ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.1 (1-13)

Icarus ••• (••••) •••–•••www.elsevier.com/locate/icarus

Composition of jovian dust stream particles

Frank Postberg a,∗, Sascha Kempf a, Ralf Srama a, Simon F. Green b, Jon K. Hillier b, Neil McBride b,Eberhard Grün a,c

a MPI für Kernphysik, Saupfercheckweg 1, 69117 Heidelberg, Germanyb Planetary and Space Sciences Research Institute, The Open University, Walton Hall, Milton Keynes, MK7 6AA, UK

c Hawaii Institute of Geophysics and Planetology, University of Hawaii, 1680 East West Road, Honolulu, HI 96822, USA

Received 21 October 2005; revised 26 January 2006

Abstract

The Cassini spacecraft encountered Jupiter in late 2000. Within more than 1 AU of the gas giant the Cosmic Dust Analyser onboard thespacecraft recorded the first ever mass spectra of jovian stream particles. To determine the chemical composition of particles, a comprehensivestatistical analysis of the dataset was performed. Our results imply that the vast majority (>95%) of the observed stream particles originate fromthe volcanic active jovian satellite Io from where they are sprinkled out far into the Solar System. Sodium chloride (NaCl) was identified as themajor particle constituent, accompanied by sulphurous as well as potassium bearing components. This is in contrast to observations of gas in theionian atmosphere, its co-rotating plasma torus, and the neutral cloud, where sulphur species are dominant while alkali and chlorine species areonly minor components. Io has the largest active volcanoes of the Solar System with plumes reaching heights of more than 400 km above themoons surface. Our in situ measurements indicate that alkaline salt condensation of volcanic gases inside those plumes could be the dominantformation process for particles reaching the ionian exosphere. 2006 Elsevier Inc. All rights reserved.

Keywords: Jupiter; Interplanetary dust; Io; Volcanism; Satellites, atmospheres

1. Introduction

In late 2000 the Cassini spacecraft, on its way to Saturn, ap-proached Jupiter for a swingby manoeuvre and observed duststreams at a distance of more than 1 AU from the gas giant.The Cosmic Dust Analyser (CDA) onboard Cassini combinesan impact ionisation detector with a time-of-flight (TOF) massspectrometer and for the first time gave access to informationabout the particle composition.

The jovian system was first recognised as a source for dustparticles during the encounter of the Ulysses spacecraft in1991/1992 when high speed intermittent streams of tiny grainswere discovered (Grün et al., 1993). The onboard dust detectorregistered several short impact bursts with a periodicity of 28± 3 days as well as of 14 ± 2 days. The radiants tended to lieclose to the line-of-sight (LOS) direction to Jupiter. Three years

* Corresponding author.E-mail address: frank.postberg@mpi-hd.mpg.de (F. Postberg).

0019-1035/$ – see front matter 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.icarus.2006.02.001

after Ulysses, the Galileo spacecraft approached the gas giantcarrying an identical dust detector. Galileo also encounteredthe jovian dust streams at Jupiter distance of more than 1 AU.In 2004 the Ulysses spacecraft again encountered Jupiter. Thistime the first streams could be detected when the spacecraft wasstill more than 3 AU away from the planet. The evaluation of alarge sample of 26 dust streams confirmed a periodicity of 26days, closely matching the solar rotation period (Krüger et al.,2005).

According to the current understanding the general mecha-nism for creating the dust streams is as follows. Once a grain ischarged positively within the plasma environment of Jupiter’smagnetosphere it will be accelerated by the outward pointingco-rotational electric field. Outside the magnetosphere, the dy-namics of the grains is governed by the interaction with theinterplanetary magnetic field (IMF) that eventually forms thedust stream. The intermittent nature of the stream detectionsis believed to be due to spacecraft traversals through layers ofcompressed solar wind that enhances the dust flux, as observed

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.2 (1-13)

2 F. Postberg et al. / Icarus ••• (••••) •••–•••

in the environment of Saturn (Kempf et al., 2005a). Zook et al.(1996) simulated stream particle trajectories for the measuredIMF throughout the Ulysses approach phase to Jupiter. It wasdemonstrated that only fast (>200 km s−1) and tiny (5–15 nm)particles were able to reach the dust detector.

Although it was soon realised that if submicron chargedparticles exist within the jovian magnetosphere they would beaccelerated and ejected (Horányi et al., 1993a, 1993b), therewas some controversy about the dust origin. Jupiter’s gossamerrings (Hamilton and Burns, 1993) and the volcanoes of the jov-ian moon Io (Grün et al., 1996) were proposed as potential dustsources. A periodogram analysis of impact rate data taken dur-ing two years of Galileo’s orbit around Jupiter implied that Io isone important source of stream particles (Graps et al., 2000)which emits a continuous flux of nanometric particles form-ing an oscillating sheet. The contribution of Jupiter’s dust ringsto the dust flux is still discussed today. With the time-of-flight(TOF) mass spectrometer of the Cosmic Dust Analyser (CDA)onboard Cassini this question could be answered. The spec-tra should provide the possibility of distinguishing between theparticle types of the two possible sources which should leavedifferent chemical fingerprints.

We first introduce the possible dust sources in Section 2.The dust detector onboard Cassini and the dataset are describedin Section 3. In Section 4 spectra, methods and results of theanalysis are presented. Qualitative and quantitative interpreta-tions lead to an identification of a source and implications forthe dust formation process on Io which is discussed in Sec-tion 5. In Section 6, all results are briefly summarised.

2. Dust in the jovian system

Jupiter’s ring system consists of three components: the mainring, the halo, and the tenuous gossamer rings (Showalter etal., 1985). A significant fraction, if not all, of the dust formingthe rings is impact-ejecta derived from the inner rocky moonsMetis, Adrastea, Amalthea, and Thebe (Ockert-Bell et al., 1999;Burns et al., 1999; Showalter et al., 2003), whose orbits are em-bedded in the ring system. It has been shown that—unlike inSaturn’s icy rings (Poulet et al., 2003; Hillier et al., 2006)—rock forming minerals are predominant in the jovian mainrings (Wong, 2004). However, little is known about the com-position of the gossamer rings and their primary sources Al-mathea and Thebe. Those dark, narrow rings with low albedocan be expected to consist mainly of rock fragments as well,but contributions from other components—like water ice ororganic material—cannot be ruled out. During Galileo’s traver-sal through the gossamer ring in 2002, several thousand dustimpacts were registered. The Galileo dust instrument was notequipped with a mass spectrometer, but the data revealed par-ticle sizes in the micron as well as in the submicron rangewith the small grains dominating the number density (Krüger,2003).

Volcanic ashes cover the surface of the moon Io whose in-terior is heated by Jupiter’s tidal forces. The amazing varietyof its colours indicate sulphur in its various molecular struc-tures (yellow, orange, red, brown, and black) as well as specific

silicate minerals. Volcanically ejected dust particles can escapeinto Jupiter’s co-rotating plasma torus where they almost in-stantly could become a stream particle (Horányi et al., 1993a;Krüger et al., 2003; Flandes, 2004). It has to be pointed outthat, while the composition of the ionian gas and plasma en-vironment has been extensively studied there is no informa-tion about the composition of its dust environment. Further-more, the genesis and composition of particles in the volcanicplumes is currently unclear (Geissler, 2005). Spectroscopicanalysis of Io’s atmosphere reveals SO2 as the main compo-nent accompanied by smaller amounts of its subproducts SO,S, and O as well as sodium chloride NaCl (Lellouch et al.,1996, 2003). The observation of emission lines in the jovianplasma torus which is supplied by Io’s gas emission gives asimilar picture: S+ and O+ are the dominant components withS2+, O2+, SO+

2 and the alkaline salt components Na+, K+,Cl+, Cl2+ in smaller amounts (Küppers and Schneider, 2000;Schneider et al., 2000). The major part of the escaping gaswhich is not ionised forms neutral gas clouds around Io. Spec-troscopy reveals Na, K, and O (Brown, 1974; Thomas, 1996).An abundant neutral S cloud is predicted but cannot be provedby optical line emission. Mendillo et al. (2004) clearly demon-strated that Io’s volcanic activities have a controlling effect onthe extent and brightness of the sodium cloud. Though silicate-type mineral components play a role in Io’s volcanic system,so far there is no observational proof for their escape from Io’sgravitation in any form (Na et al., 1998).

