Energetic electron butterfly distributions near Io

Transcription

1 JOURNAL OF GEOPHYSCAL RESEARCH, VOL. 104, NO. A7, PAGES 14,755-14,766, JULY 1, 1999 Energetic electron butterfly distributions near o R. M. Thorne x, D. J. Williams 2, L. D. Zhang 1, and S. Stone 3 Abstract. Pronounced variations in the energetic electron distribution observed by the Energetic Particle Detector during the Galileo flyby of o are described as a quasi-adiabatic response to the changing electric and magnetic field environment near the satellite. The energetic particle signatures can therefore be used to remotely sense the spatial distribution of electric and magnetic fields in the vicinity of o. Electron pitch angle distributions evolve from a normal pancake distribution (peaked at 90 ø pitch angle) in the undisturbed torus to a butterfly distribution in the strong field depression near o. The strongest flux depletions at 90 ø pitch angle result from a reduction in kinetic energy due to conservation of the first adiabatic invariant, as electrons are transported into the vicinity of o. The magnitude of the flux depletion is related to the spectral index n of the electron energy spectrum (J -0 E- ). Since the value of n tends to increase with increasing energy, the largest flux drop occurs at higher energy. n the low-speed wake region downstream of o, electrons exhibit an abrupt transition to a population which is consistent with trapping on bounce orbits within the magnetic depression near o. This trapped population, which appears in the same spatial region as intense field-aligned beams, is not a result of adiabatic transport from a source region upstream of o. The phase space density of the "trapped" electron population is reduced, compared to the background torus, and particle tracing calculations in a realistic model environment near o suggest that such electrons must be scattered into the region sampled by Galileo. Torus electrons with energies well above an MeV are excluded from a broad spatial region surrounding o. This leads to a pronounced drop in the flux of penetrating particles near o which allows the modest "trapped" electron population to be detected above the background level for energies up to 200 kev. 1. ntroduction The dynamics of the inner Jovian magnetosphere are generally dominated by the intense Jovian dipole field and corotational flow. However, during the o flyby on day 341 of 1995, the Galileo MAG (magnetic field) instrument detected a major depression of the magnetic field [Kivelson et al., 1996a], and the PLS (plasma science) instrument observed significant departures from Department of Atmospheric Sciences, University of California, Los Angeles 2Appl /ed Physics Laboratory, Johns Hopkins University, Laurel, Maryland afundamental Technologies, Lawrence, Kansas Copyright 1999 by the American Geophysical Union. Paper number 1999JA /99/1999JA tororational flow [Frank et al., 1996] over an extended spatial region within about 4 R o of o. These were accompanied by a pronounced change in the flux and pitch angle distribution of energetic particles measured by the EPD (energetic particle detector) instrument [Williams et al., 1996]. An equatorial projection of the encounter is sketched in Figure 1. Over this region, the Galileo trajectory was confined within 0.5 R o of the equatorial plane. Exceptionally low flow velocity, high plasma density, and a relatively stable, but depressed, magnetic field were found in the downstream wake (shaded) near closest approach ( UT). Highly variable magnetic fluctuations and the strongest magnetic depression occurred in the regions adjacent to the wake. These were accompanied by enhanced plasma flow and changes in the flow direction, suggesting that much of the upstream flow is smoothly deflected around o; little of the corotational electric potential is applied across the satellite itself. 14,755

2 14,756 THORNE ET AL.' BUTTERFLY DSTRBUTON NEAR O 2. Electron Observations Near o -3[... "' 17:. -4 ] [ i / [\ t [ X(o radii) Figure 1. A schematic plot of plasma convective streamlines (solid curves) and lines of constant field strength (dashed curves) adapted from the MHD model of Linker et al. [1998] with an internal o dipole field. These are used to compare convective flow and 7B drift during the Galileo encounter with o. Measured changes in plasma flow [Frank et al., 1996] are indicated by velocity vectors. An overview of the energetic electron response during the o flyby is shown in Figure 2; more detailed observational information is given in section 2. n addition to the bidirectional field-aligned beams and lowintensity, trapped-like distributions at closest approach reported earlier by Williams et al. [1996] and discussed /? flux in the vicinity of the loss cone remained essentially in more detail in a companion paper [Williams et al., unchanged, but the flux near c = 90 ø was reduced by an this issue], all electron channels exhibit a pronounced amount which is strongly correlated with the reduction flux depletion over an extended region surrounding o. in the ambient magnetic field. A comparable change The normal pancake pitch angle distributions (peaked in the pitch angle distribution was also seen during the at c = 90 ø pitch angle) which characterize the back- egress from o after 1747:30 UT. ground torus are replaced by butterfly distributions Between 1745:30 