the deep impact oblique impact cratering experiment · 2013. 7. 25. · deep impact experiment 297...

Icarus 190 (2007) 295–333www.elsevier.com/locate/icarus

The Deep Impact oblique impact cratering experiment

Peter H. Schultz a,∗, Clara A. Eberhardy a, Carolyn M. Ernst a, Michael F. A’Hearn b,Jessica M. Sunshine b, Carey M. Lisse c

a Department of Geological Sciences, Brown University, Providence, RI 02912, USAb Astronomy Department, University of Maryland, College Park, MD 20742, USA

c Planetary Exploration Group, Space Department, Johns Hopkins University Applied Physics Laboratory, Laurel, MD 20723, USA

Received 19 July 2006; revised 12 June 2007

Available online 19 July 2007

Abstract

The Deep Impact probe collided with 9P Tempel 1 at an angle of about 30◦ from the horizontal. This impact angle produced an evolvingejecta flow field very similar to much smaller scale oblique-impact experiments in porous particulate targets in the laboratory. Similar featuresand phenomena include a decoupled vapor/dust plume at the earliest times, a pronounced downrange bias of the ejecta, an uprange “zone ofavoidance” (ZoA), heart-shaped ejecta ray system (cardioid pattern), and a conical (but asymmetric) ejecta curtain. Departures from nominalcratering evolution, however, provide clues on the nature of the impact target. These departures include: fainter than expected flash at first contact,delayed emergence of the self-luminous vapor/dust plume, uprange-directed plume, narrow early-time uprange ray followed by a late-stageuprange plume, persistence of ejecta asymmetries (and the uprange ZoA) throughout the approach sequence, emergence of a downrange ZoAat late times, detachment of uprange curved rays, very long lasting non-radial ejecta rays, and high-angle ejecta plume lasting over the entireencounter. The first second of crater formation most closely resembles the consequences of a highly porous target, while later evolution indicatesthat the target may be layered as well. Experiments also reveal that impacts into highly porous targets produce a vapor/dust plume directed back upthe incoming trajectory. This uprange plume is attributed to cavitation within a narrow penetration funnel. The observed lateral expansion speedof the initial vapor plume downrange provides an estimate for the total vaporized mass equal to ∼5mp (projectile masses) of water ice or 6mp ofCO2. The downrange plume speed is consistent with the gas expansion added to the downrange horizontal component of the DI probe. Based onhigh-speed spectroscopy of experimental impacts, the observed delay in brightening of the DI plume may be the result of delayed condensationof carbon, in addition to silicates. Observed molecular species in the initial self-luminous vapor plume likely represent recombination productsfrom completely dissociated target materials. The crater produced by the impact can be estimated from Earth-based observations of total ejectedmass to be 130–220 m in diameter. This size range is consistent with a 220 m-diameter circular feature at the point of impact visible in highlyprocessed, deconvolved HRI images. The final crater, however, may resemble an inverted sombrero-hat, with a deep central pit surrounded by ashallow excavation crater. Excavated distal material observed from the Earth was likely from the upper few meters contrasted with ballistic ejectaobserved from the DI flyby, which included deep materials (10–30 m) within the diffuse plume above the crater and shallower (5–10 m) materialswithin the ejecta curtain.© 2007 Elsevier Inc. All rights reserved.

Keywords: Deep Impact; Vaporization; Ejecta; Oblique; AVGR; Comets; Experiments

1. Introduction

The Deep Impact (DI) collision was oblique: between 25◦and 35◦ from the surface horizontal as established by the shapeof nearby craters and location on the surface, shape models,

* Corresponding author.E-mail address: [email protected] (P.H. Schultz).

0019-1035/$ – see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.icarus.2007.06.006

and reconstruction of the intercept point (A’Hearn et al., 2005).The oblique trajectory also is clearly expressed by the evolutionof the initial plume, ejecta curtain asymmetry, and later curvedejecta ray systems.

Laboratory experiments provide fundamental data on the dy-namic response to materials that might be similar to a cometarysurface (e.g., Arakawa, 1999; Burchel and Johnson, 2005).While critical for crater scaling, such experiments do not allow

http://www.elsevier.com/locate/icarusmailto:[email protected]://dx.doi.org/10.1016/j.icarus.2007.06.006

296 P.H. Schultz et al. / Icarus 190 (2007) 295–333

comparisons with the actual observations of the growing cratermade by Deep Impact. The effects of strain rate, spallation,fragment size, impactor state (fragmented vs unfragmented),and post-formation collapse all affect how results from suchstudies can be reasonably extrapolated.

Experiments dealing with scaling and ejecta evolution of-ten require surrogate targets for describing the process of late-stage excavation. For example, impacts into sand targets in thelaboratory are used to trace the evolution of ejecta under thecontrol of gravity for impacts into loose particulates, e.g., re-goliths (e.g., Gault et al., 1968; Schmidt and Housen, 1987).This strategy also has been applied to much larger scales withthe assumption that prior passage of a shock with high peakpressures would have turned a solid target into a strengthless(low-strength) state. Such experiments help to establish the roleof gravity, strength, or compression on the final crater size aswell as ejecta-velocity distributions.

More recently, low-strength targets (sand) have been usedto track the evolving flow field in the gravity-controlled ejectadue to vertical and oblique impacts (Anderson et al., 2004). Inthis case, shock asymmetries and energy losses (compaction,comminution) of porous particulates are manifested in the dis-tribution of ejecta velocities (speed and direction). The flowfield within the crater translates to ballistics above and providesa measure of the coupling process and the transition to grav-ity or strength-controlled processes (e.g., Housen et al., 1983,1999). Hence, the final transient crater can be linked to the ob-served ballistic ejecta.

Excavation ceases once material motions no longer allowejecta to escape the crater rim due to either gravity (grav-ity scaling) or strength, which decelerates crater growth be-fore the gravity limit is reached (see Holsapple and Schmidt,1987; Schmidt and Housen, 1987). The limiting strength hasbeen described in a variety of ways: the angle of repose, cohe-sion, internal angle of friction, bonding forces, yield strength,among others. When the overburden (proportional to density ×gravity × projectile size) falls below the chosen measure ofstrength, the strength-limited cratering regime dominates. Nu-merous processes, however, contribute to target strength espe-cially as material motions slow on bodies with low gravity,including the potential role of inter-particle frictional drag. Ad-equately describing the conditions in the target leading to thetransition to strength-controlled growth remains a major fo-cus area for research, especially in terms of scaling to largersizes.

Targets with very high porosity (low density and compress-ibility) but low strength can result in the late stages of excava-tion encroaching into the initial stages of shock coupling (e.g.,Housen et al., 1999; Schultz et al., 2005). In such experiments,the exact match in target composition is not as important as thedepth of coupling and the physical response to the rarefactionwave off the free surface. Experiments using vertical impactsfound that crater diameter in ice targets remained unchangedwith increasing porosity, whereas depth increased (Burchel andJohnson, 2005; Schultz et al., 2005). Other experiments usinghigh-porosity silicates found that little ejecta may escape due tothe dominant role of compression (Housen et al., 1999). Vertical

impacts into similar highly porous and compressible particu-lates (but with low strength) showed that the transient crater col-lapsed due to the unstable transient profile (Schultz et al., 2005).In contrast with sand-like targets, however, decreasing the im-pact angle actually increased cratering efficiency (ejected mass,crater size) with distinctive changes in the ejecta distribution.This paradoxical result was attributed to the shallower effectivedepth of coupling and more stable transient crater profile. Con-sequently, it is important to recognize diagnostic signatures ofthis type of target before applying scaling relations.

The Deep Impact experiment represents a different andunique challenge. Without being able to see clearly the finalcrater inferences about properties of the target require com-paring the ballistic ejecta with analytical models for grav-ity and strength-controlled cratering (Richardson et al., 2005;Schultz et al., 2005; Richardson et al., 2007). But the appropri-ate excavation model needs to be identified first. Consequently,the strategy for the present contribution is to first describe eachcritical stage of cratering observed for the DI collision (initialcoupling, late-stage ejection) and compare them with a widerange of laboratory experiments designed to isolate these stagesand how they are expressed by the ejecta evolution. Because ofthe low impact angle for DI, changing styles of ejecta with timeare mapped out spatially by the ballistic ejecta. Such chang-ing styles provide qualitative but critical clues for scaling in-cluding initial coupling (plume evolution, shallow versus deepcoupling) and excavation stages (symmetric versus asymmetric,non-radial rays).

In the following discussion, we first review the general evo-lution of the event; next, describe hypervelocity impact experi-ments that can be used to interpret what happened at the comet;and last, interpret the observations of the DI collision throughcomparisons of common features.

2. The DI impact cratering experiment

Several definitions of terms are useful for the followingdescriptions of impact. For this study, the plume is an ex-panding diffuse cloud of gas and dust. A vapor plume is de-fined as a self-luminous vapor and particulate cloud (radiatingatomic/molecular and thermal emissions). The plume front cor-responds to leading edge of the plume containing the greatestnumber density of gas/particulates. The ejecta curtain repre-sents the loci of ballistic ejecta forming a relatively well definedsheet of debris. Rays are loci of ejecta following similar ballis-tic paths from a common source region but at different timesand velocities (speed, angle, and azimuth). The plane tangentto the point of impact is abbreviated as TPI (“tangent plane atimpact”), whereas the projected trajectory on this tangent planeis termed the “trajectory track line” or TTL.

