Main

There were several surprises that came out of the first perijove (PJ1) encounter (on 27 August 2016) of NASA’s Juno spacecraft with the low-altitude regions of Jupiter’s polar auroral regions. Magnetic perturbations were far weaker and more disordered than anticipated or modelled on the basis of the expected magnetic-field-aligned auroral electric currents8. Also, the downward electron energy spectra that Juno observed over the aurora did not have the anticipated strong energy peaks and, in particular, the expected peaks at energies of more than 50 keV (refs 9, 10). Electron distributions with high-energy peaks were expected on the basis of the physics derived for Earth and the characteristic energies (50–500 keV) inferred from remote spectrographic imaging of Jupiter’s aurora12,13,14.

First studies of Jupiter’s high-energy (>30 keV) auroral electrons using the Jupiter Energetic Particle Detector Instrument (JEDI)15 used relatively low-resolution measurements (30 s)9. However, the higher-time-resolution measurements (0.5–1.5 s) that we present here confirm the fundamental finding from the PJ1 encounter that no downward discrete auroral acceleration to energies of more than 30 keV was observed (Fig. 1; the global auroral context is presented elswhere16, but is similar to Fig. 2a). In Fig. 1, downward energy fluxes are shown (peaking near about 700 mW m−2) that are sufficient to account for the nominal and intense auroral intensities that are observed remotely12,13,16. Surprisingly, even with the large magnitude of the downward energy fluxes, the upward energy fluxes are often even larger (these auroral distributions are often in the form of asymmetric, bi-directional angle beams; Fig. 1d). The downward electron intensity spectra shown in Fig. 1a reveal broad energy distributions rather than sharply peaked ones as anticipated from observations of Earth’s aurora. It has been proposed that a broadband, stochastic acceleration is responsible9. The electrons measured at lower energies by the Jupiter Auroral Distribution Experiment (JADE)17 on Juno had similar features with no sharply peaked electron distributions10. Together JADE and JEDI measure the energies from 0.1 keV to 1,000 keV and thus show that no peaked electron distributions existed for the time period shown in Fig. 1 or for any time during the PJ1 encounter.

Figure 1: Energetic electron data from JEDI during Juno’s first perijove (PJ1) encounter with Jupiter (27 August 2016).
figure 1

a, Downward intensity I versus electron energy E spectra (see Methods) for the times identified in c. b, Integrated downward electron energy flux, calculated as summarized elsewhere9 (see Methods). c, Intensity versus energy distributions for electrons within 20° of the downward magnetic field direction. d, Pitch angle distributions of intensities averaged over electron energy (30–1,000 keV); the labels ‘upward’ and ‘downward’ indicate the portions of the plot (top and bottom) that represent electrons moving away from and towards Jupiter, respectively. The pitch angle is the angle between the particle velocity and the local magnetic field from the MAG instrument21. ‘Mlat’ is the dipole magnetic latitude using the VIP4 field model dipole22. R is the distance of Juno from the centre of Jupiter, in units of Jovian radius (RJ). The counting rates for these electron measurements are very high, so statistical error bars would not be visible on these plots.

PowerPoint slide

Full size image
Figure 2: Ultraviolet auroral images of Jupiter from the Juno Ultraviolet Spectrograph (UVS) instrument.
figure 2

The images contain intensities from three spectral ranges, false-coloured red, green and blue, providing qualitative information on precipitating electron energies (high, medium and low, respectively)23. An estimate of the magnetic projections of the Juno trajectory is shown (red lines), determined using the VIPAL24 magnetic field model with large uncertainties, with tick marks in steps of 1 h (see Methods). The short yellow arcs with arrows indicate the direction to the Sun when the image was taken. The blue-green lines are the average positions of the main ultraviolet aurora for the south and north, respectively12,13. a, Jupiter’s southern aurora taken during the fourth perijove (PJ4) encounter on 2 February 2017. b, Jupiter’s northern aurora taken during the third perijove (PJ3) encounter on 11 December 2016. The single orange arrow indicates approximately when the particle data were taken (Fig. 3) and is discussed in Methods in the context of magnetic mapping errors.