UT and 1747:20 UT, when Galileo (strong depletions near 90 ø pitch angle), as the mag- entered the low-speed wake behind o, the electron pitch netic field (bottom plot in Figure 2) depression intensi- angle distributions changed abruptly. ntense bidirecfies. These dominant features of the energetic electron tional field-aligned beams accompanied by a residual flux change are described in section 3 as an adiabatic trapped-like population with peak fluxes near c - 90 ø response to the changing electric and magnetic field en- were discovered in the low-speed wake '[Williams et vironment around o. Section 4 describes preliminary al., 1996]. The observed "trapped" electron fluxes are results of adiabatic particle tracing in a model magnetic well above background for energies below 200 kev, as and electric field environment; these results are used to demonstrated by the calibration which was made just account for the abrupt transition from a butterfly dis- after 1746 UT. A brief explanation for this low-intensity tribution to a "trapped" distribution in the wake region trapped electron population is given in section 4; a more downstream of o and for the possibility of an extended detailed discussion is included in a companion paper exclusion zone for extremely energetic electrons. We [Williams et al., this issue]. conclude with a brief summary of the major results and The progressive transition in the electron pitch an- a discussion of how the energetic particle signatures can be used to place constraints on models for the global electric and magnetic field environment surrounding o. The radiation environment near o includes extremely energetic particles which are able to penetrate through the shielding of the EPD detector. The instrument was therefore designed with several provisions to measure when such particles might contaminate the count rate in any specific channel [Williams et al., 1992]. Periodically, a calibration is made in which the detector is directed toward a foreground shield. Three calibration counts (near 1736 UT, 1746 UT, and 1756 UT) are clearly detectable in the E0 (15-29 kev) channel (top plot) of Figure 2. These demonstrate that the E0 channel is relatively free of contamination and that the flux of penetrating particles also exhibits a decrease by more than an order of magnitude near o. The calibration is much less noticeable in the F1 ( kev) and B1 ( MeV) channels, indicating that these two channels are strongly contaminated during the entire o transit. The higher-energy channels do, however, provide an indication of the variability of high-energy penetrating particles which are probably dominated by electrons above 10 MeV [Williams et al., this issue]. A more detailed plot of the lower-energy electron response to the changing environment within 3 Rio of o is given in Figure 3. For most of the transit past o, the EPD instrument was retained in a motor position which gave a nearly complete scan over all pitch angles twice during each satellite spin (19 s). The observed spin modulation near 1740 UT is due to the characteristic pancake distribution (peaked at c - 90 ø) of trapped energetic electrons in the torus. However, soon after 1740 UT when Galileo was about 3 Rio from o, the low-energy electron channels began to exhibit a gradual transition toward a butterfly distribution; electron gle distribution is shown in Figure 4 in which the count rate (normalized to the peak values near 1737 UT) is plotted against cos c for six complete satellite spin pe-

3 - THORNE ET AL' BUTTERFLY DSTRBUTON NEAR O 14,757 'b key --=-- 1o 1 = 0 : key - o,o {... ø" to MeV 2100 t300 - 'K':.2,/ [.-,,,,.½ - ',,... [...,./' Khurana 97 5OO 17:35 17:40 17:45 17:50 17'55 18:00 Spacecraft Event Time, UT Figure 2. (top 3 plots) An overview of changes in Energetic Particle Detector electron count rate during the Galileo encounter with o on day 341 of (bottom) Observed magnetic field variation compared to the model of Khurana [1997](solid curve) and the MHD simulation of Linker et al. [1998](dashed curve). riods at selected locations over the o encounter. The lowest-energy channels E0 and E1 are sampled every distribution with peak flux near c = 90 ø. There is a 1/3 s, and the data are relatively free of contamina- well-pronouncedrop-off at the edge of the loss cone tion. These channels provide the most accurate measure which is particularly noticeable in all E detector chanof the ambient electron pitch angle distribution with a nels. At 1741, when Galileo was approximately 3 R o resolution of about 6 ø. Channels E2 and E3, sampled from the satellite, all channel show a depression (or every 2/3 s, provide a pitch angle resolution near 12 ø and are also relatively free of penetrating background counts. Channel F0 ( kev) is partially contaminated by penetrating particles; this contamination is most serious in the undisturbed torus well away from o. The uncorrected F0 count rates therefore do not give a realistic measure of the pitch angle anisotropy in the vicinity of the loss cone (cos c => +l) in the region far from o. However, the sharp drop in penetrating particles (to values below 1000 counts/s) in the magnetic depressio near closest approach to o allows us to obtain a realistic measure of the higher-energy pitch angle distribution near 1745 UT. Near 1737 UT (left plots in Figure 4), prior to the o encounter