We find a strong dependence of apparent normal optical depth on the occultation viewing angle that can be explained by self-gravity wakes like those in the A ring, but with a much larger width/height aspect ratio. We apply the model to the entire B ring where sufficient data exist to model the azimuthal asymmetry. We use an analytic expression with the assumption of opaque self-gravity wakes as well as the simple ray-trace method of Colwell et al. (2006) with no assumptions about the wake optical depth. We analyzed the data at both 10 and 50 km resolution, the latter incorporating solar occultations which have lower inherent resolution than the stellar occultations. We found no significant differences in calculated wake properties between the different techniques and resolutions. In the next section we describe the observations and the techniques for determining ring optical depths from the measurements. In Section 3 we present the resulting optical depth profiles from the set of occultations used in this study. Our model of self-gravity wakes is explained in Section 4 together with our results on wake properties in the B ring. The α Leo occultation minimum radius was in the outer B ring and provided a detailed look at the azimuthal and radial structure of one location in the ring, revealing axisymmetric, but discontinuous wave-like structure with λ ∼ 160 m likely due to viscous overstability. This is discussed in Section 4.3. In Section 5 we summarize our results and discuss the implications for viscous evolution of the B ring. 2. Observations The Cassini Ultraviolet Imaging Spectrograph (UVIS) has three channels for observing occultations. The High Speed Photometer (HSP) is used for stellar occultations and measures the intensity of starlight between 110 and 190 nm with a 1–8 ms sampling interval (Esposito et al., 2004). The Far Ultraviolet Spectrograph (FUV) channel has the same spectral bandpass as the HSP with 1024 spectral resolution elements and 64 spatial resolution elements along the slit. At each integration period 64 spectra are obtained. A single FUV integration requires much more data than the HSP which is typically operated in a 9-bit compression mode, so the integration periods are much longer with the FUV channel than the HSP channel. The imaging capabilities of the FUV allow simultaneous measurements of the ring brightness in the same part of the spectrum that the HSP measures. For some occultations this information is used in Table 1 Occultation geometry Occultation star (rev) Date (year-day) |B| φ R (km) δ Sco (Voyager) ξ 2 Cet (A) 126 Tau (8) δ Aqr (8) α Leo (9, ingress) α Leo (9, egress) Solar (9) 126 Tau (10) σ Sgr (11) Solar (11) α Sco B (13, ingress) α Sco B (13, egress) λ Cet (28) α Sco B (29) λ Sco (29) α Vir (30) 1981-237 2004-281 2005-139 2005-141 2005-159 2005-159 2005-159 2005-175 2005-195/6 2005-196 2005-232 2005-232 2006-256 2006-269 2006-269 2006-285 28.7 14.9 21.1 12.2 9.5 9.5 21.4 21.1 29.1 21.1 32.2 32.2 15.3 32.2 41.7 17.2 104.3–107.3 81.2–79.2 73.1–85.0 59.6–54.3 112–126.8 112–97.1 78.2–78.1 155.8–150.2 133.3–119.9 74.9–75.2 25–23.5 25–26.4 78.2–78.1 57.3–75.1 6.9–32.8 122.7–132.9 72,000–142,500 109,000–135,550 70,400–141,300 60,700–169,900 114,100–131,500 114,100–176,200 66,100–168,600 103,200–144,200 86,000–146,900 54,100–116,000 101,200–155,800 101,200–146,600 74,300–144,000 79,900–149,400 88,500–143,800 64,000–151,600 Notes. Range in φ is for the B ring (92,000–118,000 km) or the fraction of the B ring spanned by the occultation. Range in R is for the entire observation rounded to 100 km. ξ 2 Ceti data are for the second of two observations of that star. δ Sco values of R only refer to the data subset retrieved from the Planetary Data System. constructing a model of the background component of the HSP signal. The Extreme Ultraviolet Spectrograph (EUV) is coaligned with the FUV and HSP, but it also includes a pick-off mirror that allows observations of the Sun without aiming the Cassini remote sensing pallet at the Sun. Because of data volume restrictions and the lower signal from stars in the EUV compared to the FUV, the EUV channel is not used during stellar occultations. The instrument and observing techniques for stellar occultations are described in Esposito et al. (1998, 2004). The HSP observed seven occultations of stars by Saturn’s rings during the first phase of Cassini’s four-year tour of the Saturn system (Esposito et al., 1998), and two solar occultations of the rings were observed by the EUV. The solar occultations are at a much lower resolution, limited by the size of the Sun on the rings as seen from Cassini (∼100 km), than the stellar occultations which are limited by diffraction to ∼ 20 m. Stellar occultations in the latter half of the Cassini nominal mission are at higher elevation angles and are therefore less diagnostic of the properties of aligned clumps in the rings, such as selfgravity wakes. We also use the Voyager PPS δ Sco occultation (Lane et al., 1982). The geometric properties of the occultations used in this study are presented in Table 1. These occultations cut the rings at a variety of angles with respect to the ring plane (B) and with respect to the local radial direction (φ, where φ = ±π/2 when the line of sight is tangent to the rings at the occultation point). The transparency of the ring along the line of sight to the star is I −b T= (1) , I0 and the normal optical depth, τn , is τn = μ ln T −1 , (2) Self-gravity wakes in Saturn’s B ring where I is the measured stellar intensity, I0 is the unocculted intensity of the star, b is the background signal, and μ = sin(B) is the projection factor to convert from line of sight optical depth to normal (vertical) optical depth. This standard expression for normal optical depth is based on a model of the rings as a planar, homogeneous medium with non-zero and finite transparency, such as a cloud. For such a medium the factor μ in Eq. (2) is necessary to correct for the length of the observed line of sight which is always greater than the vertical thickness of the ring. If the positions of the ring particles are uncorrelated and the rings are azimuthally symmetric, and if the particles are small compared to the area of the ring sampled in a single occultation integration period, then Eq. (2) gives the same value of the ring optical depth for all values of B and φ. However, the clustering of ring particles into elongated and aligned clumps, or self-gravity wakes, causes a variation of τn with viewing angle. Calculations of optical depth through numerical N -body simulations (cf. Fig. 16 in Salo et al., 2004) show the same variation of apparent normal optical depth, τn , with viewing angle that we find with our simple model (Colwell et al., 2006) and that is seen in the data described in this paper. This variation is diagnostic of the shape, spacing, and alignment of the self-gravity wakes. We present values of τn using Eq. (2) for consistency with previous discussions of ring optical depth and to illustrate the magnitude of microstructure in the rings on the inferred optical depth. The occultations presented here show large variations of τn as calculated from Eq. (2) in the B ring. The background signal, b, is dominated by sunlight scattered from the rings except in cases where the HSP field of view is within the shadow of the planet on the rings or on the unlit face of the rings. On the sunlit face of the rings (the south face during the observations described here), the signal from the rings recorded by the HSP is <2000 counts/s, with the precise value depending on the viewing geometry, the filling factor of ring material in the HSP field of view, and the proximity of Saturn to the field of view. On the unlit side of the rings the background is usually dominated by Lyman-α emission from interplanetary hydrogen which gives b ∼ 100 counts/s. The importance of the background signal in determining optical depth depends on the viewing geometry (amount of sunlit ring material in the field of view and the angular proximity of the line of sight to Saturn), the star brightness, and the optical depth of the rings. In general I0 and b must be determined by independent measurements of the signal I = I0 + b in ring gaps and I = b when the star is behind fully opaque (to the occulted star) regions of the rings. In some cases there are no opaque regions of the ring covered by an occultation, and independent determinations of the background must be made. In some cases this can be modeled with simultaneous measurements of the ring with the UVIS FUV channel (Section 2.1, for example). All occultations analyzed here included views of the star through gaps in the rings, and these observations have a background contribution from the rings on the lit face of the rings, or primarily from interplanetary hydrogen on the unlit face. The behavior of the UVIS HSP to very bright stars (I0 > 104 counts per second) complicates the determination of I0 . During cruise calibration measurements of the star α Virginis (Spica) 129 Fig. 1. Total signal measured by the HSP of the star σ Sagittarii binned to 100-ms samples illustrating the ramp-up behavior of the HSP with bright stars. After being partially obscured by the outer A ring from t = 0.25–0.32 h after the start of the observation, the star re-emerges behind the Encke Gap at a slightly lower count rate than when it was first occulted by the A ring. As the Encke Gap drifts in front of the star the measured signal restarts its ramp-up behavior. and other stars we noticed a complicated ramp-up behavior of the instrument response to the star signal. When first exposed to a bright star the instrument does not reach a constant count rate. Instead the count rate reaches a value of ∼90% of the asymptotic value in the first sampling interval (usually 2 ms) and then follows a non-linear ramp-up for a period of up to several minutes followed by a slow linear increase in the counting rate. The long time scales for this response are unusual for a photomultiplier tube. The precise shape and magnitude of the ramp-up are different for each observation and are not a simple function of the star brightness or of the time that the star has been shining on the detector. When the star has been occulted by the rings and then re-emerges into a gap, the signal recovers most of the pre-occultation rate and then resumes the ramp-up behavior (Fig. 1). This ramp-up response for the HSP is not seen for fainter stars. In practice, and below, we model b and I0 based on a combination of measurements of the occulted and unocculted star and, when available, independent measurements of the ring brightness with the FUV channel. Because the signal at any given time is a combination of the star brightness as seen by the detector, I0 , the optical depth of the ring at the location sampled at that time, and the background signal, there are several components to the uncertainty in the measurement of the optical depth. The background is not constant in general, and must be modeled from simultaneous measurements with the FUV and/or background component of the overall signal is therefore a model result with an estimated error. The star brightness is never measured without a background contribution, and the instrument behavior described above introduces an uncertainty in our determination of I0 , so our determination of I0 is also a model result with an estimated error. The estimated errors in the determination of b and I0 are “probable errors” (Bevington, 1969) and are distinct from the uncertainty due to counting statistics. These errors are based on the differences between determinations of I0 (and b) at different points in the occultation. Counting statistics alone 130 J.E. Colwell et al. / Icarus 190 (2007) 127–144 √ give the following Zσ = Z I bounds on the measured optical depth, τn : τ± = μ ln I0 − μ ln I0 exp(−τn /μ) ∓ Z I0 exp(−τn /μ) + b , (3) where Z is the number of standard deviations away from the measurement and I0 and b are measured in counts. The probable errors in b and in I0 are usually smaller than the statistical uncertainty, and they are discussed in the detailed summary of each occultation below. Ultimately, when we apply our simple geometric model of self-gravity wakes to the data, we introduce an additional error which is the deviation of the model results due to the model imperfections in describing the actual arrangement of ring particles. In these summaries we describe our determination of b and I0 and the corresponding uncertainties in τn for each occultation included in this analysis. We discuss the stellar occultations first, in chronological order, followed by the solar occultations. 