For convenience, the events of the DI collision can be di-vided into three stages: early-time (the first second), later time(from 1 to 10 s), late time (10–600 s), and closest approach(600–800 s). These times refer only to observation times af-ter the first contact with the surface as seen from the DI flyby.Laboratory experiments have the benefit of observing impactsclose-up as the ejecta leave the crater. The DI experiment, how-

Deep Impact experiment 297

(a)

(b)

Fig. 1. (a) Composite ITS images of Comet 9P Tempel 1 with inset showing the general region of impact. (b) Preliminary geologic/terrain map of impact regionwith interpreted cross-section. The point of impact appears to be in an older exhumed surface with muted hummocky relief, as indicated by boundary scarp. A lowerlevel hummocky surface occurs to the left of a boundary scarp and is interpreted as a more recently exhumed surface with numerous darker and brighter hillocks.Dashed arrows indicate downrange trajectories from early ejecta and the possible interference with topography.

ever, observed ejecta at large distances (km’s) from the crater;this requires corrections in order to make detailed comparisonsof actual stages of growth. The following section first examinesthe impact site, details the evolution of the event, and then ex-amines its effect on the surface, including estimates for cratersize.

2.1. Impact site

The ITS (impactor targeting sensor) camera provided close-up views of the impact site that proved critical for understand-ing the terrain and geology of the impact site. Fig. 1a shows ageneral context for the impact region, whereas Fig. 1b providesa preliminary geologic map and interpreted cross-section of the

DI impact site. It is based in part on the general geologic mapof the comet as described by Thomas et al. (2007). The impactsite occurs on a relatively flat surface with subtle relief that isinterpreted as an exhumed surface. Possible relics of past layersor surfaces (B, D) appear to emerge at lower elevations and canbe inferred to extend below the impact site. This interpretationis also based on the darker knobs and mesas (e.g., D) protrud-ing through the terrain B (D′) and the sequence of boundaryscarps. The fine-scale roughness of surface B has numerous,small lighter patches. This pattern contrasts with the mutedbumps on surface A and is consistent with an interpretationthat materials drape and underlying topography of surface B.The darker circular structure (C) can be reasonably interpretedas a degraded (filled in) impact crater due to its raised rim. If


(a) (b)

Fig. 2. (a) Context HRI image (deconvolved) showing the surface “trajectory track line” (TTL) defined as the probe trajectory projected on a plane tangent to thesurface at the point of impact. The determination of the impact point is based on imaging by the probe as well as MRI imaging of the first moment of impact.The regions downrange and uprange from the point of impact are identified. (b) Reconstructed perspective of the comet from the shape model at the moment ofimpact. Rectangle represents the “tangent plane at impact” (TPI). Other arrows indicate orientations with respect to the sub-Earth point (green), rotation axis (blue),sub-solar point (white), ecliptic north (yellow), celestial north (pale blue/green).

this is the case, then the impact site also might have a layer ofejecta 0.2 to 0.5 m thick just below the surface. The boundaryscarp between surface A and B circles around one side of therim of crater C, resulting in darker rim material extending togreater distances in surface A. This is consistent with exhumedrim material in B, still covered below A.

Regardless of the specific interpretation of the units and ter-rains, there appears to be physical contrasts between each layerin order to produce the well-defined levels and scarps. In ad-dition, the numerous small “brighter” patches (still quite dark)occur as diffuse areas (E). These are more common in the ex-humed surface at the lower elevation, relative to the surfacewhere DI hit. The brighter patches also occur along the baseof the boundary scarps. Although speculative, it is assumedthat there is also an underlying unit (F); this possibility willbe explored through comparisons of laboratory and DI ejectapatterns.

Consequently, the DI impact site consists of lateral and verti-cal variations and contrasts. The thermal inertia data (Groussinet al., 2006) also indicate a surface covered nearly everywherewith very fine powder, at least to the depth of the thermal wave.This result does not require that the material is strengthlessbut does indicate that it resembles a regolith that characterizesmost small bodies observed to date. The inferred layering alsoindicates that there may be weak interfaces between units, inorder to produce such well-defined contacts and scarps (e.g.,between A and B). The increased fine-scale surface roughnessand the numerous brighter diffuse patches indicate that the im-pact site is inhomogeneous at meter scales, both vertically andlaterally.

2.2. Early-time processes (first second)

2.2.1. Self-luminous plumeFig. 2 establishes a useful reference for understanding the

event in terms of Sun direction, location, and impact trajectory.The early-time flash and vapor plume rapidly evolve along thetrajectory (A’Hearn et al., 2005):

1) An initial faint “first light” (FL) uprange from the projectedpoint of impact.

2) Fading source along the trajectory that moves downrange(∼100–170 m) over the next 0.125 s after impact (AI).

3) Gradual brightening over the next 0.062 s.4) A sudden “flash” (saturated pixels) around 0.25 s after the

FL.

Over the next 0.12 s, a vapor/dust plume emerges and trav-els downrange with the leading front at a speed of ∼5.8 km s−1(projected on the celestial background) and evolves into acrescent-shaped, self-luminous leading edge (the plume front).This front also expands laterally (with respect to the trajectory)at a speed of ∼2.8 km s−1 and gradually fades. The vapor/dustplume rapidly becomes optically thin (within the first 6 framesafter FL). Other contributions detail the evolving plume pho-tometry (Ernst et al., 2006; Ernst and Schultz, 2007) and theevolving physics (Melosh, 2006).

Fig. 3 shows the sequence of the first 19 frames from FL to1.16 s later. A reference line corresponding to the trajectory pro-jected onto the tangent plane at the point of impact (termed herethe trajectory track line, TTL). The boundary of the downrangeplume front and the shadow across the surface are sketched as


Fig. 3. Medium Resolution Imager (MRI) frames 900063 to 900082 showing evolution of the first 1.18 s after impact. The dotted line corresponds to the TTL(trajectory track line) defined with Fig. 2b; direction of solar illumination is noted in the lower right. Dashed line highlights the changing azimuth of the uprangeboundary of the shadow cast by the optically thick ejecta cloud. Crescent-shaped dotted line traces the evolving downrange plume of vapor and dust. Scale bar inlower left is 1 km.

well. Fig. 3 also illustrates the rapidly changing azimuth angleof the shadow (referenced to the TTL) and the leading plumefront. The shadow is initially directed downrange (acute angleto the downrange TTL); it then becomes more normal to thetrajectory.

2.2.2. Early-time shadowsThe uprange plume is difficult to discern from DI imaging

(Fig. 3) due to its low mass, diffuseness, and foreshorteningas seen from the flyby. Its shadow, however, provides impor-tant information. In Fig. 4a, the evolution of the shadow castby the emerging opaque ejecta plume has been reoriented toresemble a laboratory experiment (as discussed in a later sec-tion). While minimum ejection angles can be calculated fromthe shape model of Tempel 1, this exercise is deferred to futurecontributions. Here the shadow provides qualitative evidencefor an initial plume dominated by downrange-directed debrisfrom the impact.

The deconvolved HRI sequence (Fig. 4b) does not have thetime resolution of the MRI (Fig. 4a) but does reveal more struc-ture in the time-averaged evolution of the opaque portions ofthe plume. In this case, the high-angle plume shadow domi-nates the low-angle, downrange component in Fig. 3 over thefirst 0.7 s. About 2 s later, the shadow has two components:a narrow central plume and an arcuate conical (or spherical)plume that appear to converge downrange. These two compo-nents could represent two superimposed stages due to the longtime exposure.

2.2.3. Image differencingSubtraction of one image at one time from another image at

an earlier time (called image differencing) clearly accentuatesrelative motions, as has been done in laboratory experiments(Schultz et al., 1985). Image differences (earlier frame sub-tracted from a later frame) accentuate both the motion and rela-tive changes in brightness (Figs. 4 and 5). The shadow evolvesinto two components at around 0.75 s: a curved boundary ex-tends around Tempel 1 into the shadow (effect of nucleus shape)and a well-defined shadow edge of the ejecta plume acute to theuprange trajectory. This evolution of the shadow indicates thatthe ejecta plume emerging from the surface is initially directeddownrange at a low angle, rapidly increases towards vertical,then evolves into a well-defined debris cloud inclined to thesurface uprange, similar to the early stages of an uprange ejectacurtain (as discussed in more detail below).

Frame differencing reveals several additional phenomena(Figs. 5 and 6). First, the axis of symmetry of the downrangeplume occurs 15◦ beyond the downrange TTL (180◦). Second,the plume front is distorted, i.e., offset to one side (∼150◦ fromTTL). Third, the differencing clearly reveals debris directed up-range along the TTL (0◦). Difference frames 70–69 through72–71 exhibit relative brightening and broadening of lighterpixels uprange (but offset from the TTL), thereby indicatingexpansion (Fig. 5).

Consequently, there was initially a cloud (or vapor plume)directed uprange during the first 0.5 s. The viewpoint from theflyby relative to incoming trajectory requires that any visibleuprange plume must be ejected back along the trajectory of the


(a) (b)

Fig. 4. (a) Evolution of the shadow cast by the emerging ejecta cloud (solid lines) during the first second (see Fig. 3). Image has been re-oriented so that theobserver is looking downrange on the tangent plane at the point of impact, similar to a perspective in a laboratory experiment. Arrow depicts solar illuminationsource direction; dotted lines represent the trajectory axis and direction orthogonal to it. The angle between the source of illumination and the curtain shadow on thesurface is termed here the “shadow emergence angle” (SEA). The initial shadow (labeled 1, frame 900069) indicates an opaque downrange cloud initially emergingat low angle with respect to the surface but rapidly increasing in angle near the base (labels 2–4 covering the first 0.25 s). See Fig. 3 for scale. (b) Evolution ofshadow imaged by the first two HRI frames (exposure times of 0.72 s). Shadow suggests a lower initial angle (above) evolving into two components in the nextframe (below). The second image has a central plume (CP) shadow that is parallel to the direction of illumination, thereby indicating a nearly vertical plume. It issurrounded by an arcuate shadow reminiscent of an expanding spherical plume (see Schultz, 1996) rising above the impact. See Fig. 2a for scale.

