On 2–3 January 2001, en route to Saturn, measurements of Jupiter's synchrotron emission were successfully carried out using the radiometer subsystem of the Cassini radar instrument7 operating at 2.2 cm (13.8 GHz). Synchrotron radiation is the relativistic counterpart to cyclotron radiation (the emission originates from the gyration motion of trapped electrons). Relativistic effects cause the continuum emission to be radiated in a narrow beam focused in the direction of the electron's motion. The spectrum of synchrotron radiation is characterized by a well defined peak frequency that depends on both electron energy and magnetic field magnitude. Observations at different frequencies (assuming equivalent magnetic field magnitude) relate to emission from electrons at different energies; however, owing to confusion with thermal emission, ground-based telescopes have difficulty observing synchrotron emission at frequencies higher than 5 GHz. Cassini's proximity to Jupiter and the excellent sensitivity of the radar instrument offered an opportunity to map and accurately measure Jupiter's synchrotron emission at a higher frequency than is possible with ground-based telescopes. The motivation was to constrain the high end of the energy spectrum of the electrons trapped in Jupiter's radiation belts. Simultaneous ground-based observations at a wide range of frequencies were carried out to obtain data corresponding to emission from 5 MeV to >50 MeV electrons. Interferometric maps were obtained with the Very Large Array (VLA) operating at 20 cm (1,400 MHz) and 90 cm (333 MHz). Single-dish total flux density measurements were obtained with the Deep Space Network (DSN) operating at 2.2, 3.5 and 13 cm wavelengths (2.2 and 3.5 cm were used for thermal emission calibration only).

The narrow beaming of the synchrotron emission couples with the non-symmetrical magnetic field and the electron pitch angle distribution to produce variations as Jupiter rotates (the beaming curve). These variations can be seen in the Cassini polarization maps shown in Fig. 1. Cassini unambiguously detected synchrotron emission at a level of 0.44 ± 0.15 Jy (jansky; total integrated flux density adjusted to 4.04 AU). The synchrotron emission was observed to be approximately 1% of the measured thermal emission from Jupiter's disk. The level of synchrotron emission measured was less than estimated from simulations using the energy spectrum and the spatial distribution from current models5,6. The results demonstrate the existence of ultra-relativistic electrons near Jupiter and extend the measured nonthermal radio spectrum, providing a new constraint for models of Jupiter's radiation belts.

Figure 1: Colour maps of Jupiter's synchrotron emission taken by the Cassini spacecraft show the distribution of the ultra-relativistic electrons in Jupiter's inner radiation belts.
figure 1

Maps of horizontal linear polarization (left side) and vertical linear polarization (right side) show bright regions near 1.4 Jupiter radii (RJ), corresponding to peak emission by electrons of energy 50 MeV. The observation was carried out shortly after the closest approach to Jupiter, at a distance of 149RJ (107 km) and a phase angle of approximately 75 degrees, using the 4-m on-axis Cassegrain reflector antenna that also serves as the Cassini primary communication antenna. The sub-spacecraft latitude on Jupiter was approximately 1.2 degrees south during the observation. The angular diameter of Jupiter was approximately 0.75 degrees. The colour scale is linear from 0 to 3 kelvin, black to yellow, as shown in the colour bar. Negative values (caused by noise and atmospheric model subtraction) have been set to zero. The 20-h observation was split into two sets of repeating raster scans, obtaining one complete rotation of Jupiter at each of two orthogonal polarizations. Each scan covered approximately 8RJ and 6RJ in east–west and north–south extent, respectively. A single scan, composed of 12 rasters separated by 0.18 degrees or approximately half-beam width (0.23 Jupiter angular diameters), required one hour to complete. Owing to the orientation of the magnetic field, emission originating near the magnetic equator is predominantly linearly polarized in the horizontal sense, while emission originating at higher latitudes is linearly polarized in the vertical sense. To remove thermal emission, a radiative transfer calculation for a nominal model of Jupiter's atmosphere was used to determine the brightness distribution across Jupiter's disk. The antenna beam pattern was determined from a composite of raster scans of the Sun and Jupiter (using an identical technique) obtained before the fly-by. The antenna gain and pointing varied less than 1% and 0.02 mrad, respectively, over the entire 20-h period of the observations. The relative distributions and amplitudes of the residuals as a function of polarization agree well with our model predictions. The variation between maps is related to the beaming curve, and has been studied extensively at 13-cm wavelength by Klein et al.22. The variation seen in the different panels shows that a form of beaming is present at 2.2 cm.

