Comet Shoemaker–Levy 9 Dust

ICARUS

121, 291–304 (1996) 0087

ARTICLE NO.

Comet Shoemaker–Levy 9 Dust JOSEPH M. HAHN Department of Physics, University of Notre Dame, Notre Dame, Indiana 46556 E-mail: [email protected]

TERRENCE W. RETTIG Department of Physics, University of Notre Dame, Notre Dame, Indiana 46556 AND

MICHAEL J. MUMMA NASA/Goddard Space Flight Center, Laboratory for Extraterrestrial Physics, Code 690, Greenbelt, Maryland 20771 Received August 8, 1995; revised March 7, 1996

Comet Shoemaker–Levy 9 (S–L 9) was imaged with the Hubble Space Telescope from January 1994, through solar opposition on April 29, and until impact in July 1994. As noted by several observers, no anti-sunward dust tails were detected east of the fragments after opposition (D. Jewitt and J. Chen 1994, Periodic Comet Shoemaker–Levy 9 (1993e), IAU Circular 5924; Z. Sekanina, P. W. Chodas, and D. K. Yeomans 1994, Astron. Astrophys. 289, 607–636; H. A. Weaver et al. 1994, Science 263, 787–791; G. P. Chernova, N. N. Kiselev, and K. Jockers 1995, in European SL-9/Jupiter Workshop (R. West and H. Bo¨hnhardt, Eds.), pp. 11–16, ESO Conference and Workshop Proceedings No. 52), a fact from which some have concluded that the S–L 9 fragments were inactive (e.g., Sekanina et al. 1994, Weaver et al. 1994, Chernova et al. 1995). However surface brightness profiles of the brighter fragments G1 , H, K, L, and S comae suggest the opposite. Images acquired during January and March 1994 show the brighter S–L 9 fragments’ azimuthally averaged brightness profiles varied as r21.160.1 in the inner r & 10 comae and r21.560.2 in the outer r * 10 comae, where r is the projected distance from a coma photocenter. These profiles are consistent with active comet fragments surrounded by dust comae disturbed by radiation pressure. However inactive comet models, which assume the dust was created during a burst of activity immediately following breakup, produce comae that remain highly elongated along the fragment axis throughout their orbit. This type of coma morphology was not observed. Simulations of an active, dust-producing comet in S–L 9’s orbit are presented. Comparisons of models with observations show that the dominant light-reflecting grains in the S–L 9 comae and tails had radii ranging between about 5 and 500 mm emitted from the fragments at outflow velocities of order 1 m/sec. The models show radiation pressure swept the dust into tails which appeared west of the fragments both before

and after solar opposition. As the fragments neared Jupiter in late June 1994, the combined effects of the jovian tide and radiation pressure distorted the dust tails into broad fans that consisted of grains larger than R * 100 mm and were oriented along the Jupiter–comet direction. The smaller R & 100 mm grains were confined to narrower tails whose projected orientation near the nucleus were approximately in the direction of Jupiter. The additional light contributed by the smaller grains caused the S–L 9 dust to appear brighter on the jovian side than the trailing side of the dust fans.  1996 Academic Press, Inc.

1. INTRODUCTION

Following a close encounter with Jupiter in July 1992, tidal forces disrupted Comet Shoemaker–Levy 9 (S–L 9) into approximately 10–21 concentrations of cometary material (Marsden 1993, Asphaug and Benz 1994, Boss 1994, Olson and Mumma 1994, Sekanina et al. 1994, Solem 1994, Sekanina 1995a). This event provided a unique opportunity to study the structure of cometary dust comae that were simultaneously perturbed by solar radiation pressure and a strong tidal field. From January through July 1994 the dust comae of comet S–L 9 were imaged with the Hubble Space Telescope (HST) as the fragments returned to impact Jupiter. From these observations we infer the history of dust production by S–L 9 and obtain estimates of its grain sizes and outflow velocities. One unusual aspect of Comet S–L 9 was its purported inactivity. The absence of any spectroscopic detections of sublimating cometary gases is generally not regarded as unusual for small icy bodies as far as 5.4 AU from the Sun (Cochran et al. 1994, Weaver et al. 1994, Weaver et al. 1995b, Stu¨we et al. 1995, Weissman 1996). However claims

291 0019-1035/96 $18.00 Copyright  1996 by Academic Press, Inc. All rights of reproduction in any form reserved.

292

HAHN, RETTIG, AND MUMMA

that S–L 9 was inactive and did not produce significant amounts of dust since its tidal breakup (e.g., Sekanina et al. 1994, Weaver et al. 1994, Chernova et al. 1995) warrant careful consideration. The conclusion that these fragments were inactive follows from a contradiction that arises when one assumes S–L 9 produced micrometer-sized dust usually associated with cometary activity. Solar radiation pressure drives fresh dust away from a fragment initially in the anti-sunward direction, creating a dust tail. The more distant parts of the tail lag behind the comet’s anti-solar vector due to Keplerian shear associated with its heliocentric motion. Smaller dust grains give rise to a straighter dust tail, whereas larger grains yield a shorter tail having greater curvature. The orientation of the tail on the observer’s sky plane depends upon the viewing geometry which is shown in Fig. 1 for the S–L 9 observations. When observing a dust tail both prior to and after solar opposition (when the Sun, Earth, and comet are nearly aligned) the observer’s view switches from one side of the anti-solar vector to the other. Figure 1 shows that after solar opposition the near-nucleus part of the S–L 9 tails can project east of the fragments while the more distant dust would appear west of the fragments due to the tail curvature. Comet S–L 9 was observed through two solar oppositions and no eastward dust features were observed (Jewitt and Chen 1994, Sekanina et al. 1994, Weaver et al. 1994, Chernova et al. 1995). Sekanina et al. (1994) reported (and we confirm in Section 4) that if S–L 9 produced substantial amounts of dust smaller than a micrometer then nearly anti-sunward dust tails would have been visible east of the fragments when observed after solar opposition. Since eastward dust features were absent from the post-opposition observations of 1993, Sekanina et al. (1994) concluded S–L 9 was inactive, that is, the observed dust was created during or soon after the comet’s breakup in July 1992 rather than via more recent dust emission. However we draw a more cautious conclusion from these observations—that the absence of any eastward dust features in the post-opposition observations indicates that dust smaller than about a micrometer were not present in sufficient quantities for detection, or were simply absent. The remainder of this analysis considers whether S–L 9 may have instead produced larger grains and asks whether this hypothesis can better explain the observed tail orientations. However, first consider the possibility that the S–L 9 fragments were inactive. If so then the dust enveloping each fragment (see Fig. 2) was likely created during a burst of activity triggered by the breakup event while still deep in Jupiter’s gravity well. Had this occurred, the planet’s strong tide would have deflected any dust flowing from the fragments in directions toward/away from Jupiter and hence along the fragment train axis. As will be shown in Section 4, column density contours of the surviving debris

FIG. 1. The viewing geometry of the Comet Shoemaker–Levy 9 observations. The sun (, Earth %, and Comet S–L 9 are projected to the ecliptic in a rotating frame in which the projected Sun–comet direction is constant. The Sun–comet–Earth phase angle is f and projected lines of sight from Earth to S–L 9 are shown for selected observing dates. The directions east/west on the sky plane are approximately left/right in the figure. As the comet moves toward Jupiter it penetrates the figure plane at an angle p508 from below during solar opposition on April 29, 1994. The projected location of Jupiter on this date is also indicated. When viewing a relatively straight dust tail (not drawn to scale) composed of small R , 1 em grains extending from the comet in the anti-sunward direction, an Earth observer will see it extending west of the comet prior to opposition and east of the comet after opposition. However dust tails composed of much larger grains will be sufficiently curved such that they always appear west both before and after opposition.