DI probe. The fact that this component is displaced slightlyaway from the impact-flyby direction also indicates that it isrising above the surface (Fig. 6B). Moreover, the sequence ofinitial brightening (frame 64), fading (65 and 66), and thenbrightening is consistent with an uprange self-luminous plumethat disperses (becoming optically thin) and is followed bylight-scattering particulates.

The emergence of the vapor/dust plume shown in Figs. 3and 5 creates a paradox. Five frames after first light (∼0.31 ssince impact), the leading edge of the vapor plume traveleddownrange about 1 km at a speed of 5.8 km s−1 (Fig. 3). Ne-glecting projection effects, this position and speed would indi-cate that the plume emerged from the surface 0.17 s after thefifth frame, or about 0.14 s after first contact. Consequently,the position and derived speed of the downrange self-luminousplume (both downrange and lateral) seem to require that thisphase must have emerged from the comet well after the timeof first contact. Either the high-speed vapor phase was some-how delayed from leaving the target, or the plume had rapidlyexpanded at first but slowed by the time it became visible in

the fourth frame. Laboratory impact experiments will be usedto assess these alternatives.

By 0.5 s after FL, the region of impact becomes progres-sively darker, thereby producing a dark pixel in the differenceframes. This observation is shown more clearly in Fig. 7 tolater times. It is clear that the successively darker frames occuroffset toward the direction of the Sun from the TTL. Conse-quently, this offset can be attributed to a shadow (photometri-cally darker) region being formed on the inside of the sunwardside of the ejecta cloud. Alternatively, a diffuse ejecta cloud isexpanding above the impact and is scattering progressively lesslight as it becomes optically thin.

2.3. Later time (1–10 s)

Over the next 9 s, the fast downrange plume decouples fromthe subsequent excavation stage of cratering and expands outof the field of view. A diffuse plume of fine ejecta (scatteredlight) expands above the impact while radial rays emerge. Up-range, both a highly foreshortened fan-shaped ejecta segmentand a shorter, narrower (collimated) ejecta cloud become evi-


Fig. 5. Image differences of successive MRI frames that highlight motion of the ejecta with time (MRI frames 900063 to 900075). Brighter areas indicate regionsthat have changed; darker regions correspond to case where preceding image was brighter. Dotted line represents the initial trajectory (TTL). Arrows in 900065minus 900064 and 900066 minus 900065 point to an uprange cloud; the offset from the TTL is consistent with an evolving plume directed back toward the incomingtrajectory but at a higher angle. Dash/dot line identifies the axis of symmetry for the downrange plume, which appears offset from the TTL. Rays in 900067 minus900066 are due to struts supporting secondary optics in the telescope. Scale bar in upper right corresponds to 1 km.

Fig. 6. Closer view of initial stages of the downrange vapor plume shown in Fig. 5 (difference frames 900070 minus 900069 and 900071 minus 900070). (A) Bright-ening uprange (900070–900069) is interpreted as an uprange plume directed back along the incoming trajectory. Square represents impact plane tangent to thesurface at the point of impact, which is determined by the ITS imaging and first light from the impact. (B) Brighter area to upper right in 71–70 could be the resultof solar reflection off a higher angle uprange ejecta cloud. Arrow identifies a kink in the downrange plume that may be the result of interference with topography.Solid outline corresponds to the boundary of the comet; dashed line indicates terminator. Dashed and dotted line represents the axis of symmetry for the downrangecrescent-shaped plume and shows that it is offset from the TTL based on the shape model.

dent (see stereo view in Fig. 8). The fan-shaped ejecta segmentextends out at an angle less than the flyby-observing angle,i.e., comparable to the initial trajectory. The more collimatedplume grows more slowly than the downrange and later ejectaplumes.

Stereo views of the ejecta provide a qualitative assessment ofthe ejecta patterns (Fig. 8). In this case, the motion of the ejectaproduces the necessary displacement for the perspective shift.The most significant insight from the stereo pairs is the con-

firmation that the uprange fan is highly foreshortened, therebyindicating that this represents ejecta directed back in the di-rection of the impactor approach trajectory. Along a given ray,the direction of ejection (azimuth) changes with time (Figs. 9aand 9b). Otherwise, rays would simply radiate from the impactpoint. Along a given uprange ray, the azimuth of the highestspeed ejecta (end of a ray) is initially directed away from theinitial impactor trajectory but evolves toward the uprange direc-tion with time. The sense of curvature (convex uprange) can be


Fig. 7. Difference frames from the MRI for later stage evolution (900072 through 900088). Dotted line corresponds to the projected trajectory. Scale bar in upperleft is 1 km.

Fig. 8. Pseudo-stereo view created by motion of ejecta 4.5 s (900926, left) and 2 s (9009023, right) after impact (left image for right eye; right, for left). Stereoreveals uprange rays, separated from the rest of the ejecta and directed back toward the flyby close to the incoming trajectory. Orientation has been changed tomaximize stereo effect.

attributed to slower initial speeds or reduced amounts of ejecta(both attributable to early-stage compression).

2.4. Late time (10–600 s)

Systems of distinct rays clearly emerge after ∼10 s (seeFig. 9). Perspective, shadows, and exposure time (and filters)all play key roles in interpreting this part of the evolution. Per-spective is important because the observed patterns representan asymmetric and incomplete ejecta cone that is highly fore-shortened (i.e., projected on the celestial background). Shadowscontribute to any interpretations because they not only are castover the surface but also project onto the interior portions of theejecta curtain. Finally, exposure time (and filters) affect the vis-ibility of distal (i.e., far from the impact) versus proximal (nearthe impact) ejecta.

At this time, ejecta form three distinct groups (Fig. 9): hemi-spherical cloud, uprange fan, and distinct ray systems. The

hemispherical cloud appears to radiate about the downrangeTTL. But because of the view angle, this system is highly fore-shortened. An uprange fan subtends an angle of ∼115◦. Be-cause of foreshortening, however, this fan more likely subtendsan angle less than 30◦ and is directed back towards the flyby.Two pronounced curving rays uprange appear to form a bound-ary for the foreshortened fan of ejecta. Many of the distinct raysare curved; consequently, extrapolations of the distal compo-nents do not converge to the same point on the surface. Forexample, the distal portions of rays in Fig. 9 appear to convergefarther uprange, whereas ray segments closer to the surface con-verge close to the point of impact. This most likely reflects thechange in azimuth of ejection as illustrated in Fig. 9.

The most significant insight from Fig. 9 demonstrates thatthe uprange fan is highly foreshortened. As viewed from thesurface, the DI flyby would appear 3.3◦ above the incoming DIprobe. This means that an ejecta ray directed back up the initial


(a) (b)

Fig. 9. (a) Close view of radial and arcuate ray system emerging from the impact point 20 s after impact (MRI frame 900934). The dark portion to one side of theTTL is consistent with shadowing. Based on the solar illumination, the shadow is from a diffuse cloud extending above the impact point cast onto ejecta. (b) Arcuaterays emanate from the impact point. Arrows indicate trajectories with the assumption that they must originate from near the impact site. Along a given ray, highestspeed ejecta (distal portions) uprange from a large zone of avoidance but subsequently become more radially symmetric. Ray curvature may be related to surfaceproperties.

Fig. 10. Evolution of ejecta rays from 12 to 476 s after impact. After 100 s, the pronounced radial rays and overall haze lessens around the impact point (Figs. 8and 9). A well-defined uprange boundary for the ejecta emerges (e.g., 195 s). By around 200 s after impact, the ejecta form a distinctive heart-shaped boundaryuprange. The uprange arcuate rays 195 s after impact appear to have detached from the rest of the system. Additionally, a downrange gap begins to appear. SeeFig. 2a for scale.

trajectory would appear to translate a small distance across thesurface of Tempel 1, just due to the small offset. For example,ejecta launched at 500 m s−1 back up the trajectory over 20 sabove the surface would appear to move only 0.58 km uprangefrom the point of impact. Uprange rays in Fig. 9 show little dis-placement uprange from 2 to 20 s, whereas lateral rays haveextended three times farther. Consequently, it appears that one

component of ejecta was directed back up the impact initial tra-jectory (toward the approaching flyby) during the early stagesof excavation (first 30 s). As viewed by the DI flyby, this cre-ated the diffuse system of rays in the foreground that graduallydissipated (became optically thin).

Rays recognized soon after impact can be seen nearly 60 slater (Fig. 10). The system of curved rays becomes more pro-


Fig. 11. Details of distal ray system 195 s after impact (MRI 900959). The 282.5 s exposure time masks detail close to the impact site but reveals ballistic raysdeveloped early in the process, now far from the comet. The uprange system (upper left) is now in the foreground and has become diffuse. Rays identified in Fig. 9but are now “detached” from the surface. Inset frames correspond to about the same time (209 s after impact, 900960) but exposed for only 50.5 s to capture detailsnear the comet. Upper inset has been contrast stretched to emphasize the well-defined conical boundary of the ejecta uprange the development of a dark ray or gapin the ejecta downrange (arrows). The lower inset is the same MRI image but stretched to optimize detail near the comet. See Fig. 2a for scale.

nounced with time as ejecta travel farther from the impact andas the view moves away from the initial trajectory line. The ini-tial uprange ray system noted in Fig. 9 has decoupled from innerejecta after ∼50 s and has become less well defined (Fig. 10).Left behind is a distinctive uprange “Zone of Avoidance” (ZoA)that persists throughout the rest of the approach. The possiblesignificance of this decoupling will be addressed in a later sec-tion. Lower speed ejecta comprise a new set of curved uprangerays, however, persisting to very late times (∼500 s AI). Thecurved rays are convex outward (with respect to the trajectory)both uprange and downrange, forming a spider-like, cardioid(heart-shaped) pattern.