Observations with the VLA were carried out on 3 January 2001 when Earth was at a jovian declination of 3 degrees. Figure 2 shows the VLA maps at 20 and 90 cm as a function of central meridian longitude. The total synchrotron emission measured was 5.15 ± 0.7 Jy at 90 cm (with this configuration the VLA cannot accurately measure the total flux at 20 cm). All four Stokes parameters were measured, although here we present only the images of total intensity, converted to brightness temperature. Observations from the DSN antennas in collaboration with the Goldstone–Apple Valley radio telescope (GAVRT) were obtained from November 2000 to March 2001. The GAVRT project also provided an opportunity for students to work with Cassini scientists, conducting ground-based observations and data analysis. Results from the DSN–GAVRT programme are shown in Fig. 3. The interpolated value of the synchrotron flux density on 3 January was 4.02 ±0.08 Jy. Short-term variations at the 10% level are evident in the data set.

Figure 2: Colour maps of Jupiter's synchrotron radiation taken with the VLA show the distribution of less energetic relativistic electrons (as compared with Fig. 1).
figure 2

The brightest regions near 1.4RJ correspond to peak emission from electrons of energy 15 MeV (a, 20 cm), and 7 MeV (b, 90 cm). The observations were made for one full jovian rotation using the VLA operating in A configuration (36 km extent) with a synthesized beam width resolution of approximately 6 arcsec at 90 cm, and 1 arcsec at 20 cm (half-power beamwidth, HPBW). Two-dimensional images were constructed every 40 degrees of central meridian longitude (CML) from data recorded at the given CML ± 40 degrees. This is a standard construction technique, and thus images from different epochs can be compared directly. After calibration, Jupiter's thermal emission was subtracted (corresponding to a blackbody disk at 350 K), and background confusion sources were removed. For comparison purposes, the VLA data were adjusted to a standard distance of 4.04 AU.

Figure 3: Variability of Jupiter's synchrotron radio emission shows evidence of long- and short-term variability.
figure 3

a, The history of long-term variations in the flux density of Jupiter's synchrotron radio emission at 13-cm wavelength. b, An expanded segment of the data taken over the past 2.5 yr that includes the Cassini fly-by period. Thermal flux density from the planet's atmosphere was subtracted from each measurement (assuming a disk temperature of 300 K) and results were normalized to a standard distance of 4.04 AU from Earth. Approximately 2,300 students and their teachers from 26 schools across the United States were part of the ground-based research team working through the GAVRT educational programme. The GAVRT data were merged with the ongoing NASA/JPL Jupiter Patrol to improve the time resolution of the data (JPL, Jet Propulsion Laboratory). The open circles are NASA/JPL Jupiter Patrol observations made with DSN antennas with apertures of 70 m and 34 m. GAVRT team observations made with 34-m-diameter antennas are represented by filled triangles. The agreement between the two data sets is approximately 1%. These events of ‘modest’ brightening appear to be intrinsic to Jupiter and are not caused by systematic errors in calibration or by discrete background radio sources that Jupiter passes in front of during its 12-yr orbital path along the ecliptic. Previous observations have suggested the presence of short-term variations in Jupiter's synchrotron radio emission23–25. SL9 indicates comet Shoemaker–Levy 9.

The new synchrotron emission radio spectrum data is shown in Fig. 4. Data from an earlier epoch at one additional wavelength8 is shown for completeness. At a frequency of 13.8 GHz (2.2 cm), the region of peak radio emission seen in the maps of Fig. 1 correspond to an average electron energy of 50 MeV gyrating in a background magnetic field of approximately 1.2 G (gauss). The VLA maps at 1.4 GHz and 333 MHz show regions of peak emission corresponding to average electron energies of 15 MeV and 7 MeV, respectively, in a 1.2-G magnetic field. Because synchrotron radiation is emitted as a continuum, a wide range of electron energies contribute to the maps at each frequency and detailed modelling is required to estimate accurately the electron energy spectrum from the full set of multi-frequency observations. Figure 1 shows significant emission at radial distances >2 RJ, suggesting much higher energy electrons (100 MeV) at the larger radial distances. At a subset of central meridian longitudes, a double radiation belt is possibly indicated with the outer belt located just outside of Jupiter's main ring near 1.8 RJ, indicative of electron absorption by ring material.