What particle signatures would we expect in our data fromthe observations stated above? A mass spectrum from a parti-cle escaping the gossamer rings should show typical features ofrock forming minerals: a strong Si+ mass line, an O+ signatureand lines of elements which typically accompany silicates (Fe,Mg, Ca, Al, Ni). As water ice or organic material might con-tribute to the gossamer rings too, H3O+ and O+, H+ as well asC+ and N+ signatures could also appear.

The composition of particles in the ionian exosphere is un-known. Deriving it from the gas and plasma analysis seems tobe appropriate at least for qualitative considerations. Thus foran Io-type particle one expects lines of S+, O+, alkali met-als (Na+, K+), and Cl+ rather than mass lines associated withsilicates or water ice. Besides the abundance in a particle how-ever, the occurrence and strength of a spectral feature dependson a number of factors which have to be considered (see Sec-tion 4.2).

3. Instrument and dataset

3.1. The mass spectrometer of the CDA

The CDA instrument has been described in greater detail bySrama et al. (2004).

The Cosmic Dust Analyser (CDA) consists of two indepen-dent instruments: the Dust Analyser (DA) and the High RateDetector (HRD). The HRD was designed to monitor high im-pact rates (up to 10,000 s−1) in dust rich environments suchas Saturn’s ring planes and will not be discussed here. TheDA (Fig. 1) has three different subsystems. The entrance grid

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.3 (1-13)

Composition of jovian dust stream particles 3

Fig. 1. A dust impact on the Chemical Analyser Target (CAT) creates a TOFspectrum. The CAT, which is composed of rhodium, is held at a potential of+1020 V. An electrically grounded grid is mounted 3 mm in front of it. Follow-ing the projectile impact, a plasma cloud expands from the target surface. Thestrong electric field (340 kV m−1) separates the plasma: while the electrons andnegative ions are collected at the CAT (QC signal), the positive ions are acceler-ated towards the grounded grid. After passage through this accelerator grid, theions drift over a distance of 0.225 m through a weak electric field (1.8 kV m−1)towards the multiplier mounted in the centre of the instrument. Depending ontheir mass-to-charge ratio (m/q) the ions reach the multiplier separated by theirtime of flight. About 50% of the incoming ions are collected by concentric gridsin front of the multiplier where their integrated charge is registered by the QIchannel. Behind these grids the passing ions enter a strong field (100 kV m−1)region for final acceleration into the multiplier. Their amplified signal (MP sig-nal) is recorded as a positive ion TOF spectrum. Since TOF = s/v ∼ √

(m/q),this ideally also represents a mass spectrum for identical ion-charges. In real-ity this equation is influenced by a broad distribution of the initial ion velocities�v0, varying flight paths �s, and plasma shielding effects. The drawing is notto scale.

which is sensitive to the charge carried by the particle (QP de-tector), the classical Impact Ionisation Detector (IID) similar toGalileo-type instruments and a time-of-flight (TOF) mass spec-trometer, which is referred to as the Chemical Analyser (CA).

The charge of the tiny jovian stream particles is far belowthe detection threshold of the QP detector grid (10−15 C), thusthis signal is negligible in this context. Depending on the tra-jectory of the particle, it either hits the central rhodium target(Chemical Analyser Target—CAT) with a diameter of 0.17 m,the surrounding gold target (diameter 0.41 m), or the innerwall of the instrument. This work deals with impacts on theCAT only, because only those provide distinct mass spectra. Ifa dust particle impacts with sufficient energy it is totally dis-sociated and ionised, forming an impact plasma of target andparticle ions, together with electrons. The instrument separatesthe—predominately singly charged—positive component of theplasma which is analysed in a TOF spectrum (see Fig. 1 andSections 4.1 and 4.2).

The high rate sample mode of the multiplier recording theTOF spectrum can be triggered either by an impact signal or bya multiplier signal itself. The spectra are digitised with 8-bit res-

olution and are sampled at 100 MHz for a period of 6.4 µs. Sincethe TOF is proportional to the square root of the mass to chargeratio of ions, this ideally also represents a mass spectrum foridentical ion-charges. Unfortunately the TOF is also influencedby a broad distribution of the initial ion velocities, varying flightpaths, and plasma shielding effects. The recording period ofthe high rate sample mode is equivalent to a mass of approx-imately 190 atomic mass units (amu) of a singly charged ion.The spectrometer is only sensitive to positive ions. The massresolution (m/�m) derived from laboratory experiments of theinstrument basically depends on atomic masses of the ions. At1 amu, m/�m is 10, increasing to 30 at 100 amu and up to 50at 190 amu, each value strongly varying with impact conditions(Stübig, 2002). In general this does not allow a detailed qual-itative analysis of complex chemical structures and only verylimited differentiation of isotopes. Nevertheless the spectra al-low classification of different projectile types such as silicates,water ice, organic- or Fe/Ni-particles.

It is important to understand that neither the CDA nor itsprogenitors were designed for detecting stream particles. TheCDA was built and calibrated for impactor masses greater than10−18 kg with velocities below 70 km s−1. This is far beyondthe properties of stream particles, which from simulations arebelieved to have masses around 10−21 kg and impact velocitiesof more than 200 km s−1 (Zook et al., 1996). As a consequencethe physical properties of an individual stream particle cannotbe derived from the impact signals.

3.2. Dataset

The dataset consists of 458 TOF spectra recorded betweenSeptember 4th 2000 and May 28th 2001. This corresponds to aJupiter distance of 1.1 AU at the inbound trajectory and about2 AU on the outbound trajectory. As Cassini is a platform forten instruments with a large number and diversity of requestedoperations, the pointing of the detector often did not allow ob-servations of stream particles. Furthermore, it is important tounderstand that only a small fraction of the particles enteringthe CDA produce a mass spectrum (a total of over 16,000 eventswere registered). During strong impact bursts only a fraction ofthe events were transmitted. The onboard software selected thisfraction via the main criterion of “multiple distinct peak occur-rence.”

To allow a statistical evaluation, similar impact conditionsand a homogeneous data set had to be ensured. Thus, 287spectra comprising the first two bursts registered on Septem-ber 4th–7th (stream 1) and October 1st–4th 2000 (stream 2),are considered in this analysis. The later impacts were recordedunder different and constantly varying impact conditions andare of very limited use for the aim of this work. Besides that,streams 1 and 2 provided the highest impact rates and showedthe best peak definitions. For stream 1 the minimum impactsignal which triggers a “valid event” (trigger threshold) was ac-cidentally set too high. Those events were triggered later by thefirst ions reaching the multiplier (Fig. 1). However, the spec-tra of stream 2 (where the thresholds were set correctly) showno apparent difference. From this we conclude that the higher

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.4 (1-13)

4 F. Postberg et al. / Icarus ••• (••••) •••–•••

thresholds during stream 1 had no influence on the detected par-ticle composition types.

The data for an individual impact consist of the TOF spec-trum, the corresponding QC-amplitudes (number of collectedelectrons at the CAT), the QI-amplitudes (number of posi-tive charges at the multiplier entrance grid), and the impacttimes. The analysis of the individual spectra is performed bya processing routine, providing an unbiased identification andintegration of features.

Due to the extreme physical properties far beyond the cal-ibration range, dynamical information of an individual streamparticle cannot be derived from the impact signals. However,some constraints for velocity and mass can be obtained. It isknown that the rise time of the impact signal is dependent onthe particle velocity (Eichhorn, 1976, 1978). This is causedby impact ejecta hitting the spherical target creating secondaryimpact ionisation events with a time delay dependent on the im-pact speed of the original particle. In fact the signal rise timesof the stream particle impacts were smaller than the signal risetimes of the fastest calibration impact measured in the Heidel-berg dust accelerator facility of about vI = 70 km s−1. Due totheir speed and size, the stream particles are likely to not pro-duce any ejecta at all. In this case the rise time of the impactsignal (QC) is dependent on the expansion time of the impactplasma, which is shorter than the time scale for possible ejectaimpacts and again is an indicator of the particle’s velocity. How-ever, with this method the impact velocity cannot be constrainedany better than to a lower limit of about vI > 100 km s−1.The spacecraft’s speed relative to the jovian system was below13 km s−1 and therefore had only limited influence on the im-pact vectors.

Many of the impacts signals are just at—or even below—thenominal trigger threshold of the target signal. That is consistentwith the very low particle masses predicted by simulations. Ifsimilar composition and speed of the particles is assumed, themasses of the impactors can be roughly compared using theamount of positive charge detected by the QI-signal.