Longer exposure times (approximately ∼200 s after impact)reveal the cardioid pattern more clearly (Fig. 11). At this time,the initial uprange ray system is decoupled from the comet (sus-pended arcuate rays noted by B and B′). Such decoupled raysprovide clues for the nature of the surface (as discussed below).The more prominent open, fan-shaped ejecta pattern convergeson the point of impact (see inset). Additionally, a downrangegap in the ejecta begins to widen (also in the inset).

Fig. 12 provides closer views of the impact region from 146to 562 s and reveals that the uprange curved rays (arrow 1) ex-tending to the impact point remain visible to at least 200 s afterimpact. The uprange, fan-shaped limit of ejecta (ZoA, arrow 2,Fig. 12D) subtends an increasing angle with time. This is due, inpart, to the changing view angle of the flyby (higher above thecrater). On the side of the TTL away from the Sun, the ejecta ap-pear slightly darker (arrow 3), which is interpreted as a shadowcast on the interior of the growing ejecta cone. If this is thecase, however, the obscuration must be a diffuse ejecta cloud(low speeds and high-angle trajectories) directly above the im-pact point due to solar elevation angle. A downrange gap (ZoA)

in the ejecta (arrow 4) subtends an increasing angle. Becausethe flyby views the excavation from the uprange direction, theshadow cast across the surface is evident (arrow 5) and extendsback to near the impact point. Two uprange rays (arrow 6) be-come evident at around 300 s (Fig. 12D) and clearly emerge by500 s (Fig. 12F). It is unclear if these uprange rays are newlyforming or just becoming more evident as the spacecraft view-point becomes more orthogonal to the TTL on the surface.

2.5. Closest approach (600–800 s)

From 200 s to the end of the approach imaging sequence,the broad ejecta cone (subtending an angle of 165◦) appears toindicate very low angle ejecta with respect to the surface hori-zontal. Stereo imaging (Fig. 13) and laboratory simulations (inlater sections) reveal, however, that this appearance likely re-sults from the combined effects of the evolving uprange ZoAand the flyby view angle.

As the flyby approached the comet, well-defined rays con-tinue to extend to the surface (Fig. 14). Uprange, the ZoA re-mains pronounced with two well-defined rays on either side ofthe trajectory and subtends 160◦, with fainter curved uprangerays subtending ∼60◦ (also Fig. 12F). A diffuse, narrow rayextends from the converging curtain base uprange; the offsetin stereo imaging for this component indicates that this mate-rial remains ballistic. Downrange, the ZoA widens and formsa “butterfly-like” pattern. The sunward-side ejecta curtain castsan obvious shadow on the interior of the opposite side as re-vealed in stereo imaging. Additionally, a high-angle plume stillemerging from crater casts a more opaque shadow across thecurtain. Fig. 15 labels key features.


Fig. 12. Detailed MRI views showing disappearance of uprange curved ray near the impact and emergence of uprange ray. Times correspond to the following:(A) 146 s (900952); (B) 209 s (900960); (C) 247 s (900965); (D) 327 s (900977); (E) 378 s (900984); (F) 562 s (901003). This sequence indicates that the anglesubtended by the uprange rays appears to gets larger with time.

Fig. 13. Pseudo-stereo view of rays mapped on well-defined rays for MRI frames 901041 (left) and 901019 (right) well after impact (∼700 s). Motion of spacecraftcreates different perspectives enabling stereoscopic views (in this case, reverse stereo where left image is for right eye; right image, for left). At this stage of growth,the persistence of rays and low ejection speeds permit viewing in stereo. This view clearly shows that the rays form a significant angle with respect to the surface atthe impact point (∼40◦–50◦). The uprange zone of avoidance (ZoA) to the left becomes more evident. See Fig. 2a for scale.

2.6. Constraints on crater size

Prior to the DI encounter, a range of possible cratering sce-narios was developed in order to define limits on the durationof imaging and resolution required to resolve the final crater(Schultz et al., 2005). The nominal model was based on gravity-scaled growth, but other possibilities included crater scalingbased on a range of models for surface materials: strength con-trolled, gravity-controlled, and compression controlled. Thesedifferent models reflected more the uncertainties in the prop-erties of the comet (bulk density, gravity, near-surface density,strength, porosity) as much as cratering physics.

The DI collision produced a crater with ejecta that evolvedthrough the entire duration of the encounter, from approach tolook back. Nevertheless, the crater could not be clearly seenthrough slow-moving ejecta above the crater during the first800 s of growth (A’Hearn et al., 2005). While disappointing,this observation can provide important clues for the nature ofthe upper 30–50 m of Comet 9P/Tempel 1. The following dis-cussion estimates the size of the final crater through differentapproaches. Discussion on possible causes for not being ableto see the final crater is deferred until after discussing relevantexperiments.


Fig. 14. Stereo view of rays mapped on well-defined rays for MRI frames 901064 (left) and 901054 (right) well after impact (∼700 s+). The flyby now viewsthe ballistic ejecta more perpendicular to the initial impact trajectory and reveals the asymmetry in the distribution. The stereo view also shows that the dark “ray”extending to the lower right is actually a shadow cast on the Sun-facing interior of the ejecta curtain by the diffuse plume rising above the impact point. Shown is areverse stereo: left image is for right eye; right image, for left.

Fig. 15. Shape model (A) compared with MRI frame 901059 (784 s after impact), (B) labeled with key features in order to aid in interpretations. Foreground (black)is distinguished from the background (white). Surface normal at the impact point corresponds to the convergence of the dark fan, which is interpreted as a shadowon the interior of the ejecta curtain. The curtain shadow on the surface extends to the lower left. Note that point of impact is offset uprange from the center ofexcavation.

Three different approaches can be used to constrain the finalcrater size: (1) time of formation based on the evolving ejectacurtain; (2) backward ray traces to the surface; and (3) estimatesderived from the total ejected mass from Earth-based telescopicobservations. Even without being able to see the crater, theoutward-advancing ejecta curtain could be used to establish thecrater formation time or limits on size as described by A’Hearnet al. (2005), Richardson and Melosh (2006), and Richardsonet al. (2007).

2.6.1. ObservationsDuring closest approach, well-defined rays extend back to

near the region of the crater. Fig. 16 allows comparing the im-

pact region from the ITS camera on the approaching probe(Fig. 17A), a deconvolved HRI pre-impact image (Fig. 17B),and a highly processed deconvolved stacked images in orderbring out detail (Fig. 17C). Fig. 17D shows a deconvolved butunstacked HRI image near the final approach sequence. At thistime (approaching 800 s), the flyby is looking from above, butoff to one side (off surface normal). The details of the im-age processing are described elsewhere (Busko et al., 2007;Lindler et al., 2007), but the value of the image process-ing can be clearly demonstrated by comparing features iden-tified in the ITS mosaic and the closest approach images(Fig. 16). Such extensive image processing may ultimatelydelineate gradients in the image, rather than actual features


Fig. 16. Close ITS mosaic (unprojected) (A) compared with deconvolved and stretched HRI close-approach image (geometrically corrected but oriented to matchthe ITS mosaic to the left) (B). The impact point is estimated on the basis the ITS image sequence. Common features are correlated and place constraints on themaximum crater size. Features uprange (toward the top) are more easily identified due to the uprange zone of avoidance. Dotted lines represent terrain features (seeFig. 1b). The bright streak offset from the incoming trajectory is interpreted as a high-angle uprange ray and ejecta cloud persisting to late time (arrow).

emerging from below the ejecta. With this caveat, such tran-sitions still may provide useful data on the limits for the finalcrater.

Fig. 16 includes mapped rays that are accentuated in theprocessed images. Several conclusions can be drawn. First,a higher albedo patch on the scarp appears to correlate withrays in the processed image. This suggests that either thispatch was excavated (or subsurface extensions of this mate-rial). Second, downrange rays appear to converge downrangefrom the “ring” center, and two rays disappear well down-range. The uprange rays, however, converge closer to the pointof impact. Third, a bright linear feature in the deconvolu-tion image and (arrow) likely corresponds to the diffuse cloudabove the impact and the uprange ray in Figs. 12–15. Andfourth, a concentric pattern emerges from the filtered decon-volution image and may indicate the base of the ejecta cur-tain advancing beyond the final crater rim. It should be noted,however, that image processing enhances gradients; conse-quently, the concentric features in Fig. 17C could be arti-facts.

In Fig. 17C, rays converge on a circular zone near (but offsetdownrange from) the point of impact, based on position of theflash and the ITS approach imaging. The diameter of the regionwhere the rays disappear is ∼300 m. While this might define themaximum diameter for the crater, it more likely represents thebase of the ejecta curtain (as demonstrated below). The darkerring (about 220 m in diameter) surrounding the bright interiorpatch (about 130 m in diameter) is more consistent with a craterrim surrounding a bright floor. If the rays are extended inwardto a point where they do not intersect (Fig. 17D), then the di-ameter of the largest ellipse (projected circle) allowed is about220 m. This should set a maximum diameter for the crater. Con-sequently, the size of the crater based on imaging ranges from130 to 220 m.

2.6.2. Ejected massPre-mission estimates for the total ejected mass were derived

from empirical scaling relations for different types of targets(Schultz et al., 2005; Richardson et al., 2005). The scaling re-lations predicted the total cratering efficiency as a function ofimpactor (velocity, angle, density, diameter) and target charac-teristics (gravity, density, physical properties). Observations ofthe dust ejected during the collision allow directly estimatingthe cratering efficiency (total displaced mass divided by the pro-jectile mass) and limits on crater size with certain assumptionsdrawn from experiments.