nearest the fragments would have remained strongly elongated along the fragment axis. Isophotes of S–L 9 imaged 1.5 years after breakup (Fig. 2) show no evidence of comae elongation in this direction, indicating most of the observed dust must have been produced while far from Jupiter. While we do not rule out the possibility of a significant post-breakup burst of dust, reports that S–L 9 did not experience subsequent activity are inconsistent with the observations. Instead it shall be shown in Section 3 that the fragments’ surface brightness profiles were similar to that expected of an active comet. When far from Jupiter and relatively undistrubed by tides, the S–L 9 brightness profiles varied as r21.1 in the inner comae and as r21.5 in the outer comae, where r is the projected distance from

COMET SHOEMAKER–LEVY 9 DUST

293

FIG. 2. Isophotes of fragments F, G1 , H, and K observed on January 25, 1994. In order to see the fainter light levels of the tails and interfragment dust, the lowest two contour levels are of the image smoothed over a 20 window. Intermediate levels are of the image smoothed over 10, and the upper three levels have been smoothed over 0.30. The elongated structure between fragments F and G1 is a smoothed star trail. This smoothing tends to square the isophotes of the fainter fragment F. The brightness of each contour level increases approximately as an exponential. The arrows indicate celestial north and east as well as the direction from the comet to the Sun ( projected onto the viewing plane (which in all these observations is close to perpendicular) and Jupiter J.

the coma photocenter. Such profiles are consistent with an active comet that steadily produced dust that was perturbed by radiation pressure (Jewitt and Meech 1987). That circularly symmetric dust comae were seen surrounding each of the brighter S–L 9 fragments, that their inner comae brightness profiles were observed to vary nearly as r21, and that the dust comae did not appear elongated along the fragment train axis when far from Jupiter, all indicate the fragments had been continuously replenishing their dust comae. Below we inspect HST observations of Comet S–L 9 in order to resolve the controversy generated by the dust tail observations, which were interpreted by some investigators to mean the fragments were inactive, and the dust comae brightness profiles which indicate active replenishment. Comet S–L 9 was observed with the HST during the three months occurring both before and after solar opposition on April 29, 1994. With these observations we report on changes in the coma/ tail morphology as the fragments approached Jupiter. By examining their brightness profiles and comparing the images to models of comet dust tails, we quantify the nature and history of dust production by S–L 9 and estimate the likely range of dust outflow velocities and grains sizes that dominated the reflected comae light. 2. HST IMAGES OF COMET SHOEMAKER–LEVY 9

The HST images of Comet Shoemaker–Levy 9 analyzed here were obtained during the final six months of its orbit about Jupiter (see also Weaver et al. 1995b). These images were acquired soon after the HST optics were corrected in December 1993, and the coverage began on January 25,

1994 when the comet was about 4 3 107 km from Jupiter on the return leg of its orbit, or about 80% of its apojove distance. The last-look images of a given fragment were often taken a few hours before impact which took place July 16–22, 1994. The comet was imaged with the Wide Field Planetary Camera 2 (WFPC2) in the F702W filter, ˚. roughly an R filter having a band pass of 6000 to 8000 A Integration times were typically 30 min. The usual pipeline processing has been applied to all images (e.g., Baum 1994, Burrows 1994, Holtzman et al. 1995): the bias and dark currents have been subtracted, the images flatfielded, and cosmic ray hits repaired. The WFPC2 instrument has two types of CCD detectors: the Wide Field (WF) camera which samples the sky at 0.1 arcsec/pixel and the Planetary Camera (PC) which has 0.046 arcsec/pixel. The HST point spread function full width at half-maximum is about 0.06 arcsec so the PC images are slightly oversampled and the WF images are undersampled by a factor of 1.7. For a target at a geocentric distance of about 5 AU, a PC pixel element subtends about 170 km and about twice that for the WF. The nearly complete train of comet fragments was observed during three observing runs in January, March, and May 1994. Due to the increased extent of the fragment train as it neared impact only portions of the train were imaged during June and July. For this reason we concentrate our analysis on only those brighter S–L 9 fragments that were imaged when both far from and near to Jupiter. Bright fragments having close companions such as the Q1 –Q2 pair had entangled brightness distributions that are more difficult to interpret; we refer the reader to Weaver et al. (1995b). This then leaves us with the following fragment

294

HAHN, RETTIG, AND MUMMA

program sample G1 ,1 H, K, L, and S, which we submit to a detailed examination below. We have inspected the dust comae/tail morphologies and brightness distributions of all the major fragments imaged with the HST and will often compare them to the program sample. All the cometary comae and tails were observed to undergo marked evolution as they approached Jupiter (Rettig et al. 1994a, Colas et al. 1995, Jewitt 1995, Rettig et al. 1995, Weaver et al. 1995b). Except for the unusual cases noted below, the evolution of most of the fragments generally paralleled each other as they approached Jupiter. For this reason we only show graphical results of our analysis for a single fragment (often H) on each observation date and then comment upon its similarities and differences to other members of the fragment train. The exceptional cases mentioned above are of course the time-varying fragments such as T, which faded from view on approach to Jupiter (Weaver 1994), fragments J and M which disappeared prior to these observations, and fragments G2 , P1 , and P2b , all of which separated from other fragments before fading (Sekanina et al. 1994, Weaver et al. 1995b, Sekanina 1995a). These fragments are not treated here. There is considerable debate concerning the nature of the solid material embedded in the center of the S–L 9 comae. Most treatments of S–L 9 assume a single solid body existed within each coma (e.g., Scotti and Melosh 1993, Weaver et al. 1994), whereas tidal breakup models indicate the fragments may have been fractured rubble piles (Asphaug and Benz 1994, Solem 1994). It has also been suggested the S–L 9 fragments were instead gravitationally bound swarms and/or reaccreted clumps of numerous bodies that condensed from the available debris after tidal breakup (Olson and Mumma 1994, Rettig et al. 1996b). We also note a recent examination of these HST images reports clusters of up to eight faint bodies hidden within p0.20 of the comae photocenters (Sekanina 1995b). However when we refer to a comet/nucleus/fragment, it is simply for convenience and no conclusion about the structure of an S–L 9 fragment is implied other than that it is treated here as a single unresolved source of dust. Figure 2 shows isophotes of a section of the S–L 9 fragment train composed of fragments F, G1 , H, and K’s dust comae as they appeared on January 25, 1994. During its final circuit about Jupiter, the long axis of S–L 9’s highly eccentric orbit was oriented approximately perpendicular to the direction of the Sun and inclined about 508 below Jupiter’s orbit plane, so in these images Jupiter always appears in the approximate direction of celestial northeast. The fragments themselves were always aligned toward Jupiter (Sekanina et al. 1994) with fragment A in the lead and W trailing. For this observation the Sun–comet–Earth 1

The literature often identifies this fragment simply as G prior to its splitting before May 7, 1994.