Uncalibrated spectra of events with a low impact signal al-ways have an additional time offset of (0.50 ± 0.04 µs) whichrepresents the TOF of H+. These spectra have not been trig-gered by the impact itself but by H+ ions exceeding the thresh-old at the multiplier. As a consequence, no H+ mass line ap-pears in these spectra although it is an abundant component ofthe impact plasma (Fig. 2). The presence of H+ is a known con-comitant phenomenon of high velocity impacts (Krueger, 1996;Stübig, 2002; Kempf et al., 2005b) (see Section 4.2 for details).As a positive side effect, all events whose impact signal at thetarget is insufficient to trigger the high rate sample mode, stillhave a chance to deliver a spectrum if the event is triggeredby the ubiquitous H+ ions at the multiplier. For a high statis-tical quality it is necessary to mix nominally triggered spectrawhich show a H+ feature with those that are triggered by themultiplier and thus do not have this mass line. As a conse-quence, the H+ features have not been taken into account forthe total spectrum area when relative abundances were cal-culated. This has no influence on the implications of our re-sults.

4. Results

4.1. Identification of ion species

In the following the term “occurrence of features” deter-mines in how many spectra a certain feature appears, while“abundance” refers to the amount of ions of a certain speciesin a spectrum. The “total ion charge” is derived from the QI-amplitude which represents the ion charge of all species mea-sured at the ion grid in front of the multiplier. Dependent ontheir initial energy, approximately 30% of the ions produced atan impact are focussed here by the electrical fields. About 50%of those incoming ions are registered by the grid and deliver the“total ion charge” which in general is proportional to the num-ber of ions produced by an impact (Reber, 1997; Srama, 2000;Stübig, 2002). The ions that pass through the grid form the TOFspectrum.

Table 1 lists properties of the 9 most distinct spectral fea-tures. In general only the 4 most abundant (F1, F3, F5, F9) andF8 show a well defined shape. The weaker features are oftenirregular or barely separable from background noise in individ-ual spectra. This can be improved by coadding and creating sumspectra. As can be seen in Fig. 3 this leads to improved peak de-finition and better signal-to-noise-ratio at the price of a slightlydecreased resolution.

The assignment of masses to this kind of TOF spectrum isnot a simple task. Considering the target peak of rhodium (F9)as an example, it can easily be seen that the ions of one speciesare distributed over a wide range of the TOF-scale (Figs. 2and 3). This primarily is due to their initial angular and energydistribution in the impact plasma. Thus, in a strict sense theapplication of a mass scale to the abscissa is not correct. Forthis work, the mass-calibration was performed using a cross-correlation method. The relative peak positions of four refer-ence species usually present (C+, Na+, K+, Rh+) were used asa template. The mass resolution derived from laboratory exper-iments as described in Section 3.1 could be achieved in manycases.

Except for the broad features F2 and F7, the atomic masslines shown in Table 1 could be identified by the position ofthe peak maximum. The average maximum position of F2 cor-responds to a mass of 5.2 amu. Since no isotope with a massof 5 amu exists, F2 is likely to represent a multiply chargedspecies. Doubly charged ions are expected to appear at a TOFequivalent of half the mass of the singly charged species. Thereis no clear evidence for B+ (isotopes with 10 and 11 amu) inthe spectra, thus a strongly shifted C2+ mass line is the bestcandidate. This assumption is supported by the fact that thereis a clear anti-correlation between the abundance of C+ ions(F3) and F2. Doubly ionised atoms obviously do not stay longenough inside the plasma to recombine. Favoured by their highcharge to mass ratio C2+ ions probably leave the impact plasmaearlier than the singly charged ions used for mass calibration.This makes a substantial shift towards lower TOF plausible (seealso next paragraph).

Especially in strong spectra, a broad feature with varyingand irregular shape extending from 2.65 to 2.9 µs appears (F7).

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.5 (1-13)

Composition of jovian dust stream particles 5

Fig. 2. Examples of two spectra obtained on 2nd October 2000 which have been filtered and calibrated. The upper spectrum is caused by a strong impact(QI-amplitude ≈ 280,000 ions), the lower spectrum reflects a weak impact (QI-amplitude ≈ 40,000 ions). While the upper spectrum shows most of the typicalfeatures as described in Table 1, the weaker spectrum only shows features of carbon (F3), sodium (F5), and the target material rhodium (F9). This event was trig-gered by the multiplier and thus does not show a hydrogen feature. Note, that in contrast to the upper spectrum the first peak in the lower spectrum is an artefact

which indicates the start of the high sample rate mode of the multiplier and not a hydrogen peak.

Table 1Dominant spectral features

Feature 〈tmax〉 (µs) 〈�t〉〈tmax〉 Occurrence Extent (amu) Origin

F1 0.52 2.1% 287 (100%)a 0.9–1.3 H+F2 1.09 5.0% 109 (38%) 4.2–6.0 C2+ (?)F3 1.66 0.5% 287 (100%) 9.0–14.5 C+F4 1.89 1.2% 228 (79%) 14.5–16.5 O+F5 2.32 0.3% 281 (98%) 20–25 Na+F6 2.56 1.6% 190 (66%) 26–30.5 Si+F7 2.77 2.5% 168 (59%) 30.5–37 S+/Cl+F8 3.01 0.6% 225 (78%) 38–41 K+F9 4.84 0.5% 287 (100%) 70–110 Rh+

Note. The second column refers to the average position of the feature’s max-imum and column 3 list its relative variation. Column 5 assigns the commonextent of the feature as indicated by the arrows in Fig. 3 with respect to an idealatomic mass scale.

a Only a part of the spectra show a H+ feature (F1), but the recording of theothers is triggered by H+, so we state an occurrence of 100%.

In other spectra there are one or more small and narrow peakswhich extend only over a part of the TOF-range stated above.

The well defined shape of F7 displayed in Fig. 3 can only beachieved by adding a number of spectra. An analysis of the po-sitions of peak maxima reveals that it is composed of more thanone mass line (Fig. 4). They are most frequent at tmax ≈ 2.70 µs,tmax ≈ 2.80 µs, and tmax ≈ 2.87 µs. Those times are consis-tent with mass lines of sulphur ions (S+, 32 amu) and slightlyshifted mass lines of chlorine ions (Cl+, isotopes with 35 amu(75%) and 37 amu (25%)). Except for Cl+ isotopes, a shift to-wards lower TOF can be observed for the possible C2+ feature(F2) and the O+ mass line (F4). These three features repre-sent the ions with the highest ionisation energies of all observedspecies (Table 2). These ions have a greater chance to recom-bine with electrons which favours survival of those that canescape the impact plasma very quickly. This causes a shift ofthe maximum towards a lower TOF relative to the ions used formass calibration. In most cases, the shifted Cl+ main isotopeat about 2.80 µs (≈34 amu) dominates F7. In contrast to itsless abundant neighbours it can often be integrated with an ac-ceptable error and is used for quantitative correlation tests (seeSection 4.2).

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.6 (1-13)

6 F. Postberg et al. / Icarus ••• (••••) •••–•••

Fig. 3. Sum spectrum of 30 typical TOF spectra. Nine major peak features can be identified (see Table 1). The lower spectrum shows a stretched y-scale. The arrowsshow the common extent of the feature (Table 1, column 5).

Fig. 4. Distribution of peak occurrences in the TOF-range of F6 and F7. Bothfeatures cover a relatively broad range of possible atomic masses. In contrast toF6, the peak maxima of F7 cluster at certain positions. As described in the textthis leads to an identification of sulphur and two chlorine isotopes representedby F7. Though F6 might represent more than one mass line as well, it cannotbe resolved.

Four main species which are familiar from calibration exper-iments can be found in almost every spectrum and on averagerepresent about 95% of the detected ions. Along with the targetions Rh+ (F9) and H+ (F1), the spectra are dominated by theC+ line (F3), which is believed to be due to target contamina-tion (see Section 4.2), and the alkali species Na+ (F5).

As a consequence of the high energy densities after the im-pact, the formation of molecular or cluster ions is unlikely.Thus, we consider all the distinct features shown in Table 1and Fig. 3 to be basically of atomic origin. Except F2, thereis little indication for multiple charged ions as well. This is a

little surprising considering the energy densities involved, butin good agreement with previous results from impact ionisationinstruments at Comet Halley (Brownlee and Kissel, 1990) witha relative impact speed of over 60 km s−1. It appears that only asmall fraction of the kinetic energy is used for ionisation whilemost of the energy is used for excavation and vaporisation. Fur-thermore most of the initially multiply charged ions obviouslyrecombine with electrons inside the impact plasma before theyare accelerated towards the multiplier.