2.1. ξ 2 Ceti occultation Cassini observed the occultation of the star ξ 2 Ceti by the rings on 2004-280 to 2004-281 (October 6–7) with the UVIS HSP. The Far Ultraviolet Spectrograph channel (FUV) also observed the occultation which was the most distant stellar occultation to be observed by Cassini (Table 1). The large distance of Cassini from Saturn and the foreshortening of the rings resulted in only 5–7 spatial elements of the FUV detector filled with ring material, and the star signal in two of these detector rows. This poor spatial coverage limits the utility of the FUV observation in providing an independent measurement of the background. The speed of the star across the rings projected in the radial direction varied between 0.88 km/s at the inner edge of the C ring and 1.03 km/s in the outer A ring where the observation ended. The total duration of the occultation from the cloud tops to the F ring was 23.5 h. The occultation was observed in two parts separated by a downlink of spacecraft data. The first observation covered the inner portion of the C ring; the outer C ring data were lost due to downlink problems. The second observation began while the star was behind the dense core of the B ring at R = 109,000 km and continued to near the outer edge of the A ring at R = 135,500 km. The large distance of the spacecraft from Saturn for this occultation also meant that a large portion of the ring system was within the HSP 6 mrad field of view. The star is north of Saturn’s ring plane so at the time of the occultation Cassini was observing the lit face of the rings. Because the star is relatively dim and the sampling intervals in the ring plane are closely spaced we can estimate the star brightness from the change in the signal at edges of optically thick rings and ringlets. The Huygens ringlet at R = 117,800 km in the Cassini Division has sharp edges and is optically thick, and the outer edge of the B ring has a peak in optical depth that brings the measured signal down to the background level. The measured difference in I between the Huygens ringlet and the Huygens gap interior to the ringlet gives I0 = 1430 ± 10 counts/s. With I0 fixed, the background in the known clear gaps in the rings is given by b = I − I0 . The background level drops by one third between the outermost gap in the Cassini Division at R = 120,300 km and the Encke Gap at R = 133,600 km. We set b in the B ring at the minimum count level observed when the star was behind the opaque core of the B ring at R = 105,000 to 110,000 km. We then modeled b with a linear decrease from the outer Cassini Division across the A ring, with a value in the Encke Gap set by I − I0 there. A constant value of b in the B ring part of the observation is expected given that the HSP field of view is filled by the illuminated optically thick B ring. As a check on the linear model we calculated the fraction of the HSP field of view that is filled by ring material, weighted by optical depth using the Voyager PPS occultation profile (Lane et al., 1982). We then tied the resulting curve to the gaps where the background value can be directly measured. The resulting curve differs from the linear model by less than 20 counts/s out of 600–800 counts/s depending on the location. In the multiple gaps of the Cassini Division the background derived from the HSP field of view filling factor was a slightly worse fit to the observations than the linear model because the filling factor model had an increasing background where the data were consistent with a constant or decreasing background. The constant value of b = 800 ± 10 counts/s in the B ring core is consistent with the data out to R = 115,000 km where I − b becomes negative. The reflected sunlight measured with the FUV across the rings is consistent with this picture, but is contaminated by starlight and therefore not used in refining the background model. For the B ring regions studied here we therefore adopt a linear decrease in b from 800 counts/s at R = 114,000 km to 772 counts/s at R = 118,000 km with a probable error in b estimated at 10 counts/s across the B ring. The data with this background model subtracted are shown in Fig. 2 binned by 50 points (0.4 s) and smoothed to approximately 20 km resolution. Below we compare measurements at 50 km resolution; for this occultation that resolution corresponds to about 50 s of data. We estimate the probable error in our model value of b at that resolution to be ±200 counts or comparable to the statistical uncertainty in the sig- Fig. 2. HSP data from three occultations at 50 km resolution, with the background removed (see text). Self-gravity wakes in Saturn’s B ring √ nal of I . In the optically thick B ring the probable error in I0 at this resolution is much smaller and does not contribute significantly to our estimated error of the optical depth or ring transparency. We set the minimum measurable transmitted signal Imin = I − b = 200 counts at 50 km resolution based on the probable error in b; this corresponds to a maximum normal optical depth, τmax = 1.51. At this resolution the statistical uncertainty in the optical depth is τ < 0.03 at local optical depth minima and τ < ±0.2 everywhere that τ < τmax . As τ approaches τmax the upper error bar becomes infinite because the probable error in b prevents us from distinguishing that signal from zero transparency. 2.2. 126 Tauri occultations The star 126 Tauri (126 Tau) was observed by Cassini on its orbital revolutions (revs) 8 and 10 of Saturn. The 126 Tau rev 8 occultation (126Tau8 hereafter) was an egress occultation that spanned the entire ring system with a data dropout in the middle of the C ring. The occultation was observed from ∼20 RS resulting in a 9-h track across the rings with a radial sampling interval in the ring plane of 3 m in the inner C ring to 5.5 m in the F ring. The occultation was also observed with the FUV channel on the lit face of the rings, and the long duration of the occultation provided a high signal-to-noise map of the ring reflectance in the FUV (Fig. 3). This map was used in conjunction with HSP measurements of the signal in gaps and opaque regions of the rings to derive a background model. The data binned by 500 points to 1-s resolution in time are shown in Fig. 4 with our background model. With this background model we determine I0 = 4244 counts/s. We bin the data to 50 km resolution, or about 200 s in the outer B ring. In the B ring, where the background is nearly constant, we estimate the probable error of b to be 20 counts/s. Setting Imin = 20 counts/s, τmax = 1.93, with uncertainties in the optical depth of τ < 0.01. The rev 10 occultation of 126 Tau (126Tau10) was an ingress occultation. The star passed behind Saturn in the inner B ring at R = 103,000 km. There is a data dropout in the A ring ramp in the outer Cassini Division. Cassini revs 8 and 10 have the same geometry, and the 126Tau10 occultation is the ingress portion of the 126Tau8 egress occultation. Scattered light is less in 126Tau10 because it is on the night side of the planet while 126Tau8 egress was on the day side. The orientation of the FUV spectrograph slit on the rings for the 126Tau10 occultation is more nearly radial across the rings than for 126Tau8 in which the detector rows immediately adjacent to those with the star signal cut nearly radially across the rings. Because the distance from Saturn was large for both occultations (∼23 RS for 126Tau10), the resolution from row to row of the detector is ∼1400 km in 126Tau10. The scattered light measurements of the rings with the FUV in 126Tau10 are consistent with a nearly constant background across the B ring and a linear decrease across the A ring. Because of the low resolution of the FUV measurements and significant row-to-row variations in the scattered light mea- 131 surements due to the viewing geometry, we model the background from the HSP measurements of gaps and opaque regions in the rings. We find I0 = 4550 counts/s with a probable error of 30 counts/s. The difference in I0 between revs 8 and 10 is real. We could not find a single value for I0 that fit both occultations. This may be due to instrumental effects: we have observed different count rates of the same star in different calibration observations suggesting long-term unmodeled variability in instrument sensitivity, perhaps related to the ramp-up behavior described above. The background model consists of 7 linear segments from the middle B ring to the F ring. In the B ring region studied here, the background increases from b = 1730 counts/s at R = 103,213 km to b = 1920 counts/s at R = 108,000 km, followed by a decrease to b = 1770 counts/s at R = 116,500 km and then another increase to b = 1835 counts/s at R = 118,000 km. This is followed by a monotonic decrease in b across the Cassini Division and A ring. We estimate the probable error of b to be 40 counts/s. Setting Imin = 40 counts/s, τmax = 1.7 with uncertainties in the optical depth of τ < 0.06 in the outer B ring. 2.3. δ Aquarii occultation The star δ Aquarii (δ Aqr) was observed on 2005-141 (rev 8) following the first 126 Tau occultation and an occultation of the star α Virginis (Spica) that did not reach the B ring. This is the faintest UV star occultation observed by UVIS in the initial phase of the Cassini mission. However, the dependence of apparent optical depth on viewing angle discussed below is such that light from δ Aqr at B = 12.2 degrees and φ ≈ 55 degrees was able to penetrate all of the A ring as well as much of the inner and outer portions of the B ring. Because of the relatively low count rate from the star the accuracy of the background determination has a stronger effect on the derived optical depths than in occultations of the brighter stars. The FUV channel observed the δ Aqr occultation providing an independent