when Galileo was over 4 Rio from the center of o, all measured electrons exhibit a normal pancake flattening) of the pitch angie distributio near c = 90 ø. The "butterfly" distributions are clearly evident in all channels at 1743 UT when Galileo is within 2 R o of the satellite. Near closest approach (1745 UT, when Galileo moved to within 1.5 R o of the satellite), but prior to entry into the low-speed wake, the butterfly distributions become very pronounced with a flux minima at c = 90 ø and peak flux near cos c m All uncontaminated channels (E0-F0) exhibit similar butterfly distributions, but the location of the peak flux changes throughou the o encounter. During the egress from o, the pitch angie distributions gradually relax toward a pancake shape, although a remnant of the butterfly distribution persists until well after 1755 UT (> 4 Rio distance). This asymmetry between the ingress and egress is probably

4 14,758 THORNE ET AL.' BUTTERFLY DSTRBUTON NEAR O 1000o 1oo 40 17:45 17:50 Spacecraft Event Time, UT Figure 3. Electron count rate measured by EPD near Galileo closest approach to o on day 341 of Butterfly distributions are apparent at distances less than 3 Rio, and a distinct trapped population occurs in the wake region. related to residual magnetic field distortion associated with the Alfv n wings of the interaction between the plasma flow and o. 3. Adiabatic Transport Near o Let us consider the adiabatic motion of energetic electrons, carried along in the plasma flow past o, which are also subject to gradient and curvature drifts because of the spatially varying magnetic field environment. For electrons with c = 90 ø, the drift velocity can be expressed as 1.4 x 105 (72-1) VB x B km/s (1) 7 B3 where B is the ambient magnetic field measured in nanoteslas, VB is the gradient in the magnetic field measured in nt/rio and? = 1 + Ek/mc 2, where Ek is the electron kinetic energy and mc 510 kev is the electron rest energy. n the unperturbed Jovian dipole field (B =1850 nt) near the orbit of o (L = 5.9), VB 24 nt/rio, and v m (72-1)/7 km/s. This electron drift occurs in a direction opposite to the corotation flow. For electron energies below a few MeV, the gradient drift speed is well below the convective drift past o (v 57 km/s). However, the strong magnetic gradients associated with the field depression near o [Kivelson et al., 1996a, b] can enhancelectron drift velocities by more than an order of magnitude at distances within 2 Rio of the satellite. The enhanced field gradients together with the observed reduction of convective flow in the wake allow gradient drift motion to dominate for higher-energy electrons close to o. For example, if the field depression were primarily due to an internal o dipole magnetic field, with a polar surface field (m 1300 nt) suggested by Kivelson et al. [1996a, bl, the electron gradient drift velocity near closest approach (1.5 / o) would be v 51(? 2-1)/7 km/s. This drift would exceed the observed plasma flow speeds in the wake region (v < 10 km/s) for energies above 60 kev. The dominance of gradient drift motion near o implies the existence of a stagnation point in the net flow and the possibility of an exclusion zone (which is inaccessible from the upstream flow) in the vicinity of o. This conclusion would apply regardless of the source of the field depression, as long as the field gradients are large near o. The Larmor radius of energetic electrons pe x 103(72-1)i/2/B(nT) km is typically below 1 km, much smaller than the scale size for measured magnetic field gradients near o. Energetic electron bounce time along the field is less than 10 s which is short in comparison to the convective drift time past o. Under such conditions, electron trajectories in the vicinity of o can be determined from the conservation of total energy

5 THORNE ET AL.' BUTTERFLY DSTRBUTON NEAR O 14,759 17:37:30 17:41:00 17:43'13 17:45:07 17:49:02 17:50:38 ß. ß ß o ' o.,' ' - - ' o.o oo.o, i/o.o,,,, o.o / t. o t ' ' 1.0 i i , - ß '' ', ß 6 / ' ' o.o oo. o.o/ o.o,,., /o.o o ß ß ,' 0.6,' 0. ' 0.6,',' O kev o.o, i, o.o a.o o.o o.o. o, a 1.0 ' ß ' 7'" '!... 'l '... [ l l.0,',',,',' ',', 0.6,' 0.6 " 0.6 ' 0.6,' ',' kev...!.o... o.!.o -- 1.o i.o o kev o.o,,, o.o 0.6 ' 0.0 i i o.o/, s, - o i - o -1 o -i o - o x o Cosine Figure 4. Pitch angle distributions measured by EPD over six complete satellite spin periods during the encounter with o on day 341 of The dashed vertical lines are estimates of c * defined by equation (5) at five locations with significant field depression. /-/(x)-- c(2ra/ Bm(X)q- ra2c2) 1/2 -e (x) (2) and the first two adiabatic invariants p2 / - 2raBm - f where p and Pll are the electron momentum and its component along B, respectively, Bm(X) is the mirror point magnetic field, (F(x) is the electric field potential, and the integral (for J2) is taken along the electron bounce orbit. n the reference frame where o is stationary, the electric field E = -V ' = -v x B determines the plasma flow velocity v(x). Upstream of o, E m 190 kv/r[o, but this field is reduced substan- (3) tially in the low-velocity wake region near o (Figure 1). f the spatial distribution of the magnetic and electric field were known, (2)-(4) would uniquely describe the trajectory of electrons as they move from one spatial location to another. Since this information is unavailable, we simply use the conservation laws (equations(2), (3), and (4)) to evaluate the change in electron energy and pitch angie in the vicinity of o, compared to that in the upstream "source" region. A direct comparison will be made with observations along the Galileo trajectory. Well upstream of o, all electrons follow bounce trajectories along the undistorted Jovian dipole field between magnetic mirror points where the field Bm > Bequator 1850 nt. The large magnetic field depression near o (Figure 2) requires that those electrons with mirror points near the equatorial plane (J2 m 0 upstream of o) must experience a reduction in Bm

