Ion Partitioning in the Hot Io Torus' The Influence of

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 90, NO. A12, PAGES 12,065-12,072, DECEMBER 1, 1985 Ion Partitioning in the Hot Io Torus' The Influence of Outgassing MIGUEL A. MoREno Department of Earth and Space Sciences and Institute of Geophysics and Planetary Physics University of California, Los Angeles WILLIAM I. NEWMAN Department of Earth and Space Sciences and Department of Astronomy University of California, Los Angeles MARGARET G. KIVELSON Department of Earth and Space Sciences and Institute of Geophysics and Planetary Physics University of California, Los Angeles Calculations of ion partitioning in the hot outer Io torus have failed to account for certain observed features. Notably, the low concentration of OIII measured in the first half of 1981 has been difficult to reconcile with abundances of other ions inferred from Voyager 1 and 2 measurements in One possiblexplanation invokes a two-state plasma torus and suggests time-dependent changes of bulk composition. In this paper we propose an alternative time-independent model (time-independent in a mean-value sense) but introduce the ratio of sulfur to oxygen in the neutral source as a free parameter. The recent evidence that there exist distinct sources of S2 and SO2 justifies thisintroduction of independent sources of S and O. In our calculations, we also varied the relative abundances of hot (1 kev) and thermal electrons, the neutral injection rate and characteristic radial diffusion time scale, and the characieristic ion temperatures. We integrated the rate equations over time until a steady state was achieved. For a pure SO: source, we found no solutions qualitative!y compatible with observations. A solution in qualitative agreement with observations emerges when the injection rate of neutral S: is'about 60% the rate of SO2. The deficiency of OIII observed in 1981 is obtained a model c6nsistent with other aspects of composition and dynamical features of the torus observed 'in 1979 at the time of the Voyager encounter by including of order 0.05% of 1 kev electrons and using a radial diffusion time of order 35 days. The intensity of EUV spectral lines attributed to OII and OIII can be produced by our ulfur-rich torus model if collision strengths of lines near 833 / are taken near the upper limit of their range of uncertainty. 1. INTRODUCTION Volatile material from sources internal to Io provides neutral gases (SO2 and S2) to the Io torus, probably through the mechanism of sputtering. Following dissociative detachment, the resulting sulfur and oxygen atoms are ionized by charge exchange and electron impact [Johnson and Strobel, 1982] and accelerated to corotation speed [e.g., Hill, 1979; Goertz, 1980; Barbosa et al., 1983]. The resulting ions reside in a torus localized near Io's orbit at a jovicentric distance of 5.9 R s (R s is the radius of Jupiter), confined to roughly 1 R s from the equator; electrical neutrality is provided by electrons with density near 2000 cm-3 at a temperature of 5 ev. A small population (< 1%) of 1 kev electrons is also observed [Scudder et al., 1981;Sittier and Strobel, 1984]. The ionization states attained are regulated by nonlinear interactions dictated by a variety of competing processes. Electrons modify charge states primarily through collisional ionization and dielectronic recombination. These production and loss mechanisms are coupled to other processes such as charge exchange and radial diffusion. The main diffusion process is flux tube interchange [Richardson and Siscoe, 983] through which plasma is ultimately lost from the torus, Additional losses result from charge exchange and recombination Copyright 1985 by the American Geophysical Union. Paper number 4A /85/004A-8051 $ ,065 processes that can produce atoms traveling tangentially away from the torus at the c0rotation speed. In our model of the hot outer torus, viewed as a region of uniform plasma density, we incorporate these and other competing processes. We seek to reproduce observed Values of ion partitioning using model parameters consistent with observed electron density and temperature, and inferred ion source strengths and radial diffusion rates. Data on ion partitioning are sparse and are obtained from intermittent observations during the Voyager flybys and thereafter. Previous attempts to model the measured results have assumed a pure SO2 neutral source. No single model has explained the results at different epochs, and time variable behavior has been invoked to account for the different observations [Brow net al., 1983]. In this paper, we take a different approach and ask whether by relaxing the constraint on the source composition, we can find a time-independent model (time-independent mean values over small numbers of rotation periods) consistent with the data to within its uncertainty. The assumption of time independence of the bulk torus composition is supported by the stability of the ultraviolet observa- tions from IUE, the International Ultraviolet ExPlorer satel- lite [MOos and Clarke, 1981] over an extended interval be- tween March 1, 1979, and May 2, '1980, including the times of both Voyager 1 and Voyager 2 encounters with Jupiter [Moos et al., 1985]. The time-independent model is also appealing on aesthetic grounds (e.g., Occam's razor). We argue that the observed parti.tioning requires that the