measure of the background. The occultation was observed from the unlit face of the rings which kept the background signal below even that of the faint star; the background signal is dominated by Lyman-α emission from interplanetary Hydrogen shining through the rings. Observations of dark sky as well as the shadowed rings with the HSP have a mean counting rate due to Lyman-α of approximately 150 counts/s (Chambers et al., 2007). The image of the occultation in the FUV also demonstrates that starlight is passing through the inner and outer regions of the B ring as well as the A ring (Fig. 5). To model the background we averaged and smoothed the signal in the rows of the FUV detector on either side of the two rows with the star signal to get the dependence of the background on radial position in the ring plane. We then scaled this curve to the HSP signal, I , in the opaque part of the B ring at 100,000–105,000 km where we assume the star signal is completely attenuated. This is a safe assumption given that occultations of much brighter stars observed by UVIS (Section 2.5) as well as Cassini Radio occultations (Marouf et al., 2005) were unable to penetrate this dense core of the B ring. We then determine the star signal I0 = 515 counts/s from the sig- 132 J.E. Colwell et al. / Icarus 190 (2007) 127–144 Fig. 3. FUV image of the rings made during the 126Tau8 occultation. The star is centered in the spectrograph slit which is oriented vertically and time increases to the right. The color image was made by assigning red to Lyman-α emission, and green and blue to longer wavelengths in the FUV where water ice does not absorb. The rings are blue in this scheme with the sky red due to interplanetary hydrogen. The star is visible in the central few rows. The vertical bands are due to bleeding of the bright star signal across the spatial dimension of the FUV detector. Scans of the ring brightness above and below the rows with star signal were used to model the shape of the background in the HSP occultation data. Fig. 4. HSP data from the 126Tau8 occultation binned to one second intervals (1.5 to 2.8 km radial resolution) together with the background model (dashed line) constructed from the scattered light measurements with the FUV and fits to opaque and clear regions in the rings (see text). The increase in the background close to the planet is due to scattered light from Saturn. This is the most complicated background model of the occultations discussed here due to the relatively high b/I0 ratio and the changes in b during the occultation. nal I = I0 + b in known gaps in the Cassini Division and A ring and between the A and F rings. Statistical fluctuations in I as well as probable errors in both I0 and b mean that the quantity I − b can be negative. We select a positive minimum detectable signal of Imin = I − b that at 50 km resolution results in a maximum optical depth of 0.95. While smaller values of Imin give a nominally larger τmax , Fig. 5. FUV image of the δ Aqr occultation. The occultation slit is oriented vertically and time increases from left to right. There are 164 records with a duration of 60 s each in the observation. The star was centered in the slit and the signal from the star is seen on the central two of 64 spatial elements of the imaging spectrograph. The planet is the pink feature at left. The color scheme is the same as in Fig. 2. The observation was of the unlit face of the rings, so the opaque B ring is dark. The star signal is visible everywhere except the dense core of the B ring. The faint blue line at the right is scattered sunlight from the unresolved F ring. The C ring is also visible in scattered sunlight, shown in blue in this image, just interior to the dark B ring. Fig. 6. FUV image of the α Leo occultation. The occultation slit is oriented vertically and time increases from left to right. There are 160 records with an integration period of 60 s each. The color scheme is the same as in Fig. 2. The star emerges from behind the planet at left before being occulted by the A ring which is in the shadow of the planet. The occultation path crosses the shadow boundary near the outer edge of the Cassini Division before making a grazing cut across the outer B ring. The star signal is clearly seen through the B ring. Also obvious is the asymmetry between the transmitted starlight in the A ring on egress (right edge of image) and ingress (portion in shadow; Colwell et al., 2006). The vertical bands are due to bleeding of the bright star signal across the spatial dimension of the FUV detector. the uncertainties at those larger values of tau are so big as to render the measurement meaningless. At our selected value of Imin = 5.5 counts/s, the uncertainty in optical depth at τ = τmax is τ ∼ 0.4. However, most of the areas of the B ring chosen for this study have a one-σ uncertainty in τn of τ = ±0.02 Self-gravity wakes in Saturn’s B ring to 0.03, although some locations were at the maximum value of τn = 0.95. Those measurements were not included in the analysis below. 2.4. α Leonis occultation The α Leonis (α Leo, Regulus) occultation on Cassini orbital rev 9 was a grazing occultation that penetrated to a minimum ring plane radius of 114,150 km (Fig. 6). The egress (outbound) portion of the occultation was observed out to a ring plane radius of 131,550 km. The mean signal, I , in the gaps in the A ring and the Cassini Division varies between 4.5 × 104 and 4.8 × 104 counts/s. The gap signals increase with time indicating that the variation is mainly due to the instrument response to a bright star and not background variations. These count rates are significantly higher than the minimum count rates observed in the more optically thick regions of the A ring of 300 counts/s on ingress (when the occultation was in the shadow of the planet, Fig. 6). During both the ingress and egress cuts of the B ring the occultation was not in shadow and the background is higher. However, the B ring was not opaque to this star, and on egress the A ring was particularly transparent due to the viewing angle with respect to the self-gravity wakes (Fig. 6; Colwell et al., 2006). This prevents a direct measurement of the background signal for the latter part of the occultation. The signal, I , is lower in the Cassini Division gaps in the egress portion of the occultation than in the ingress portion, even though the background signal must be higher due to the illumination geometry. We attribute this to the instrument response time variation described above. In the outer Cassini Division in the egress portion of the occultation the signal, I , has recovered to nearly the same level that it had at that location in the ingress portion. We adopt I0 = 4.6 × 104 counts/s with a probable error of 1 × 103 counts/s for the B ring and egress portions of the occultation, though I0 = 4.7 × 104 counts/s is more appropriate for the ingress A ring occultation. This makes a difference of 3.6 × 10−3 in τn . We estimate the background b = 300 counts/s on ingress to a ring plane radius of R = 122,000 km followed by a linear increase with decreasing radius to b = 1500 counts/s at R = 117,600 km. We estimate the probable error in the background to be 100 counts/s in the shadow (b = 300 counts/s) and 500 counts/s when b = 1500 counts/s, and we set Imin = 500 counts/s. At 50 km resolution the maximum statistical uncertainty in τn in the B ring is τ = 0.02 and τ < 0.01 across most of the A ring. 2.5. σ Sagittarii occultation This occultation spanned the ring system from beyond the F ring across the A and B rings to the outer C ring at 85,955 km. The star σ Sagittarii (σ Sgr) is south of the ring plane so the unilluminated face of the rings was observed. The path of the occultation passed behind the most optically thick region of the B ring between 105,000 and 110,000 km. The signal there provides a measurement of the background of b = 1000 ± 100 counts/s. The FUV channel recorded data for this 133 occultation at a time resolution of 60 s with the star near the central row of the 64 spatial pixel FUV detector. The FUV signal is dominated by light from the star even in rows far from the star due to bleeding of the star signal. Any scattered light from the rings is not separable from the starlight in the FUV data. The background may be less on the more transparent portions of the ring, but here we analyze the optical depths in the optically thick B ring. The total signal is much greater than the background for this occultation, so even changes as large as a factor of two in the background have a small effect on the calculated optical depth in those more tenuous rings. The mean signal, I , just before the star is occulted by the outer edge of the A ring, is 1.200 × 105 counts/s, and in the Huygens gap just prior to immersion behind the B ring the signal is 1.199 × 105 counts/s. With b = 103 ± 102 counts/s we adopt I0 = 1.19 × 105 counts/s for this occultation. The maximum value of τn is τmax = 3.43, set by our assumption of the uncertainty in the background and requiring I − b to be greater than that uncertainty. With this large estimate of the uncertainty in b, the 1-σ statistical uncertainty in τn is ∼0.08 at the maximum values of τn and is less than 0.01 throughout the rest of the B ring. At the specific positions in the B ring we analyze in detail below, the 1-σ error bars in τn are τ = 0.004 or less. 2.6. α Scorpii occultations The α Scorpii revolution 13 (α Sco 13) occultation track was a chord that traversed the outer portion of the ring system to a minimum ring plane radius of 101,172 km, just interior to the dense B ring core. The Cassini VIMS instrument was the primary instrument for α Sco observations, and UVIS took data as a “rider.” UVIS detects light from α Sco B, the distant (3 arc-s) binary of α Scorpii (Antares), while Antares itself is detected by VIMS.1 We do not have FUV data of this occultation. The star is relatively dim in the UV so the HSP ramp-up effect is small. Like the σ Sgr occultation, this one was observed from above the unlit face of the rings. The background is significantly smaller for α Sco, however, due to less scattered light from Saturn. In the B ring core, the minimum signals are b = 120 counts/s and b = 100 counts/s on ingress and egress, respectively. We adopt b = 110 counts/s with a probable error of 20 counts/s for the B ring portion of the occultation. The greater transparency of the Cassini Division and A ring allow Lyman-α emission from interplanetary Hydrogen to add to the background signal, possibly increasing it by as much as another 100–200 counts/s. With this determination of the background we estimate the star brightness from measurements in the Cassini Division and A ring gaps of I0 = 3375 counts/s with a probable error of 200 counts/s. We set Imin = 100 counts/s resulting in τmax = 1.87. The statistical error in the optical depth is τ < 0.11 in the B ring and τ < 0.03 across most of the A ring at 50 km resolution. We 1 The identification of the star providing the signal in UVIS data as the secondary star in the binary α Scorpii system was not made in Colwell et al. (2006), but the difference in ring plane radii of the tracks of the two stars (∼0.2 km) is much less than the resolution of the data presented in that paper. 