6 14,760 THORNE ET AL.' BUTTERFLY DSTRBUTON NEAR O and thus lose energy (in order to conserve both/ and J2) as they are carried toward o. This can occur as an electron gradient drifts across surfaces of constant electric potential in the vicinity of o. Electrons must move to lower electric potential (away from Jupiter) as B n is lowered. The reduction in energy will be largest for those electrons with pitch angles near 90 ø, since these should experience the largest change in B n near o. Electrons which mirror away from the field minimum (J2 > 0) have a larger mirror point field Bm- B/sin 2 a. Since the strong field depression is confined close to o, at any location along the Galileo trajectory there will be a critical pitch angle a* defined by 17:37 17:40 17:43 17:45 sin 2 a* - B/Bs (5) where the electron mirror point field strength is equal to the equatorial field Bs well upstream of o. Over a range of equatorial pitch angles centered on a - 90 ø, which satisfy the condition sina > sina*, electrons must lose kinetic energy (to conserve/ ) near o, since Bm is less than the mirror point field Bm,S in the upstream source region. Conversely, for sina << sin a*, the electron mirror point occurs well away from o in a region where the field distortion is small; for such electrons, kinetic energy can be approximately conserved. The value of a* can be estimated from the ratio be- Since these electrons must also mirror close to o, the second invariant (4) will be small, and the source region for such electrons must lie close to the magnetic equator (as 90 ø) in the upstream flow. Figure 5 indicates that it is reasonable to model the electron en- 1o 3 lo i i i i i i i i Energy, key,oo Figure 5. Omnidirectional electron differential flux measured at four locations ranging from 4 to 1.5 Rio from o. is the phase space density at constant/. Under the assumption that fu and/ are conserveduring transport, relation (6) may be used to place an upper limit on the electron differential flux measured by the Galileo EPD in a fixed-energy channel near o: tween the observed magnetic field variation near o and the anticipated unperturbed Jovian field as specified by the model of Khurana [1997] at the appropriate L shell (see Figure 2). The resulting values of a*, plotted as vertical dashed lines in Figure 4, are clearly related to the peak flux in the observed butterfly distributions. The location of this peak flux is independent of energy; it is simply controlled by the change in magnetic field The approximations used to obtain (7) are only valid strength along the electron trajectory. An explanation for pitch angles that satisfy the condition sin a > sin a*, for this correlation is given below. and over this limited range J(E) < Js(E). The reduc- Considering first those electrons near o which sat- tion in flux should increase as the magnetic field is reisfy the condition sin a > sin a*, for nonrelativistic elec- duced in the vicinity of o, and the largest flux drop trons, the kinetic energy E near o can be related to that should occur for pitch angles near 90 ø. As a = a*, in the upstream flow (Es) through the conservation of (7) indicates that the particle flux shourd reach a max- / (equation(3)) and the requirement that Bm,S Bs; imum value comparable to that in the upstream source region J(a*) - Js (as = 90 ø). The distributionshown ES S ---- Bm,s g m - <B. Bs = ( ss)sin2 B 1 (6) in Figure 4 clearly indicate that the peak flux at any given location occurs for a a*. During the approach to o, the observed flux in all electron channels tends Js(E) J(E) = (.E Es )n(p)2 PS -- < ( S B ) n+l sin2(n+l) 1 a (7) to decrease with decreasing L shell. The decrease in flux is probably associated with scattering loss in the torus. However, over an extended region between 1743 and 1750 UT, the peak flux observed in all E chanergy spectrum in the undisturbed torus by a power law nels remains remarkably constant. We interpret this as distribution Js(E) = JoE -n. The spectral index varies evidence for a roughly constant source Js(E) in the upbetween n = 1 for the E0 channel and n = 1.5 for the E3 stream flow. The upstream source properties adopted channel. Contamination by penetrating particles limits for model fits near o are listed in Table 1. Using our ability to accurately determine the energy spectrum these source parameters and measured magnetic field above 100 kev. The differential flux during transport strength [Kivelson et al., 1996a, b], the predicted flux past o can be expressed as J(p) = p2fy (p), where f, (p) variation (7) provides a reasonable description of the