2 12,066 MORENO ET AL.' S 20UTGASSING AND ION PARTITIONING IN IO'S TORUS TABLE 1. Observed Ion Partitioning and Previous Theoretical Predictions In Situ Observations Reduced Data* Voyager 1 Flyby at 6 Ground Voyager 1 Observations EUV Constant February-April Observations Isothermal Thermal Theoretical Species Plasma Speed Predictions;[ SI 7.8 SII SIII õ 1198 SIV oi II õ iii < 4 < i¾ 0.15 ne, cm Ion partitioning values are given in ions per cubic centimeter. *Brown et al. [1983]. tbagenal and Sullivan [ 1981]. ;[Including 0.005% kev electrons and charge exchange. The characteristic loss time is 100 days for this model, and the OIII ion temperature is 250 ev. This model was developed to explain the post-voyager torus observations. õfor this case, OII and SIII cannot be separately identified, and [OII]+ 2[SIII] = source be rich in sulfur and cite evidence that the material force the ratio of total sulfur to oxygen ions to agree with the sputtered from Io's surface is not pure SO2 as generally as- UVS ratio of The same value of x implies [SIII]= 74 sumed in ion partitioning calculations but a mixture of S2 and and [OII]- 952, giving the value 470/74/170 for the ratios SO e. Previous models have incorporated many of the features [SII]/[SIII]/[SIV], markedly different from the ratios obwe consider. Here, we review the pertinent models and explain why we think they must be rejected on the basis of nonspectroscopic evidence regarding plasma transport properties. Our model is consistent with such plasma properties as well as tained from UVS analysis (120/170/115) or from other available observations. Thus the alternative PLS analysis also fails to agree with UVS abundances, and the differences have not been reconciled, though we understand that the investigators with the surface observations noted. are attempting to do so. In fact, Bagenal [1985] has reported In the next section, we present and discuss the measurements of ion abundances in the torus plasma, noting some ambiguities and identifying the features that are well established. Next we review physical processes important to the understanding of the ion partitioning calculation. We then agreement between compositions inferred from PLS and UVS for measurements near 5.75 R s. Given the uncertainties in the data, we focus on those features qualitatively present in all three data sets, and those supported quite unambiguously by a single measurement. For present our model which allows the source strengths of neutral example, all the data sets are consistent with [SIII]> [S1V], sulfur and oxygen to vary independently. We then discuss the an inequality not satisfied by the values emerging from some results of our model in the context of observational con- of the early theoretical treatments. An additional feature that straints and conclude with a discussion of spectroscopi constraints. we believe to be well established, at least at the time of observation, is the low density of OIII. In 1981, Brown et al. [1983] searched for the OIII 5007 ]k line in the warm torus and their 2. ION PARTITIONING: OBSERVATIONS upper limit of < 4 cm-3 appears to us to be a firm observa- Data on ion partitioning are obtained from ¾oyager 1 measurements and from intermittent ground observations. tional result. No analogous observations are available for the Voyager 1 epoch (1979). Representative values are provided in Table 1, from which it is OII abundances do not seem to be unambiguously esapparent that there are some discrepancies among the data tablished, and here, too, time variations could provide a parsets. Temporal Variability may explain some of the differences tial explanation for the range of values. As well, there are between ground-based and Voyager 1 data but cannot explain considerable uncertainties in the measurements. Ion densities the differences between the Voyager 1 UVS (the ultraviolet inferred from P LS depend on assumptions regarding thermal spectrometer ) and PLS (the plasma detector) results. For ex- properties and, for some assumptions, do not distinguish difample,.the PLS analysis [Bagenal and Sullivan, 1981], assuming an isothermal plasma, gives for the ratio of total sulfur ions to oxygen ions a value of 3, to be compared with the ferent ions with the same mass per unit charge. Probably the plasma distribution is not accurately represented as either isothermal or with constant thermal velocity [cfi Barbosa et al., UVS ratio of order 0.7 to 0.8. On the other hand, the disagree- 1983], but these assumptions have been used in the data ment may imply that the alternative PLS analysis of Table 1 analysis. UVS results depend on identifying lines that may be based on equal thermal velocities for all ions is to be preferred. associated with more than one ion or that merge with other In this case, one cannot obtain the sulfer-to-oxygen ratio for, as the PLS investigators note, one cannot distinguish OII and SIII. One can, however, require that [OII]= 1100x and [SIII]= 1100(1- x)/2 with 0 _< x < 1. Selecting x , we nearby lines. In particular, Brown et al. [1983] have argued that a spectral feature at 833 in the Voyager UVS spectrum is evidence for large OII abundance (Table 1, column 3). In section 6, we discuss some uncertainties in the interpretation

3 MORENO ET AL.' S 20UTGASSING AND ION PARTITIONING IN IO'S TORUS 12,067 of the 833 spectral feature and suggest that it can be modeled without assuming such a large OII abundance. Groundbased observations are corrected by subtracting solar continuum radiation, a process that contributes to the measurement errors. Thus, OII abundances must be regarded as having large uncertainties. 3. PHYSICAL PROCESSES Outgassed volatiles from Io's interior provide the neutral matter that serves as the source of the torus ions. Previous studies assumed a neutral source with 2[-S] = [-O] such as would be produced by full dissociation of SO:. The source strength is still controversial [Shemansky, 1980; Brown and Shemansky, 1982; Cheng, 1982; Smyth and Shemansky, 1983; Brown et al., 1983], with proposed injection rates varying from 1027 to 1029 ions/s. We believe that recent evidence suggests that SO2 is not the only source of sulfur in the torus but that there is an independent additional source of S2. The evidence has been presented by McEwen and Soderblom [1983], who compared Voyager 1 and 2 images of Io with laboratory observations of S2 and SO: and concluded that two types of volcanic plumes occur on Io. Small volcanic plumes and associated surface deposits are formed of SO:, whereas the less frequent larger plumes and their associated surface deposits are composed of pure sulfur. Neither SO2 nor S2 is injected directly from the plumes into the torus, since maximum ejection velocities (of order 1 km s- ) are less than the escape velocity