Many more small features than shown in Table 1 can beobserved, but except for a few signatures they do not showa statistical accumulation at a specific region and disappearwhen a number of spectra are added to form a sum-spectrum.However, some of them show up more or less regularly andare worth mentioning. A tiny feature occasionally appears at2.10 µs which is consistent with the hydronium ion (H3O+,19 amu) indicating the presence of traces of water. Anotherregion where small, often irregularly formed features are some-times present is on or preceding the leading flank of therhodium feature (F9). Between 3.45 and 3.8 µs they mightbe associated with the doubly charged target species Rh2+(51.5 amu) or other metal ions such as Cr+ (52, 53 amu), Mn+(55 amu), Fe+ (56 amu), Ni+ (58, 60 amu). We note that,in this work, we cannot determine whether these low signifi-cance features are artefacts, contamination, or particle compo-nents.

Fig. 5 shows how the occurrence of peaks of differentspecies depends on the total ion charge which is a good indi-cator of the impact energy.

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.7 (1-13)

Composition of jovian dust stream particles 7

Fig. 5. Occurrence of mass lines as a function of the total ion charge(QI-amplitude). Carbon and rhodium features (F3, F9) can be observed in everyspectrum. The occurrence of some features clearly depends on the total ioncharge, while others (F6 and F4) do not show a correlation. It has to be pointedout that the overall signal-to-noise ratio of a spectrum depends on the total ioncharge. Thus, the assumption that the occurrence decrease of F2, F7, and F8features is due to an overlay of background noise is consistent.

4.2. Identification of particle ions

Impacts with high energy densities typical of those due to thetiny, high velocity stream particles can neither be reproducedin the laboratory nor have they been simulated on computers.It can be expected that for an impact of nanoparticles withspeeds above 50 km s−1 only a small fraction of the producedions are in fact particle material (Hornung and Kissel, 1994;Hornung et al., 2000). The majority of the plasma ions originatefrom the target material (F9) which contains impurities and sur-face contamination (e.g., F1, F3 (see below)). Due to the smallnumber of particle atoms, species with low abundance or highionisation energies might be overlain by background noise. So,particularly at small total ion charges, a species that is presentin the particles might not be visible in the spectra at all. An-other aspect known from impact experiments is that even withsimilar composition, velocity, and mass of the projectiles, theresulting spectra often appear quite different because of vary-ing local impact conditions on the target. As a consequence,the identification of features associated with the particle is dif-ficult. Only statistical considerations based on a large sample of

impacts can provide a reliable insight into the particle compo-sition.

H+ appeared as a very abundant feature (F1) in our spec-tra. This phenomenon has been frequently observed after highvelocity impacts (Krueger, 1996; Stübig, 2002; Kempf et al.,2005b) where those ions have been ascribed to target contami-nation. All metals can solve hydrogen that occupies interstitialsites or vacancies in the host metals lattice or adsorb it on thesurface. The large features observed (F1) represent far too manyions to consider hydrogen surface adsorption or particle mater-ial as the only source. On the other hand, due to the negativeenthalpy of solution rhodium cannot solve large amounts ofhydrogen atoms in an equilibrium state at low ambient pres-sure (Wipf, 2001). A comprehensive introduction in this fieldis given by Fukai (2005). A possible explanation would be anentry of protons through the solar wind when the target is notshielded by a magnetic field or the spacecraft. With impactspeeds of several hundred kilometres per second the protons areimplanted about 50 nm under the surface. Due to the low tar-get temperature (≈220 K) the diffusion coefficient of hydrogenin the metal lattice is very low. This might lead to an accumu-lation close to the surface where the protons can be releasedthrough a high velocity impact. Another possible source couldbe hydrocarbon contamination of the target. The phenomenonis examined and discussed in further detail by Postberg et al.(2006, in preparation).

When discussing the C+ feature, we consider the impactorsto be high velocity particles with a mass of about 10−21 kg.Those tiny particles are composed of too few atoms to create apeak area of the size of the observed C+ feature (F3), thus theC+ ions cannot be solely of particle origin. As in the spectra ofKronian stream particles (Kempf et al., 2005b) it is the dominat-ing feature especially at the lowest total ion charges (see Fig. 2).In those faint spectra the area of F3 often amounts to signifi-cantly more than 50% of the total spectrum area, representingmany times more atoms than the particles were likely to con-tain. Its increasing relative abundance at low total ion chargesand several other indicators clearly imply C bearing compoundsto be an abundant surface contaminant. The details of this andother contamination phenomena of the CA target are discussedby Postberg et al. (2006, in preparation). The phenomenon isin principle known from high velocity impacts at the CDA cal-ibration experiments at the Heidelberg dust accelerator facility(Reber, 1997; Stübig, 2002). However, the abundance of C+ instream particle spectra obtained at Jupiter and Saturn are aboveall expectations from laboratory experiments and are the sub-ject of further investigation.

Alkali metals like sodium and potassium are well known sur-face contaminants, too. But, as with carbon (F3), the high ionabundance and the high occurrence of the alkali features (F5,F8) is unusual. Alkali signatures can be identified in 98% ofthe spectra. Analysis of the surface of the rhodium plate whichwas used for the CA target shows that the surface is not ho-mogeneously covered with alkali metals and only about 30%(Na) or 15% (K) of the area is affected (Postberg et al., 2006, inpreparation). This is clearly contradictory to an interpretationof all Na and K as contaminants in impact spectra measured

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.8 (1-13)

8 F. Postberg et al. / Icarus ••• (••••) •••–•••

Fig. 6. Detailed view of the part of a coadded spectrum after the rhodium peak,the trailing flank of which is seen on the left. If spectra with a distinct F7 signa-ture (31–37 amu) are co-added, features of significance (F10 and F11) appearin that area. The TOF is with respect to the maximum of the rhodium signature.

Fig. 7. Here only a fraction of the spectra of which Fig. 6 is composed, areco-added. Grey line = peak maxima between 31 and 33 amu. Black line = peakmaxima between 33.5 and 37 amu. Only fractions of F10 and F11 show up.Those features likely represent different target-projectile cluster ions. The vari-ation in the 3-σ line is due to a different number of co-added spectra.

at Jupiter. Measurements of stream particles originating fromSaturn with similar impact properties show an alkali feature inless than 50% of all spectra (Kempf et al., 2005b). Furthermore,the latter show very weak signatures of only traces of Na+ andK+: the average proportion of alkali ions in the Kronian duststream spectra is more than a magnitude lower than in spec-tra recorded at Jupiter. The tiny alkali features seen in Kronianstream spectra likely represent the alkali surface contaminationwhich clearly favours sodium to be a substantial particle com-ponent of the jovian dust streams.

Another method to identify major particle components isthrough target-projectile cluster ions. For decades, laboratoryexperiments have demonstrated that the most abundant particleions may form clusters with the target material (Knabe, 1983).Unfortunately, this is only effective at low energies, so in thecase of jovian stream particles these types of mass lines cannotbe expected to be very noticeable. Nevertheless, even an iden-tification of faint target–cluster mass lines is a strong indicatorfor a major particle component, provided that the correspond-ing projectile mass line shows up in the spectrum as well. Thishas already been demonstrated with Kronian stream particles(Kempf et al., 2005b).

Peaks which are consistent with target-projectile-ions ex-ceeding a 3-σ significance of the noise amplitude can barely

Fig. 8. Correlation between the abundances (as a percentage of the total spec-trum area) of the species responsible for F3 (sodium) and F8 (potassium; brokenline) and chlorine (solid line). For the latter, only the part of F7 features whichare dominated by Cl+ ions are taken into account (maximum between a TOFof 2.75 and 2.85 µs, equivalent 33 to 35.5 amu).

be found. However, when certain spectra are co-added, the sumspectrum reveals some weak features of this kind. Adding allstream 1 spectra that show clear F7 features (30 to 37 amu)reveals two new peaks as can be seen in Fig. 6. They occur be-tween 0.45 and 0.55 µs (F10), and between 0.65 and 0.85 µs(F11) after the rhodium peak maximum. F11 corresponds to atarget-projectile cluster ion of the form RhX+, with X beingthe projectile species of a mass between 30 and 37 amu. Withthose target-projectile-features of 3–5σ significance, a correla-tion can be proved: If only the subset of the spectra with F7signatures that show peak maxima between 31 and 33 amu areused for the sum spectrum, the corresponding RhX+ featuregets narrower with a maximum 0.69 µs (≈31 amu) after thetarget feature (Fig. 7, grey spectrum). The sum of spectra withmaxima between 33.5 and 37 amu shows an equivalent localmaximum 0.8 µs (≈36 amu) behind the rhodium peak (Fig. 7,black spectrum). So the species that form the small, broad F7signature all have “mirror peaks” as a target-projectile-ion. Thisimplies that those ions, namely Cl+ and S+, originated frommajor particle components. It is remarkable that F10 is not ap-parent in the grey spectrum of Fig. 7, while it is pronouncedin the black spectrum of Fig. 7 with a maximum at 0.51 µs(≈23 amu) after the Rh+ peak, which is consistent with aRhNa+ cluster ion. The possible conjoint appearance of RhCl+and RhNa+ cluster ions indicate a correlation of the alkali metalwith chlorine in the particle.