7 THORNE ET AL.: BUTTERFLY DSTRBUTON NEAR O 14,761 Table 1. Source Parameters for Model Fits Near o Channel E0 E1 E2 E3 Js, c/s 9.2 x 10 a 1.7 x x x 104 n J( r/2)/js pitch angle distributions observed in the vicinity of o for values of sin c > sin c *. Figure 6 shows a comparison between the observed distribution in each E channel at 1745 UT and our predicted variation (7) (bold dotted curves), based on the measured value for the local magnetic field depression B nt and the assumption of a source field Bs = 1875 nt corresponding to a source region just inside the orbit of o. The location of this upstream source region is consistent with the expectation of electron gradient drift across surfaces of constant (x) in the direction away from Jupiter. Electrons with sinc << since* mirror at a location where ] rn is well above the equatorial field strength in the upstream source location. For these electrons, the net kinetic energy change will be small, since the mag- netic mirror points lie in a region with little field distortion. As a consequence, the electrons will follow trajectories close to the surfaces of constant ()(x). During the transport past lo, the electron pitch angle must change from that in the upstream source location according to the conservation of/ ; sin c - (] /] s) 1/2 sin c s. n or- der to model the distribution near o, an assumption must be made for the pitch angle distribution Js(c ) in the upstream source region. n the absence of direct observations, we assume that the source region pitch angle distribution (light dashed curves in Figure 6) is similar to the pancake distributions observed prior to o encounter (i.e., near 1737 UT), and we normalize the c ø flux to the adopted values listed in Table 1. The predicted pitch angle variation of such particles near 1745 UT is shown by the bold dashed curves in Figure 6. This provides a good fit to the observed dis- tributions over an extended region near each loss cone, confirming our assumptions on the source distribution. The assumptions employed above to derive model fits to the pitch angle distribution become invalid for pitch angles near c *. A treatment which also carefully conserves the second invariant (4) would be required to model the observed variation near c *. This would entail 1.0,.., ½ 1737UT '.. { 1737UT ' :'; ': ß 1745 UT o.o, E0 ).o[ E o t -'- ', 11.o 0.0,.½. //1737UTx,..: ' '"'"' l,7 t.. ;'; '.. k..'"/: 'tl E UT , UT -1 0 x O- 180 o E Cosine Figure 6. Model fits to the pitch angle distribution observed by the Galileo EPD near 1745 UT for sin c > since* (bold dotted curve) and sinc << since* (bold dashed curve) compared to observations (light dotted or solid curves). The light dashed curves are measured pitch angle distributions at 1737 UT.

8 .. 14,762 THORNE ET AL.' BUTTERFLY DSTRBUTON NEAR O 4 2.5MeV/Gauss 4 5.0MeV/Gauss // >" o 0 ' X(o radii) X(o radii) 100MeV/Gauss 6000MeV/Gauss X(o radii) X(o radii) Figure 7. Drift trajectories for energetic electrons with c - 90 ø based on the model B(x) and ß (x) surfaces shown in Figure 1. (top plots) ower energy electrons (/ < 5 MeV/G) do not have direct access to the wake region behind o. (bottom plots) Relativistic electrons (/ > 100 MeV/G), are excluded from an extended region surrounding o, leading to the pronounced drop in the penetrating background flux (E > 10 MeV). a detailed knowledge of the magnetic field strength variation along field lines that pass near o. Such information is currently unavailable, but it could be modeled by specifying the processes responsible for the magnetic depression observed by Galileo. Several physically distinct tron flow around o, at a location where the bounceaveraged gradient and curvative drift speed is comparable to the E x BB 2 convective flow. The location of this stagnation point is strongly dependent on the electron energy (or magnetic moment) and the global processes could contribute to the net field depression distribution of the electric and magnetic field variation [Khurana, et al., 1997], including an internal dipole field near o. For sufsciently high electron energies, a region of o [Kivelson et al., 1996a], diamagnetic effects associ- of closed drift orbits (confined near o) can occur, which ated with enhanced plasma pressure [Frank et al., 1996; is inaccessible by adiabatic transport from the upstream Hill and Pontius, 1998], and "AlDen wing" current sys- region. A particle tracing code has been developed to tems associated with the iteration between the plasma consider the access of electrons to the spatial region near flow and o [Wolf-Gladrow et al., 1987; Neubauer, 1998]. o. Adiabatic electron drift trajectories for equatorially mirroring (c = 90 ø) electrons have been computed, 4. Electron Drift Trajectories Near o based on the model magnetic field gradients and plasma flow velocities obtained from a recent MHD simulation The combination of strong magnetic field gradients near o and reduced plasma flow in the wake behind o can allow a stagnation point to occur in the net elec- [Linker et al., 1998] of the environment near o. Surfaces of constant magnetic field strength, which define the gradient drift trajectories for equatorially mirroring