PowerPoint slide

Full size image

More recent perijove passes have revealed the presence of high-energy (>50 keV), discrete, downward auroral acceleration in the form of ‘inverted V’ structures that are embedded in other ongoing processes, for about 50% of the main auroral crossings (Fig. 3; from the fourth perijove (PJ4) encounter; Fig. 2a shows the qualitative global auroral context). The intensity (I) versus energy (E) spectra show energy beams with positive slopes and that increase faster than E+1 (a true energy beam has a positive slope in the phase-space density with a shape of I/E in the non-relativistic regime18; see Methods). The peak in the intensity distributions is inferred to represent the electric potentials along magnetic field lines that have coherently energized the electrons2. For the clean inverted-V distributions, the intensities decrease rapidly above the energy peak. However, spectrum (3) in Fig. 3a might represent a transition from one process to another, or just the simultaneous occurrence of the two different processes. The peak in this intensity spectrum is marginally suggestive of a discrete acceleration process. However, a larger portion of downward energy flux comes from broadband additions, extending to energies both below and above the peaked structure. The highest downward energy fluxes come from spectrum (4) in Fig. 3a, which displays no evidence of discrete auroral acceleration.

Figure 3: Energetic electron data during Juno’s fourth perijove (PJ4) encounter with Jupiter’s low-altitude polar regions (2 February 2017).
figure 3

As Fig. 1; the auroral context is shown in Fig. 2a.

PowerPoint slide

Full size image

One example of particularly energetic downward acceleration was observed during a northern polar pass within an arc-like structure located poleward of the nominal main auroras (Fig. 2b). Contrary to the case of the PJ4 encounter, such a location does not map to the middle magnetosphere but far into the outer magnetosphere19. Within this structure a downward electron beam was observed that peaked at about 400 keV, more than an order of magnitude larger than the largest values observed at Earth (Fig. 4)11. In this extreme case the downward energy fluxes (approximately 70 mW m−2) were more comparable in magnitude with the maximum downward energy fluxes that were observed elsewhere for this region (150–300 mW m−2; the entire event is not shown). However, even here the coherent discrete auroral acceleration was not the prime acceleration mechanism.

Figure 4: Energetic electron data during Juno’s third perijove (PJ3) encounter with Jupiter’s low-altitude polar regions (11 December 2016).
figure 4

As Fig. 1; the auroral context is shown in Fig. 2b.

PowerPoint slide

Full size image

At Earth, the broadband or Alfvénic aurora represents stochastic acceleration that is caused by particle interactions with various downward-propagating Alfvén waves3,4 at radial positions of 4–5 Earth radii20, generating transitory aurora at the poleward boundary of the auroral oval. At low altitudes the downward fluxes will be stronger than the upward fluxes4, opposite to what is often, although not always, observed at Jupiter.

Another form of stochastic acceleration at Earth is the mostly upward acceleration that occurs in regions of downward electric currents1,2. Much weaker downward components are revealed, cutting off at much lower energy than does the upward component. These downward fluxes are exceedingly weak, yielding ‘black aurora’. These regions have qualitative similarities to those observed at Jupiter. But at Jupiter the downward intensities extend in energy to very high energies (>1,000 keV), and have sufficient downward energy fluxes to power the most intense auroral emissions in the Solar System. We do not know whether the similarities are coincidental or result from similar physical processes. The magnetic perturbations are too weak and disordered to allow for decisions on the basis of electric current polarity8.

The transition in Fig. 3a from the coherent acceleration regions (spectra (1) and (2)), through spectrum (3), which combines stochastic and coherent acceleration, to the region dominated by stochastic acceleration (spectrum (4)) is intriguing. It is possible that auroral acceleration begins as a coherent discrete acceleration process like Earth’s, but then instability is initiated as the drivers become stronger, resulting in the broadening of the distributions through stochastic processes. This and other possibilities will be pursued in future studies as additional data becomes available.

Methods

Energy flux

Although summarized elsewhere9, here we provide additional information regarding the method of estimating the downward energy flux (in mW m−2) of electrons that impinge onto Jupiter’s atmosphere to generate aurora. The data are filtered for measurements that have pitch angles centred within 15° of the downward magnetic field direction to estimate the average intensities in the geometric loss cone. The geometric loss cone is the cone that is centred on the magnetic field direction and contains the velocity vectors of those particles that will strike the atmosphere before the magnetic mirror force can reflect them back to the upward direction. We then perform the summation over all n, where n represents the JEDI energy channels, and In is the particle intensity from each channel, En is the central energy of each channel and ΔEn is the energy band pass of each channel. The coefficient π is the area-projection-weighted size of the loss cone just above the atmosphere, used on the basis of the rough estimate from the observations that the downward geometric loss cones are fully populated. We assume that as the particles move along the magnetic field line from the spacecraft to the atmosphere there is no retarding electric field below the spacecraft that would decrease the energy of the electrons on their way to the atmosphere. Hence, the intensities in the loss cone just above the atmosphere are greater than or equal to the intensities at the spacecraft.