TABLE I Observation Parametersa Date (1994)

Db (AU)

r c (106 km)

f d (8)

January 25 March 31 May 18 June 26

5.40 4.50 4.41 4.80

210.5 25.6 3.7 9.5

June 27 July 4 July 12 July 14

4.82 4.91 5.03 5.07

July 19 July 20

5.14 5.16

July 21

5.17

40.0 31.2 22.8 (H)e 12.1 (S) 13.2 (K) 12.2 (Q1 ) 10.1 (Q1 ) 6.81 (G1 ) 3.92 (H) 4.49 (S) 5.78 (K) 1.37 (L) 0.86 (Q1 ) 1.01 (S) 1.90 (S) 1.49

9.5 10.0 10.5 10.6

10.7 10.7

10.7

a Provided by P. W. Chodas (1995, private communication). b The comet’s geocentric distance. c The comet’s jovicentric distance. Note that the fragment train length grows from p3 3 105 km on January 25 to about p1.5 3 106 km on June 26. d The Sun–comet–Earth phase angle. e Fragment ID.

phase angle was near its maximum of f 5 210.58 with the minus sign indicating a pre-opposition observation. Table I lists all the important viewing parameters and Fig. 1 sketches the observing geometry. Contours of the comae on January 25 are nearly circularly symmetric within about 10 of their photocenters. Beyond this distance the isophotes become progressively elongated in the projected anti-sunward direction due to the combined action of solar radiation pressure and the grains’ orbital dynamics, causing dust tails to extend from most of the fragments. The comet fragments were also embedded in a sheet of dust whose southern edge traces the brighter members of the fragment train. Though faint, its light contribution affects coma photometry beyond about 50 from the comae photocenters, and in these images any tail structure is completely lost in the background beyond about 15–200. Note that ground-based observations during 1993 revealed tails as long as 19 (Scotti 1993a,b; Jewitt and Chen 1994). Later observations were at smaller phase angles as the Earth approached the Sun–comet line (see Fig. 1). On March 31, 1994, f 5 25.68, and the tails remain oriented in the projected anti-sunward direction (Fig. 3a), approximately west–northwest. The next observation on May 18 (Fig. 3b) occurred after opposition and at our smallest phase angle f 5 3.78. The dust tails were considerably less

COMET SHOEMAKER–LEVY 9 DUST

295

FIG. 3. Isophotes of (a) fragments H and K on March 31, 1994, (b) H and K on May 18, (c) H on June 26, and (d) G1 on July 14. These images have been smoothed in a manner similar to that described in the legend to Fig. 2. In all the observations the northwest side of the fragment train axis is dusty while the background-limited southeast side appears free of dust. The narrow elongated structures are star trails. The shaded areas in (a) indicate the image regions from which sunward and tailward brightness profiles are calculated (Fig. 6), and the profiles in Fig. 7 are generated from the shaded regions in (d).

distinct as the line of sight was nearly along the Sun–comet direction. Rather than displaying anti-sunward dust tails projecting approximately east and into the dust-free region of space, as would be expected for an active comet emitting micrometer-sized or smaller dust (e.g., Sekanina et al. 1994), the isophotes still show tails pointing in approximately the same westward direction as in the previous observations. Ground-based observers imaged the comet as it passed through solar opposition on April 29, 1994

and no changes in the appearance of the comae/tails were detected (Chernova et al. 1995). Nor did ground-based observations obtained after the comet’s previous solar opposition on March 26, 1993 reveal any eastward dust features. Instead, the dust tails were always aligned roughly west of the fragment train (Scotti 1993a,b; Sekanina et al. 1994 and references therein). From the absence of any anti-sunward dust features, Sekanina et al. (1994) and Chernova et al. (1995) concluded the fragments were inac-

296

HAHN, RETTIG, AND MUMMA

tive, while Jewitt (1995) suggested the tails were formed by a mechanism other than radiation pressure. Note also that the interfragment dust sheet had not changed its aspect between the pre- and post-opposition observations. If this sheet were simply a plane of dust stretching away from the fragment axis in the anti-sunward direction, then in the post-opposition May images it would have been seen on the southeast side of the fragment line. Since it was seen northwest of the fragments, this sheet must have been warped such that the downstream dust always appeared northwest of the comet fragments both before and after solar opposition. Similar to the dust tails, this curvature was likely due to Keplerian shear which caused it to lag west of the anti-solar vector. The next observation of S–L 9 on June 26 (Fig. 3c shows fragment H) was at a phase angle of f 5 9.58 so the viewing geometry was nearly the mirror image of the January observation but with the comet now three times closer to Jupiter. Again, evidence for a bright anti-sunward tail is absent and the downstream portion of the dust sheet remained northwest of the fragment train. The isophotes clearly reveal the influence of the jovian tide which elongated the coma in directions to and away from Jupiter (see also Weaver et al. 1995b). Figure 3d shows fragment G1 as seen on July 14 when it was 3.9 3 106 km from Jupiter (part of fragment H’s coma was clipped by the detector edge so an adjacent fragment is shown). A faint contribution by additional dust lies north of the fragment. Nearly all of the fragments observed with the HST in June and July, objects K, L, Q1 , Q2 , R, S, U, V, and W, showed a morphology very similar to the contour plots of fragments H and G1 in Figs. 3c and 3d, differing only in total brightness. West et al. (1995) labeled the dust features extending from the fragments in the direction opposite to Jupiter (Figs. 3c and 3d) as ‘‘normal’’ tails for their position angles agree well with the tail orientation predicted by Sekanina et al. (1994) for an inactive fragment. West et al. also identified the dust features aligned toward Jupiter as ‘‘anomalous’’ tails. However the simulations presented in Section 4 will show these features were instead tidally broadened dust fans that in projection appear very foreshortened. Rettig et al. (1994a) and West et al. (1995) pointed out that the dust surface brightness on the jovian side of the dust fans were brighter than the trailing dust. This is seen more clearly in Fig. 4 which shows a contrast-enhanced image of fragment G1 on July 14. Contrast-enhanced images have also been produced for fragments H, K, L, and S, and generally reveal similar dust structures over similar spatial scales. 3. S–L 9 BRIGHTNESS PROFILES

Characteristics of the S–L 9 dust may be obtained from these images by inspecting surface brightness profiles of the

FIG. 4 The contrast-enhanced image of fragment G1 on July 14, 1994, produced by subtracting from the G1 image the mean brightness ¯ 1 . The mean brightness image G ¯ 1 is formed from the G1 image image, G by replacing each pixel with the mean value of all those pixels lying in a pixel-wide ring concentric with the coma photocenter. The difference ¯ 1 )/ image is formed and displayed in units of the mean brightness, (G1 2 G ¯ 1 . The coma photocenter is at the origin. Areas of light/dark indicate an G excess/deficit of dust relative to the mean dust column density at a given radial distance from the photocenter. To facilitate viewing, the gray-scale image has been smoothed over a 0.20 window and the contour has been smoothed over 10. The coma brightness asymmetry in the jovian direction is real rather than an artifact of the image processing. For instance, shifting the assumed coma photocenter location by a full pixel (in excess of the fractional pixel uncertainty in the photocenter locations of the program fragments) does not significantly alter the appearance of the contrast-enhanced images nor the coma brightness asymmetry.

fragments’ comae and tails. From these profiles we extract estimates (or limits) on the dust grain sizes and ejection velocities. When the comet was far from Jupiter (i.e., at the time of January and March 1994 observations) the innermost azimuthally averaged profiles generally fall off as r2n where r is the projected distance from a coma photocenter and n typically ranges from 1.0 to 1.2. Weaver et al. (1995a) characterized the inner comae with a power law index n 5 1.16, amending a previous estimate (Weaver et al. 1995b). Under ideal conditions an n 5 1 index is anticipated in the inner coma of a comet experiencing steady and isotropic dust emission (Wallace and Miller 1958), and deviation from unity simply reflects size sorting of the grains by radiation pressure. At sufficiently large r, an n 5 1.5 power law will develop asymptotically (Jewitt and Meech 1987). Fragment H’s azimuthally averaged brightness profile is extracted from Fig. 3a (March 31, 1994) and displayed in Fig. 5. We refer to this profile as azimuthally averaged