9 THORNE ET AL.- BUTTERFLY DSTRBUTON NEAR O 14,763 electrons near o, are shown by the dashed curves in Figure 1. These were obtained by adding the effect of the spatial variation in the Jovian dipole field to the MHD model field variation [Linker et al. 1998] for the case in which o is assumed to have an internal field with a magnetic moment suggested by Kivelson et al. [1996a]. Electric field drift past o is taken directly from the MHD simulation, and the surfaces of constant electric potential are shown by the solid curves in Figure 1. The adopted MHD model provides a realistic simulation of the observed plasma flow and magnetic variation along most of the Galileo trajectory. However, the current model of Linker et al. [1998] is unable to account for the large magnetic gradients encountered on either side of the wake (Figure 2). The preliminary electron trajectory calculations described below will therefore be subject to revision (primarily in the region close to o) as better field models become available. A more exten- sive discussion of the sensitivity of the particle tracing calculation to variations in the magnetic and electric field environment will be considered in a separate study (R. M. Thorne et al., manuscript in preparation, 1999). On the basis of the model fields described above, the drift trajectories of electrons with c : 90 ø are shown in Figure 7 for four distinct values of the first invariant, which correspond to electron kinetic energies of 45 kev, 85 kev, 1 MeV, and 10 MeV in the region upstream of o. Lower energy electrons (2.5 MeV/G and 5.0 MeV/G) have direct access (along open drift orbits) to the Galileo trajectory (shown by the bold solid line) prior to 1744 UT and following 1748 UT. The phase space density fu should be conserved along these open drift trajectories, and the analysis outlined in section 3 can therefore be used to describe the variation of the flux as electrons are carried from the upstream source into the field depression near o. However, these lowenergy electrons would not have direct access to an extended region of the downstream wake since the electron drift trajectories either intercept or lie very close to o. n the absence of scattering (which would allow electrons to move from one drift trajectory to another), one should expect a decrease in fu in the downstream wake region. The spatial extent (along the Galileo orbit) of this region without direct access from the upstream flow is dependent on electron energy, and it should also be influenced by the adopted field model which becomes less accurate close to o (Figure 2, bottom plot). Although precise comparisons should be deferred until better field models become available, the EPD data clearly exhibit a decrease in the phase space density of lower energy electrons over an extended spatial region in the downstream wake. The top plot of Figure 8 exhibits the variation in phase space density along the Galileo orbit between 1730 and 1800 UT for electrons with c m 90 ø. The gradual decrease in f, as Galileo moved inward toward Jupiter is consistent with the radial variation described by Ye and Armstrong [19931 using earlier Voyager data. The radial gradient probably e 6.0 r. 9.0[ ed) 8.o 7.0 K :30 17:36 17:42 17:48 17:54 18:00 Spacecraft Event Time Figure 8. (top) The phase space density for equatorially mirroring (c m 90 ø) electrons near o exhibits a modest depletion in the wake region for/ < 5 MeV/G. The plot for/ - 30 MeV/G is dominated by the penetrating background flux which exhibits a much larger depletion surrounding o. (bottom) The phase space density for electrons with mirror points far from o exhibit a large depletion in the wake region behind o. represents a quasi-equilibrium balance between inward radial transport (from a source at higher L) and weak precipitational loss in the torus. The sharp increase in phase space density near 1734 UT is caused by a rapid (inward moving) interchangevent which has been described by Thorne et al. [1997]. The feature of interest to the present study is the abrupt decrease (and subsequent recovery)in fu for / MeV/G and 5.0 MeV/G as Galileo passed through the region of the wake (shown shaded) behind o between about 1744:30 and 1747:30 UT. The general location of the depletion in fu is roughly consistent with the expectations of particle tracing (compare the top plot of Figure 8 with the top two plots of Figure 7). However, the decrease in fu is relatively modest, and the lower-energy (< 10 MeV/G) particle fluxes remain well above background throughout the wake. This suggests that particles can gain access to this region by scattering across surfaces of constant H(x) (see equation(2)). The large-amplitude magnetic field fluctuation near entry into and exit from low-speed wake [Kivelson et al., 1996b] could provide a mechanism for such scattering. Extremely high energy (> 10 MeV) electrons are a serious source of contamination throughout most of the