Qualitative and quantitative correlations between mass linesare an indicator for a common source. For that reason, correla-tion analysis has been a major field of research during the statis-tical examination of the data. As shown in Fig. 8 there are clearquantitative correlations between Na+ and K+ as well as Na+and Cl+. If both elements are present, the extremely reactivesodium and chlorine immediately form NaCl, which supportsthe assumption of NaCl being the common parent molecule ofNa+ and Cl+. The correlation between Na+ and K+ stronglyindicates that potassium is part of the particle, too.

Calibration experiments with the CDA flight spare at theHeidelberg dust accelerator facility demonstrated that at high

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.9 (1-13)

Composition of jovian dust stream particles 9

Fig. 9. Relative abundance (as a percentage of the total spectrum area) of theion yield of sodium, oxygen and silicon features in dependence on the totalion charge (QI-amplitude). Only spectra where a distinct peak of the speciesis observed are considered. It is obvious that the proportion of Na+ increaseswhile the proportion of Si+ and O+ decreases.

velocities the contribution of surface contaminants to the re-sulting mass spectra decreases with increasing particle mass(Stübig, 2002). This is due to the fact that the amount of surfacecontamination scales with the projected grain area while the to-tal charge scales with the grain volume (in a regime where theparticle becomes completely ionised). The same applies to tar-get material: while the amount of particle ions increases linearlywith particle mass mp the generated target ions only increase

with m2/3p (Krueger, 1996). This leads to a steadily increasing

particle proportion with increasing particle mass for a givenvelocity. As can be seen in Fig. 9, the relative abundance ofNa+ in the spectra quadruples from less than 3% at low totalion charges to 11.5% at the highest total ion charges. Such adistinct increase of a particle-species indicates that for the ob-served impacts the varying total ion charges basically dependon variations of the particle mass. If the particle velocity wereresponsible for the variation of total ion charge, this would leadto a relative decrease of particle ions with increasing total ioncharge.

Despite the frequent observation of O+ and Si+ mass lines—which could imply the presence of silicates—there is little in-dication that those ions represent particle material. O+ and Si+show neither a significant mutual correlation as should be ex-pected for silicates nor a correlation with any of the other pos-sible particle materials identified above. Furthermore, Si+ doesnot have an associated target–cluster-ion RhSi+, as in the caseof the Kronian stream particles (Kempf et al., 2005b). In ad-dition, the relative abundance of Si+ and O+ in the spectradecreases with increasing total ion charge as can be seen inFig. 9. For low total ion charges the features of Si+ and Na+ onaverage are of about the same magnitude, whereas for the high-est ion charges the Na+ projectile ions are over ten times moreabundant than the Si+ ions. Finally, there is only weak evidencefor mass lines of metals that typically accompany rocky miner-als (Fe, Mg, Ca, Al, Ni), which in general are much more likelyto be detected as positive ions than Si or O. Silicon is a knownimpurity of the rhodium plate material (Tack, 1992. Test Re-port 101 3 L001. Technical report, Heraeus GmbH, Hanau) and

tends to accumulate in “islands” on and below the rhodium sur-face (Lura, F., 1997. Cassini CDA/TDT Schlussbericht. Techni-cal report, DLR GmbH, Berlin-Adlershof). This is in agreementwith the fact that the occurrence of Si+ signatures is basicallyindependent of the total ion charge. Nevertheless, as a resultof TOF-SIMS analysis of the rhodium target surface only about5% of its volume shows silicon impurities (Postberg et al., 2006,in preparation), which is certainly not enough to explain the oc-currence of the mass line. However, if silicates are part of theparticle then their abundance appears to be quite independentof the projectile masses.

An absolute pure metal surface even of a noble metal likerhodium does not exist when exposed to reactive substances.Oxygen from the Earth’s atmosphere could have oxidisedrhodium and formed a monolayer of Rh2O3 on the metal sur-face representing a plausible source for the observed O+ ions.However, in many spectra the feature appears to be too big tobe solely caused by a thin Rh2O3 layer. As in the case of Si+the great abundance of O+ could be explained if a fraction ofthe ions were of particle origin.

The rare occurrence of a weak feature consistent with watercould be due to contamination or a rare particle component.

5. Discussion

The following section is divided into four parts. In the firstsubsection mainly qualitative aspects of particle componentsare considered to derive the source of the stream particles.Quantitative estimates which are associated with greater uncer-tainties are discussed in Section 5.2. In Section 5.3 an estimatefor the particle masses derived from the number of ions in thespectra is demonstrated. Thus far the particle composition hasbeen discussed based solely on the spectra analysis. The fi-nal subsection is about the implications of those results for theparticle formation in volcanic plumes on Io. Findings from re-search of Io’s atmospheric chemistry are combined with ourresults and a particle formation process, as well as a refinedpicture of the particle composition, are suggested.

5.1. Composition and source of particles

As a consequence of the results presented above, we canidentify sodium chloride (NaCl) as an important particle com-ponent. Because of the quantitative correlations Na+ ⇔ K+,NaCl is accompanied by K+. As an extremely reactive elementpotassium has to be bound in a compound (KXy), maybe KCl.Except for some of the faintest signals we observe sodium andpotassium ions in every spectrum, indicating that the vast ma-jority of the particles are of the NaCl/KXy composition-type.The fact that the Cl+ mass line occurs in fewer spectra thanthe alkali metals does not necessarily mean that Cl− is not theirassociated anion in the particle. Its high ionisation energy (Ta-ble 2) allows an immediate recombination of many initial Cl+ions. Thus, especially for small total ion charges it is quite rea-sonable that a Cl+ line could be overlain by background noise.It is not surprising as well, that there is a weaker indicationfor target-projectile-ions of the type RhNa+ and RhK+ than for

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.10 (1-13)

10 F. Postberg et al. / Icarus ••• (••••) •••–•••

Table 2Ionisation energies and electron affinities of some selected elements

Element E0 −→ E+ (eV) E+ −→ E2+ (eV) E0 −→ E− (eV)

Carbon 11.25 24.35 −1.60Oxygen 13.62 35.12 −1.46Sodium 5.14 47.27 −0.55Silicon 8.16 16.34 −1.39Sulphur 10.36 23.34 −2.07Chlorine 12.97 23.81 −3.62Potassium 4.34 31.63 −0.50Iron 7.91 16.19 −0.16Rhodium 7.46 18.03 −1.14

RhS+ and RhCl+. For alkali ions it is energetically much morefavourable to “stay alone” than for elements with high ionisa-tion energies (see Table 2).

The occurrence and area of mass lines consistent with S+ isquite low. Nevertheless, in laboratory experiments target clus-ter ions have only been observed with major particle and targetcomponents. Thus the presence of corresponding cluster ionsindicates that sulphur plays an important role in the elementalcomposition of at least some particles. S can either be presentin its elemental form or, considering the other possible particleelements, probably in an alkali compound or in SOx .

Whilst the signatures of Si+ and O+ are consistent with tar-get contamination as a source of these ions, the abundance andfrequent occurrence of both species is difficult to explain solelywith impurities. This leads to the conclusion that these speciesare probably present in the particles themselves.

As a consequence of C+ and H+ being an abundant targetcontaminant, carbon and hydrogen can neither be excluded norconfirmed to be part of the elemental composition of jovianstream particles.