COMET SHOEMAKER–LEVY 9 DUST

FIG. 5. The azimuthally averaged brightness profile of fragment H on March 31, 1994. Each data point is the mean coma brightness in a 0.10-wide ring-shaped aperture of radius r concentric with the coma photocenter. The brightness profile is normalized to the value at the r 5 0 photocenter, and the height of the plotting symbol indicates the data uncertainty. Since the axes are logarithmic, the r 5 0 data point is placed at r 5 sA pixel 5 0.050 for this WF image. The lower line is the uncertainty in the background subtraction which suffers from light pollution by spatially varying light sources such as the interfragment dust and star trails beyond r p 100 (see Fig. 3a). A r2n power law is fitted to the inner coma (0.2 # r # 0.60) with n 5 1.1 6 0.1 determined by a least-squares fit weighted by the data uncertainty. The fit to the outer coma (1.5 # r # 4.00) yields n 5 1.5 6 0.2.

since each data point in the figure is the mean coma brightness within a pixel-wide ring concentric with the coma photocenter. The flattening of the r & 0.150 profile is a consequence of the telescope’s point spread function and does not represent the true coma profile. In the inner 0.15 & r & 10 coma region, the azimuthally averaged profile for fragment H and many of the other bright fragments observed between January and March 1994 fall off with a power law index n 5 1.1 6 0.1. Similar results are found for the dimmer fragments though with larger uncertainty. Generally the isophotes of all the brighter fragments appear circularly symmetric within r & 10 of their photocenter (Figs. 2 and 3a). Beyond r * 10, the comae develop a bow in the projected sunward direction, the isophotes (Figs. 2 and 3a) lose their circular appearance, and the azimuthally averaged brightness profiles fall off as a power law with n 5 1.5 6 0.2 (Fig. 5). These profiles are consistent with S–L 9 as an active comet.2 2

Note that the earlier S–L 9 observations in 1993 show that the fragments’ comae brightness varied as pr20.7 beyond r * 10 of their photocenters (from Fig. 6 of Jewitt (1995)), indicating that dust production by S–L 9 previously had a more complex history.

297

Information on the dust spatial distribution is obtained from sunward and tailward brightness profiles, examples of which are given in Fig. 6 for fragment H on March 31. These profiles use only those pixels appearing in the two shaded wedge-shaped apertures indicated in Fig. 3a having opening angles of 408. Each data point in Fig. 6 is the mean coma brightness within a pixel-wide arc concentric with the coma photocenter and bounded by the sunward/tailward apertures. The tailward profile falls off as pr21 while the sunward profile decreases slightly faster than pr21 in the inner r & 0.750 comae. Beyond r Q 0.750 the sunward profile rapidly drops to zero and the contour maps (Figs. 2 and 3a) of this region lose their circular appearance. All the fragments on our program list (G1 , H, K, L, and S) as well as fragments Q1 and R generally obey these trends. The observed spatial scale of a dust coma is determined by the grain sizes and outflow velocities of its dominant light-scattering dust. Grains ejected at higher velocities and/or relatively large dust will have a greater radial extent than any smaller/slower grains which are quickly driven tailward by radiation pressure. Thus a measurement of the projected ‘‘radius’’ of a coma, i.e., the projected distance from the coma photocenter to its sunward apex, will pro-

FIG. 6. Brightness profiles in the sunward and tailward directions for fragment H on March 31, 1994. These profiles are extracted from the shaded wedge-shaped apertures of Fig. 3a. The characteristic photocenter–bow distance r c* Q 0.750 lies where the profiles differ by a factor of 2, and the curves first begin to significantly diverge beyond rm * in Q 0.150. The sunward coma signal is lost in the noise beyond rm * ax p 40 in the dust-free sector southeast of fragment H. A r21 curve is also indicated. Error bars in the tailward profile are comparable or smaller than the plotting symbol. The profiles and the coma measurements are relatively insensitive to the choice of aperture opening angles # 808.

298

HAHN, RETTIG, AND MUMMA

TABLE II Measurements of Comae Distorted by Radiation Pressure Fragment

Date (1994)

r *min (arcsec)

r *c (arcsec)

rm * axa (arcsec)

G1 H H K K L L S S

March 29 January 25 March 31 January 25 March 31 January 25 March 31 January 24 March 31

0.35 0.10 0.15 0.4 0.5 0.15 0.1 0.20 0.15

1.35 0.70 0.75 0.8 1.0 0.5 0.5 0.75 1.0

5.0 5.0 4.0 4.0 3.5 3.0 3.5 2.2 3.7

a

These are observed lower limits.

vide useful constraints on the fragment’s dust sizes and outflow velocities. Since a comet emits dust having a range of sizes and velocities, this distance is not defined precisely. However an order of magnitude estimate of the characteristic photocenter-sunward apex distance, r c* , is obtained by identifying this distance to be where the intensity of the sunward brightness profile is half that of the tailward profile (Fig. 6). This distance approximately divides the circularly symmetric inner coma region from the elongated isophotes of the outer comae/tails in Figs. 2 and 3a. Table II summarizes the comae measurements r c* for the program fragments observed on January 25 and March 31, 1994. Table II also lists r m * in , the distance beyond which the sunward profile first begins to significantly deviate from the tail’s approximately r21 profile, or the distance beyond which the dust comae first start to show deviations from circular symmetry. Among the five fragments inspected, * in Q 0.1 to 0.50. We note this this distance ranges from r m deviation from circular symmetry may have an impact on photometric estimates of the S–L 9 fragment sizes. Two independent measurements of the S–L 9 sizes have been obtained by subtracting the light contributed by model dust comae from these HST images, and from the residual light the fragment sizes or upper limits were estimated (Weaver et al. 1995b, Sekanina 1995b). Both models assumed a priori the S–L 9 dust comae were circularly symmetric. Whether the observed deviations from circular symmetry might alter the fragment size measurements/ limits or the number of fragments reported hidden within a single coma (e.g., Sekanina 1995b) deserves consideration. Also listed in Table II is the distance beyond which the sunward coma signal of each fragment is lost in the background-subtracted noise, r m * ax , which ranges from 2 to 50, and is a lower limit on the projected sunward extent of the S–L 9 comae. These quantities r c* , r m * in , and r m * ax will be used in Section 4 for identifying the range of the S–L 9 dust radii and outflow velocities.

Later observations of S–L 9 demonstrate the increasing influence of the planetary tide as the comet fragments approached Jupiter (Figs. 3c, 3d). These isophote maps are of fragments H and G1 imaged on June 26 and July 14, 1994 at respective jovicentric distances of 1.2 3 107 and 3.9 3 106 km. To illustrate the development of the coma appearance, consider the initially radial flow of dust generated by a fragment. The grains will be deflected to or away from Jupiter by the planetary tide as radiation pressure pushes anti-sunward and nearly along an observer’s line of sight. The net result is a fan of dust oriented in the Sun–comet–Jupiter plane. However the dust flow remains radial and isophotes appear circular very near the fragment where tides/radiation pressure have not yet deflected the grains. Since the radius of the inner coma region scales with outflow velocity, its measurement will serve as a useful indicator of the dust outflow velocities. The effective inner coma radius of fragment G1 , obtained from its brightness profiles on July 14 (Fig. 7), is calculated in the same manner as for the sunward/tailward profiles of Fig. 6. The profile labeled ‘‘jovian’’ is the average of the two profiles extracted from the dark gray regions of Fig. 3d pointing to and away from Jupiter, and the perpendicular profile is the average of the profiles ex-

FIG. 7. Jovian and perpendicular brightness profiles of the tidally distorted coma of fragment G1 on July 14, 1994. The jovian profile is the average of the two profiles pointing to and away from Jupiter in the dark gray region of Fig. 3d, and the perpendicular profile is the average of the two light gray profiles. The distances r tmin p 0.10 and rtc Q 0.70 denote where the coma isophotes lose curcularity. Since the Earth–comet– Jupiter angle is about 1008 for this observation an observer’s line of sight is nearly perpendicular to the tidal axis, making projection effects relatively unimportant.