Various Earth-based observers reported ejected masses rang-ing from 5 × 105 kg (Sugita et al., 2005) to 7 × 106 kg (Kelleret al., 2005). Analysis of Spitzer Space Telescope observationsyield a minimum ∼8 × 106 kg dust mass of and 6 × 105 kgof water (Lisse et al., 2006). The observed ejected mass fromEarth, however, cannot establish the size of the largest frag-ments (clumps), which could comprise the greatest mass. More-over, estimates based on these observations represent onlya small fraction of the total ejected mass from a gravity-controlled crater.

For solid targets, the ratio of ejected mass to displacedmass comprising the crater represents only about one half(e.g., Schultz et al., 1981). For weakly bonded (sand-like) tar-gets, this fraction reduces to ∼1/3, whereas for highly porousand compressible targets, it can reduce to 1/5 (Schultz et al.,2005) or less (Housen et al., 1999). By definition, gravity-controlled growth retains most (90%) of the ejected mass within5 crater radii (see Housen et al., 1983, 1999). Consequently, to-tal ejected mass based on Earth-based observations could be 50times less than the total displaced mass for the crater (neglect-ing the contribution by ices).

If the density of the excavated mass is 0.3 g cm−3, then thetotal cratering efficiency (total displaced mass scaled to pro-jectile mass) can be estimated. The cube root of the derived


Fig. 17. Comparison of features in ITS approach image (A), deconvolved HRI image (B), filtered/processed HRI image (C), and closer view of the impact site (D).All images have been geometrically corrected to approximately match. Dot in (B) indicates point of impact based on ITS imaging. “A” through “E” are benchmarkfeatures. “Se” represents shadow on interior of curtain (optically thick base) away from the Sun. “Sc” is the shadow of high-angle interior plume on Sun-facinginterior ejecta curtain. A subtle but obvious concentric pattern emerges at the center of convergence of ejecta rays (C) as outlined in (A). The “x” in (C) indicates animage defect. Outer dotted ring in (A) is interpreted as the base of the ejecta curtain (corresponding to the tips of the arrows in (C)). The middle ring is the borderof the dark region in (C). The white ring in (A) corresponds to the interior bright area in (C). If the border of the dark region corresponds to the crater rim, thenthe crater ranges from ∼160 to 170 m in diameter. (D) HRI (9001037) negative (left) and positive (right) images showing convergence of rays near the impact site(unrectified). Features E and F are identifiable in the heavily processed (deconvolved and stacked) HRI images (C). At this stage (772 s after impact), rays remainclearly defined but disappear in a region about 220 m in diameter. Ray curvature may indicate an evolving flow field. Dot corresponds to point of impact, which isuprange from the region of ray convergence and the bulk of the ejecta at late stage.

cratering efficiency provides a first-order estimate of crater di-ameter (and depth) after introducing a shape factor. This addi-tional shape factor is needed in order to match observed craterdiameters with values derived from taking the cube root of thecrater volume. Table 1 gives the range in the derived cratersizes. These values reasonably agree with pre-encounter esti-mates based on gravity scaling of oblique impacts into a highlyporous target (Schultz et al., 2005). If a greater fraction of thetotal mass was actually observed (i.e., never returned to thecrater), then the derived diameters would be reduced. In real-ity, crater diameter and depth need not be simple ratios due toeffects of substrate properties (discussed below).

2.7. Summary

Collision by the DI probe produced ejecta with seven com-ponents stretched out over 800 s.

1) Early-time downrange plume that expanded while travelingdownrange at a velocity slightly less than the initial impactvelocity.

2) Downrange diffuse plume composed of vapor-entrained de-bris.

3) Narrow plume directed uprange out of the initial penetra-tion zone creating a diffuse ray system in the foreground.


Table 1Estimates of crater size from observed ejecta

Ejectedmassa

(kg)

Cratering

efficiencyb

(log)

Diameter (m)c

(rim–rim)Depth (m)(from rim)

Pumice Sand Pumice Sand

1 E7 5.608 250 300 42 605 E6 5.307 200 240 33 481 E6 4.607 115 140 19 28

a Range from Sugita et al. (2005), Meech et al. (2005), Lisse et al. (2006).b Assumes: density = 0.3 g/cm−3 (based on cratering style and ejecta curtain

advance); ejected mass (observed): total displaced mass (crater) = 1/3; 80%return to comet (not visible from Earth); and a 5:1 diameter ratio (sand) and 6:1(pumice).

c Diameter and depth is derived from the cube root of cratering efficiencyusing a shape factor for different target types in order to match correspondingdiameters and depths in experimental craters.

4) Low-angle, uprange ejecta fan related to excavation flowmodified by the initial stages of penetration.

5) High-angle plume composed of low-speed ejecta above theimpact that lasts throughout the approach sequence (identi-fiable in stereo images and in its shadow cast on the innersurface of the ejecta curtain).

6) Late-stage bilaterally symmetric butterfly pattern with agrowing zone of avoidance uprange (transparency of theejecta curtain) and downrange (expressed as a transparentwedge). The net result of these components is a cardioid-shaped ejecta fan flanked downrange by two rays and splituprange by a smaller ray.

7) In the last images, ejecta rays converge on a center thatis offset downrange from the point of impact (determinedfrom the FL and the ITS approach imaging).

Each ejecta component from the Deep Impact collision canbe observed in oblique impacts in laboratory experiments. Theirrelative importance at different stages provides clues for the na-ture of the comet surface and subsurface, as discussed in thenext section.

3. Laboratory experiments: Initial coupling

The different stages of the Deep Impact collision cannotbe fully simulated in laboratory experiments due to obviousdifferences in scale (360 kg versus 0.3 g in the laboratory),impact velocity (10.2 versus 6 km s−1), and gravity (980 ver-sus 0.04 cm s−2). Independent variables affecting initial cou-pling include not only speed and mass but also the contrast inimpedance (density and sound speed). The impactor/target im-pedance contrast in the laboratory can be reasonably matchedwith the DI experiment. Both laboratory experiments (Schultzet al., 2005) and code calculations (O’Keefe et al., 2002) indi-cate that high-velocity impacts into highly porous targets willpenetrate to great depths before fully coupling.

Various phenomena and processes recorded in laboratory ex-periments have been replicated in hydrocodes (e.g., Pierazzoand Melosh, 1999) and recognized on planetary surfaces atmuch broader scales (e.g., Schultz and Anderson, 1996). Mate-rial motions created by an impact reflect a fundamental process

that should occur over a wide range of scales with the appropri-ate scaling relations as is evident in ejecta ray systems on theMoon. The Deep Impact collision, therefore, provides a rare op-portunity to directly compare impact processes at very differentscales while in formation.

The evolution of the initial vapor/dust plume and subsequentejecta were anticipated in a series of experiments before the DI-encounter (e.g., Schultz et al., 2005) and in models based onanalytical approximations from experiments (Richardson et al.,2005). The strategy in the present study is to use laboratory-scale results for reconstructing selected observations from theDI collision. Table 2 provides a summary of key observations,experimental strategies, and relevant figures.

3.1. Interpreting and comparing experiments

The impact process involves interacting chemical and phys-ical processes acting over a wide range of spatial and temporalscales. There are two different time scales to be considered. Thefirst addresses the early stages of energy/momentum transfer(coupling) and scales the time of an observation to the time(t ) for the projectile to travel its own diameter (a) with im-pact speed (v), i.e., τ = t/(a/v). The second applies to the latestages of crater formation as the crater grows to its final dimen-sions, i.e., time (t ) scaled to end of crater growth (Tc).

Laboratory experiments are designed to observe the impactprocess in detail at close range so that high-speed ejecta are cap-tured close to the growing crater. The DI flyby, however, couldonly observe features at large distances (>300 m), well awayfrom the point of impact. As a result, there is a time differencebetween stage of crater growth and the observed ballistic ejecta,especially at early times. Laboratory experiments also have theadvantage of knowing when the crater has finished forming,whereas DI-flyby imaging was unable to document this clearly.Consequently, these two scaled times are only used as a generalguide (Table 3).

The following discussion first explores the key features dur-ing the initial moments (within the first frame of Fig. 18) cap-tured in hypervelocity impacts in the range of laboratory exper-iments (see Table 4). Observations of the DI collision are theninterpreted in the context of these results. Second, later stages(subsequent frames) of excavation are examined in the labo-ratory. Such experiments provide insight for the Deep Impactexperiment through visualization of the process from differ-ent perspectives, detailing the earliest stages at much higherframing rates (smaller values of τ ) and with different instru-mentation (spectroscopy), and constraining source depths forthe observed vapor through experimental approaches.

An overview of ejecta evolution for different target types isillustrated in Fig. 18. Laboratory experiments using imaging at500 frames per second have an inter-frame exposure (τ ∼ 1600)only twice as long as the DI-MRI camera (τ ∼ 790). In bothcases, the vapor plume rapidly evolves before it becomes visibleand emerges between the first and second frame.

Porosity and compressibility affect the evolution of theejecta from vaporization to excavation. One predicted outcomeof the DI collision reflected the effects of an under-dense,


Table 2Selected experimental strategies for interpreting results from the DI collision

Feature/process DI examples(figures)

Experimentalstrategyc

Experiments(figures)

Initial stagesInitial vapor/dust plume 3, 5 High-speed imaging, photometrya 18–23

Evolving plume composition b High-speed spectroscopy 24, 25Faint “first light” 3, 5 High-speed spectroscopy 18, 25Excavation depth 4–5 High-speed imaging 23

Excavation stageEarly uprange plume 4–6, 9 Under-dense targets 26–29bEarly, low-angle rays 4, 8–10 Perspective views 23, 26B, 28Early-high-angle cloud 8, 9 Imaging 18, 23Initial uprange rays 8, 9, 12 Layered target 32b, 23Disappearance of UR rays 10, 12 Layered target 23, 32bCurved rays 8–12 Evolving ejecta (PIV) 32, 36Uprange ZoAc 10–14 Oblique impact 28–32Late, downrange ZoAc 10–14 Layered target 28, 32bLate, uprange ejecta cloud 8–14 Highly porous target 18Late, crater obscuration 14, 15 Porous target 18Cardioid pattern 10 Evolving flow field 32b, 36Excavation depths 38 Layered targets 33, 34Crater size 16, 17 Imaging, scaling 35, T1Offset impact pt. 15, 17 Highly porous target 18, 23, 27

a Also see Ernst and Schultz (2007).b This observation is noted in A’Hearn et al. (2005).c All experiments used 0.635 cm Pyrex spheres impacting at 4.9 to 6 km s−1 at an angle of 30◦ from the horizontal (specific data given in figure captions).