134 J.E. Colwell et al. / Icarus 190 (2007) 127–144 estimate the systematic error in τn due to the errors in the determination of b and I0 to be <0.02. The α Scorpii revolution 29 (α Sco 29) occultation was an ingress occultation that spanned the ring system from the F ring to the inner C ring. The occultation was observed from the unlit face of the rings, so the background level is low. FUV data were obtained for this occultation. Serendipitously, another hot star (HD 148605) was in the FUV field of view on spatial pixel number 56 approximately 23 mrad from α Sco. α Sco itself, as expected, was not in the FUV field of view due to imperfect alignment of the HSP, FUV, and VIMS fields of view. Given the unlit-face viewing geometry of the occultation, the FUV signal is dominated by interplanetary Lyman-α transmitted through the rings. In the optically thick core of the B ring the signal reaches a mean minimum level of only 50–60 counts/s. Although the background may be higher behind the Cassini Division and C rings, the signal in the outer C ring gaps is nearly identical to that between the A and F rings. Rather than adopt an uncertain radius-dependent model for both I0 and b whose differences cancel to match the observed overall signal, we adopt a constant b = 60 counts/s and I0 = 3600 counts/s, and we estimate the probable errors as 20 and 50 counts/s respectively based on the variation in the total signal between C ring gaps and Cassini Division gaps. At 50 km resolution we have τmax = 2.77 with uncertainties in τ of ±0.3 at those large optical depths, and τ < 0.1 for τ < 2 and τ < 0.01 for τ < 1. 2.7. λ Ceti occultation The λ Ceti (λ Cet) occultation track traversed the entire main ring system from the F ring through the C ring. FUV data were obtained for this occultation and they show a decrease in the FUV signal to a minimum in the central part of the B ring that is likely due to increased Lyman-α transmission through the A and C rings than through the more opaque B ring. The star is relatively dim in the UV so instrumental effects are small. The star is north of the ring plane, so at the epoch of this observation the lit face of the rings was observed. The observation is on the night side of Saturn, so there is not much scattered light from Saturn itself. The HSP field of view is within the shadow of the planet on the rings between the middle C ring and the Encke Gap. In the B ring core, the minimum signals are I = 46.5 counts/s between 105,000 and 107,000 km. This average value is skewed upward by narrow regions of the core that are not totally opaque. We adopt b = 45 counts/s between 100,000 and 108,000 km and have a linear increase in b to 105 counts/s at R = 85,000 km and then a steeper linear increase to 190 counts/s at 74,300 km (the inner extent of the occultation). Exterior to the B ring core we adopt a linear increase in the background to 109 counts/s at 140,000 km. With this background model the ringlet at the Titan 1:0 nodal resonance is totally opaque at 10 km resolution, but the gaps surrounding it show a higher unocculted star signal than in the outer portion of the ring system. Part of this increase may be the HSP instrumental response, and part of it may be due to errors in the simplified background model presented here. We adopt I0 = 2450 counts/s at R > 85,000 km, and a linear increase in I0 to 2525 counts/s at the inner extent of the occultation. Based on the spread of the signal in the opaque core of the B ring and the Titan 1:0 ringlet we estimate the probable error in b to be 20 counts/s. From a similar inspection of the signal in the gaps across the ring system we estimate the probable error in I0 also to be 20 counts/s. We adopt Imin = 20 counts/s which gives τmax = 1.27 at 50 km resolution. With our estimates for the probable error in b and I0 we get a systematic error uncertainty in normal optical depth of ±0.05 everywhere τn < 1, with τ < 0.15 at higher optical depths. The statistical errors are comparable (within a factor of two, at this 50 km resolution) to the systematic errors for this occultation. 2.8. λ Scorpii occultation The λ Scorpii revolution 29 (λ Sco 29) occultation was an ingress occultation that spanned the ring system from the F ring to the outer C ring. The occultation was observed from the unlit face of the rings, so the background level is low. FUV data were obtained for this occultation, but the spectra all along the slit are dominated by starlight because the star is particularly bright and because the rings on the unlit face are quite dark. Consequently the FUV data are not helpful in estimating the background contribution to the HSP signal. The background is likely higher in the optically thin regions of the ring where interplanetary Lyman-α can shine through. The optically thick core of the B ring has a minimum mean count rate of 100 counts/s. Dark sky measurements with the HSP show 100–150 counts/s due to Lyman-α emission. We adopt b = 100 counts/s for this occultation in the B ring, and b = 150 counts/s elsewhere, with a probable error in b of 50 counts/s. The high count rate for this star results in a varying baseline level for the unocculted star brightness, I0 . The magnitude of this variation is larger than the anticipated variation in the background signal described above. We therefore adopt a time-dependent value for I0 that is fit to the measured signal minus the background in the gaps in the rings. The model value of I0 varies between 2.74 × 105 counts/s in the B ring to 3.02 × 105 counts/s in the Cassini Division, with I0 = 2.87 × 105 counts/s in the A ring. We estimate the probable error in I0 to be 1.5 × 103 counts/s. Setting a minimum detectable signal at 50 counts/s we have τmax = 5.79, and both statistical and systematic uncertainties in τ of less than 0.2 in the central B ring and less than 0.01 in the A and C rings and lower optical depth regions of the B ring. 2.9. α Virginis occultation The rev 30 α Virginis (Spica, α Vir 30) occultation track traversed the entire main ring system from the F ring through the C ring. At the time of this occultation the star was on the lit side of the rings so it was observed from the unlit side. This reduces the background of scattered sunlight from the rings. Poor weather over the Deep Space Network antennas in Goldstone, California, resulted in 104 data dropouts in this observation. Most of the dropouts are in the data for the outer A ring and F ring re- Self-gravity wakes in Saturn’s B ring gion, the Cassini Division, and the outer B ring with a loss of approximately 12% of the data. FUV data were obtained for this occultation, but due to the brightness of α Vir there is significant bleeding across the spatial pixels. Combined with the low background level on the unlit face of the rings this prevents the FUV data from being useful in determining a background model. The minimum mean signal in the opaque parts of the B ring core is 140 counts/s. The background likely increases behind the more transparent regions of the rings where interplanetary Lyman-α can shine through, but b = 140 counts/s is in the range of typical dark sky count rates for the HSP. In the optically thick region between 100,000 and 101,000 km the minimum mean count rate is 180 counts/s, but this count rate could include some signal from the star. We adopt b = 160 counts/s for the entire occultation with a probable error of 40 counts/s to allow for the possibility of enhanced Lyman-α transmission through the A and C rings. In the central B ring, adopting b > I simply means those regions are effectively opaque for this occultation. Inspection of the light curve shows that significant signals are at levels much higher than the minimum signals, so there is virtually no loss of information in taking b = 160 counts/s instead of 140 counts/s. The HSP instrumental response to bright stars (the rampup behavior) results in a changing baseline value of I0 across the occultation. We fit linear segments between the clear gaps in the rings to create a model of I0 as a function of ring plane radius, R. The resulting model varies between 5.14 × 105 counts/s in the Encke Gap and 5.44 × 105 counts/s in the inner C ring. Because there are no gaps in the B ring a constant value of I0 across the B ring is assumed with I0 = 5.34 × 105 counts/s. We adopt a probable error of 103 counts/s for I0 , and τmax = 2.82 at 50 km resolution. With our estimates for the probable error in b and I0 we get a systematic error uncertainty in normal optical depth of less than ±0.008 everywhere τn < 1, with τ < 0.05 everywhere τn < 2. Statistical uncertainties (1σ ) in τ are ±0.23 in the optically thick B ring core and ∼0.01 elsewhere. 135 most of the disk of the Sun outside the EUV solar port field of view. As a result the integrated signal over all EUV wavelengths was only 7500–8000 counts per 4-s integration period. We assume a constant value of I0 = 1741.5 counts/s based on a background determination in the B ring core, and then fit a third order polynomial to the signal I0 + b measured beyond the A ring, in gaps in the Cassini Division, and interior to the C ring. We then construct an optical depth profile using Eq. (2). The line of sight to the Sun moved 40–50 km in radius in the ring plane during the 4-s integration periods. The distance to the rings along the line of sight varied from 2.6 × 105 km in the C ring to 2.1 × 105 km at the F ring. The Sun therefore was a disk 210–260 km in diameter on the sky at the distance of the rings. There was virtually no projection effect in the radial direction, so the projected size of the Sun in the radial direction was also 210–260 km. However, most of the disk of the Sun was outside the UVIS field of view due to the pointing error, reducing the effective size of the observed disk of the Sun. The flux in the rev 11 solar occultation (solar11) from the entire solar disk was approximately 105 counts/s. Assuming a uniformly bright solar disk this gives an observed area of the Sun in solar9 of approximately 1.7% of the full solar disk, so the integration period sets the resolution at ∼50 km. We set Imin = 10 counts/s for the solar9 occultation resulting in τmax = 1.89. The statistical uncertainty τ = 0.19 at τn = τmax , and τ < 0.05 in the optical depth minima in the outer B ring discussed below. The solar11 occultation extended from the atmosphere to the outer B ring at R = 116,000 km. This observation had the full solar disk in the EUV solar port field of view. The limiting resolution is the projected size of the Sun in the ring plane, or 240–260 km. We determine I0 = 3.95 × 105 counts per 4-s integration period from the measurement of the unocculted Sun interior to the C ring and a background determination of b = 750 counts per 4-s integration from the signal in the B ring core at R = 105,000–110,000 km. We set Imin = 100 counts/s based on our estimated uncertainty in b. This results in τmax = 2.5. Because of the high counting rate, the statistical uncertainty τ < 0.004 at R > 111,500 km, and τ = 0.03 at τn = τmax . 