COMET SHOEMAKER–LEVY 9 DUST

TABLE III Measurements of Tidally Distorted Comae Fragment

Date (1994)

r tmin (arcsec)

r tc (arcsec)

G1 G1 H H K K L S S S

June 26 July 14 June 26 July 14 June 27 July 19 June 27 July 4 July 14 July 21

0.25 0.1 0.5 0.2 0.3 0.05 0.4 0.6 0.25 0.1

3.0 0.7 2.2 0.6 3.0 0.25 2.5 1.8 1.2 0.3

tracted from the lighter gray regions. The jovian profile decreases approximately as r21 while the perpendicular profile falls off faster. The profiles coincide with each other t only out to about r min Q 0.10 from the coma center, and this distance is the boundary of the innermost coma region that appears undisturbed by tides (note this width is near the resolution limit of the HST). Beyond r ct Q 0.70 the perpendicular profile is fainter than the jovian profile by a factor of 2. As in our examination of radiation pressure effects, we take this length to be the distance beyond which the typical coma grains were deflected from a radial outflow into a tidally distorted flow. Table III lists the coma measurements obtained for several fragments imaged in June and July 1994, and these will be used in conjunction with dust tail models to estimate the S–L 9 outflow velocities.

4. DUST TAIL MODELS

A cometary dust tail is simulated by numerically following the motion of dust particles continuously produced by a comet in orbit about Jupiter. Comparing the model results to the HST images will provide additional evidence that S–L 9 was indeed an active comet, and comparisons to the comae measurements (Tables II and III) will yield estimates of the dust sizes and outflow velocities. This model is guided by the HST observations of comet S–L 9 which impose the following restrictions. On January 25 and March 31, 1994, the model dust tail is required to lie west of the comet with a sunward coma apex in the east when viewed from Earth, as in Figs. 2 and 3a. Specifically the sunward apex should lie r c* p 10 east–southeast of the nucleus for most of the dust grains, but with some grains traveling at least as far as r m * ax p 50 where the coma signal is lost (Table II). On May 18, after solar opposition, the tail should still appear west with an eastern coma edge that extends no more than p50 (Fig. 3b). By June 26 the

299

tail should broaden into a fan that is aligned to/away from Jupiter (Fig. 3c). The circularly symmetric inner coma region should steadily shrink upon approach to Jupiter (e.g., Table III) and the dust fan should be preferentially brighter in the direction of Jupiter. By July 14 all these features should become even more pronounced (Figs. 3d and 4). We employ a simple model of the Solar System which considers only the gravity of the Sun and Jupiter. Both are treated as point masses and the Sun is assumed stationary rather than orbiting the system barycenter. Over the two years spanning the comet’s final orbit, Jupiter’s orbit is treated as a planar ellipse given by its osculating orbit elements on August 1, 1993 (Donat 1993). Its motion in time is computed from the elliptic orbit equation expanded to second order in Jupiter’s eccentricity (Smart 1961). Starting from positional data for fragment H on July 14, 1993 (P. W. Chodas 1995, private communication) when it was furthest from Jupiter, we first integrate the comet’s motion backward in time to October 14, 1992, about three months after breakup, using a fourth-order Runga–Kutta integrator (Press et al. 1989). We start dust production on this date and emit grains in random directions as the comet orbits Jupiter. The particles’ motions are integrated forward in time with their positions recorded for each observation date. The dust are subject to jovian and solar gravity as well as radiation pressure appropriate for dark absorbing grains larger than a micrometer. Comet gravity is ignored and the model assumes a dust grain mass density of 1 gm/cm3. To test the model’s accuracy, our computed orbit for fragment H is compared to the orbit calculated by P. W. Chodas who employed a more sophisticated solar system model and algorithm. The errors in our computation are insignificant until the fragment nears Jupiter. By May 18, 1994, the fractional errors in its computed jovicentric position are smaller than p0.1% and they grow to p1% on July 14 when the fragment was imaged 4.5 3 106 km from Jupiter a few days before impact. For our purposes this model computes sufficiently accurate comet/dust orbits. However these positional errors are not due to an accumulation of integration errors. To show they are indeed small, we integrate the comet’s orbit backward in time from apojove to about 10 months prior when it was half the jovicentric distance. From that position the comet is integrated forward in time until July 14, 1994 just prior to impact. Its position is compared to a single forward integration from apojove to July 14, 1994. When viewed from p5 AU away the relative errors are about 0.10. Since this is near the resolving limit of the HST, we conclude that accumulated integration errors do not cause significant numerical drift of the relative positions of the model comet and its associated dust. Model dust tails have been computed for grain radii 0.1 em , R , 1 cm and outflow velocities V 5 0 (i.e., dust

300

HAHN, RETTIG, AND MUMMA

blown from the nucleus by radiation pressure), 0.3, 1.0, and 3.0 m/sec. Each dust tail simulation emits grains of a single radius R at a single velocity V. Since the width and orientation of a model dust tail depends sensitively on R and V, the range of dust sizes and outflow velocities that dominant the light-reflecting grains in the S–L 9 comae may be determined by comparing a suite of tail simulations to the images. The grain size/velocity classes that produce model comae and tails consistent with the entire time sequence of S–L 9 images are easily identified. Similarly, those simulations producing comae too large or tails too wide or having an incorrect orientation are set aside—such grains were not produced in sufficient quantities for detection or were simply absent. Selected simulations are displayed in Fig. 8 in order to demonstrate the range of dust sizes needed to reproduce the S–L 9 comae. Note, however, that the true S–L 9 grain size/velocity distribution should be regarded as a superposition of many individual dust tails, with each grain size/velocity class contributing to the observed comae light in a manner dependent upon their (undetermined) abundance. The first simulation, Fig. 8a, shows the top view of a model dust tail composed of R 5 500 em dust grains ejected at velocity V 5 1.0 m/sec computed for three observation dates January 25, May 18, and July 14, 1994 (results for the intermediate March and June observations are not shown). On each date the position of all the grains are projected to the plane that passes through the Sun and the fragment with the x axis parallel to Jupiter’s orbit plane. The view is looking down from the approximate direction of celestial north, the z axis points from the comet in the direction opposite to the Sun, and to an Earth observer the positive x direction is approximately west. Figure 8b shows the evolution of the tail appearance as seen from Earth. It is evident from both figures that in January (prior to solar opposition) and May (after opposition) the model dust tails trail from the comet to the west. From this simulation we conclude that the absence of anti-sunward (i.e., eastward) tails in the May 18, 1994 image is no longer evidence against the comet’s activity. By July 14, Figs. 8a and 8b show the dust has spread out to a fan that lies along the comet-Jupiter direction with a projected cross section resembling the elongated coma morphology seen in the HST image, Fig. 3d. In all the simulations tidal effects do not become evident until mid-June and well after solar opposition which indicates jovian gravity was not a factor in the tails’ earlier westward orientation. Figure 8b also shows that the extent of the model coma in the direction east–southeast of the fragment on May 18 is about r p 40 from the coma photocenter. This distance is similar to the observed spatial scale of fragments H and K’s comae (Fig. 3b) as well as the other program fragments. Since this distance scales with the dust outflow velocity,