The strong indication for sulphur being part of many of theparticles confirms Io as a source for the dust streams. NaClis a known component of Io’s volcanic ejecta as well. It isevident that it does escape from the moon’s gravitational in-fluence, forming the major source for Na and Cl species inthe ionian system (Fegley and Zolotov, 2000; Küppers andSchneider, 2000; Schneider et al., 2000; Moses et al., 2002a;Lellouch et al., 2003; Mendillo et al., 2004). Since sodiumand especially the highly volatile chlorine are supposed to beonly minor components of interplanetary minerals (Küppersand Schneider, 2000), it is unlikely they are major componentsof the jovian ring particles. Thus, our results support Io beingthe source for the vast majority, maybe all the stream particlesdetected far outside the jovian magnetosphere. A minor contri-bution of Jupiter’s gossamer rings with an upper limit of about5% of the observed impacts, is consistent with our data. Thisestimate is done under the assumption that gossamer ring par-ticles would not have an abundant alkali or sulphur componentand do mainly consist of rock forming minerals.

5.2. Quantitative estimations of the particle composition

All previous applications of impact ionisation have shownthat many unknown parameters have influence on the relative

ion yield of the different species in an impact plasma. This re-flects the complexity of the impact process. Thus, derivation ofquantitative statements about the particle’s composition is as-sociated with major uncertainties. However, the task gets lesscomplicated if we consider all particle atoms to be ionised inthe impact plasma. As a matter of fact, with impact veloci-ties and particle masses of the jovian streams stated above,the energy densities are high enough. We have reason to as-sume that the entire particle material was initially convertedinto mainly positive atomic ions and electrons (Hornung, 2004,private communication). As a consequence, the influence ofthe various ionisation efficiencies of atomic species is reduced.In an environment with such energy densities matrix effects,which reflect the interaction of different molecular or atomicspecies, are suppressed as well. But still it has to be pointedout, that the quantitative considerations made in the followinghave a speculative component.

The ionisation energy of atoms (Table 2) is one known keyparameter for the relative ion yields (Kissel and Krüger, 1987).It is an important indicator for the formation and recombina-tion probability of ions and electrons inside the impact plasmabefore they can be accelerated by the external electric field.

The influence of this parameter for the relative ion yieldsof jovian stream particles can be observed by comparing theion yields of Na+ and Cl+. If we consider NaCl as the parentmolecule, the abundance of both species in the particles can beassumed to be roughly identical. This assumption is supportedby the finding that volcanically emitted NaCl is the dominantsource for Na and Cl ions in the gas of the upper Io atmosphere,the plasma torus and the neutral cloud (Fegley and Zolotov,2000; Küppers and Schneider, 2000; Schneider et al., 2000;Moses et al., 2002a; Lellouch et al., 2003). In the spectra about6–10 times more Na+ than Cl+ appears (Fig. 8). Consideringthe extreme difference of ionisation energies between the al-kali and halogen species, this ratio is surprisingly low. Previouscalculations for high velocity impacts would imply Na+/Cl+ratios of 50/1 or more (Krueger, 1996). This is due to the factthat in all previous cases much lower energy densities wereconsidered, where the ionisation efficiencies clearly favour theformation of Na+. In contrast, all jovian stream particles areassumed to be initially ionised regardless of their ionisation en-ergy, which in this case only has influence on the recombinationof ions.

On average Na+ is about 5 times more abundant than K+(Fig. 8). As the latter has the lowest tendency to recombine(Table 2), this ratio is likely to be exceeded in the particle, indi-cating K bearing components to be only a minor companion ofNaCl.

The results imply that the proportion of sulphur is less thanthe proportion of chlorine for the observed particles. Cl has ahigher ionisation energy than S, nevertheless the F7 feature ingeneral represents more Cl+ than S+ and the maxima consistentwith Cl+ occur more frequently (Fig. 4). Although unknownmatrix effects might have an unfavourable influence on the re-lease of S+, it is unlikely that they could compensate the consid-erable difference of ionisation energies. This finding is in con-trast to the estimated gas emission into Io’s atmosphere where

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.11 (1-13)

Composition of jovian dust stream particles 11

sulphuric components—especially SO2—are clearly dominant(Lellouch et al., 2003). This phenomenon is extensively dis-cussed in Section 5.4.

A part of the observed Si+ and O+ might be of particle ori-gin. However, the absence of RhSi+ target-projectile ions andappropriate metals, imply that silicate minerals represent onlya minor component. The yield of Na+, K+, and Cl+ clearlyenlarges with increasing total ion charge. In contrast, the occur-rence and the abundance of the Si+ and O+ features show onlylittle dependency on the total ion charge (Figs. 5 and 9). Thus,if silicate signatures are not exclusively due to contamination,their contribution to the particle seems to be basically indepen-dent of its mass. This interpretation would be consistent with asilicate rich condensation nucleus in the particle.

Due to the omnipresent huge contaminant-signatures of car-bon and hydrogen in the spectra, it cannot be ruled out that theyplay a role in the particles’ elemental composition in any pro-portion. Molecules composed of those elements (e.g., water ororganic compounds) are not known to play a relevant role inIo’s volcanic activities, its atmosphere or exosphere (Zolotovand Fegley, 2000; Schaefer and Fegley, 2005b). Thus it can beexpected that their contribution to the dust streams is small ornegligible.

The picture we finally get from the quantitative considera-tions inferred from the spectra is as follows: NaCl is the mainconstituent of the observed jovian stream particles, while potas-sium bearing components play a minor role. Sulphur or sul-phurous components are other important constituents. No def-inite conclusion can be drawn regarding the presence of waterice. Silicates or rocky minerals could be a particle component,representing a minor contribution, possibly located in the graincores. A more refined estimation can be given when consider-ations about condensation mechanisms in volcanic plumes aretaken into account (Section 5.4).

5.3. Estimation of particle masses

Under the assumption that the vast majority of the observedNa+ is of particle origin and the recombination of alkali met-als is negligible, we can estimate a lower limit for the numberof sodium atoms in the particle. Dependent on their initial en-ergy about 15% of the ions released by the impact plasma aredetected by the ion collector grid (QI-channel) (Srama, 2000;Stübig, 2002). This is almost equivalent to the number of ionsreaching the multiplier. An average QI-signal represents about100,000 ions. Taking a correction for the abundance of H+ions (Reber, 1997; Stübig, 2002) into consideration we detecta mean value of about 2500 Na+ ions in the spectra. This cor-responds to 16,000 Na atoms in an average dust grain. Further-more it can be assumed that the Na atoms are basically boundin NaCl molecules as the most abundant particle constituent byfar. Thus it yields a good estimation for particle masses. Thesemolecules with a mass of 58 amu each, represent a total massof about 1.5 × 10−21 kg in an average dustgrain. Correspond-ing to the area of the Na+ peak, the variation of that value isabout one order of magnitude up and down which is equivalentto particle radii ranging from 5 to 25 nm (assuming a density

of 2200 kg m−3). This is in very good agreement with the par-ticle masses derived from simulations on basis of Ulysses data(Zook et al., 1996).

5.4. Implications for dust formation in Io’s atmosphere

In contrast to the gas components the knowledge aboutdust formation and composition in Io’s atmosphere and ex-osphere is basically speculative. NaCl is the parent moleculefor the vast majority of the Na and Cl species detected inIo’s gas and plasma environment (Fegley and Zolotov, 2000;Küppers and Schneider, 2000; Schneider et al., 2000; Moses etal., 2002a; Lellouch et al., 2003). Furthermore, its abundanceis dependent on the volcanic activity on the moon (Lellouchet al., 2003; Mendillo et al., 2004). However, it is less abun-dant than species derived from SO2 gas, the main compo-nent of the ionian atmosphere. The current estimation forthe disc averaged proportion of NaCl in the atmosphere is0.3%, though higher concentrations probably occur in volcanicplumes (Lellouch et al., 2003). The Na+, Cl+ concentration inthe plasma torus amounts to 2–6% each (Küppers and Schnei-der, 2000; Schneider et al., 2000). So obviously the finding thatNaCl is more abundant than sulphur in stream particles needsa mechanism that explains the inversion of these proportions.To understand what affects the particle’s composition we haveto understand the dust formation and the escape mechanismson Io.

Krüger et al. (2003) showed a correlation between the oc-currence of the high plumes of the Pele-type (named after thelargest ionian volcano) and dust stream impacts on the Galileodust detector. Dust particles in the plume reaching a sufficientaltitude are charged by the capture of ionospheric electrons andsubsequently accelerated through the magnetosphere. Only par-ticles which reach the top of the plumes at an altitude of almost400 km have a chance to escape (Flandes, 2004). The flighttime from the vent to this height is about 15–20 min. Thus theCDA spectra imply NaCl to be the most abundant component innanometre-sized particles at the top of Pele-type plumes 20 minafter the ejection from the vent.