10 14,764 THORNE ET AL.: BUTTERFLY DSTRBUTON NEAR O 17:45:26 17:45:46 17:46:49 17:47:09 17:47:27 1.o!! i i i! ', 9 i i i i i i.,.-..' ',.'--- ' ',,' ' '. i.',' ' ',,',' ', ' 0.5 ',,.,' ', ",' -. ',1 ' ', - ', ' ', ',,,',, ',,' '..,,. 'J,.,',,.',,, ',. o.o / ",,, / /,,, / 2.0 ß,,'" ß ' ',,'', 0.5 ',,' ''' ',,' '1 ß,' ',. -.,' ',... ' " ß -. o - -.,',' ' ',',' ',',', ß '1 ''', 0.0 / i,,. ' i,,, ',,,,''',,,' ',,'.-'*... ß,. -.oo ,' " ' "' '" "' '",tt i,, -, o.o -1 o 1-1 o -1 o i -1 o i -1 o 1 x0-180 o Cosine Figure 9. Trapped low-energy electron population with (sin > sin *) observed by EPD in the o wake region on day 341, Fluxes fall to tow levels over the pitch angie range between * (indicated by dashed vertical lines) and the Jovian loss cone (cos 4-1) where intense beams are observed. o torus. The background flux of penetrating particles from the upstream region) for the highest-energy (> 10 tends to dominate EPD count rates for electron energies MeV) electrons. For the MHD simulation fields adopted above 100 kev. Drift trajectories of such high-energy for this study, the exclusion zone for/ = 6000 MeV/G electrons are strongly influenced by the enhanced mag- extends between 1741 and 1753 UT along the Galileo netic field gradients surrounding o. The bottom two trajectory. The background flux of contaminating (> plots of Figure 7 exhibit electron drift trajectories for 10 MeV) electron should be substantially reduced over / = 100 MeV/G and 6000 MeV/G which correspond this region. The large reductions in the count rate of to kinetic energies of MeV and 10 MeV, respectively, channels F1 and B1 (Figure 2), which are dominated by in the undisturbed torus upstream of o. Closed drift penetrating particles, and the pronouncedrop in f, for orbits surrounding o and orbits which intercept the / m 30 MeV/G (also mostly dominated by penetrating satellite (shown as dotted curves) indicate the existence electrons) as exhibited in the top plot of Figure 8 are of an extended "exclusion zone" (which is inaccessible consistent with the anticipated spatial extent of this

11 THORNE ET AL.: BUTTERFLY DSTRBUTON NEAR O 14,765 exclusion zone. High-energy particles could still gain access to the region of closed drift orbits by nonadiabatic processes which cause scattering across surfaces of constant/-/(x) during transport past o. The quasi adiabatic response of the low-intensity flux of penetrating electrons to the magnetic field variations around closest approach to o, as indicated by the third plot of Figure 2, suggests that a small fraction (few percent) of highenergy electrons in the o torus are, indeed, injected onto closed drift orbits in the wake region, but their flux is very modest. Another distinct feature of the wake region behind o is the pronounced reduction in flux for all low-energy electrons measured with pitch angles over the range between the intense field aligned beams and sin c = sin c * as shown in Figure 9. Such electrons, which would mirror far from o and thus spend little time in the region of field depression, are expected to be carried along drift trajectories where ()(x) m constant (section 3). n the absence of significant cross (x) gradient drift such electrons would not have direct access to the downstream wake. The lower plot of Figure 8 shows the variation in phase space density for electrons with mirror points far from o. Such electrons exhibit a drop in f, by almost 2 orders of magnitude in the region of the downstream wake (compare this with the much more modest depletions for electrons with c 90 ø (top plot)). t therefore appears that electrons with mirror points far from o are subject to little scattering into the wake, presumably because of the limited time spent on their bounce path within the field depression near o. The low-energy electron distribution in the wake therefore has the characteristics of a population which is "trapped" on bounce orbits within the field depression near o. Over the range sin c > since*, the "trapped" electron population in the wake initially (second col- umn of Figure 9) retains some evidence of the butterfly distributions which characterize the region near o but outside of the downstream wake. However, as described above, the pitch angle distributions observed in the wake do not respond adiabatically, and there appears to be evidence of a progressive depletion of elec- trons between c : 90 ø and c c * as Galileo moved farther into the wake (compare columns 2, 3, and 4 of Figure 9). Since the electrons observed close to the exit from the wake are expected to follow drift trajectories that pass closer to o as Galileo traversed the wake (Figure 7), we conclude that the erosion in flux as could be induced by collisions with the satellite or its atmosphere. A more detailed discussion of the "trapped" electron distributions in the wake region is given in a companion paper [Williams et al., this issue]. 5. Conclusions Variations in the energetic electron distribution during the Galileo passage by o have been described in terms of an adiabatic response to the disturbed elec- tric and magnetic field environment near the satellite. Electron pitch angle distributions evolve from the characteristic pancake distribution (peaked at c - 90 ø) in the background torus to a butterfly distribution in the strong magnetic depression near o. Strong flux depletions for pitch angles near 90 ø are caused by a reduction in the particle kinetic energy, associated with the conservation of the first adiabatic invariant, as electrons are carried by the flow into the field depression near o. Over an extended spatial region in the low-speed wake behind o, low-energy electrons ( 100 kev), which remain above background levels throughout the encounter, exhibit an abrupt transition to a pitch angle distribution consistent with trapping within the magnetic field depression surrounding the satellite. Those electrons which mirror far from o should roughly conserve kinetic energy and therefore follow trajectories along surfaces of constant electric potential. Such par- ticles would consequently not have direct access to the wake, and