and most of the volcanic debris falls back to the surface of Io. This point has been confirmed by observations of Io's surface I-Strom et al., 1981]. With direct injection excluded, sputtering of surface materials is regarded as the probable mechanism for transfer of neutrals from Io to the torus. Indeed, Lanzerotti et al. [1982] have found in laboratory studies that SO: frosts typical of Io's surface are eroded by charged particles, through Cheng [-1982] has questioned whether the allowed rate is compatible with ihe required torus source rate. Linker et al. [-1985] have shown that the radial profile of total flux tube content places restrictions on the source of neutral particles sputtered from the surface. They argue that the constraints are more readily satisfied by a dominant S: source, although this is not required and some SO2 is also probable. If surface sputtering is the principal process of importance in populating the torus, then the ratio of sulfur to oxygen in the source may differ significantly from the ratio 2 inherent in a pure SO2 source. Once introduced into the torus, the neutral gases are readily dissociated and soon are ionized by charge exchange and electron impact. Three principal loss mechanisms act to produce a steady state population. Ions are lost by the process of centrifugally driven flux tube interchange diffusion wherein plasma at different radial distances exchange position as a consequence of Rayleigh-Taylor instabilities. Centrifugal potential energy is converted to kinetic energy and this energy is fed into the interchange motion I-Sonnerup and Laird, 1963]. The time scale characterizing this process has not been firmly established but current estimates range from 20 to 150 days [Richardson and Siscoe, 1983; Smyth and Shemansky, 1983]. Probably the most pertinent evidence on radial diffusion rates comes from radial flux profiles of ions, both thermal [Siscoe et al., 1981] and energetic [Thomsen et al., 1977]. Thorne [1983] shows that in the vicinity of Io, the corresponding radial diffusion time scales ( 17 days for the Siscoe et al. estimates and > 2 days for the Thomsen et al. estimate) are consistent with the radial profiles of energetic protons if strong pitch angle scattering is assumed. Ions are also lost by recombination and by charge exchange between corotating ions and neutrals approximately at rest with respect to Io. These two processes produce neutrals moving on average at the local corotation velocity in a direction locally tangento the torus. No longer bound to field lines by electromagnetic forces, these neutral atoms exist from the torus in a time of order one hour, i.e., the time to traverse the torus tangentially. The escape time is sufficiently short that there is a low probability of a second ionization within the torus and consequently both charge exchange and recombination are appropriately regarded as loss processes. Although change-exchange is incorporated in these calcu!ations, we conclude that it affects the partitioning very little because of the low overall densities (see section 4.3). For the steady state situation, sources and losses must balance. Hill [1980] has shown that measured departure of the azimuthal drift velocity from the corotation velocity implies a source strength of 1030 amu s-x. For a mean ion mass of 20 amu, this corresponds to 5 x 1028 ions s-. F. Bagenal (personal communication, 1983) estimates that an average n e cm -3 over a volume Vr = 8 x 103 cm 3 provides a reasonable description of the bulk torus. Table! suggests an average of approximately 1.5 electrons/ion. These numbers lead to a mean loss time of 12 days in the stead y state. Both charge exchange and radial transport contribute to the loss, roughly equally, so the diffusive loss time may well be as long as 30 days, but it would be difficult to reconcile longer time scales with the above observations. The loss time of 17 days [Thorne, 1983] implies an average ion density in the bulk torus of order 0.6 n e and the corresponding source strengt h S(i) - 5 x 1028 ion s- x implies that the process of ion pickup supplies of order 4 x 10 2 to the torus if each new ion contributes 500 ev. Thus, the numbers we favor also allow the pickup process to drive the UV radiation at the observed [Shemansky, 1980] level of 3 x 10 2 W. Electrons of the torus are important ionizing agents, their effectiveness depending sensitively on their temperature and density. From in situ measurements [Ba tenal and Sullivan, 1981' Warwick et al., 1979] as well as Voyager UV data [Strobel and Davis, 1980], we have selected a representative electron density of order 2000 cm-3 and thermal energy of 5 ev at 5.9 Rj. It has been pointed out to us (F. Bagenal, personal communication, 1984) that if one takes the Voyager 1 PLS contour map [Ba tenal and Sullivan, 1981], the volum6 of the torus with n e greater than 2000 cm-3 is occupied by only 3% of the torus electrons, and tl e volume with rt e greater than 1000 cm-3 is occupied by 25% of all torus electrons. As these regions include both the regions of most intense emitted radiation and the in situ PLS measurements to which we refer in Table 1, we believe that a model with n e cm-3 is appropriate for comparison witl observations. The revised plasma densities of Ba tenal et al. [1985] modify the regions in which specific density levels are observed, but are still compatible with a representative density of 2000 cm-3. Scudder et al. [1981] and Sittler and Strobel [1984] point out that in the hot torus there is, in addition to the thermal electrons, a non-maxwellian tail in the electron distribution corresponding to a small population, say 1%, of electrons with characteristic energy near 1 kev. (The exact energy of the suprathermal electron component has no significant implications for ion partitioning since the relevant ionization rates are quite insensitive to energy for energies above 400 ev.) The