Table 3Comparing DI and laboratory time scales

Deep Impact Laboratory experiment

Early time processesa

High-speed video High-speed imaging(500 fps) τ = 1570 (6000 fps) τ = 130

MRI (early): τ = 790 1 frame/2 MRI frames 6 frames/MRI frameHRI: τ = 9180 5.8 frames/1 HRI frame 71 frames/HRI frame

Late-time processesb

High-speed video High-speed imaging(500 fps) (6000 fps)

Deep Impact 50 frames = Tc 720 frames = Tc(800 s = 4Tc)

a “Fps” refers to “frames per second. Late-stage dimensionless time = t =t/Tc ∼ [V/g]1/6 for gravity-controlled growth (crater volume = V ; gravita-tional acceleration at surface = g; total time for crater formation = Tc). As-sumes Tc = 200 s for D; Tc = 0.12 s for laboratory experiment.

b Early-stage dimensionless time = τ = t/tp = t/(a/v) = dimensionlesstime defined as actual time, t , for impactor diameter, a, impact velocity = v.

compressible, unconsolidated target (UR models in Schultz etal., 2005). Such targets are compressible at all scales (bulkpacking and individual grains, e.g., snow). They contrast withunder-dense, porous targets (PR model) that may be porouson a macro-scale but not compressible at the grain scale (e.g.,typical quartz sand targets). The UR model produced a faintinitial flash, high-angle ejecta cloud, and asymmetric ejecta.Moreover, impacts into UR targets were more efficient (largercraters) at lower impact angles than vertical impacts, contraryto results for PR models.

As noted in Table 4, perlite targets (or mixtures with per-lite) are used to represent the UR models. This material is usedinstead of snow or porous ice because snow (or similar frozenmaterials) creates challenges in the laboratory due to chang-ing conditions during the time required to achieve near-vacuumconditions (bonding of grains as they freeze). Ultimately, thegoal of the experiments is to understand the effect of poros-ity and compressibility on styles of excavation. Peak pressuresare high only in a region around the impact point, much smallerthan the size of the crater. Crater excavation occurs under exten-sion (rarefaction off the free surface). Consequently, it shouldmake little difference if the target is made of snow or perlite dur-ing excavation for the preliminary comparisons being soughthere.

3.2. Observed processes

3.2.1. Self-luminous vapor/dust plumeQuestions raised Features observed during the initial stagesof cratering observed from the DI flyby raised several specificquestions:

a) What is the fast-moving vapor/dust plume observed duringthe first second in the MRI?

b) Why do the speed and position of the downrange plumeresult in a mismatch between the known impact point andthe inferred point of emergence (derived from the speed andlocation of the plume leading front)?

c) What is the significance of the faint uprange-directedplume?


Fig. 18. Selected images from oblique (30◦ for from the horizontal) hypervelocity impact experiments into three different target types: (Left column) pumice powderoverlain by a 1.8 cm thick layer of ground perlite powder with a thin (100 µm) layer of graphite in-between; (Middle column) thick ground perlite powder; (Rightcolumn) 0.4 cm layer of fine sugar over thick, ground perlite powder (see Tables 3 and 4). All impacts used 0.418 cm-diameter Pyrex spheres at ∼5.7 km s−1 (arrowindicates trajectory from right). The pumice has a density of around 1.3 g cm−3; ground perlite, ∼0.4 g cm−3; and fine sugar granules, ∼0.8 g cm−3. The cameraviewed the impact from a port 45◦ above and offset to the side about 45◦ (away from the trajectory axis). Dashed lines correspond to axes along and the orthogonalto the trajectory; dots indicate point of impact. As target porosity increases (density decreases), several changes occur. First, the initial impact flash decreases (firstframe, left to right). Second, the point of impact coincides with the final uprange rim (middle). Third, increasing porosity results in both a high-angle diffuse plumeblocking the view of the crater from above and a collimated uprange plume controlled by the transient crater (middle and left).

d) What accounts for the delay between the “first light” andthe first saturated image?

e) Why does the self-luminous component appear to movedownrange offset from the initial trajectory?

f) How should the observed spectral emissions depend on ob-servation times and location?

g) What is the depth of origin for the initial vapor/dust plume?

While not all of these questions can be definitively answeredat this point, each question can be addressed in laboratory ex-periments by isolating processes, replicating the viewing posi-tion, and selecting the correct target for the same impact angleas the DI probe.

Overview from experiments The observed self-luminousplume immediately after the DI collision evolved from ateardrop shape to a diffuse, optically thin cloud that traveledrapidly downrange while expanding laterally, decoupled fromthe rest of the cratering process (see Figs. 3–6). A similarevolution is observed in laboratory experiments for hyperve-locity oblique impacts (5–6.5 km s−1) into volatile-rich tar-gets (Schultz and Gault, 1990; Schultz, 1996) as illustrated inFig. 19. The inter-frame time interval in Fig. 19 corresponds toτ ∼ 23, much shorter than the τ ∼ 790 for the MRI sequences(i.e., only ∼3% of the scaled inter-frame for MRI). This sub-section first reviews the early-time processes before exploringimplications for the DI experiment.


Table 4Summary of experimentsa

Experiment (density, porosity) Purpose Figures

Powdered pumice (1.3 g cm−3; 43%) Evolution ejecta shadow 26A, 35Sieved perliteb (0.2 g cm−3; 90%) Evolution ejecta, flash 23AGround perlite (0.4 g cm−3; 70%) Evolution ejecta, flash 18, 32aSugar (0.8 g cm−3) over ground perlite Evolution ejecta shadow 26BPumice powder with sieved perlitea (1.8 cm

thick over thin 100 µm graphite layer)Evolution ejecta, flash 18, 28

Sugar over ground perlite Evolution ejecta, flash 18Dolomite block (2.8 g cm−3;


ever, clearly revealed in the DI collision. Absence of this com-ponent may reflect the porosity of the upper surface or strongatomic/molecular (rather than thermal) emissions. Impacts intohighly porous targets in laboratory experiments result in greaterpenetration depths before full coupling, in contrast with solidtargets. As a result, the jetting phase may mix turbulently withsubsequent vaporization before emerging from the target. Whilethe decoupled jetting plasma still occurs, the mass is greatly re-duced. Nevertheless, this component may contribute further tothe momentum-driven vapor component.

Reverse plume Computations (O’Keefe and Ahrens, 1977;O’Keefe et al., 2002) and hypervelocity experiments (Schultz,1996) recognize a vapor plume that is temporarily containedby the growing transient cavity and has been termed cavita-tion. An extreme example of this process occurred during theShoemaker–Levy collision. After each body penetrated deepinto the jovian atmosphere, vapor and condensates emerged ina collimated plume that raced back out the rarefied and heatedatmosphere (e.g., Ahrens et al., 1994; Boslough et al., 1994).Observations from Earth, however, captured ballistic conden-sates re-entering the jovian atmosphere at large distances.

Momentum-driven vapor An oblique impact creates an asym-metric shock front that generates vapor traveling downrangealong the initial velocity vector. Because this retains informa-tion on the momentum of the collision, it is termed here the“momentum-driven” vapor component. The momentum-drivencomponent represents a combination of several processes(see Schultz, 1996; Pierazzo and Melosh, 1999): downrange-directed shock heated target/projectile, turbulent mixing withthe jetting phase, and the evolution from non-isentropic expan-sion to isentropic expansion. The vapor cloud expands whiletraveling downrange at 80 to 140% of the initial impact speed(depending on the nature of the target) in an inertial frame thatis decoupled from the rest of the impact, as shown in Fig. 19.

Hemispherically expanding cloud: The fourth vapor compo-nent expands above the impact point and is related to vapor tem-porarily contained and then released from the cavity (Schultz,1996). Due to a downrange component, however, the expansionappears “hinged” near the impact point, i.e., the net downrangeexpansion rate is greater than the uprange expansion rate (seeFig. 19). The vapor cloud expanding above the cavity can bequite complex because it is a multiphase mixture (both gas anddust).

Consequently, vapor generated by an impact initially can-not be treated simply as a spherically expanding isentropic gasfrom a point source. Rapid phase changes, different compo-nents feeding the vapor over a finite time, and the consequencesof an evolving shock front in response to asymmetric shocksall occur within the first frame. The DI-MRI imaging sequencecould not resolve these early stage components, yet all con-tribute to the integrated observations (imaging and spectra). Thetotal amount of vaporization represented by these various com-ponents, however, represents a very small fraction of the totalejected mass (discussed below). Most of the ejected mass wouldhave been subjected to very low peak pressures with little ele-vated temperature.