Figure 4: Jupiter's nonthermal radio spectrum indicates a softer than expected electron energy spectrum at high electron energies.
figure 4

The observed flux density of Jupiter's synchrotron emission observed with Cassini (13.8 GHz) is plotted alongside simultaneous measurements obtained using the DSN (2.3 GHz) and the VLA (0.333 GHz). Previous measurements at 6 cm (5 GHz) are shown for completeness8. All measurements are from January 2001 except for the 6-cm data from 1994. Error bars on the data are indicative of the preliminary nature of the results. Further analysis, including a recent Cassini beam calibration, is expected significantly to reduce the error bars on the data. The ratio between the integrated nonthermal flux densities measured at 2.2 cm and 13 cm is S2.2/S13 = 0.105±0.035.

The softer energy spectrum implied by the Cassini high-frequency observations indicates fewer ultra-relativistic electrons (50 MeV) than predicted by current radiation-belt models. Current models of Jupiter's radiation belts use electron energy spectra based on in situ data from spacecraft (Pioneer3 and Galileo probe4) and ground-based synchrotron observations. Although the in situ measurements suggested the existence of electrons greater than 20 MeV, the measurements lacked sufficient energy resolution to constrain the energy spectrum at high energies. Simulations of synchrotron emission using the Cassini-derived softer energy spectrum failed to match the intensity level of the lower range of frequencies observed. The preliminary modelling analysis suggests that both a softer energy spectrum and an increase in the number of electrons below 20 MeV are required to match the synchrotron emission spectrum in Fig. 4. The results suggest that the models of Jupiter's radiation belts5 will probably require more modifications than were previously recommended9 to estimate accurately the hazards to spacecraft venturing close to Jupiter.

The results reported here have direct effects on our understanding of Jupiter's radiation belts. The detection of synchrotron emission at 13.8 GHz implies that a steady source of 50 MeV electrons must exist. This challenges current theories that have difficulty identifying a source even for the 20-MeV electrons that are known to exist from previous observations. Secondly, the DSN–GAVRT 13-cm result confirms that the synchrotron emission can vary on relatively short timescales (days), which implies the existence of a process that can rapidly affect the distribution or number of 20-MeV electrons trapped in Jupiter's strong magnetic field.

The most widely accepted theory postulates that the relativistic electrons originate in the solar wind or in the outer jovian magnetosphere and diffuse inward, gaining energy through conservation of the first and second adiabatic invariants10,11,12. This can account for 10-MeV electrons near the peak of the synchrotron zone. However, additional acceleration mechanisms are required to explain higher energies. Efforts to model radial diffusion at Jupiter have failed to reproduce accurately a synchrotron emission radio spectrum12. Additional sources of electron energization that are known to occur in the Earth's magnetosphere include local stochastic acceleration by wave–particle interactions13,14, and acceleration by field-aligned potentials in the auroral zone15. Intense wave activity may also be responsible for rapid time variations in synchrotron emission16. Alternative mechanisms, which combine radial diffusive transport with wave–particle interactions, such as recycling and magnetic pumping17,18,19,20, have also been proposed. Interaction between a very fast interplanetary shock and the Earth's magnetosphere has been proposed as a source of drift resonance acceleration to explain the sudden appearance of relativistic electrons in the terrestrial ‘Van Allen’ radiation belts21. A similar process could exist at Jupiter, although analogies between the terrestrial and jovian magnetosphere are not always appropriate.

Thus, the inner radiation belts of Jupiter and Earth both contain extremely high-energy electrons that require substantial acceleration by processes other than adiabatic radial diffusion. These two magnetospheres are arguably the most explored examples of planetary magnetospheres in the Solar System and in many ways represent the archetype for magnetospheric physics. The fact that both magnetospheres contain electrons that require acceleration processes beyond radial diffusion suggests that this may be a fundamental property of all magnetospheres. This implies that our theoretical understanding of the global properties of magnetospheres may need revision and that the importance of local acceleration mechanisms may currently be underestimated.