4 12,068 MORENO ET AL.' S 20UTGASSING AND ION PARTITIONING IN IO'S TORUS suprathermal electrons, though present in small proportions, lieve they favor. The others assume the same unrealistically can have a profound influence on the ion partitioning high electron density and either imply the same % or predict [Moreno et al., 1982] since 5 ev electrons are essentially in- unacceptably large ratios of [SII-I to [SiII-I and too high an capable of producing the doubly or more highly ionized states abundance of OI. easily created by the more energetic electrons. Two processes that oppose the effect of ionization are radiative and dielec Rationale for a New Model tronic recombination. The long diffusion time scale required by the B-83 model 4. MODELS OF ION PARTITIONING scaled to 2000 cm-3 and the correspondingly low source rate 4.1. Previous Models as well as absence of energetic electrons are all incompatible with Voyager 1 observations. Brown et al. [1983] recognized Models of the Io torus plasma incorporating some or all of this point and argued that their results applied to a different the mechanisms described above have been evaluated by sev- state of the Io torus in the post-voyager epoch. We, on the eral authors [Shemansky, 1980; Brown et al., 1983]. The She- other hand, ask whether there is a steady state torus model mansky [1980] model was designed to fit ion partitioning that that can explain both Voyager 1 and post-voyager 1 observations. is substantially different from present estimates (D. E. Shemansky, personal communication, 1984). Furthermore, no We shall demonstrate that a model based on a neutral charge exchangeffects were included. Therefore, we shall not source with twice as much oxygen as sulfur predicts large OIII discuss that model in this paper. abundance (i.e., > 100 cm-3) over a wide range of parameters. To compare the model of Brown et al. [1983] with evidence, However, if we remove the constraint on neutral source comwe consider not only the partitioning among ionic states but position, we find it possible to reduce the OIII abundance but also whether model parameters such as ne and % are accept- retain other features of the ion partitioning that are well esable. tablished by the Voyager observations. The Brown et al. [1983] model (Table 1, column 5), which In section 4.3, we describe our model and the approach to we refer to hereafter as B-83, assumes that n e = 5000 cm-3. calculations. The following section analyzes discrepancies be- This high density was not representative of the radiating por- tween model and.experimental abundances, and argues that tion of the torus which, as noted in column 2 of Table 1 was they do not exceed the uncertainties in the data. characterized by ne = 2000 cm-3 or possibly even less. Conse quently, we must determine how to scale the B-83 model to a A New Model lower density. There exists a homology transformation which permits a scaling of the ion density so that relative ion populations are Our model assumes that there are independent sources of S and O, that thermal electrons with a density 2000 cm-3 and a temperature of 5 ev are supplemented by a small percentage preserved. In particular, let us suppose that all ion species are of 1 kev electrons, and that the system has reached steady changed by a factor of r/, since the electrons are assumed to be state in the presence of losses characterized by a time %. The derived from the resident ion population. We observe then that both collisional ionization and total recombination terms rate coefficients used for our calculations are tabulated in Tables 2-4. Parameters of the model are the relative source are changed by a factor of r/2 (since they vary linearly with strengths of S and O, the percentage of hot electrons, and %. respecto electron and ion populations). Since charge ex- In the previous sections, we have stressed the uncertainties change terms vary as the square of ion density, the charge of some of the parameters (e.g., temperatures, diffusion time exchange term also scales as r/2. Recalling that tangential scale) that enter the calculations. Our first step was to carry losses of ions are associated with recombination or charge out a parameter search within the range permitted by Voyager exchange terms, it follows that the tangential loss term scales 1 observations (i.e., less than 1% hot electrons, and % < 40 as r/2. In order for the equilibrium solution to scale in a days) to determine whether a model similar to that of Brown manner which preserves all ion abundance ratios, the remain- et al. [1983] with a source in the ratio two O to one S, a ing diffusion and source terms must also scale as r/2. As the diffusion term varies with the diffusion rate and the ion density, it follows that the diffusion rate D must be proportional TABLE 2. Charge Exchange to r/(and the corresponding diffusion time to r/-x). Moreover, the source rates must vary as r/2. Reaction Rate Reference* Using this scaling argument, we now examine the B-83 O + + S-- S + +O 8.80 E-09 1 S 3+ + S-- S 2+ + S E-09 1 model of Table 1. The ratios of abundances are not affected by S } S O e 3.00 E scaling but the abundances themselves must be reduced by a S--} S + + O E factor of 0.4. The loss time, %, must now be replaced by S 2+ + S--} 2S E [100/0.4] days = 250 days. As we have noted, such a large S + S+--}S + +S 1.76 E value for % is not supported by other diffusion studies for the S 2+ + S--} S + S E Voyager epoch. To reproduce low OIII abundances while re S +--}O + +S E-09 2 O + S+--}O + +S 0.06 E taining high sulfur ionization states, the B-83 model also re- S 3+ + O--} S E quires that there be no hot (> 100 ev) electrons in torus. (Hot S 2+ +O--}S + +O E-09 1 electrons minimize the effects of small differences in ionization S 2+--}O + + S E potentials between corresponding ionization levels of S and O O --} E O+O +--}O ++O 1.30E-08 1 Elimination of the hot population reduces drastically the OIII O 2+ + O--}O + O E-09 1 abundance.) This assumption about hot electrons is incompat- S 3+ +O--}S 2++O E-08 1 ible with Voyager 1 observations. Table 1 gives only one of the models presented by Brown et al. [1983], the one we be- Read 8.80 E - 09 as 8.80 x *References' 1, Johnson and Strobel [1982]' 2, Brown et al. [1983].