2.10. Solar occultations 3. Radial structure of the B ring Solar occultations are observed with the EUV channel of the UVIS using a pick-off mirror 20 degrees away from the main optical boresights to avoid direct solar pointing (Esposito et al., 2004). The mirror is a section of a cylinder designed to disperse the image of the Sun along the spatial direction of the detector to prevent too high a flux of photons onto the detector. We include data from the first two ring solar occultations observed by UVIS in revs 9 and 11. There are no strong spectral features in the main rings in either solar occultation so we have summed the signal at all wavelengths in the EUV channel. The angular diameter of the Sun is 1 mrad at Saturn, and this limits the spatial resolution of these occultations. The rev 9 solar occ (solar9 hereafter) spanned the entire ring system from beyond the F ring to inside the C ring. The optical depth of the inner D ring is too low to be detected by UVIS at the moderate value of B = 21.45 degrees of this occultation. The solar9 occultation suffered from a pointing error that placed The σ Sgr and λ Sco occultations probed the highest optical depths of the stellar occultations discussed here because they are the brightest stars at higher values of B observed in the first part of the Cassini tour. The σ Sgr occultation incidence angle is comparable to that of the Voyager PPS observation of the δ Sco occultation. The angle between the line of sight and the radial direction, φ, is also similar between these two occultations. The calculated normal optical depths are virtually indistinguishable between the Cassini σ Sgr occultation and the Voyager δ Sco occultation (Esposito et al., 1983) across the entire ring system to within measurement uncertainties. Maximum optical depths are higher for σ Sgr due to the greater signal and therefore smaller detectable transparencies. At the higher elevation angle B = 41.7 degrees of the λ Sco occultation, apparent normal optical depths are higher, while occultations at lower values of B show a reduced normal optical depth (Figs. 7–10). 136 J.E. Colwell et al. / Icarus 190 (2007) 127–144 Fig. 7. Normal optical depth profiles for the inner B ring from four Cassini UVIS occultation observations at 10 km radial resolution. The apparent normal optical depth increases with increasing incidence angle, B (see Table 1). At lower optical depths the increase slows or stops at B ∼ 30◦ . Fig. 8. Three occultation profiles at 10 km radial resolution showing rapidly changing optical depth in the B ring core. The region is characterized by large fluctuations in optical depth with maximum values that are opaque to occultations. In these regions the variation in optical depth with occultation angle B is smaller than in the outer core where the valleys in optical depth are not as deep. At R > 104,000 km the low optical depth valleys become more widely spaced and have larger minimum optical depths than in the regions 99,000–100,000 and 101,000–104,000 km. In the A ring, there is a strong correlation between optical depth and φ indicating that the alignment of self-gravity wakes controls the apparent transparency of the ring (Colwell et al., 2006). This was seen most dramatically in the α Leo occultation which spanned most of the A ring at two different values of φ (but identical B for the same star) with a factor of two difference in normal optical depth. The two occultations of 126 Tau at different φ also showed a strong correlation with the transparency of the ring with alignment of the line of sight with self-gravity wakes. The B ring optical depth shows a weaker dependence on φ and a strong correlation with B. This correlation is on top of the large radial variations in optical depth across the B ring. The inner B ring from R = 92,000 to R = 99,000 km is characterized by optical depths of τn = 0.5–2.0. There is a nearly featureless valley of relatively low optical depth at R = 94,450–95,350 km, and the long Janus/Epimetheus 2:1 density wave train at R = 96,200–96,750 km (Fig. 7). There is an enhancement in the optical depth in the valley at R = 95,200 km that is 50–100 km in width in the otherwise featureless valley in this region. This enhancement is most clear in occultations at higher values of B. This may indicate a change in the ring particle properties at that location which affect the formation of self-gravity wakes. However, the amplitude of the enhancement (<0.05) is too small for differences in the wakes to be identified with our model and the data sets presented here. Beginning at R = 98,830 km the optical depth increases and enters a regime of large fluctuations on a range of spatial scales. Peak normal optical depths exceed 5 in this region, and transitions between the peaks and the neighboring valley optical depths of ∼1 usually span 10–50 km but take less than 1 km in some cases. This region extends out to R = 110,000 km (Figs. 8 and 9). At R = 104,000 km the number of valleys decreases and the optical depth in the valleys increases out to R = 108,000 km (Fig. 8). Horn and Cuzzi (1996) found quasi-periodic structures across the B ring, including this region, in analyses of Voyager imaging data, consistent with these optical depth fluctuations. Local minima in the optical depth, although near the maximum detectable values for some occultations, correlate between the occultations (Figs. 8 and 9). Minima in the normal optical depth in the σ Sgr and λ Sco profiles are τn = 2–3, and maxima exceed the measurement limit (Fig. 8). The mean optical depth begins to decline at R = 108,000 km as the valleys broaden, and the maximum optical depths drop to τn ∼ 3 at R > 110,000 km with the exception of the peak at R = 116,400 km. The outer B ring (R > 110,000 km) has complicated variations in optical depth with more gradual transitions and lower peak optical depths than the region at R = 99,000–110,000 km (Figs. 9 and 10). In this region the dependence of transparency on φ can be seen by comparing the two occultations of 126 Tau which had nearly orthogonal values of φ (Fig. 10). The ingress and egress portions of the α Leo occultation show a smaller divergence as the difference in φ increases from zero at the turnaround point to ∼25 degrees at the outer edge of the B ring (Fig. 10). The difference in φ between the ingress and egress branches of the α Sco occultation is ∼65 degrees, but the apparent optical depth is high in both cases and the differences between the two branches, where measurable, are small. 4. Aligned structures in the B ring 4.1. Self-gravity wake model The strong dependence of apparent normal optical depth on the inclination angle B is illustrated in Fig. 11 (see also Figs. 2, 7–10). The factor of μ in Eq. (2) corrects for the longer line of sight of an observation through a semi-transparent medium than would be seen from normal incidence (B = 90 degrees). This correction assumes that the optical depth scales with path length. If the medium is entirely opaque or entirely transparent, then path length is irrelevant and the factor of μ in Eq. (2) introduces a trend like the one seen in Fig. 11. We find that Self-gravity wakes in Saturn’s B ring 137 versely correlated with overall optical depth. High optical depth regions have tightly packed wakes with little empty space between them, and lower optical depth regions have more widely spaced self-gravity wakes. We model the B ring observations with the same model that was used to interpret the azimuthal transparency asymmetry due to self-gravity wakes seen in occultations by the A ring (Colwell et al., 2006). This model consists of regularly spaced slabs of normal optical depth τwake , length L, and width W separated by relatively clear gaps of normal optical depth τ gap and gap width S. Both the self-gravity wakes and the gaps are assumed to have a height H , and the orientation of the self-gravity wakes is φwake as measured from the local radial direction. With the simplification that τwake = ∞ and L = ∞, a simple analytic expression for the transparency is given by Fig. 9. Optical depth profiles at 10 km radial resolution showing the transition from the optically thick core to the complex structure in the outer B ring at about 110,000 km. In most occultations the ring remains mostly opaque out to R = 110,000 km as in the δ Aqr, λ Cet, and 126 Tau occultation profiles shown here. Gaps and plateaus in the profiles occur when there is no detectable signal from the star. Occultations at higher elevation angles (B) and with bright stars such as σ Sgr and λ Sco reveal complex structure within the core that resembles the outer B ring region, but at a larger average optical depth. T = exp(−τn /μ) [S/W − H /W | sin(φ − φwake )| cot B] exp(−τgap /μ). (4) = S/W + 1 We fit the data at both 50 and 10 km resolution using the analytic expression in Eq. (4) and the simplified ray tracing technique of Colwell et al. (2006) that allows for finite τwake . Solar occultation data are only included in the 50 km resolution fits. We fit the model parameters to the data by minimizing the quantity D where D= N 1 (τn,i − τm,i )2 , N (5) i=1 Fig. 10. Six Cassini UVIS optical depth profiles of the outer B ring at 10 km radial resolution showing variations in apparent normal optical depth with viewing angle. The α Leo occultation only penetrated to R = 114,150 km. The two 126 Tau profiles (black and green curves) have different optical depths due to a variation in φ indicating aligned structures in the ring. A smaller φ dependence is seen in the α Leo profiles. A data gap in the 126 Tau (10) data is responsible for the straight segment from 115,500 to 115,800 and near 116,400 km. at virtually all locations in the B ring the line of sight optical depth, τ = τn /μ = − ln(T ) is roughly independent of B. The ring cannot be entirely opaque, however, or no light would be seen through it at any viewing angle. The observations can be explained by alternating regions of totally opaque clumps and intervening gaps of low optical depth. The observed transparency in a given viewing geometry then depends primarily on the fraction of the ring plane that is obstructed by opaque clumps. From this simple argument one can anticipate the result of our more detailed modeling presented below, namely that the separation of self-gravity wakes in the B ring is in- where τn,i is the measured normal optical depth, τm,i is the model normal optical depth, and N is the number of usable observations. Using the uncertainties in τn described in Section 2 we can also calculate a reduced χ 2 statistic. However, because our model is an idealization (perfectly parallel and periodic structures with rectangular cross-section) there are nonstatistical errors in our fit to the data due to the assumptions of the model that complicate interpretation of the value of χ 2 . We find comparable results for the model parameters whether we minimize D or χ 2 . Similarly, changing the optical depth to transparency in Eq. (5) produces no significant changes. The parameter D is equal to χ 2 for the case where the measurement errors are equal. That is, we treat each measurement with equal weight, even though some have larger formal measurement errors as outlined in Section 2. This procedure effectively gives the error associated with the model simplifications outlined above greater importance than the measurement errors, by giving each measurement the same weight. Figs. 12–14 illustrate the sensitivity of D on different parameters for the model fit to the data at one location in the outer B ring. Like the results for the A ring (Colwell et al., 2006), our model essentially places an upper limit on H /W and a lower limit on τwake . We are able to more tightly constrain S/W and τgap , though there are still relatively large uncertainties for those parameters and for φwake (Section 4.2). Although this model oversimplifies the ring structure, it does capture the basic behavior of the ring transparency with viewing geometry allowing us to deduce properties of clumps of 138 J.E. Colwell et al. / Icarus 190 (2007) 127–144 Fig. 11. Measured normal optical depths (asterisks) at 6 locations in the B ring at 50 km resolution with the best-fit self-gravity wake model (diamonds) using the analytic expression of Eq. (4). The dominant trend is for an increase in apparent normal optical depth with elevation angle B. Spread in the data at similar values of B are due to the different values of φ of the observations. particles within the B ring. More sophisticated models will be needed when additional observations are available that probe the rings at new geometries and which are able to penetrate the optically thick B ring with greater signal. The best opportunities to see through the densest parts of the B ring with stellar occultation measurements begin in mid-2008 when Cassini will be on highly inclined orbits around Saturn. 