the model’s agreement with the data indicates that the typical light-reflecting grains were ejected at velocity V p 1 m/sec. Increasing V enlarges the coma spatial scale proportionately, so simulations of faster V $ 3.0 m/sec grains fail to agree with the HST data since they produce comae too large. At the higher velocities, grains larger than p5 em contribute comae extending too far eastward * ax . 50) in January through May, and by June and July (r m the tidally distorted dust fans have inner comae radii (r ct . 100) much larger than is recorded in Table III. The model indicates that only grains larger than p100 em emitted with V p 1 m/sec could have avoided being swept away by radiation pressure and allow the tidally broadened fans to develop upon approach to Jupiter, as seen in the June and July images (Figs. 3c, 3d, see also Rettig et al. 1996a). However a wider distribution of sizes is necessary to simultaneously explain all of the observed dust features. Comparisons of the January–March images with simulated comae indicate that smaller dust grains with radii 5 em & R & 100 em ejected at V p 1 m/sec are also needed to contribute comae with a characteristic sunward bow distance extending out to r c* p 10 (Table II). However grains much smaller than a few micrometers could not have been strong contributors to the comae column density for such grains would have produced narrow anti-sunward tails extending east-southeast of the fragments in observations acquired after solar opposition. This effect is demonstrated in Fig. 8b which shows the appearance of an R 5 0.5 em tail on May 18 and we find smaller grains would project even further east–southeast. Since these anti-sunward features were absent in the observations (see Fig. 3b), we conclude that S–L 9 did not produce detectable quantities of dust much smaller than about 5 em. This was first noted by Sekanina et al.(1994) who used this argument to show that grains smaller than a micrometer were also absent in ground-based images of S–L 9 acquired after the previous solar opposition in late March 1993. Also, we find little evidence for grains much larger than 1 mm ejected at velocity V . 1 m/sec. Such grains are relatively immune to radiation pressure and, if produced in detectable numbers, would have extended the comae radii beyond the p50 observed during January through May. As noted in Section 3, all the program fragments observed in June and July exhibited tidally distorted comae or fans that were brighter on their leading side in the direction of Jupiter (see Fig. 4). The model indicates this brightness enhancement was due to the smaller 5 & R & 100 em grains present in the S–L 9 dust distribution as demonstrated by the colored tails in Fig. 8c for an R 5 100 em, V 5 0.3 m/sec dust model. This velocity was chosen for display purposes only for S–L 9 may have emitted dust at higher velocities which would simply have widened the tails. In May and all observation dates prior, simulated

COMET SHOEMAKER–LEVY 9 DUST

301

FIG. 8. (a) An R 5 500 em, V 5 1 m/sec dust tail simulation. The comet lies at the origin and the various plotting symbols indicate the location of each dust grain projected onto the x–z plane, where z points anti-sunward from the comet and the x direction parallels Jupiter’s orbit plane. The positive x axis is the approximate direction of west and the direction above the plane is approximately celestial north. Model dust tails are displayed for January 25, 1994 (yellow), May 18 (blue), and July 14 (red). In (a–d) the dated lines indicate the direction to Jupiter projected to the figure plane. (b) The view of the three tails in (a) as seen from Earth on the indicated dates, with the directions north and east given by the arrows. The orange tail is composed of R 5 0.5 em grains emitted at V 5 1 m/sec calculated for May 18. For dust smaller than a few micrometers, the near-nucleus dust tails project east–southeast of the comet while the more distant tail region (hidden by the blue tail) curves behind the comet to the west; the sense of this curvature is indicated by the white curve. (c) The colored dust tails are for R 5 100 em grains ejected from the nucleus at V 5 0.3 m/sec as seen from Earth on the indicated dates. The white curves passing through the tails are dust ejected at V 5 0, and a higher outflow velocity simply widens the tails. The tail on January 25 (not shown) appears similar to the May 18 tail. The dotted curves are the July 14 syndynes for dust having the indicated radii. (d) A ‘‘burst’’ model where all the dust grains are released during the 90 days following the comet’s breakup in July 1992. These R 5 1 cm grains have outflow velocities uniformly distributed between 0 and 1 m/sec.

302

HAHN, RETTIG, AND MUMMA

tails for dust radii 5 & R & 100 em lie west of the fragment. By late June the combined effects of radiation pressure and the increasing jovian tide cause the projected tails to rotate on the sky plane. By July the grains in the nearnucleus tail regions are confined to narrow tails with the dust most concentrated in the approximate direction of Jupiter. Superimposing the July tail in Fig. 8c over the large R 5 500 em dust simulation (Fig. 8b) indicates that the jovian side of the dust fan would be preferentially brighter than the trailing dust when there is a distribution of grains between 5 & R & 500 em. Prior to solar opposition, dust tails of all grain sizes extended west of the fragments with an orientation similar to that of the May 18 tail shown in Fig. 8c. But as the comet neared Jupiter the combined radiation pressure/ tidal force segregated the dust tails according to their grain size. This is illustrated by the dotted curves in Fig. 8c, which shows the orientations of various dust streams as seen from Earth on July 14. Each curve represents a stream of dust with the indicated radius ejected at zero velocity, i.e., zero-velocity syndynes. These syndynes also trace the tail axis for grains emitted at nonzero velocities. Comparison of these syndynes to the July 14 observation indicates that most of the light-reflecting dust must have been larger than 5 em since smaller grains would project too far into the dust-free region of space in Fig. 3d. This result is consistent with the May 18 image analysis which also shows an absence of dust smaller than about 5 em. In addition the July 14 syndynes indicate that the smaller 5 & R & 100 em grains would also have been spread out over a region to the north and northwest of the fragments by the combined radiation pressure and tidal force. This effect is evident in the July image, Fig. 3d, which shows a faint dust haze northwest of the comet fragments. To further strengthen the argument that S–L 9 was indeed an active comet, we have also investigated ‘‘burst’’ models which assume all the observed dust was created during or shortly after the comet’s tidal breakup in July 1992 and that the fragments were largely inactive ever since. In this scenario radiation pressure would have swept the comae of all grains smaller than about 1 cm during the final two-year orbit. Figure 8d shows the resulting coma on May 18, 1994 for a simulation of R 5 1 cm grains emitted only during the first three months since breakup. Not shown are any smaller grains that would contribute to a dust tail west of the fragment. While there is a wide range of grains sizes and emission velocities available for investigation, the common feature of all burst models are inner comae that remain highly elongated along the fragment train axis, a morphology that would persist throughout its entire orbit. We find no evidence of this elongation in the January–May images of the fragments’ inner comae (Figs. 2, 3a, 3b). Also, burst models do not produce surface brightness profiles that vary as r21 except under special

circumstances.3 Further, burst models do not produce elongated comae that are brighter in the jovian direction just prior to impact as seen in Fig. 4. While the burst scenario appears unable to account for most of the S–L 9 comae dust, we note that a post-breakup burst may have contributed to the debris seen in the interfragment region. To remain along the train axis during the final two-year orbit, this material is required to have been larger than about 1 cm since smaller grains would have been swept downstream into the dust sheet by radiation pressure (Fig. 2). From these comparisons of simulated dust comae/tails to the HST images, we conclude that S–L 9 continuously ejected dust of radii 5 & R & 500 em at an outflow velocity V p 1 m/sec. 5. DISCUSSION & SUMMARY