5 MORENO ET AL.: S 20UTGASSING AND ION PARTITIONING IN IO'S TORUS 12,069 TABLE 3. Dielectronic Recombination TABLE 5. Ion Partitioning From Model Calculations Reaction Rate Reference* Percent 1 kev S + + e-- S + hv E S 2+ + e-- S + + hv E S 3+ + e- S 2+ + hv E S ' + + e- S 3+ + hv E s(o)/s(s) e--, 0 + hv E SII e- O + + hv E SIII e hv E SIV Read E - 11 as x 10- x. OI *References: 1, Johnson and Strobel [1982](with n e = 1900 cm -3 at OII 5.9 Rj); 2, D. E. Shemansky (personal communication, 1982). OIII OIV SI /3 1/4 1/8 1/16 1/8 1/ neutral injection rate of 3 x 1027 s-1 and charge exchange could reproduce the principal observed abundances. We were particularly interested in determining whether the observed [SIII]/[SIV] > 1 and [SII]/[SIII] _< 1 could be obtained from parameters that also led to a low OIII abundance. If we assumed that the neutral source provided 2 oxygen atoms for each sulfur atom, we found that it was not possible to obtain [SII]/[SIII-I _< 1, [SIII]/[SIV] > 1 and a very low concentration of OIII. Representative results for 1% and 0.03% kev electron densities are tabulated in the last two columns of Table 5. Clearly models with a pure SO2 source have provided valuable insight into torus processes, but they do not yield low values for OIII abundance and, therefore, would be inconsistent with a time-independent model of average torus properties. To reduce the abundance of OIII, one must invoke either (1) a mechanism that selectively removes neutral O prior to its initial ionization, or (2) a source of neutrals in which the ratio of O to S is less than 2. No exotic mechanism of the sort required by option 1 has been suggested. Option 2 is consistent with the surface-sputtering arguments discussed above. We have already presented arguments for the view that some of the reported levels of OII (say in column 3 of Table 1) may be high. If the total oxygen density were diminished, the relative populations of different levels might remain close to those reported without violating firmly established ion density measurements. This argument leads us to accept the concept that the source may provide neutral O and S independently. We next studied the sensitivity of ion partitioning to S(O)/S(S), the ratio of the injection rate of neutral oxygen to that of neutral sulfur; values ranging from 2 to 1/32 allowed us to test the complete range from a source with the relative neutral composition of SO, to a source representative of sur- TABLE 4. Electron Impact Ionization Rate Reaction 5 ev Reference e+s S + +2e 5.000E-09 1 e+s +-- S 2+ +2e 3.210E-10 1 e+s 2+- S 3+ +2e 1.526E-11 1 e+s 3+-- S ' + +2e 2.100E-13 1 e + O- O + d- 2e E-09 1 ed-o +--O 2 + d-2e 8.947E-12 1 e+o 2+-- O 3+ +2e1.210E-13 1 Rate 1 kev Reference E E E E E E E Read E - 09 as x 10-9 *References: 1, Johnson and Strobel [1982] (the rates from Table 3 at 5.9 Rj are divided by their electron density at 5.9 Rj, i.e., n e = 1900 cm-3); 2, Lotz [1967, Table 4]; 3, W. Lotz (personal communication to W.I.N., 1975). T a, days Ne, cm These models are also based on a cold electron temperature, T -- 5 ev, and an ion temperature, T ev. face deposits. The electron density remained fixed at 2000 cm -3 and the bulk electron temperature was fixed at 5 ev. The relative abundance of hot electrons (1 kev) was varied (0.05, 0.1, 0.5 and 1.0%) through a range consistent with uncertainties indicated by Scudder et al. [1981] and variations in the data of Sittler and Strobel [1984]. Also, source rates, S(S), or, equivalently, diffusion time scales were varied from 1026 to 5 X Charge exchange between ions and neutrals occurs with relative velocity near the corotation velocity in Io's frame. Consequently, rate coefficients for sulfur ions at 560 ev and oxygen ions at 280 ev are appropriate and were used (see Table 2). We also explored possible charge exchange reactions between neutral oxygen and protons only to find that the influence of these reactions is negligible. As a means of determining the equilibrium ion partitioning, we integrated the rate equations in time using a Runge Kutta fourth order integration scheme [Dahlquist and Bjorck, 1974] until a steady state was achieved. (Integration of the rate equations was performed as a numerical device for finding the solution to a set of ill-conditioned nonlinear equations that did not respond well to quasilinearization or invariant imbedding solution schemes. These computations are not meant to be interpreted in the sense of a "simulation," although our code would be readily adapted to that role and could be used to study the effect of radial variations in torus properties.) We then determined the electron density implied by the calculated ion distribution. Keeping the diffusion time and all other parameters fixed, we varied the percentage of kev electrons (relative to the 5 ev electrons), thus obtaining a relation between the number density of all electrons and the input fraction of hot electrons. We then varied the hot electron fraction until the total number density of electrons was 2000 cm -3. This approach provided the self-consistent results tabulated in Table RESULTS The results provided in the first six columns of Table 5 are restricted to cases for which the parameters chosen produced predicted partitionings that accorded with the principal features of experimental observations tabulated in Table 1. As noted above, these features are [SII]/[SIII] < 1, [SIII]/[SIV] > 1, and low density of OIII. Calculations confirmed anticipated aspects of the effect of varying parameters.