4.2. Self-gravity wake properties in the B ring The number of useful measurements of τn varies across the B ring as different occultations have τ > τmax at different locations and not all occultations have full radial coverage (Table 1). We are able to fit the observations with the selfgravity wake model across the B ring with the exception of R = 100,000–101,000 km, and R = 104,000–110,000 km due to the high optical depths in those regions. We calculate the best-fit values of τgap , S/W , H /W , and φwake using leastsquares minimization (Eq. (5)) for all points in the B ring for which there are at least eight UV occultation measurements of the optical depth with τn < τmax and for which D < 0.1. This corresponds to a mean error in the optical depth of 0.1. When using the ray-tracing method (Colwell et al., 2006) we also get best-fit values of τwake , although for most of the ring these values are indistinguishable from infinity (i.e., opaque wakes). An exception is the relatively low optical depth regions in the inner B ring. Nevertheless, even in the valley at R = 95,000 km where τn ∼ 1 (Fig. 21), the mean calculated τwake = 4.9. Thus, assuming τwake = ∞ is a reasonable approximation across the B ring. We checked this by comparing the results for self-gravity wake parameters using the analytic expression of Eq. (4) with the results from the ray-tracing method at both 10 and 50 km resolution. We do not see systematic differences between the various techniques in the B ring. For example, the calculated values of τgap are shown in Fig. 15 for the two techniques. If the B ring structure were azimuthally symmetric ringlets of large optical depth separated by nearly empty gaps, that Self-gravity wakes in Saturn’s B ring Fig. 12. Contours of D (Eq. (5)) for the self-gravity wake model fit to the 14 observations of the B ring optical depth at R = 113,225 km (Fig. 11, lower left) as a function of S/W and τgap . The particular solution for S/W and τgap found by the automated minimization of D is shown by the asterisk, and the values of D for the two minimum contours are labeled. Fig. 13. Contours of D (Eq. (5)) for the self-gravity wake model fit to the 14 observations of the B ring optical depth at R = 113,225 km (Fig. 11, lower left) as a function of φwake and τgap . The particular solution for φwake and τgap found by the automated minimization of D is shown by the asterisk. The minimum (heavy line) contour has D = 0.012, and the next contour (dashed) has D = 0.016. This broad minimum in D makes it difficult to tightly constrain φwake . would explain the observed trend of τn with sin(B) (Fig. 11). Aligned structures in the rings introduce variations in τn with the observing angle φ. This variation is seen in the scatter of points at equal values of B in Fig. 11. The best fit values of φwake across the B ring and the A ring using the ray-trace method are shown in Fig. 16. Although there is considerable scatter in the values, the data clearly show that structures in the B ring are not azimuthal and are canted from the azimuthal direction at about the angle expected for trailing self-gravity wakes. There is a clear trend for the cant angle (π/2 − φwake ) to decrease from the inner to outer B ring, while the trend reverses in the A ring (Colwell et al., 2006). The outer B ring has higher optical depth, however the correlation between φwake and τm (the maximum value of τn measured at any given location) for all locations in the B ring is not statistically significant. Nevertheless, the difference in values of φwake between the inner 139 Fig. 14. Contours of D (Eq. (5)) for the self-gravity wake model fit to the 14 observations of the B ring optical depth at R = 113,225 km (Fig. 11, lower left) as a function of S/W and H /W . The particular solution for S/W and H /W found by the automated minimization of D is shown by the asterisk. The minimum (heavy line) contour has D = 0.012, and the next contour (dashed) has D = 0.026. Fig. 15. Calculated values of τgap using the analytic expression of Eq. (4), which assumes τwake = ∞, and using the ray-trace method of Colwell et al. (2006). The ray-trace results have been median filtered by 5 points to make the comparison between the two sets of results more clear. and outer B ring may reflect limitations of the model (infinitely long wakes, for example) that are emphasized in the outer B ring where the low-B α Leo occultation measured much lower optical depths than those at even slightly larger values of B. The finite line-of-sight optical depth at B = 9.5 degrees in the α Leo occultation may be the result of light passing over the ends of self-gravity wakes. The dominant trend in S/W is a correlation with total optical depth (Fig. 17). Large values of optical depth can only be achieved if the self-gravity wakes are tightly packed. In other words the normal optical depth of the ring is in large part determined by the fraction of the ring surface area that is blocked by the clumping of particles into self-gravity wakes. Regions with lower optical depth have more open space between the selfgravity wakes while the more opaque regions have the clumps more tightly spaced. If they have large spaces between them, then in order to have large overall optical depths there would 140 J.E. Colwell et al. / Icarus 190 (2007) 127–144 Because S/W appears to be primarily determined by τm , which we use as a proxy for the true normal optical depth (i.e., τn for B = π/2), we can fit the results shown in Fig. 17 with an exponential to get a usable value of S/W for any region of the B ring: W (6) = 0.28e1.616τm . S Values of τgap show large scatter, like those of H , with a mean value of 0.25, and 90% of calculated values of τgap are less than 0.42 and greater than 0.08. There are no correlations between τgap and either position in the ring, total optical depth, or other self-gravity wake parameters. 4.3. Small scale structure in the α Leo occultation Fig. 16. Best fit values of φwake across the B ring (diamonds) and A ring (asterisks) using the ray-trace method at 10 km resolution and median filtered by 5 points. Fig. 17. Inverse correlation between S/W , the relative spacing between self-gravity wakes, and the maximum measured optical depth. High optical depths require tightly packed self-gravity wakes (W/S 1). need to be a significant τgap . This would in turn make the ring optical depth sensitive to path length and destroy the actual strong dependence of apparent τn on B. Values of H /W are widely scattered between 10−3 and 0.2, consistent with the weak dependence of D on H below some critical upper limit (Fig. 14). The mean value of H /S for the B ring is 0.2, with scatter that is consistent with the uncertainty in the determination of H at each location. That is, to the limits of this model, one can assume H /S = 0.2 and use the values of S/W determined by the model fit and shown in Fig. 17. A similar trend in H /W with total optical depth was seen in the A ring (Colwell et al., 2006). Fig. 18 shows H /W and τgap for the A and B rings, where A ring values have been updated from Colwell et al. (2006) with the new occultations presented here. Gap optical depths are higher in the B ring than in the A ring, but because the wake spacing is much higher in the A ring the gaps contribute a larger fraction of the overall optical depth than in the B ring. The turnaround of the α Leo occultation in the outer B ring provided the first opportunity to study one radial location in the rings with a nearly azimuthal track along the ring. The small change in ring plane radius from point to point results in an effective increase in the resolution of radial structure, and the ingress and egress portions of the occultation track near the minimum radius provide two measurements under identical observing conditions of the same ring plane radii at different longitudes. The increased signal to noise at fine radial resolution allows us to directly observe ring structures at scales of ∼100 m. The turnaround radius of this occultation is R = 114,149.32 km (Fig. 10). The accuracy of the geometry solution is determined by a number of factors including the spacecraft trajectory reconstruction and the pole of Saturn and is estimated to be typically ∼1 km for the stellar occultation results presented in this paper. The precision of the calculation is better than 1 m. Here we provide the more precise values to give the relative positions of features at the turnaround, but a more accurate geometric solution that adjusts the pole of Saturn, the spacecraft trajectory, the position of the star, and the timing of the data is beyond the scope of this paper. There are 3.20 s of data (1600 points) taken just in the innermost 680 m of the occultation, between R = 114,149.32 km and R = 114,150.00 km, and there are 6.36 s of data interior to R = 114,152.00 km. Near the turnaround, the relative radial positions are affected only by the uncertainty in the timing of the data which is comparable to the integration period (2 ms). We bin the α Leo data to 0.1 s for analysis of azimuthal structure near the turnaround radius. The inner 2.7 km are shown in Fig. 19. The turnaround radius is within a local minimum in total optical depth (Fig. 10), and shifts in the data in an attempt to make the small scale radial structure correlate between ingress and egress destroy the large scale correlation between ingress and egress seen in Fig. 10. The timing uncertainty for the data shown in Fig. 19 is much less than a single point, so the correlation between the inbound (ingress) and outbound (egress) portions of the occultation is accurate as plotted. While some areas show a positive correlation, such as at R = 114,150.1–114,150.2 km, there are also several regions of width ∼0.1 km that are anti-correlated