Comet Shoemaker–Levy 9 was imaged with the HST both before and after solar opposition on April 29, 1994. No bright dust tails were seen in the projected anti-sunward direction when observed after solar opposition as would have been expected had S–L 9 continuously produced dust smaller than about a micrometer. From the absence of any anti-sunward dust tails in these observations some investigators have concluded that S–L 9 was inactive. However this analysis shows the evolution of the S–L 9 dust comae and tails are well explained by the continuous production of dust larger than about 5 em. Between January and March 1994 the azimuthally averaged brightness profiles of S–L 9 varied as pr21.1 in the inner r & 10 comae and pr21.5 in the outer comae. The spatial dependence of these brightness profiles are consistent with S–L 9 as having been a continuously active comet. In contrast we find burst models, which assume that most of the observed dust was produced during or shortly after breakup, tend to produce comae that remain highly elongated along the fragment axis throughout its orbit. This type of coma morphology was not evident between January and May 1994, indicating that the observed dust was emitted long after breakup. Dust tail models indicate that the dominant lightreflecting grains in the S–L 9 comae had radii ranging between about 5 and 500 em emitted from the fragments at outflow velocities of order 1 m/sec. By early July the combined effects of radiation pressure and the jovian tide had distorted the dust tails into fans composed of larger R * 100 em grains oriented along the Jupiter–comet vector. The smaller R & 100 em grains occupied narrow dust tails whose near-nucleus regions were oriented ap3 For example, a burst producing N grains has a coma column density N/4 r Vt when viewed a time t later provided the dust outflow velocities v are uniformly distributed over an interval 0 # v # V.

303

COMET SHOEMAKER–LEVY 9 DUST

proximately in the projected direction of Jupiter. Since the smaller grains were most concentrated on the jovian side of the fragments’ dust fans, their contribution caused this region to appear brighter than the trailing side of the fans. There is a distinct absence of dust smaller than about 5 em as well as any grains emitted at velocities faster than about 3 m/sec. It is uncertain if these grains were truly absent from S–L 9 dust distribution, or, perhaps, that the small grains achieved high enough velocities such that their column density contribution was too small for detection. We also note that the observed range of dust sizes is significantly larger than the p1–10 em grains that are usually the dominant visible light scatterers found in most cometary comae (Fulle 1987, Gru¨n and Jessberger 1990). However the few comets orbiting beyond 5 AU that have had their grain size distribution determined, e.g., Schwassmann–Wachmann 1 (Fulle 1992) and Chiron (Fulle 1994), also exhibit a range of grain sizes similar to that of S–L 9. One can ask whether cometary gas emission at 5.4 AU was sufficiently vigorous to eject the observed dust grains. The largest grain that may be emitted by a comet is obtained by equating its surface gravity to the gas drag force, which yields Rmax 5 9CD Amu Zv/32fGRn rn rd , where CD 5 2 is the drag coefficient for free molecular flow, A is the gas molecular weight, mu is one atomic mass unit, Z is the gas production rate per area, v is the gas velocity at the surface, G is the gravitation constant, Rn is the nucleus radius, and rn , rd are the densities of the nucleus and dust (Gombosi et al. 1986). A recent thermal model of an Rn 5 1 km comet having Halley-like properties predicts a water production rate Z p O (1012 ) molecules/ cm2 /sec for a pure water-ice comet at S–L 9’s heliocentric distance (Weissman 1996). The gas velocity is v 5 Ï3kT/Amu 5 2.1/ ÏA km/sec, where k is the Boltzmann constant and T 5 170 K is the blackbody temperature at the subsolar point on a nonrotating comet. Setting rn p 0.5 gm/cm3 and rd p 1 gm/cm3, the largest grain lifted by outgassing is Rmax p O (1) em, indicating water production is too feeble to eject the large S–L 9 dust. However sublimation of more volatile species may be sufficient. Models of pure CO and CO2 ice comets yield Z p O (1017) molecules/cm2 /sec at the subsolar point on a nonrotating comet (D. C. Boice 1995, personal communication). These production rates should be regarded as upper limits for they are obtained from a comet model that treats the nucleus as a pure unmantled ice sphere, whereas real S–L 9 fragments may be composed of refractories and ices mixtures whose surface layers had steadily devolatilized since tidal breakup. While we acknowledge the need for a more realistic nucleus model, a plausible estimate may be obtained by multiplying Z by the volatiles’ fractional abundances. Since cometary abundances of CO and CO2 are of order a few percent (Mumma et al. 1993), Rmax is

of order a few millimeters and greater than the observed S–L 9 dust sizes. While it appears possible that a CO or CO2 outflow may account for the emission of large dust by S–L 9, the observed absence of micrometer-sized dust typical of many other comets is puzzling for such small grains are easily entrained in a gas flow. However it is also possible that we were seeing the signature of an entirely different dust production mechanism. It has been suggested that the S–L 9 fragments were not solid bodies but instead were selfgravitating swarms of cometesimals that formed after the 1992 tidal breakup event (Olson and Mumma 1994, Rettig et al. 1996b). The source of the observed dust production might have been due to the ejection of dust generated by colliding cometesimals. However it remains to be seen whether existing observations of Comet Shoemaker–Levy 9 can distinguish a gas-driven dust outflow from the emission of collisionally generated dust.

ACKNOWLEDGMENTS The authors thank Robert Williams, Director of STScI, who provided us with additional time to observe Comet Shoemaker–Levy 9, and Hal Weaver for sharing the S–L 9 data set. We also thank Paul Chodas and Don Yeomans for providing S–L 9 orbit data, Matt McMaster and Krista Rudloff for their advice on the image processing, Greg Sobczak for his assistance, and Anita Cochran and Paul Weissman for their conscientious reviews. Support for this work was provided by NASA through Grant 5624.21-93A from the Space Telescope Science Institute which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA Contract NAS5-26555. Support to M. J. Mumma was provided under RTOP 196-41-54 by the Planetary Astronomy Program of NASA’s Solar System Exploration Division.

REFERENCES ASPHAUG, E., AND W. BENZ 1994. Density of Comet Shoemaker–Levy 9 deduced by modeling breakup of the parent ‘‘rubble pile.’’ Nature 370, 120–123. BAUM, S. (Ed.) 1994. HST Data Handbook. STScI publication. BOSS, A. P. 1994. Tidal disruption of periodic comet Shoemaker–Levy 9 and a constraint on its mean density. Icarus 107, 422–426. BURROWS, C. J. (Ed.) 1994. Wide Field and Planetary Camera 2 Instrument Handbook. STScI publication. CHERNOVA, G. P., N. N. KISELEV, AND K. JOCKERS 1995. Imaging photometry and color of Comet Shoemaker–Levy 9. In European SL-9/Jupiter Workshop (R. West and H. Bo¨hnhardt, Eds.), pp. 11–16. ESO Conference and Workshop Proceedings No. 52. COCHRAN, A. L., A. L. WHIPPLE, P. J. MACQUEEN, P. J. SHELUS, R. W. WHITED, AND C. F. CLAVER 1994. Preimpact characterization of P/Comet Shoemaker–Levy 9. Icarus 112, 528–532. COLAS, F., L. JORDA, J. LECACHEUX, J. E. ARLOT, P. LAQUES, AND W. THUILLOT 1995. Pre-impact observations of Shoemaker–Levy 9 at Pic du Midi and Observatoire de Haute Provence. In European SL-9/Jupiter Workshop (R. West and H. Bo¨hnhardt, Eds.), pp. 23–28. ESO Conference and Workshop Proceedings No. 52.