6 12,070 MORENO ET AL.' S 20UTGASSING AND ION PARTITIONING IN IO'S TORUS For example, the 5 ev background electrons produce prin- spectra of Shemansky and Smith [1981] use 336 cm-3 for OII, cipally singly ionized sulfur and oxygen. As the relative abun- almost two orders of magnitude larger than the upper limit of dance of hot electrons increases for fixed source strength, 4 cm -3 set by B-83 based on later (May 1981) observations. higher ionization states appear and the population density of IUE observations, contemporaneous with the Voyager epoch the highest occupied ionization states increases, though the [Moos and Clarke, , show that the OIII density was well increase is limited by the reverse process of dielectronic re- below 336 cm-3 even in Those IUE measurements decombination. If the relative abundance of 1 kev electrons is tected the 1664 ]i line of OIII and established an upper limit reduced to 0.03% and the diffusion time falls to 27.5 days, (as of 110 cm 3. Subsequently, Moos et al., [1983], citing analysis in the last column of Table 2), the OIII abundance drops but of problems with detector response provided by Hackney et al. other features of the partitioning diverge significantly from ], revised the OIII density downward by a factor of 3. observations ([SII]/[SIII] > 1). The different columns of Table 5 illustrate how the ion partitioning changes with changes in the percentage of 1 kev The revised new abundance of 37 cm-3 is almost an order of magnitude lower than the density used to synthesize the EUV spectrum at the same period. We are not aware of other diselectrons (0.03, 0.05, 0.1 and 1.0) and the ratio of source cussions of this inconsistency. strength [S(O)/S(S)= 1/3, 1/4, 1/16, 1/8 and 2]. For a fixed Here we consider conditions under which our model, domipercentage of 1 kev electrons, the self-consistent diffusion time nated by sulfur ions, can also satisfy the constraints imposed increases the source becomes richer in sulfur. Naturally, the by the intensity of the 833 ]i feature. We argue that the obconcentration of oxygen ions decreases as the source becomes served intensity is compatible with our model provided that richer in sulfur. Also, for a fixed ratio of sulfur to oxygen in uncertainties of at least a factor of 2 in collision strengths are, the source, the self-consistent diffusion time decreases as the acknowledged. percentage of 1 kev electrons increases. It is relatively straightforward to reproduce qualitatively the A satisfactory description of the torus, consistent with the shape of the 833 ]i feature by superimposing the OII and OIII well-established features of observations including a low den- emissions near 833, SIII emissions at 825 ]i, and SIV emissity of OIII, is achieved for source-strength ratios between sions at 816,. The shape is relatively insensitive to details of 1/16 and 1/3, and 0.03% to 0.05% kev electrons; rd is of order the emission spectrum because the instrumental resolution of 35 days, as in the first columns of Table 5. the Voyager EUV spectrometer is 30 ]i and the spectrum is Our models and the B-83 model provide estimates of neu- acquired in steps of 10 ]i [Shemansky and Smith, 1981]. Figure tral sulfur density as well as the density of ionized species. The 1 illustrates this point by comparing the data with a simulaneutral sulfur cloud has been detected by Durrance et al. tion from Shemansky and Smith (1981) and with a simulation [1983], but they have not published estimates of density. A based on our model of Table 5 [S(O)/S(S)= 1/3]. Evidently, confirmation of these rocket measurements of UV spectra in the shape of the line is quite well modeled by either approxilight of model predictions would be worthwhile, especially be- mation. The intensity of the feature, however, requires either a cause the B-83 model (column 5 of Table 1) predicts a much large density of emitting ions or a large collision strength for smaller neutral sulfur density than the cm-3 we predict. the principal emitters. The collision strength of SIII at 825 ]i Additional measurements would be important. is 0.41 [Shemansky, 1980], sufficiently low that no reasonable For our preferred parameters with % = 35 days, the corre- density of SIII can provide a substantial contribution to the sponding source strength S(S) is 6.8 x 1027 s-. The injection intensity of the feature. The highest density of SIV discussed in rate obtained is consistent with the requirement S(S)= any torus model is less than 200 cm -3 and with a collision ni'vt'/%, where ni' is the total ion density (1200 cm-3) and strength of 2.2 for the 816 ]i emission [Bhadrand Henry, is the volume of the dense part of the torus (approximately 1980], this species can make, at most, a comparatively unim x 8 X 103 cm-3), and % is the residence time for an ion portant contribution. Turning to oxygen ions, we note that (of order 3 x 106 s). A range of