between the ingress Self-gravity wakes in Saturn’s B ring Fig. 18. Gap optical depths and H /W ratio for self-gravity wakes across the B and A rings using the ray-trace method at 10 km resolution and median filtering by 5 points. The A ring values can be compared to Colwell et al. (2006) where H /W values show more scatter due to fewer occultations used to derive model parameters. H /W , like S/W , are inversely correlated with optical depth. Fig. 19. T from data binned to 0.1 s of the α Leo occultation near the minimum ring plane radius in the outer B ring sampled by this occultation. and egress profiles in the fluctuations in optical depth (Fig. 19). For the inner 2.7 km shown in Fig. 19 the correlation coefficient is −0.02. When the data are binned into equally spaced bins in ring plane radius, the correlation coefficient between the ingress and egress profiles is −0.001. We searched for a phase lag by shifting the data in time resulting in shifting the minimum radius point on the light curve. No significant correlation was found for any possible phase lag; larger shifts of the minimum radius point destroy the ingress–egress symmetry of large scale circular features in the rings and so were not explored. The power spectra of both the ingress and egress branches, however, do show a strong peak at a wavelength of 160 m. This is the only wavelength above the noise (Fig. 20). When the data are binned in equally spaced bins in ring longitude the power spectrum does not show any dominant wavelength. The line of sight distance to the occultation point in the rings at the minimum ring plane radius is 3.12 × 105 km. The spectrum of α Leo convolved with the HSP quantum efficiency as a function of wavelength shows that 75% of the photons detected by the HSP were in the wavelength range of 132– 141 167 nm. The radius of the Fresnel zone on the plane of the sky is therefore λFr = 6.8 m using an effective wavelength of 150 nm. Projected onto the rings the Fresnel zone becomes an ellipse with its long axis in the φ direction and λFr /μ = 41 m long, and its short axis equal to λFr . At this point in the occultation φ = 112 degrees meaning that the Fresnel ellipse is rotated 22 degrees clockwise from the local azimuthal direction as seen from north of the rings. The diffraction-limited resolution is therefore sin(φ − π/2) × 41 = 15 m in the radial direction and 38 m in the longitudinal direction. Interior to R = 114,200 km the change in radius from one 2 ms integration period to the next is less than 1.5 m. The track in the azimuthal direction in one 2 ms integration period, however, is 50 m long for R < 114,200 km. In 2 ms the ring particles travel 38 m along their orbits around Saturn in the opposite direction as the occultation path, increasing the azimuthal smear relative to the ring particles to 88 m. The angular diameter of α Leo is 1.3 milli-arcsec (Radick, 1981), and at the line of sight distance to the occultation in the B ring of ∼3.1 × 105 km the projected size of the star in the sky plane at the rings is less than 2 m. So the effective radial resolution near the minimum radius is diffraction limited to 15 m. Thus the radial wavelength of 160 m is a resolved radial length scale at this location in the ring. Because the radial velocity of the occultation track is so slow for the data in Fig. 19, in the time it takes to span one 160 m radial wavelength, the occultation has sampled several km in the azimuthal direction. These structures must therefore be nearly azimuthal themselves and have much longer azimuthal extent than self-gravity wakes. Similar wavelength structure has been observed in power spectral analysis of all UVIS stellar occultations with sufficient radial sampling (Sremčević et al., in preparation) and Cassini Radio Science (RSS) occultations (Thomson et al., 2006). The sub-km structure seen in the highest resolution images (Porco et al., 2005) might also be connected with the 160 m waves observed here. These azimuthally symmetric undulations co-exist with the self-gravity wakes. This periodic wave-like azimuthal structure is likely due to viscous overstability in regions of relatively high optical depth (Schmidt et al., 2001; Salo et al., 2001). Because the model described here does not include both self-gravity wakes and axisymmetric waves, such as produced by viscous overstability, locations in the rings where the overstability waves are prominent may have different wake properties than those described here. 5. Discussion We find that clumping in Saturn’s B ring produces variations in the apparent normal optical depth of the same magnitude that is seen in the A ring. Although it has long been recognized that the azimuthal brightness asymmetry of the A ring is most easily explained by aligned structures, it is a surprise that these aligned structures persist throughout much of the B ring. Recent theoretical attempts to model the formation of largescale structures in the B ring have treated the rings as a viscous fluid with transport coefficients such as the viscosity and thermal diffusivity that can be modeled as monotonically increasing 142 J.E. Colwell et al. / Icarus 190 (2007) 127–144 Fig. 20. Fast Fourier Transform in ring plane radius space of ingress (blue) and egress (red) data from the α Leo occultation for the inner 6 km of the occultation (R < 114,115.3 km). The peaks in the ingress and egress data are at a radial wavelength of 160 m. Fig. 21. Predicted normal optical depths for measurements at B = 90 degrees at 10 km resolution compared to the λ Sco occultation, which measured the highest overall optical depths in the B ring at most locations. functions of the local number density of particles (Salo et al., 2001; Schmidt et al., 2001). These models predict a viscous overstability that produces azimuthally symmetric radial density variations that slowly grow in wavelength. Salo et al.’s (2001) direct N -body simulations have demonstrated the initial development of the viscous overstability at optical depths appropriate for the A ring. A viscous fluid model is needed to extend the small scale N -body results to larger scales. The presence of aligned self-gravity wakes in the B ring places limits on the applicability of a viscous fluid model for the B ring. In particular, a viscous fluid model can only be justified on radial scales that are larger than the characteristic wavelength of the gravity wakes. The viscous and thermal transport coefficients of such a fluid must therefore represent the spatially averaged transport caused by the wake structures. N -body simulations of self-gravity wakes in the A ring suggest that the effective viscosity is dominated by the gravitational interactions between the self-gravity wakes and that particle collisions play a secondary role (Salo, 1995; Daisaka et al., 2001; Griv et al., 2003). These authors argue that the scale of the self-gravity wakes is governed by Toomre’s critical wavelength for axisymmetric Jeans instabilities (Toomre, 1964), which scales linearly with the surface mass density of the rings. Based on scaling arguments, Daisaka et al. (2001) argue that the effective kinematic viscosity of the rings in the presence of self-gravity wakes scales as the square of the Toomre wavelength and thus increases as the square of the surface mass density. This scaling has been used in the context of modeling the spreading rate of the protolunar disk (Ward and Cameron, 1978; Kokubo et al., 2000; Takeda and Ida., 2001). However, it may not make sense to apply this scaling to Saturn’s B ring where the optical depth within the self-gravity wakes is substantially larger than unity. The inferred structures in the B ring are highly flattened, tightly packed, trailing spiral self-gravity wakes, consistent with theoretical predictions and numerical simulations for lower-optical depth rings. The variation of optical depth with the azimuthal viewing angle, φ, is smaller than in the A ring, and there is a stronger dependence on the elevation angle, B. This paints a picture of broad, flat sheets of particles that are nearly opaque with relatively empty space between them. The sheets are loosely organized into the trailing spiral density enhancements that give the azimuthal brightness asymmetry observed in the A ring. The φ-dependent asymmetry is stronger than that of the B ring. This may explain why a much weaker asymmetry has been observed in the B ring in previous observations made at a limited range of the elevation angle, B (e.g., Nicholson et al., 2005). These clumps persist across the B ring, although measurements in the dense core are limited and preclude a model fit at all locations in the central B ring. While these structures are qualitatively the same as the self-gravity wakes in the A ring, they are apparently wider relative to the ring thickness than their A ring neighbors. The height to width ratio of ∼0.1 in the B ring implies that the B ring self-gravity wakes are monolayers of the largest (r ∼ 5 m; Zebker et al., 1985) particles, with the smaller particles filling in the spaces between the larger particles. The resulting higher collisional viscosity of the B ring self-gravity wakes likely produces structures that persist for many orbital periods in contrast to the A ring where self-gravity wakes form and break apart on the orbital timescale. The viscous spreading timescale for the B ring would be four times shorter than for the A ring if viscosity continues to scale with the square of the surface mass density. We suggest instead that the B ring self-gravity wakes are strongly nonlinear structures with a characteristic wavelength that does not necessarily follow the scaling for marginal Jeans instabilities given by Toomre’s critical wavelength. The azimuthal extent of individual self-gravity wakes in the B ring is unknown and is difficult to estimate from N -body simulations due to the combination of large simulation cells needed and the high collision rates. If the B ring self-gravity wakes do indeed persist for many orbits and have a lower azimuthal frequency than their A ring counterparts, they will contribute a smaller rate of angular momentum transport by gravitational interactions between the wakes than would be implied by an extrapolation of the scal- Self-gravity wakes in Saturn’s B ring ing hypothesized by Daisaka et al. (2001) based on the Toomre critical wavelength. In addition, the high density of particles within the B ring self-gravity wakes raises the possibility that the shear rate is substantially reduced below the Keplerian rate due to finite yield stresses between tightly packed inelastic particles. This has been suggested in a recent model for the origin of large-scale structures in the B ring (Tremaine, 2003). The net result could be a shear-banded structure where the shear rate exceeds the Keplerian value in the low-density regions between the wakes. Exactly how the effective rate of angular momentum transport varies with the average surface density on scales much larger than the scale of the self-gravity wakes in the B ring is currently unknown and deserves further theoretical study. A convincing explanation for the large-scale irregular radial structure in the B ring must take into account the existence of these broad