304

HAHN, RETTIG, AND MUMMA

DONAT, W. 1993. The Astronomical Almanac. U.S. Government Printing Office, Washington, DC. FULLE, M. 1987. A new approach to the Finson–Probstein method of interpreting cometary dust tails. Astron. Astrophys 171, 327–335. FULLE, M. 1992. Dust from short period comet P/Schwassmann– Wachmann 1 and replenishment of the interplanetary dust cloud. Nature 359, 42–44. FULLE, M. 1994. Spin axis orientation of 2060 Chiron from dust coma modeling. Astron. Astrophys 282, 980–988. GOMBOSI, T. I., A. F. NAGY, AND T. E. CRAVENS 1986. Dust and neutral gas modeling of the inner atmospheres of comets. Rev. Geophys. 24, 667–700. GRU¨ N, E., AND E. JESSBERGER 1990. Dust. In Physics and Chemistry of Comets (W. F. HUEBNER, Ed.), pp. 113–175. Springer-Verlag, Berlin. HOLTZMAN, J. A., J. J. JESTER, S. CASERTANO, J. T. TRAVGER, A. M. WATSON, G. E. BALLESTER, C. J. BURROWS, J. T. CLARKE, D. CRISP, R. W. EVANS, J. S. GALLAGHER III, R. E. GRIFFITHS, J. G. HOESSEL, L. D. MATTHEWS, J. R. MOULD, P. A. SCOWEN, K. R. STAPELFELDT, AND J. A. WESTPHAL 1995. The performance and calibration of WFPC2 on the Hubble Space Telescope. Publ. Astron. Soc. Pac. 107, 156. JEWITT, D. 1995. Pre-impact observations of comet Shoemaker–Levy 9. In European SL-9/Jupiter Workshop (R. West and H. Bo¨hnhardt, Eds.), pp. 1–4. ESO Conference and Workshop Proceedings No. 52. JEWITT, D., AND J. CHEN 1994. Periodic Comet Shoemaker–Levy 9 (1993e). IAU Circular 5924. JEWITT, D. C., AND K. J. Meech 1987. Surface brightness profiles of 10 comets. Astrophys. J. 317, 992–1001. MARSDEN, B. G. 1993. Periodic Comet Shoemaker–Levy 9 (1993e). IAU Circular 5800. MUMMA, M. J., P. R. WEISSMAN, AND S. A. STERN 1993. In Protostars and Planets III (E. H. Levy and J. I. Lunine, Eds.), pp. 1177–1252. Univ. of Arizona Press, Tucson. OLSON, K. M., AND M. J. MUMMA 1994. Simulations of the breakup and dynamical evolution of Comet Shoemaker/Levy 9 employing a swarm model. Bull. Am. Astron. Soc. (4), 1574–1575. PRESS, W. H., B. B. FLANNERY, S. A. TEUKOLSKY, AND W. T. VETTERLING 1989. Numerical Recipes in C. Cambridge Univ. Press, Cambridge. RETTIG, T., J. HAHN, S. TEGLER, M. MUMMA, AND M. DISANTI 1994a. Periodic Comet Shoemaker–Levy 9 (1993e). IAU Circular 6028. RETTIG, T. W., J. HAHN, G. SOBCZAK, AND M. J. MUMMA 1995. Deep coma imaging of Comet Shoemaker–Levy 9 to characterize dust outflow velocities and grain sizes. In The Collision of Comet P/Shoemaker–Levy 9 and Jupiter (H. Weaver, Ed.), IAU Colloquium 156, p. 92. [Abstract] RETTIG, T. W., G. SOBCZAK, AND J. M. HAHN 1996a. Dust outflow velocity for Comet Shoemaker–Levy 9. Icarus 121, 281–290. RETTIG, T. W., M. J. MUMMA, G. J. SOBCZAK, J. M. HAHN, AND M. DISANTI 1996b. The nature of Comet Shoemaker–Levy 9 sub-nuclei from analysis of pre-impact HST images. J. Geophys. Res. Planets 101, 9271–9282.

SCOTTI, J. V. 1993a. Periodic Comet Shoemaker–Levy 9 (1993e). IAU Circular 5725. SCOTTI, J. V. 1993b. Periodic Comet Shoemaker–Levy 9 (1993e). IAU Circular 5745. SCOTTI, J. V., AND H. J. MELOSH 1993. Estimate of the size of Comet Shoemaker–Levy 9 from a tidal breakup model. Nature 365, 733–735. SEKANINA, Z., P. W. CHODAS, AND D. K. YEOMANS 1994. Tidal disruption and the appearance of periodic comet Shoemaker–Levy 9. Astron. Astrophys. 289, 607–636. SEKANINA, Z. 1995a. The splitting of the nucleus of Comet Shoemaker– Levy 9. In European SL-9/Jupiter Workshop (R. West and H. Bo¨hnhardt, Eds.), pp. 43–55. ESO Conference and Workshop Proceedings No. 52. SEKANINA, Z. 1995b. Evidence on sizes and fragmentation of the nuclei of Comet Shoemaker–Levy 9 from Hubble Space Telescope images. Astron. Astrophys. 304, 296–316. SMART, W. M. 1961. Celestial Mechanics. Wiley, New York. SOLEM, J. C. 1994. Density and size of Comet Shoemaker–Levy 9 deduced from a tidal breakup model. Nature 370, 349–351. STU¨ WE, J. A., R. SCHULZ, AND M. F. A’HEARN 1995. NTT Observations of Shoemaker–Levy 9. In European SL-9/Jupiter Workshop (R. West and H. Bo¨hnhardt, Eds.), pp. 17–22. ESO Conference and Workshop Proceedings No. 52. WALLACE, L. V., AND F. D. MILLER 1958. Isophote configurations for model comets. Astron. J. 63, 213–219. WEAVER, H. A. 1994. Periodic Comet Shoemaker–Levy 9 (1993e). IAU Circular 5973. WEAVER, H. A., M. F. A’HEARN, C. ARPIGNY, P. D. FELDMAN, P. LAMY, K. J. MEECH, K. S. NOLL, T. E. SMITH 1995a. Temporal and spatial variations among the SL9 fragments. Bull. Am. Astrom. Soc. 27, 1115. WEAVER, H. A., M. F. A. A’HEARN, C. ARPIGNY, D. C. BOICE, P. D. FELDMAN, S. M. LARSON, P. LAMY, D. H. LEVY, B. G. MARSDEN, K. J. MEECH, K. S. NOLL, J. V. SCOTTI, Z. SEKANINA, C. S. SHOEMAKER, E. M. SHOEMAKER, T. E. SMITH, S. A. STERN, A. D. STORRS, J. T. TRAUGER, D. K. YEOMANS, B. ZELLNER 1995b. Hubble Space Telescope (HST) observing campaign on Comet Shoemaker–Levy 9. Science 267, 1282–1287. WEAVER, H. A., P. D. FELDMAN, M. F. A’HEARN, C. ARPIGNY, R. A. BROWN, E. F. HELIN, D. H. LEVY, B. G. MARSDEN, K. J. MEECH, S. M. LARSON, K. S. NOLL, J. V. SCOTTI, Z. SEKANINA, C. S. SHOEMAKER, E. M. SHOEMAKER, T. E. SMITH, A. D. STORRS, D. K. YEOMANS, AND B. ZELLNER 1994. Hubble Space Telescope Observations of Comet P/ Shoemaker–Levy 9 (1993e). Science 263, 787–791. WEISSMAN, P. R. 1996. If it quacks like a comet. . . . Icarus 121, 275–280. WEST, R. M., R. N. HOOK, AND O. HAINAUT 1995. A morphological study of SL-9 CCD images obtained at La Silla (July 1–15, 1994). In European SL-9/Jupiter Workshop (R. West and H. Bo¨hnhardt, Eds.), pp. 5–10. ESO Conference and Workshop Proceedings No. 52.