parameters close to those B-83 used 7.4 for the collision strength of the OIII line at 834 preferred parameters can give qualitatively acceptable results ]i and 4.3 for the OII line at 833 ]i. With these collision but, as remarked previously, substantially higher or lower in- strengths and our model for S(O)/S(S)= 1/3, we obtain 55% jection rates predict sulfur ion partitionings quite inconsistent of the intensity obtained for the B-83 model scaled downward with observations. by a factor of 0.4 as discussed in section SPECTROSCOPIC CONSTRAINTS The failure of our model to reproduce the intensity of the 833 ]i feature as successfully as does B-83 may not be a In the previousections, we described our time-independent serious flaw, because collision strengths are poorly established. model of Io torus plasma and identified some parameters for Indeed, in the last two decades, estimates of the collision which the model yields ion abundances consistent with a strength of the OII 833 ]i line have varied between 1 [Shegroup of observational constraints that we consider to be mansky, 1980] and 10 [Davis et al., 1975]. Let us note that all firmly established. Additional constraints are provided by the these calculations assume thermal equilibrium, which is evi- Voyager 1 and 2 EUV spectra which reveal an intense and dently an unrealistic approximation [Barbosa et al., 1983]. If persistent spectral peak near 833 / [e.g., Strobel and Davis, all the collision strengths were a factor of 2 larger than those 1980]. used for B-83, our model would reproduce the observed inten- Interpretations of these spectra suggest OII as the dominant sity. In view of the sensitive dependence of the calculated inionic constituent [e.g., Brown et al., 1983]. In particular, OII tensity on an ill-known set of parameters, it seems to us and OIII produc emissions near 833 ]i, and these intense premature to exclude the sulfur-rich models of this paper on lines are used to account for the 200 R of the spectral feature. spectroscopic grounds. In fact, B-83 state that "the very bright and persistent 833 ]i EUV line cannot be reasonably attributed to any species other 7. CONCLUSIONS than OII and OIII." This is consistent with their model, domi- We conclude that a biomodal source model, together with a nated by oxygen ions. Indeed, the synthesized Voyager-EUV small percentage of hot electrons and a relatively short diffu-

7 MORENO ET AL..' S 20UTGASSING AND ION PARTITIONING IN IO'S TORUS 12, I I ' REFERENCES O 0 o I 8OO 85O 9OO Hackney, R. L., K. R. H. Hackney, and Y. Kondo, Spectral anomalies in low-dispersion SWP images, in Advances in Ultraviolet Astron- Wavelength (Angstroms) omy: Four Years of IUœ Research, edited by Y. Kondo, J. M. Mead, and R. D. Chapman, p. 335, U.S. Government Printing Office, Washington, D.C., Fig. 1. Spectral profile 805/!to 855/!t. In this diagram we super- Hill, T. W., Inertial limit on corotation, J. Geophys. Res., 84, , impose three figures: (1) The observed spectrum (solid line), (2) the best fit synthesized spectrum of Shemansky and Smith [1981] (dashed Hill, T. W., Corotation lag in Jupiter's magnetosphere: A comparison line) with OIII as the main emitter at 833 /!t, and (3) the present of observation and theory, Science, 207, , analysis (dotted line), which represents the partitioning of column 1 Johnson, R. E., and D. F. Strobel, Charge exchange in the Io torus on Table 5, where OII is the principal emitter at 833/!t and the source and exosphere, J. Geophys. Res., 87, , is predominatly sulfur, i.e., the source rate ratio S(O)/S(S) of oxygen to Lanzerotti, L. J., W. L. Brown, W. M. Augustyniak, R. E. Johnson, sulfur is 1/3. The instrumental response is built into the model calcu- and T. P. Armstrong, Laboratory studies of charged particle erolations to represent the contributions of blended lines in the 833 /It sion of SO2 ice and applications to the frosts of Io, Astrophys. J., 259, , feature. Synthesized spectra 2 and 3 both duplicate the observed spectra except in the ranges where the departure is evident in the diagram. Linker, J. A., M. G. Kivelson, M. A. Moreno, and R. J. Walker, Inward displacement of Io's hot plasma tours: Explanation and consequences for sputtering sources, Nature, 315, , Lotz, W., Electron-impact cross-sections and ionization rate coefficients for atoms and ions, Astrophys. J. Suppl., 14, , sion time scale, can account for the principal features of all the McEwen, A. S., and L. A. Soderblom, Two classes of volcanic plumes observations in Table 1 without invoking significant time on Io, Icarus, 55, , variability. In particular, the OIII deficiency observed in 1981 Moos, H. W., and J. T. Clarke, Ultraviolet observations of the Io torus from the IUE observatory, Astrophys. J., 247, , can be obtained in a model that produces other aspects of the Moos, H. W., S. T. Durrance, T. E. Skinner, P. D. Feldman, J. L. torus ion partitioning