In smoothed ultraviolet Voyager images of the Pele plumea bright top appears (Strom and Schneider, 1982). The au-thors suggested that this top is caused by a concentration ofparticles. Simulations of Pele-type plumes with entrained parti-cles demonstrated that only spherical particles with sizes below10 nm track the volcanic gas flow (Zhang et al., 2004). Further-more it was shown that those particles could reach the shockand concentrate there (Moore et al., 2003). The condensationof particles on refractory nuclei inside the plume is likely. Dueto their fluffy structure they can track the gas flow much bet-ter than compact spherical particles. Thus particularly thosecondensed particles can reach high altitudes even if their sizeexceeds 10 nm (Zhang et al., 2004). Like most authors, Zhang etal. (2004) assumed the possible particles in the plume to consistmainly of condensed or liquid SO2 because of its great abun-dance in the atmosphere and on the surface. This assumptionhowever, is neither required to explain current observations nora thermochemical necessity.

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.12 (1-13)

12 F. Postberg et al. / Icarus ••• (••••) •••–•••

The swiftly changing equilibria of condensed and gas phasein a volcanic plume are not easy to model. Moses et al.(2002a, 2002b) considered chemical kinetics, photolysis, dif-fusion and condensation of volatiles outgassed from a Pele-type vent with a temperature of 1440 K in a one-dimensionalmodel. They showed that due to their high condensation tem-perature NaCl and KCl are abundant condensates 20 min af-ter outgassing from the vent. About 5% NaCl and KCl ofthe initial amount in the gas-phase form stable condensateswith low vapour pressure. Under these ambient conditions sul-phuric components (except a minor contribution of Na2SO4)are still far from condensation. In their simplified model Moseset al. assumed the vapour to condense about preexisting con-densation nuclei of refractory materials, present proportionalto the atmospheric pressure. Thermochemical equilibrium con-siderations of the alkali and halogen species in volcanic gasesfrom silicate magmas on Io at 1760 K were accomplished bySchaefer and Fegley (2005a). It was shown that below 1000 KNaCl and KCl are by far the most abundant condensates fora wide range of pressure and temperature. Na2SO4 condensesas a minor component and solid KCl converts to K2SO4 in aSO2 atmosphere below 620 K. When temperature drops be-low 500 K sulphur condenses as S8, still far above the con-densation temperature of SO2 which is about 115 K depend-ing on the atmospheric pressure. However, in the model ofMoses et al. (2002a) S8 contributes only a negligible propor-tion of the condensed phase. So the alkali sulfates Na2SO4and K2SO4 could represent the main sulphuric components ob-served in the TOF spectra with the latter being the main potas-sium bearing component. Fegley and Zolotov (2000), Moseset al. (2002a), Lellouch et al. (2003), and Schaefer and Fe-gley (2005a) consider a sodium/potassium ratio in volcanicejecta of about 10/1. This ratio is in good agreement with ourestimations for stream particles. The declining proportion ofO+ with increasing total ion yield (Fig. 9) can be explainedwith the finding that a part of the observed O+ is due to tar-get contamination and/or a silicate core. Anyway it indicatesthat alkali sulfates cannot be a very abundant particle compo-nent.

Sulphur species like SO2 or SO definitely condense whencontacting Io’s cold surface. Condensation probably also occursin shocks or colder regions of the plume at greater horizontaldistance from the vent. However, the results imply that thereare probably no pure SO2 condensates with a size of 5–25 nmpresent in Pele-type plume tops and thus these condensates donot appear to play an important role within the first minutes af-ter outgassing from the volcano’s vent. An alternative scenariowould be that SO2 condenses on particle surfaces as the finallayer of the condensation cascade. Subsequent sputtering of thestream particle’s surface on its voyage through the ionosphereand the plasma torus of Io would lead to dissociation and evap-oration of the highly volatile SO2 frost.

6. Conclusion

The Cassini spacecraft flew by the jovian system in late2000. The Cosmic Dust Analyser onboard was able to record

TOF mass spectra of positive ions released after impact eventsof jovian dust stream particles. We present the results of acomprehensive statistical analysis of 287 datasets obtained insitu at a Jupiter distance of about 1 AU. An extraction ofthe chemical composition of the particles has been accom-plished.

• We identify sodium, chlorine, sulphur, and potassium asparticle components. The continuous occurrence of pro-nounced alkali features is in contrast to the tiny alkali sig-natures occasionally observed in spectra of Kronian streamparticles (Kempf et al., 2005b). This rules out that the alkalifeatures recorded in the jovian system only represent con-tamination of the target surface. Due to a clear quantitativecorrelation NaCl is the parent molecule of the majority ofthe recorded Na+ and Cl+ ions. Silicates might be anotherconstituent. From this chemical fingerprint Io’s volcanoesare identified as the source for the vast majority of, and pos-sibly all, stream particles detected.

• NaCl is the main constituent of the observed jovian streamparticles, while potassium plays only a minor role. Sulphuror sulphurous components are other important constituents.If silicates or rocky minerals are present in a particle, theyrepresent only a minor contribution, probably located in thegrain cores.

• The size and mass of the particles predicted by simulationscan be confirmed. The mass of an average particle is es-timated at about 1.5 × 10−21 kg which is equivalent to aparticle radius of 12 nm.

• Considering NaCl as the most abundant particle compo-nent contrasts with (rather than contradicts) observations ofgaseous components in the ionian atmosphere, the plasmatorus and the neutral cloud where sulphur species are dom-inant while alkali and halogen species are minor compo-nents. We suggest a condensation cascade inside eruptingvolcanic gases probably around pyroclastic silicate cores.Due to their high condensation temperatures sodium andpotassium salts condense prior to sulphuric compounds.The resulting particles escape from the top of Pele-typeplumes probably before the abundant SO2 could condense.If the conditions allow SO2 to become solid on the particlessurface, subsequent sputtering or heating in the ionosphereand the plasma torus yields dissociation and evaporation ofthe highly volatile SO2 frost.

• The results of the thermochemical considerations bySchaefer and Fegley (2005a) are qualitatively and quanti-tatively in good agreement with our findings. A speculativepicture of the particle considering both results can be givenas follows. The main constituent is NaCl. Na2SO4 andK2SO4 are minor constituents each representing 5–10%of the NaCl mass. Depending on the condensation temper-atures sulphur molecules might have a small contributionon the surface. The particle is probably condensed arounda silicate core whose mass has only a limited dependencyon the total particle mass.

ARTICLE IN PRESS YICAR:7897JID:YICAR AID:7897 /FLA [m5+; v 1.56; Prn:21/03/2006; 11:08] P.13 (1-13)

Composition of jovian dust stream particles 13

References

Brown, R.A., 1974. Optical line emission from Io. In: Exploration of the Plan-etary System. Proceedings of the IAU Symposium, Torun, Poland, Septem-ber 5–8, 1973. D. Reidel, Dordrecht/Boston, pp. 527–531.

Brownlee, D.E., Kissel, J., 1990. The composition of dust particles in the en-vironment of Comet Halley. In: Mason, J.W. (Ed.), Comet Halley: Inves-tigations, Results, Interpretations. Dust, Nucleus, Evolution, vol. 2. EllisHorwood/Prentice–Hall, Chichester/Englewood Cliffs, NJ, p. 89.

Burns, J.A., Showalter, M.R., Hamilton, D.P., Nicholson, P.D., de Pater, I.,Ockert-Bell, M.E., Thomas, P.C., 1999. The formation of Jupiter’s faintrings. Science 284, 1146–1150.

Eichhorn, G., 1976. Analysis of the hypervelocity impact process from impactflash measurements. Planet. Space Sci. 24, 771–781.

Eichhorn, G., 1978. Primary velocity dependence of impact ejecta parameters.Planet. Space Sci. 26, 469–471.

Fegley, B., Zolotov, M.Y., 2000. Chemistry of sodium, potassium, and chlorinein volcanic gases on Io. Icarus 148, 193–210.

Flandes, A., 2004. Dust escape from Io. Geophys. Res. Lett. 31, doi:10.1029/2004GL020046. L16802.

Fukai, Y. (Ed.), 2005. The Metal–Hydrogen System: Basic Bulk Properties,second ed. Springer, Berlin.

Geissler, P.E., 2005. Volcanic plumes and plume deposits on Io. Lunar Planet.Sci. 36. Abstract 1875.

Graps, A., Grün, E., Svedhem, H., Krüger, H., Horányi, M., Heck, A., Lam-mers, S., 2000. Io as a source of the jovian dust streams. Nature 405, 48–50.