self-gravity wakes throughout the B ring. Radial structures formed by viscous overstabilities are not excluded in this picture, but the radial scale on which they can occur should be larger than the self-gravity wake scales. In the example presented here in the α Leo occultation (Fig. 19) the overstability radial scale is ∼160 m (Fig. 20) which is roughly a factor of two larger than the Toomre length scale for self-gravity wakes at that location. At regions in the B ring where the optical depth, and presumably the surface mass density, are higher, the self-gravity wake scale could be significantly larger and any overstability structures would be restricted to correspondingly larger scales as well. The self-gravity wake parameters allow us to predict the apparent normal optical depth (that is, the optical depth calculated from Eq. (2)) of the B ring that would be measured at any value of B and φ. The predicted normal optical depths (measured at B = 90 degrees) are 1.18 times the largest values measured by stellar occultation, which at most radii are the measurements from the λ Sco occultation at B = 41.7 degrees. The predicted true normal optical depths and the λ Sco optical depth profile at 10 km resolution are shown in Figs. 21 and 22 for the inner and outer B ring. We were only able to fit the wake model to a few locations in the low-τ valleys in the middle B ring, so they are not plotted. In contrast, the model fits virtually the entire A ring where optical depths are generally lower. The predicted true normal optical depth for the A ring is shown in Fig. 23 with the λ Sco profile and the maximum measured normal optical depth. The λ Sco optical depths are not necessarily the maximum values because the wider spacing of the self-gravity wakes in the A ring reduces the effect of increasing τn with B. The predicted true normal optical depth in the inner half of the A ring (R < 130,000 km), where self-gravity wakes are most prominent, is 1.19 times the maximum optical depths measured using the data presented in this paper, nearly identical to the B ring ratio of 1.18. The ratio drops to 1.07 at R > 130,000 km where self-gravity wakes are less well-organized (Colwell et al., 2006). There are no obvious trends between the self-gravity wake parameters and location in the B ring other than the trend in φ. Overall, B ring optical depth appears to be related to how closely packed these dense self-gravity wakes are packed to each other (Fig. 17). We cannot rule out the possibility that this 143 Fig. 22. Predicted normal optical depths for measurements at B = 90 degrees at 10 km resolution compared to the λ Sco occultation, which measured the highest overall optical depths in the B ring at most locations. Each location with a self-gravity wake model solution is marked by a + on the red curve. Regions of very high optical depth have no solution because there are not enough measurements. The region from R = 114,500 to 115,200 km is unusual in that the optical depth is not extreme, but the self-gravity wake model failed to find solutions that met our criteria for a good fit. Fig. 23. Predicted normal optical depths for measurements at B = 90 degrees at 10 km resolution compared to the λ Sco occultation and the largest measured normal optical depth at each location. trend is an artifact of the model in that it may be that close packing of the wakes is the only way for this regular geometric model to produce higher optical depths. However, it is difficult to reconcile the observations with an alternative explanation (such as a locally homogeneous distribution of closely packed particles). Light from faint stars is seen through the B ring at very low incidence angles (B) where the inferred normal optical depth is much greater than unity. Virtually empty gaps in the ring are the only explanation we have for this transparency in low-B observations. The gaps between self-gravity wakes fit the requirements of the observations. Photometric models of the rings must include the wake structure. The wakes make the optical depth for both incident and emerging rays from the rings dependent on both B and φ. 144 J.E. Colwell et al. / Icarus 190 (2007) 127–144 For example, the observer may be looking across the wakes and see a relatively large ring cross-section, but if the Sun direction is parallel to the wakes there will be a small illuminated crosssection and the ring would appear darker than an observation at the same phase, incidence and emission angles but with the Sun shining across the wakes. Geometry-dependent optical depths can be estimated from the self-gravity wake model and parameters presented here for arbitrary viewing geometries. To a lesser extent the wake structure will also affect the multiple scattered light. Further observations at a range of values of B and φ will enable tests of models with fewer constraints than the one described here and will help refine the variations of the fine structure in the B ring with other global properties of the ring, such as composition, surface mass density, and location. Acknowledgments This work was supported by NASA through the Cassini project. G.R.S. was supported by NASA under grant NNG06GG46G issued through the Office of Space Science Planetary Geology and Geophysics Program. We thank Jürgen Schmidt and an anonymous referee for detailed reports that greatly improved this paper. We thank Brad Wallis, Jeff Cuzzi, Phil Nicholson, and the Cassini Rings Working Group for their invaluable assistance in making these observations a reality. References Bevington, P.R., 1969. Data Reduction and Error Analysis for the Physical Sciences. McGraw–Hill Book Company, New York. Camichel, H., 1958. Mesures photométriques de Saturne et de son anneau. Ann. Astrophys. 21, 231–242. Chambers, L.S., Cuzzi, J.N., Asphaug, E., Colwell, J.E., Sugita, S., 2007. Hydrodynamical and radiative transfer modeling of meteoroid impacts into Saturn’s rings. Icarus, submitted for publication. Colombo, G., Goldreich, P., Harris, A.W., 1976. Spiral structure as an explanation for the asymmetric brightness of Saturn’s A ring. Nature 264, 344–345. Colwell, J.E., Esposito, L.W., Sremčević, M., 2006. Gravitational wakes in Saturn’s a ring measured by stellar occultations from Cassini. Geophys. Res. Lett. 33, doi:10.1029/2005GL025163. L07201. Cuzzi, J.N., Lissauer, J.J., Esposito, L.W., Holberg, J.B., Marouf, E.M., Tyler, G.L., Boischot, A., 1984. Saturn’s rings: Properties and processes. In: Greenberg, R., Brahic, A. (Eds.), Saturn. Univ. of Arizona Press, Tucson, pp. 73–199. Daisaka, H., Hidekazu, T., Shigeru, S., 2001. Viscosity in a dense planetary ring with self-gravitating particles. Icarus 154, 296–312. Dones, L., Porco, C.C., 1989. Spiral density wakes in Saturn’s A ring? Bull. Am. Astron. Soc. 21, 929. Dones, L., Cuzzi, J.N., Showalter, M.R., 1993. Voyager photometry of Saturn’s A ring. Icarus 105, 184–215. Dunn, D.E., Molnar, L.A., Niehof, J.T., de Pater, I., Lissauer, J.J., 2004. Microwave observations of Saturn’s rings: Anisotropy in directly transmitted and scattered saturnian thermal emission. Icarus 171, 183–198. Esposito, L.W., O’Callaghan, M., Simmons, K.E., Hord, C.W., West, R.A., Lane, A.L., Pomphrey, R.B., Coffeen, D.L., Sato, M., 1983. Voyager photopolarimeter stellar occultation of Saturn’s rings. J. Geophys. Res. 88, 8643–8649. Esposito, L.W., Colwell, J.E., McClintock, W.E., 1998. Cassini UVIS observations of Saturn’s rings. Planet. Space Sci. 46, 1221–1235. Esposito, L.W., and 18 colleagues, 2004. The Cassini Ultraviolet Imaging Spectrograph investigation. Space Sci. Rev. 115, 299–361. French, R.G., Salo, H., McGhee, C., Dones, L., 2007. HST observations of azimuthal asymmetry in Saturn’s rings. Icarus, in press. Gehrels, T., Esposito, L.W., 1981. Pioneer fly-by of Saturn and its rings. Adv. Space Res. 1, 67–71. Griv, E., Gedalin, M., Yuan, C., 2003. On the stability of Saturn’s rings: A quasi-linear kinetic theory. Mon. Not. R. Astron. Soc. 342, 1102–1116. Hedman, M.H., Nicholson, P.D., Salo, H., Wallis, B.D., Burratti, B.J., Baines, K.H., Brown, R.H., Clark, R.N., 2007. Self-gravity wake structures in Saturn’s A ring revealed by Cassini–VIMS. Astron. J. 133 (6), 2624–2629. Horn, L.J., Cuzzi, J.N., 1996. Characteristic wavelengths of irregular structure in Saturn’s B ring. Icarus 119, 285–310. Kokubo, E., Ida, S., Makino, J., 2000. Evolution of a circumterrestrial disk and formation of a single moon. Icarus 148, 419–436. Lane, A.L., Hord, C.W., West, R.A., Esposito, L.W., Coffeen, D.L., Sato, M., Simmons, K.E., Pomphrey, R.B., Morris, R.B., 1982. Photopolarimetry from Voyager 2—Preliminary results on Saturn, Titan, and the rings. Science 215, 537–543. Lumme, K., Irvine, W.M., 1976. Azimuthal brightness variations of Saturn’s rings. Astrophys. J. 204, L55–L57. Lumme, K., Esposito, L.W., Irvine, W.M., Baum, W.A., 1977. Azimuthal brightness variations of Saturn’s rings. II. Observations at an intermediate tilt angle. Astrophys. J. 216, L123–L126. Marouf, E., French, R., Rappaport, N., McGhee, C., Wong, K., Thomson, F., 2005. Cassini radio occultation results for Saturn’s rings. Eos (Fall Suppl.) 86 (52). Abstract P31D-04. Nicholson, P.D., French, R.G., Campbell, D.B., Margon, J.-L., Nolan, M.C., Black, G.J., Salo, H.J., 2005. Radar imaging of Saturn’s rings. Icarus 177, 32–62. Porco, C.C., and 34 colleagues, 2005. Cassini imaging science: Initial results on Saturn’s rings and small satellites. Science 307, 1226–1236. Radick, R.R., 1981. The angular diameter of Regulus from the 28 March 1980 CTIO occultation. Astron. J. 86, 1685–1689. Reitsema, H.J., Beebe, R.F., Smith, B.A., 1976. Azimuthal brightness variations in Saturn’s rings. Astron. J. 81, 209–215. Salo, H., 1995. Simulations of dense planetary rings. III. Self-gravitating identical particles. Icarus 177, 287–312. Salo, H., Schmidt, J., Spahn, F., 2001. Viscous overstability in Saturn’s B ring. I. Direct simulations and measurement of transport coefficients. Icarus 153, 295–315. Salo, H., Karjalainen, R., French, R.G., 2004. Photometric modeling of Saturn’s rings. II. Azimuthal asymmetry in reflected and transmitted light. Icarus 170, 70–90. Schmidt, J., Salo, H., Spahn, F., Petzschmann, O., 2001. Viscous overstability in Saturn’s B ring. II. Hydrodynamic theory and comparison to simulations. Icarus 153, 316–331. Takeda, T., Ida, S., 2001. Angular momentum transfer in a protolunar disk. Astrophys. J. 560, 514–533. Thompson, W.T., Lumme, K., Irvine, W.M., Baum, W.A., Esposito, L.W., 1981. Saturn’s rings—Azimuthal variations, phase curves, and radial profiles in four colors. Icarus 46, 187–200. Thomson, F., Marouf, E., French, R., Rappaport, N., Salo, H., Tyler, L., Anabtawai, A., 2006. Statistical modeling and characterization of microstructure in Saturn’s rings. In: AGU Fall Meeting. Abstract P34A-05. Toomre, A., 1964. On the gravitational stability of a disk of stars. Astrophys. J. 139, 1217–1238. Tremaine, S., 2003. On the origin of irregular structure in Saturn’s rings. Astron. J. 125, 894–901. Ward, W.R., Cameron, A.G.W., 1978. Disc evolution within the Roche limit. Proc. Lunar Sci. Conf. IX, 1205–1207. Zebker, H.A., Marouf, E.M., Tyler, G.L., 1985. Saturn’s rings—Particle size distributions for thin layer model. Icarus 64, 531–548.