3.2.2. Plume evolutionThe emergence and evolution of the vapor plume from an

oblique impact reveal clues for the observed sequence for DIand the possible nature of the target. Fig. 20 includes a time-exposed view revealing the self-luminous components of thevapor plume for a high-porosity target. Fig. 21 illustrates thetime-resolved evolution of vapor (and dust) phases resultingfrom 5.7 km s−1 impacts at 30◦ into a mixture of dolomite andsand. In contrast with Table 3, each microsecond in Fig. 21corresponds to τ ∼ 1.22, which represents only 0.15% of thetime covered by the first MRI image (0.062 s). These imagesare from a short-exposure (250 ns) thermal infrared camera,which captures entrained thermal sources, rather than the ac-tual vapor phases. The sequence reveals that the vapor forms a

Fig. 20. Time-exposed stereo view (reverse) of a 30◦ impact into a 0.635 cm sieved perlite layer over powdered pumice at 5.2 km s−1 (no lights). Self-luminouscomponents create streaks that define trajectories, even though unresolved in time. While the uprange self-luminous plume is not identified in high-speed imaging,it is captured here. Targets composed entirely of sieved perlite exhibit a lower uprange angle self-luminous plume.


Fig. 21. Evolution of self-luminous plume of vapor and dust during the first 8 µs after oblique (30◦) impact by 0.635 cm Pyrex into a mixture of sand and dolomite(1:1 by weight). View is from above with impact point noted by small circles. The images were taken with a thermal imaging camera and demonstrate the rapiddownrange motion tied to vapor-driven expansion.

well-defined plume leading a slower moving (downrange speedand lateral expansion) diffuse cloud of entrained hot dust andgas.

Consequently, experiments demonstrate that the initial vaporplume is not simply a spherically expanding gas. Oblique im-pacts produce an asymmetric shock front (Dahl and Schultz,2001), even at great distances from the impact point (wellbeyond the crater cavity). The resulting vapor retains someof this downrange-directed flow as recorded in experiments(Schultz, 1996; Wrobel et al., 2004; Schultz et al., 2006a) andin hydrocodes (e.g., Pierazzo and Melosh, 1999). For exam-ple, a 5 km s−1 impact can produce a vapor plume initiallytraveling at 8–12 km s−1, depending on the density (porosity)of the target (Schultz et al., 2006b). Vapor phases are con-firmed by the presence of atomic and molecular emissionsusing high-speed spectrometers that isolate gas and plasmasunrelated to the jetting process (Eberhardy and Schultz, 2005;Schultz et al., 2006a).

The initial downrange speed of the vapor plume is not con-stant but rapidly decreases within 10 projectile diameters fromthe point of first contact (Fig. 21). This momentum-driven com-ponent results from: the downrange-directed shock, partial con-tainment of the vapor prior to free expansion, and turbulentmixing with impact jetting (e.g., Sugita and Schultz, 1999) atthe contact between the irregular-shaped DI impactor and sur-face materials. The gas then evolves from non-isentropic expan-sion to isentropic expansion. After sufficient time, this can beviewed as an expanding vapor cloud traveling downrange in aninertial frame that is decoupled from the rest of the impact (e.g.,Schultz, 1996; Sugita and Schultz, 2001).

While the downrange motion preserves part of the horizon-tal component of the impactor, lateral growth represents ther-modynamic gas expansion (Fig. 21). The initial self-luminousplume evolves into two components: an irregular front (likelythe result of Kelvin–Helmoltz instabilities at first contact) and abrighter interior plume. The instabilities at first contact can leadto rays at greater distances (but would not account for the raysmuch later). The leading edge results from the greater numberdensity of the emitting particles and gas within the outer edge of

Fig. 22. Evolution of vapor cloud front in Fig. 21. Lateral expansion and down-range speeds of the gas front slow with time. The initial slowdown (the first10 µs) is attributed to rapid changes in the vapor (non-isentropic to isentropicexpansion) and the conversion of internal energy into kinetic energy of expan-sion. The vapor plume expands while traveling downrange at a speed propor-tional to the horizontal component of the impact velocity (vi cos θ ). At muchlater times (50–100 µs), vapor expansion asymptotically approaches a constantvalue (umax) that can be related to the internal energy of the cloud.

an expanding shell, also recognized in Fig. 21. The inner plumeis the result of delayed emergence of the vapor from the cavityand vaporization from downrange ricochet debris.

Expansion rates over the first 30 µs rapidly decrease withtime (Fig. 22). Between 1 and 2 µs, the radial expansion rateis 9.0 km s−1 but decreases to 5.5 km s−1 (2 to 5 µs) and5.1 km s−1 between 5 and 8 µs. The underlying physical processfor the initial deceleration of the vapor plume front is due toshock rarefactions created by the gas/vacuum interface, sim-ilar to that observed for explosions in a vacuum (Ahrens etal., 1971). This results in the redistribution of momentum andkinetic energy from the plume front back into the gas. In ad-dition, unsteady, non-adiabatic processes (ionization and phasechanges) initially may produce a greater than expected initial


expansion of the leading edge of the gas (Chen et al., 1995)and could play a similar role here. Eventually adiabatic ex-pansion controls expansion as the internal energy in the gas isconverted to kinetic energy of expansion (along with lower den-sities and temperatures). The effect of the residual atmospherein the chamber (∼0.5 Torr) does not contribute to decelerationuntil much later in time, after the gas has expanded to muchgreater dimensions (Schultz, 1996).

In Fig. 22, it is interesting to note that the downrange hor-izontal component of velocity (cosine of the initial impactvelocity = 4.9 km s−1) added to the radial gas expansion rategives a net expansion of 13.9 km s−1, similar to the initial down-range plume speed of 12.7 km s−1. This should be expected ifthe downrange plume represents the combination of gas expan-sion along a moving inertial frame with a speed similar to theinitial impactor momentum.

3.2.3. Inferences from vapor expansion in laboratoryexperimentsInitial temperatures The temperature of the initial gas phasebetween 2 and 8 µs can be derived from the observed expan-sion of the vapor plume (Fig. 22) by adapting the gas expansionequation for a highly compressed gas (Zel’dovich and Razier,1967):

(1)RT/μ = 12u2exp(γ − 1)2/2γ,

where T is temperature (K); R, the gas constant; uexp, the ex-pansion velocity; μ, the gas molecular weight; and γ , the ratioof specific heats. If individual atoms comprise the gas phasebetween 2 and 5 µs AI, then the value of γ is 1.67 and μ is∼18. The resulting temperature is ∼4200 K for an average lat-eral expansion speed between 2 and 8 µs of 5.3 km s−1. If thegas is instead composed of CO, then γ is 1.5 and μ is ∼28. Theresulting temperature of the gas then becomes slightly cooler,∼3900 K. These assumptions about composition will be ad-dressed below using high-speed spectroscopy. Both results areconsistent with color temperature measurements for compara-ble impacts into silicate powder (Ernst and Schultz, 2004) andrelative intensities of atomic emission lines (Sugita et al., 1998).

Vaporized mass In laboratory experiments, lateral expansionrates asymptotically approach a constant value after about100 µs and reflect the complete conversion of the initial internalenergy of the gas to kinetic energy of expansion. The maximumexpansion speed, umax (the gas front), then is simply related tothe energy of the vapor plume (Ev) by the following relation-ship (Zel’dovich and Razier, 1967):

(2)Ev = 12mvu

2∞,

where u∞ is the constant mean expansion velocity of an isen-tropic plume of mass, mv . This expansion velocity is relatedto the maximum speed of the leading front of the vapor plume(umax) by the following:

(3)u∞ = umax[(γ − 1)/2γ ]1/2,

where γ ranges from 1.28 for CO2 vapor, 1.37 for H2O, and 1.5for CO.

Impact angle affects the expansion speed of the vapor plume(expressing peak pressure, temperature, and vaporization as dis-cussed in Schultz, 1996). The value of umax has been mea-sured in laboratory impact experiments for different targets foroblique impacts: ∼2.2 km s−1 for dolomite (2.0 km s−1 fordry ice and 1.8 km s−1 for water ice) for an impact speed of5.5 km s−1 and slightly lower impact angle. The combinationof Eqs. (2) and (3) with the observed energy in the vapor cloudalso allows deriving the total vaporized mass. In laboratory ex-periments, the observed vaporized mass approached 3 to 5 pro-jectile masses for impacts into dry ice and carbonates (Schultz,1996). As discussed below, this same strategy can be applied toDI provided assumptions are made about the energy partitionedto the initial vapor plume.

Effect of porosity and layering As porosity increases, experi-ments indicate that the initial vapor plume splits into downrangeand uprange self-luminous components. The downrange plumeis significantly reduced in intensity (e.g., see Fig. 18), as furtherdocumented in the photometry (Ernst and Schultz, 2003, 2007).Increasing target porosity also decreases the expansion speed ofthe emerging vapor plume (Fig. 22). While this decrease is inpart due to reduced shocked mass, high-speed imaging showsthat it results from delayed release of vapor from the initial pen-etration cavity (also see Schultz et al., 2005).

Deeper impactor penetration results in a higher emergenceangle of the plume (i.e., energy/momentum transfer and releasedeeper in the target). As shown in Fig. 22, the downrange speedat 5 µs (calculated between first contact and 5 µs) decreaseswith increasing porosity. It decreases from 13.8 km s−1 for animpact into solid dolomite block; for powdered dolomite, to12.0 km s−1; for a dolomite and sand mixture, to 10.1 km s−1;and for a perlite and dolomite mixture, to 6.0 km s−1.

The initial low ejection angles for the DI vapor plume sug-gests that the impact was well coupled within several projectilediameters (Figs. 3 and 4). This could indicate a denser (or morecompetent) layer closer to the surface. As a test, quarter-spaceexperiments were performed. This experimental design allowslooking inside the growing crater cavity (e.g., see Schultz etal., 2005; Piekutowski, 1977). Here, layered porous particu-lates were used with different density contrasts. This strat-egy provides a useful comparison with results in Figs. 18, 20,and 22. Three experiments addressed first-order effects of ex-treme porosity, buried volatiles, and density contrasts at depth.For reference, sand targets result in a hemispherical transientcavity with the vapor phase rapidly expanding above the sur-face (see Schultz et al., 2005).