Energy spectra

The intensity versus energy spectra shown in Figs 1a, 3a and 4a were modestly corrected to take into account the fact that a portion of the higher-energy particles fully penetrate the detectors. A preliminary version of the correction procedure is summarized elsewhere9. These corrections do not change the fundamental character of the spectral shapes. The fractional penetration of the higher-energy particles yields an additional high-energy efficiency of the form

where 0 < ε < 1 is the efficiency and EkeV is energy in kiloelectronvolts. Each electron that penetrates the detector leaves behind a fraction of its energy, called the ‘minimum ionizing’ energy. The distribution of minimum ionizing energy that is deposited is parameterized as

where

and the factor of 35.47 factor normalizes the distribution to a unit area.

These equations are combined with a parameterized spectral function (equation (1) in ref. 25) to fit the higher energies and thereby reproduce the JEDI measured responses by means of a free-parameter optimization procedure. This process is visualized elsewhere9, but here the accuracy of the procedure (using the parameters of equations (1)–(3)) is substantially improved. The procedures described here are unnecessary for the spectra that have strongly peaked energy distributions, with the exception that the corrected efficiency factor (equation (1)) is used at the highest energies.

Magnetic mapping

Juno magnetic field measurements show that the configuration of the magnetic field close to the planet is substantially different from what prevailing magnetic field models can reproduce26. When an existing magnetic field model is used to map the position of the spacecraft onto the atmosphere, as was the case for Fig. 2, substantial uncertainties are expected. This point is illustrated here using Fig. 2b and Extended Data Fig. 1. Extended Data Fig. 1 is identical to Fig. 4b–d except that the timescale is expanded to include the entire northern polar pass. The major bump that occurs in Extended Data Fig. 1a, centred at about 15:40 universal time (ut), occurs about half way (in terms of time) between the main auroral crossings near 15:00 ut and 16:20 ut. This feature clearly corresponds to the auroral arc structure in Fig. 2b that appears in the central regions of the polar cap, poleward of the main auroral oval. Yet, the bump at 15:40 ut occurs at a time along the projected trajectory in Fig. 2b (indicated with a small orange arrow) that is well separated from the image of the arc. This problem is one of magnetic mapping; when an accurate model of the magnetic field close to the planet becomes available later in the Juno mission, we expect that the position of the spacecraft at 15:40 ut will map much more accurately to the position of the arc structure. This discussion is intended to illustrate why the auroral images at this point in time provide only qualitative global contexts to the in situ particle measurements. The images are also obtained at different times from those of mapped Juno crossings of the various features.

Phase space density

As detailed in standard text books18, the behaviour of statistical distributions of objects such as atoms or charged particles is governed by Liouville’s theorem. When the interactions between the objects is weak (collisionless), Liouville’s theorem simplies to the collisionless Boltzmann equation, known as the Vlasov equation when applied to electrified gases called plasmas. The Vlasov equation governs the behaviour of ‘phase space density’ (PSD) distributions, which are the densities of particles in the six-dimensional space comprising configuration space (x, y, z) and momentum space (px, py, pz). PSD(p), where p is the momentum vector, contains the signatures of acceleration processes on particle populations. The directional intensity of particle populations (I(E)) is generally used in observational studies because its shape better reflects the raw instrumental measurement parameters (compared to the PSD) and because it is useful in determining the important parameters of particle and energy flux onto a surface or atmosphere. But, the shape of directional intensity profiles can be misleading with regard to the processes that are acting on a population of particles. Such profiles can have peaked distributions even if there are no coherent acceleration processes acting on the populations (see, for example, Fig. 1a). There is a fairly easy conversion between the I(E) and PSD(p): PSD = I/p2, which in the non-relativistic regime is proportional to I/E. The clear identification of coherent acceleration of electrons by magnetic-field-aligned electric fields occurs only in the PSD, or equivalently in the I(E) distributions with positive slopes that are greater in magnitude than the slope of Em with m > 1.

Data availability

The data presented here are available from the Planetary Plasma Interactions Node of NASA’s Planetary Data System (https://pds-ppi.igpp.ucla.edu/). All data are in the form of column-labelled, .csv, ASCII flat files.