this can lead to an abrupt transition from a butterfly distribution to a "trapped" population within the low-speed wake region. We demonstrate that the phase space density of electrons with mirror points well away from o is reduced by almost 2 orders of magnitude in the wake region. The observed electron variations in this region are far from adiabatic, and the phase space density of the "trapped" population with c - 90 ø is also depressed compared to values in the surrounding torus. The "trapped" population appears in the same region that intense electron beams were observed [Williams et al., 1996], but the two populations are thought to have an entirely different origin as discussed by Williams et al. [this issue]. A particle tracing code has been used to follow electron trajectories under the influence of the disturbed plasma flow and enhanced magnetic gradient drifts due to the strong magnetic field depression near o. Using the most realistic E and B fields obtained from an MHD simulation of the region near o [Linker et al., 1998] we demonstrate that equatorially confined (c - 90 ø) lowenergy (Ek 100 kev) electrons should not have direct access to an extended region of the wake behind o from a source region in the upstream flow. nstead, the "trapped" electron distribution observed in the downstream wake must be injected into the region by scattering. The process responsible for scattering into the wake must be eificient since the phase space density of the "trapped" electron population is reduced by less than a factor of 3 below values in the surrounding torus. n the future, we plan to undertake more detailed modeling of adiabatic trajectories for particles with finite first and second adiabatic invariants. This tech- nique should allow us to place further constraints on the spatial distribution of magnetic and electric fields around o, and thus address the fundamental question of whether the field depression observed by Galileo is partially due to an internal onian magnetic field [Kivelson et al., 1996a] or whether it is caused by diamagnetic

12 14,766 THORNE ET AL.: BUTTERFLY DSTRBUTON NEAR O effects associated with the plasma interaction with the satellite. Acknowledgments. This study has benefited from numerous informative discussions with M. G. Kivelson. The MHD model field parameters used in our particle tracing were provided by J. A. Linker and K. K. Khurana. F. Xiao assisted with some of the graphic presentations. This work was supported by the subcontract from the Applied Physics Laboratory, Johns Hopkins University to the primary NASA grant G1770. Janet Luhmann thanks Frank Crary and another referee for theif assistance in evaluating this paper. References Frank, L. A., W. R. Patterson, K. L. Akerson, V. M. Vasyliunas, F. V. Coroniti, and S. J. Bolton, Plasma observations at o with the Galileo spacecraft, Science, 27, 394, Hill, T. W., and D. H. Pontius Jr., Plasma injection near o, J. Geophys. Res., 103, 19,879, Khurana, K. K., Euler potential models of Jupiter's magnetospheric field, J. Geophys. Res., 102, 11,295, Khurana, K. K., M. G. Kivelson, and C. T. Russell, nteraction of o with its torus: Does lo have an internal magnetic field?, Geophys. Res. Lett., 2J, 2391, Kivelson, M. G., K. K. Khurana, R. J. Walker, C. T. Russell, J. A. Linker, D. J. Southwood, and C. Polansky, A magnetic signature at o: nitial report from the Galileo magnetometer, Science, 273, 337, 1996a. Kivelson, M. G., K. K. Khurana, R. J. Walker, J. Warnecke, C. T. Russell, J. A. Linker, D. J. Southwood, and C. Polansky, o's interaction with the plasma torus: Galileo magnetometer report, Science, 27J, 396, 1996b. Linker, J. A., K. K. Khurana, M. G. Kivelson, and R. J. Walker, MHD simulations of o's interaction with the plasma torus, J. Geophys. Res., 103, 19,867, Neubauer, F. M., The sub-alfv nic interaction of the Galilean satellites with the Jovian magnetosphere, J. Geophys. Res., 103, 19,843, Thorne, R. M., T. P. Armstrong, S. Stone, D. J. Williams, R. W. McEntire, S. J. Bolton, D. A. Gurnett, and M. G. Kivelson, Galileo evidence for rapid interchange transport in the o torus, Geophys. Res. Lett., 24, 2131, Williams, D. J., R. W. McEntire, S. Jaskulek, and B. Wilken, The Galileo energetic particles detector, Space Sci. Rev., 60, 385, Williams, D. J., B. H. Mauk, R. E. McEntire, E. C. Roelof, T. P. Armstrong, B. Wilkin, J. G. Roederer, S. M. Krimigis, T. A. Fritz, and L. J. Lanzerotti, Electron beams and ion composition measured at o and in its torus, Science, 274, 401, Williams, D. J., R. M. Thorne, and B. H. Mauk, Energetic electron beams and trapped electrons at o, J. Geophys. Res., this issue. Wolf-Gladrow, D. A., F. M. Neubauer, and M. Lussem, o's interaction with the plasma torus: A self-consistent model, J. Geophys. Res., 92, 9949, Ye, G., and T. P. Armstrong, Electron distributions in the inner Jovian magnetosphere: Voyager observations, J. Geophys. Res., 98, 21,253, S. Stone, Fundamental Technologies, 2411 Ponderosa, Suite A, Lawrence, KS (stone@ftecs.com) R.M. Thorne, and L. Zhang, Department of Atmospheric Sciences, University of California, P. O. Box , Los Angeles, CA (rmt@atmos.ucla.edu; zhang@atmos.ucla.edu) D.J. Williams, Applied Physics Laboratory, Johns Hopkins University, Laurel, MD (djw@ap 1 comm.j hu ap 1. ed u ) (Received December 1,1998; revised February 15,1999; accepted March 8, 1999.)