The first experiment used a uniform target of perlite, whichwas sieved to remove particle sizes less than about 2.3 mmacross (resulting in a bulk density of 0.2 g cm−3). The secondexperiment buried a volatile-rich layer (0.67a where a is theprojectile diameter) below a porous perlite layer (∼5a). Thisstrategy assessed the possible effect of a highly porous surfacelayer on the initial vapor plume (Fig. 23B). The perlite layerin this case was finely ground resulting in a greater fraction of


Fig. 23. Quarter-space experiments (see text) covering the first 0.6 ms of an impact by a 0.48 cm Pyrex sphere into different types of high-porosity targets oflow-density particulates at ∼5.4 km s−1 (30◦). (A) Perlite sieved to remove grains smaller than about 500 µm to create a very low density (0.2 g cm−3) andhigh-porosity (90%), low-strength particulate target. Most of the high-temperature silicates are driven below the surface and line the deep penetration funnel. A faintreverse plume also emerges out of the penetration funnel, back up the initial trajectory (not visible here; see Fig. 20). (B) Ground perlite target with thin layer ofdolomite (5.4 km s−1). “FP” represents finely ground perlite (0.4 g cm−3); “D,” a thin layer of powdered dolomite; “MP,” mixed perlite (with


tile radius), however, reduces penetration for an oblique impactand results in shallower coupling. This produces a well-defineddownrange plume, which is more consistent with observationsfor DI. Even in this case, the intensity of the visible vapor plumeis reduced due to less compressed mass in front of the projec-tile and due to incandescent ejecta initially hidden deep belowthe surface (e.g., see Fig. 20).

The narrow penetration funnel collimates ejecta into a nar-row cone emerging from the point of impact and directed backalong the incoming trajectory. Without vaporization, ejectaemerge in a narrow cone. As the cavity opens, however, the re-verse plume (back along the trajectory) evolves to higher anglesfrom the surface (Fig. 18, middle). The evolution of this reverseplume (angle, duration), therefore, provides important clues forthe initial coupling and porosity of the target.

3.2.4. Nature of the early vapor plumeStrategy Spectral measurements were made in different por-tions of the expanding vapor plume, as shown in Figs. 18–21.These experiments did not use simulants of a cometary sur-face because the impact velocities available in the labora-tory (5–6 km s−1) are less than the speed of the DI collision(10.2 km s−1). The lower available impact velocities limit va-porization of silicates. Available high-speed spectrometers alsolimit the wavelength range to the visible in the laboratory exper-iments, in contrast with the infrared range for DI. Instead, theexperimental series used easily vaporized materials that can beobserved and analyzed. This strategy permits probing the evolv-ing compositions at different times and in different directions.The following specific questions can be addressed:

1) Are the observed vapor components for DI (A’Hearn et al.,2005) primary materials or recombination products?

2) What is the nature of the faint “first light” and delayedbrightening for DI?

3) What is the source depth for the spectral emissions ob-served by DI?

4) What is the source depth for the most distal (highest veloc-ity) materials?

Dolomite (calcium, magnesium silicate) has proven to bea useful surrogate for tracing evolving compositions, temper-atures, and processes within impact-generated vapor phases(e.g., see Sugita et al., 1998, 2003; Sugita and Schultz, 1999;Eberhardy and Schultz, 2005). These surrogates allow deter-mining the relative proportion of emitting atomic (e.g., Ca, Mg)and molecular species (CaO, MgO, CO2, C2) at impact ve-locities available in the laboratory. They also can be used toexamine the effects of porosity, mixtures, and surface layers onvapor conditions and evolution.

The surrogate targets here include three oblique impacts:a solid dolomite block, dolomite powder with a thin layer ofsugar, and dolomite powder with a thin layer of sugar and adusting of graphite. The solid dolomite target provides a ref-erence for the evolving compositions in different directions andover different times. The second and third experiments assessed

the effect of a porous target (same composition) but with thinlayers of carbon compounds.

Evolution of vapor composition The slit of the Deep ImpactIR spectrometer was oriented perpendicular to the trajectoryand offset downrange from the impact point. First results de-scribed the initial plume passing in a pixel exposed for 0.72 s(A’Hearn et al., 2005; Sunshine et al., 2006). A similar config-uration was reconstructed in a series of laboratory experimentsin order to aid interpretations of the DI impact. One series as-sessed the evolution of the vapor components using a surrogatetarget composed of easily volatized and disassociated species.The second series contrasted these results with a surface layerof complex organics (in this case, powdered sugar).

The image in Fig. 24a includes a reference for the different“fields of view” (FOV) by telescopes used to probe the vaporplumes. Because exposures lasted 50 µs, each spectrum rep-resents a time-integrated passage of the vapor plume passingthrough a 2.5 cm FOV at 8 km s−1 (Fig. 21). This exposure timerepresents τ = 46, in contrast with τ = 9180 for the HRI-IRspectrometer. While dramatically different, most of the atomicand molecular emissions come from the leading front of theplume passing the FOV, just as in the case of DI.

The spectrometers used telescopes coupled by fibers so thatthe spectra were recorded on common CCD’s (2 on one, 3 onanother). Four telescopes viewed each event from two 45◦ portspositioned uprange (but on each side of the trajectory axis∼45◦) and looking downrange; a fifth telescope was positionedin the 45◦ port looking directly downrange.

Fig. 24b contrasts the spectral evolution of the plume per-pendicular to the trajectory and positioned ∼10 cm downrange(FOV’s 6, 4, and 2) for an impact into a solid block of dolomite.Slightly off axis (FOV 4), atomic lines become weaker relativeto the molecular lines but a weak Mg emission line at 418 nmappears. Still farther (FOV 2) from the initial trajectory, the va-por is almost entirely composed of molecular emission lines(CO, C2).

These spectra reveal that gas phases directly downrange con-tain both atomic and molecular disassociation and recombina-tion products (Ca, CO, C2, C3) and a relatively weak thermalsignal. One set of lines along the downrange trajectory (FOV-6)is tentatively identified as high-pressure bands that are part ofthe Swan band system that only occur in special circumstancessuch as C + C2O = C2 + CO (468, 466, 437, 435, 409 nm de-grading to the violet, with the strongest at 468 nm). Fartheraway from the downrange trajectory axis, the thermal compo-nent is less.

The second experimental series added a thin (projectile ra-dius) layer of powdered sugar over dolomite powder (Fig. 25).The goal was not only to assess the effect of organics; it wasalso to understand the contribution of this layer relative to thesubstrate. Sugar simulates the possible contribution by complexassemblages of C and H. The layer of sugar significantly sup-pressed the radiance immediately after impact, but the inferredvapor expansion speed within the first 5 µs was comparableto the speed produce by an impact into the dolomite substratealone. This result confirms that a faint plume in the visible does


(a)

(b)

Fig. 24. (a) Reference image from above (see Fig. 21) showing size and lo-cation of spectral measurements for an oblique impact (5.9 km s−1) into soliddolomite (arrow at top indicates direction). Each dash-dotted semi-circle rep-resents different stages of expanding vapor plume as it travels downrange andpasses the different fields of view (FOV’s) over the first 50 µs. This sequenceisolates the effect of both direction (off trajectory axis) and time (lateral va-por expansion). (b) Atomic emissions disappear by the time the vapor reachesFOV-2. Near off-axis (FOV-4), and on-axis (FOV-6) indicate lateral expansionof same component. Very different spectra for FOV-2 indicate either that thelaterally expanding plume cooled or that this represents a different component(i.e., different initial PT conditions) away from the trajectory. Conditions ofimpact: 0.635 cm Pyrex at 5.89 km s−1 at 0.37 Torr.

not indicate an absence of vapor but the absence of significantthermal background emission. In both cases, the thin surfacelayer significantly suppressed spectral emission lines derivedfrom the underlying dolomite powder.

Photometric observations (Ernst et al., 2006; Ernst andSchultz, 2007) indicate that the faint self-luminous vapor plumefrom the sugar layer increases in brightness with time, in con-trast with vapor phases from dolomite. This brightening is notdue to the dolomite substrate but is the result of strong thermalsources emerging in the plume. While impacts into dolomite ex-periments also yield byproducts of the intense heat of impact,they do not form significant solid phases over the first 50 µs.Superimposed images of the vapor from the sugar layer clearly

reveal this brightening as the plume travels downrange 35 µsafter impact (Fig. 25a).

The spectral evolution for the same experiment but with adusting of graphite on top (Fig. 25b) exhibits similar brighten-ing with distance (time). The spectrum taken close to the impactpoint but off axis (FOV-1) doubled in radiance at a given wave-length at FOV-6 (along the TTL) and FOV-4 (along the sameline but opposite sides of the trajectory). The delayed brighten-ing is attributed to two processes. First, the thin layer of darkgraphite was ejected uprange and obscured the initial emis-sion spectra. The thermal imaging camera shows this cooler(darker) component uprange (inset in Fig. 25b). Second, latercondensation of carbon from the vaporized sugar after expan-sion (Fig. 25a) resulted in late “blooming” in radiance.

In contrast with results for impacts into just dolomite,Fig. 25b reveals that molecular CO and C2 absorption linesoccur close to the impact (FOV-1). These result from coolergases expanding along the line of sight, consistent with theinset image. The molecular absorptions (CO and C2) and thepresence of Mg and MgO emission lines (although weak) in-dicate that the cooler components expand above (and uprangefrom) the impact. These cooler components are derived fromthe substrate, rather than the thin surface layer.

Such experiments demonstrate that the viewing position af-fects interpretations of the spectra, a result relevant to interpret-ing the DI results from the DI-flyby, Earth-based telescopes,and other space probes (e.g., Rosetta). The following re

the deep impact oblique impact cratering experiment · 2013. 7. 25. · deep impact experiment 297...

Documents