Grün, E., and 22 colleagues, 1993. Discovery of jovian dust streams and inter-stellar grains by the Ulysses spacecraft. Nature 362, 428–430.

Grün, E., and 21 colleagues, 1996. Constraints from Galileo observations onthe origin of jovian dust streams. Nature 381, 395–398.

Hamilton, D., Burns, J., 1993. Ejection of dust from Jupiter’s gossamer ring.Nature 364, 695–699.

Hillier, J.K., Green, S.F., McBride, N., Schwanethal, J.P., Srama, R., Kempf,S., Postberg, F., McDonnell, J.A.M., Grün, E., 2006. The composition ofSaturn’s E ring. Science. Submitted for publication.

Horányi, M., Morfill, G., Grün, E., 1993a. Mechanism for the acceleration andejection of dust grains from Jupiter’s magnetosphere. Nature 363, 144–146.

Horányi, M., Morfill, G., Grün, E., 1993b. The dusty ballerina skirt of Jupiter.J. Geophys. Res. 98, 21245–21251.

Hornung, K., Kissel, J., 1994. On shock wave impact ionisation of dust parti-cles. Astron. Astrophys. 291, 324–336.

Hornung, K., Malama, Y., Kestenboim, K., 2000. Impact vaporisation and ion-isation of cosmic dust particles. Astrophys. Space Sci. 274, 355–363.

Kempf, S., Srama, R., Horanyi, M., Burton, M., Moragas-Klostermeyer, G.,Roy, M., Helfert, S., Grün, E., 2005a. Discovery of high-velocity duststreams originating from Saturn. Nature 433, 289–291.

Kempf, S., Srama, R., Postberg, F., Burton, M., Green, S.F., Hillier, J.K.,McBride, N., McDonnell, J.A.M., Moragas-Klostermeyer, G., Roy, M.,Grün, E., 2005b. Composition of saturnian stream particles. Science 307,1274–1276.

Kissel, J., Krüger, F.R., 1987. The organic component in dust from Comet Hal-ley as measured by the PUMA Mass Spectrometer on board VEGA 1. Appl.Phys. A 326, 301–317.

Knabe, W., 1983. Massenspektrometrische Untersuchung der Ionenbildungbeim Einschlag schneller Staubteilchen. Ph.D. thesis, Heidelberg Univer-sity.

Krueger, F.R., 1996. Ion formation by high- and medium-velocities dust im-pacts from laboratory measurements and Halley results. Adv. Space Res. 17,71–75.

Krüger, H., 2003. Jupiter’s Dust Disc, An Astrophysical Laboratory. Shaker-Verlag, Aachen.

Krüger, H., Geissler, P., Horányi, M., Graps, A., Kempf, S., Srama, R.,Moragas-Klostermeyer, G., Moissl, R., Johnson, T., Grün, E., 2003. Jov-ian dust streams: A monitor of Io’s volcanic plume activity. Geophys. Res.Lett. 30. 3-1–3-4.

Krüger, H., Graps, A., Flandes, A., Forsyth, R.J., Hamilton, D.P. Horányi, M.,Grün, E., 2005. Ulysses jovian latitude scan of electromagnetically inter-acting dust streams. Planet. Space Sci. Submitted for publication.

Küppers, M., Schneider, N.M., 2000. Discovery of chlorine in the Io torus.Geophys. Res. Lett. 27, 513–516.

Lellouch, E., Strobel, D.F., Belton, M.J.S., Summers, M.E., Paubert, G.,Moreno, R., 1996. Detection of sulphur monoxide in Io’s atmosphere. As-trophys. J. 459, L107–L110.

Lellouch, E., Paubert, G., Moses, J.I., Schneider, N.M., Strobel, D.F., 2003.Volcanically emitted sodium chloride as a source for Io’s neutral cloudsand plasma torus. Nature 421, 45–47.

Mendillo, M., Wilson, J., Spencer, J., Stansberry, J., 2004. Io’s volcanic controlof Jupiter’s extended neutral clouds. Icarus 170, 430–442.

Moore, C., Zhang, J., Goldstein, D.B., Varghese, P.L., Trafton, L., 2003. Mod-elling of particulates and condensates in Io’s Pele-type volcanic plumes.Lunar Planet. Sci. 34. Abstract 2102.

Moses, J.I., Zolotov, M.Y., Fegley, B., 2002a. Alkali and chlorine photo-chemistry in a volcanically driven atmosphere on Io. Icarus 156, 107–135.

Moses, J.I., Zolotov, M.Y., Fegley, B., 2002b. Photochemistry of a volcani-cally driven atmosphere on Io: Sulphur and oxygen species from a Pele-typeeruption. Icarus 156, 76–106.

Na, C.Y., Trafton, L.M., Barker, E.S., Stern, S.A., 1998. Note: A search for newspecies in Io’s extended atmosphere. Icarus 131, 449–453.

Ockert-Bell, M.E., Burns, J.A., Daubar, I.J., Thomas, P.C., Veverka, J., Bel-ton, M.J.S., Klaasen, K.P., 1999. The structure of Jupiter’s ring system asrevealed by the Galileo Imaging Experiment. Icarus 138, 188–213.

Poulet, F., Cruikshank, D.P., Cuzzi, J.N., Roush, T.L., French, R.G., 2003. Com-positions of Saturn’s rings A, B, and C from high resolution near-infraredspectroscopic observations. Astron. Astrophys. 412, 305–316.

Reber, O., 1997. Bestimmung der Energie- und Winkelverteilung von Primäri-onen beim Einschlag von Hochgeschwindigkeits-Mikrometerpartikeln aufOberflächen und deren Anwendung auf Einschlagssensoren an Bord inter-planetarer Raummissionen. Ph.D. thesis, Heidelberg University.

Schaefer, L., Fegley, B., 2005a. Alkali and halogen chemistry in volcanic gaseson Io. Icarus 173, 454–468.

Schaefer, L., Fegley, B.J., 2005b. Predicted abundances of carbon compoundsin volcanic gases on Io. Astrophys. J. 618, 1079–1085.

Schneider, N.M., Park, A.H., Küppers, M.E., 2000. Spectroscopic studies ofthe Io torus during Galileo encounters: Remote plasma diagnostics and thedetection of Cl++. Bull. Am. Astron. Soc. 32. Abstract 35.03.

Showalter, M.R., Burns, J.A., Cuzzi, J.N., Pollack, J.B., 1985. Discovery ofJupiter’s “gossamer” ring. Nature 316, 526–528.

Showalter, M.R., Burns, J.A., de Pater, I., Hamilton, D.P., Horanyi, M., 2003.Recent Hubble observations of Jupiter’s ring system. In: AAS/Division forPlanetary Sciences Meeting 35. Abstract 11.08.

Srama, R., 2000. Vom Cosmic-Dust-Analyzer zur Modellbeschreibungwissenschaftlicher Raumsonden. Ph.D. thesis, Technische UniversitaetMünchen.

Srama, R., and 43 colleagues, 2004. The Cassini Cosmic Dust Analyser. SpaceSci. Rev. 114, 465–518.

Strom, R.G., Schneider, N.M., 1982. Volcanic eruption plumes on Io. In: Satel-lites of Jupiter. Univ. of Arizona Press, Tucson, AZ, pp. 598–633.

Stübig, M., 2002. New insights in impact ionisation and in time-of-flight massspectroscopy with micrometeoroid detectors by improved impact simula-tions in the laboratory. Ph.D. thesis, Heidelberg University.

Thomas, N., 1996. High resolution spectra of Io’s neutral potassium and oxygenclouds. Astron. Astrophys. 313, 306–314.

Wong, M.-H., 2004. Groundbased near-IR spectroscopy of Jupiter’s main ring.In: AAS/Division for Planetary Sciences Meeting 36. Abstract 07.11.

Wipf, H., 2001. Solubility and diffusion of hydrogen in pure metals and alloys.Phys. Scripta T94, 43–51.

Zhang, J., Goldstein, D.B., Varghese, P.L., Trafton, L., Moore, C., Miki, K.,2004. Numerical modelling of ionian volcanic plumes with entrained par-ticulates. Icarus 172, 479–502.

Zolotov, M.Y., Fegley, B., 2000. Volcanic degassing of hydrogen compoundson Io. Lunar Planet. Sci. 31. Abstract 1186.

Zook, H., Grün, E., Baguhl, M., Hamilton, D., Linkert, G., Liou, J.-C., Forsyth,R., Phillips, J., 1996. Solar wind magnetic field bending of jovian dust tra-jectories. Science 274, 1501–1503.