measured during Voyager 1 encounter Bertaux, and M. C. Festou, IUE spectrum of the Io torus:identifi- and in subsequent ground-based observations. Previous discussions of the full set of observations have been forced to assume major temporal changes post-voyager 2 encounter. The diffusion lifetime obtained in our model is in good agreement with more direct assessments recently provided. The bimodal source gains strong support from surface measurements of S and SO2 distributions, a feature ignored by all previous models. Acknowledgments. We thank D. E. Shemansky, J. Trauger, and A. Chutjian for useful conversations. We are especially grateful to D. E. Shemansky for running his program using our parameter so that we could check our model calculation. This work was partially supported by NASA grant Institute of Geophysics and Planetary Physics publication The Editor thanks F. Bagenal and another referee for their assistance in evaluating this paper. Bagenal, F., Plasma conditions inside Io's orbit, J. Geophys. Res., 90, , Bagenal, F., and J. D. Sullivan, Direct plasma measurements in the Io torus and inner magnetosphere of Jupiter, J. Geophys. Res., 86, , Bagenal, F., R. L. McNutt, Jr., J. W. Belcher, H. S. Bridge, and J. D. Sullivan, Revised ion temperatures for Voyager plasma measurements in the Io plasma torus, J. Geophys. Res., 90, , Barbosa, D. D., F. V. Coroniti, and A. Eviatar, Coulomb thermal properties and stability in the Io plasma torus, Astrophys. J., 274, , Bhadra, K., and R. J. W. Henry, Oscillator strengths and collision strengths for SIV, Astrophys. J., 240, 368, Brown, R. A., and D. E. Shemansky, On the nature of SII emission from Jupiter's hot plasma torus, Astrophys. J., 263, , Brown, R. A., D. E. Shemansky, and R. E. Johnson, A deficiency of OIII in the Io plasma torus, Astrophys. J., 264, , Cheng, A. F., SO 2 ionization and dissociation in the Io plasma torus, J. Geophys. Res., 87, , Dahlquist, G., and A. Bjorck, Numerical Methods, Prentice-Hall, Englewood Cliffs, N.J., Davis, J., P. C. Kepple, and M. Blaha, Distorted wave calculations for multiply charged nitrogen and oxygen, J. Quant. Spectrosc. Radiat. Transfer, 15, 1145, Durrance, S. T., P. D. Feldman, and H. A. Weaver, Rocket detection of ultraviolet emission from neutral oxygen and sulfur in the Io torus, Astrophys. J., 267, L125, Goertz, C. K., Io's interaction with the plasma torus, J. Geophys. Res., 85, , cation of the 5S 2-- 3P2,! transition of SIII, Astrophys. J., 275, L19-L23, Moos, H. W., T. E. Skinner, S. T. Durrance, P. D. Feldman, M. C. Festou, and J. L. Bertaux, Long-term stability of the Io high- temperature plasma torus, Astrophys. J., 294, , Moreno, M. A., W. I. Newman, and M. G. Kivelson, A self-consistent model of ion partitioning in the Io torus, Eos Trans. AGU, 63, 1062, Richardson, J. D., and G. L. Siscoe, The non-maxwellian distribution of ions in the warm Io torus, J. Geophys. Res., 88, , Scudder, J. D., E. C. Sittier, and H. S. Bridge, A survey of the plasma electron environment of Jupiter: A view from Voyager, J. Geophys. Res.,86, , Shemansky, D. E., Mass-loading and diffusion-loss rates of the Io plasma torus, Astrophys. J., 242, , Shemansky, D. E., and G. R. Smith, The Voyager 1 EUV spectrum of the Io plasma torus, J. Geophys. Res., 86, , Sittier, E. C., Jr., and D. F. Strobel, Io plasma torus electrons: Voyager 1 (abstract), Eos Trans. AGU, 65, 1038, Siscoe, G. L., A. Eviatar, R. M. Thorne, J. D. Richardson, F. Bagenal,

8 12,072 MORENO ET AL.: S 20UTGASSING AND ION PARTITIONING IN IO'S TORUS and J. D. Sullivan, Ring current impoundment of the Io plasma torus, J. Geophys. Res., 86, , Smyth, W. H., and D. E. Shemansky, Escape and ionization of atomic oxygen from Io, Astrophys. J., 271, , Sonnerup, B. U. O., and M. J. Laird, On magnetospheric interchange instability, Geophys. Res., 68,!31-139, Strobel, D. F., and J. Davis, Properties of the Io plasma torus inferred from Voyager EUV data, Astrophys. J., 238, L49-L52, Strom, R. G., N.M. Schneider, R. J. Terrile, A. F. Cook, and C. Hansen, Volcanic eruptions on Io, J. Geophys. Res., 86, , Thomsen, M. F., C. K. Goertz, and J. A. Van Allen, A determination of the L dependence of the radial diffusion coefficient for protons in Jupiter's inner magnetosphere, J. Geophys. Res., 82, , Thorne, R. M., Microscopic plasma processes in the jovian mag- netosphere, in Physics of the Jovian Magnetosphere, edited by A. J. Dessler, Cambridge University Press, New York, Warwick, J. W., J. B. Pearce, A. C. Riddle, J. K. Alexander, M.D. Desch, M. L. Kaiser, J. R. Thieman, T. D. Carr, $. Gulkis, A. Boischot, C. C. Harvey, and B. M. Pedersen, Voyager 1 planetary radio astronomy observations near Jupiter, Science, 204, , M. G. Kivelson, M. A. Moreno, and W. I. Newman, Department of Earth and Space Sciences, University of California, Los Angeles, CA (Received March 26, 1984; revised June 10, 1985; accepted June 25, 1985.)