The operation of the scalar sensor is considerably more complicated than the two fluxgate sensors. In this section we detail the process used during the LEGA. The three Juice Earth gravity assists (EGAs) present crucial opportunities for SCA as they are the only times during the cruise phase when SCA can make a direct measurement of the external field magnitude. By this we mean that the CDSM is not suitable for measurements below 100 nT with high accuracy and so in the solar wind an auxiliary DC field of order 2000 nT is applied via the auxiliary coil solenoid fitted around the sensor optical axis. This lifts the field magnitude in the volume around the sensor way above the 100 nT limit, thus enabling detection of the external field component along the SCA optical axis with high accuracy. To compare fluxgate and SCA data in the solar wind, one must first project the measured vector from the fluxgates onto the SCA optical axis as it is only sensitive to the component of the field strength along this direction.

The duration of the lunar pass was too short for SCA to heat up and stabilise, and with relatively low fields in the magnetotail, the auxiliary coil would be required. For these reasons it was agreed not to operate the SCA during the lunar pass. For the close approach to Earth, the challenge was to switch on while still in a low field environment with the auxiliary coil enabled and then switch off the auxiliary coil once the external field increased to a point where it was comfortably above the detection limit. The approach taken for this was as follows.

SCA was switched on while still in a relatively low field environment, the cell and laser diode were heated to their operational temperatures, and the auxiliary coil was switched on to apply a constant DC field. Resonance sweeps were performed while operating in n2 (longitudinal) resonance mode . This is the standard approach to get SCA into nominal field measurement mode. Then, as the external field increased, the auxiliary coil field was correspondingly ramped down such that the field in the vicinity of the SCA sensor remained above 100 nT at all times in order to avoid losing lock. The timing of the auxiliary coil ramp down profile was estimated in advance by field modelling. As the angle between the external field and the optical axis strayed out of the n2 operation cone, the SCA was switched to n3 (transverse mode) at the start of the auxiliary coil ramp down. This was performed successfully, marking the first time in the mission that the n3 mode was used. When the auxiliary coil current decreased to zero, SCA was free running in n3 mode and measuring the true external field magnitude. This is the first time the SCA was operated nominally, meaning that the angle of the magnetic field was taken from a fluxgate sensor, and based on this the resonance was chosen. This covered approximately four hours centred on the closest approach. There were two mode switches during the LEGA, from n2 to n3 and then back again. They happen at around 60° magnetic field angle, 0–60°n2, and 60–90°n3.

At external fields below 100 nT, a firm Zeeman resonance lock is not possible under all operating conditions. Here, generating valid science data within the SCA specifications is also not possible . As a consequence, SCA was switched back on with a constant auxiliary coil field some time later. SCA remained in that state from that point until the end of the J-MAG operational period. The timeline of SCA operation was executed as expected from the modelling predictions.

We start by examining the performance of each of the sensors within Earth's magnetic field. Figure a shows the magnetic field magnitude measured by each of the three sensors in the two hours around closest approach. Artificial offsets of +100 and +200 nT have been added to IBS and OBS, respectively, in order to make all three traces visible. At this scale, the three measurements lie exactly on top of each other when plotted without offsets applied. Panel b shows a zoom on the few minutes around peak field, but with no artificial offsets. OBS and IBS data are shown at a cadence of 16 Hz, while SCA is 1 Hz. We can see that all signals are very clean, follow the expected pattern at close approach, and at the peak in field magnitude, the fluxgate sensors agree within 3 nT (out of ∼4100 nT) of the scalar sensor.

Figure  shows the magnetic field vector measured by J-MAG during the Earth flyby in GSE coordinates (thin, solid lines). Overlaid (thick, transparent lines) are the results of modelling the geomagnetic field expected to be seen over the flyby trajectory. The internal component of the field model comes from the latest version of the International Geomagnetic Reference Field model IGRF-14; see. Externally-driven perturbations arising from solar wind interaction with the magnetosphere are then added using the Tsyganenko 2001 empirical model seefor details, parametrised using solar wind conditions from OMNI during the flyby (i.e. those in Fig. ).

As discussed in Sect. , by comparing J-MAG observations with the modelled field, we can make an initial calibration estimate. Specifically, we look to see if there is any potential misalignment of the sensors with respect to the defined science coordinate systems. Initially, we see a maximum difference between the observations and model of around 40 nT. By performing a minimisation, we determined that a small rotation of less than 1° in each axis can reduce the maximum difference between the measured and modelled field at any point to ∼20 nT, which we consider very good agreement in this high field environment. It is believed that much of the remaining discrepancy may be physical, instances where the models do not accurately capture the changing conditions, particularly on the edges of the flyby period. This small rotation has been applied to the data displayed in Fig. , as well as subsequent Figs.  and which explore observations of the outer magnetosphere and solar wind. This aspect of calibration is still preliminary and will be refined in future Earth flybys (see Sect. ).

The alignment correction applied here is needed to address several potential sources of error. The Juice magnetometer boom is very long (10.6 m) and comprises three hinged segments. While excellent for electromagnetic cleanliness, this comes with increased risk of mechanical instability that may cause sensors to change their displacement or rotate relative to each other and the spacecraft, meaning measurements are not truly made in their nominal defined frames. There also may be uncertainties in the comparison with field models. Though the empirical models are parametrised with solar wind conditions during the flyby, they will not capture all transient field deviations caused by magnetospheric dynamics and processes such as field aligned currents. The modelling comparison also relies on accurate spacecraft ephemeris to reconstruct Juice's path through the Earth field. It is understood that the position uncertainty during flybys is generally of order 10 km, reducing to 10 s of metres at closest approach. A sensitivity analysis for this and future Earth gravity assists is planned for future work.

All of this can result in the J-MAG sensor frames appearing to be being misaligned with their definitions, meaning the rotation calculation here may represent more of an effective alignment correction than true rotation. By monitoring this discrepancy over subsequent Earth flybys and usage of the JACS coil, we will be able to establish whether this misalignment remains constant (and can therefore be corrected by rotation), or whether it changes and represents an uncertainty arising from other factors. In this particular flyby, it is possible that the limited operation window and steep temperature gradients are likely to have a greater effect on measurement to model deviations than misalignment.

Figure  shows the position (in GSE coordinates) of Juice relative to the Moon during the lunar phase of the LEGA. Also shown are the positions of the two spacecraft of the Acceleration, Reconnection, Turbulence and Electrodynamics of the Moon's Interaction with the Sun mission ARTEMIS;. These two probes were formerly part of the Time History of Events and Macroscale Interactions during Substorms mission THEMIS;, launched in 2007 in Earth orbit to study the dynamics of Earth's magnetosphere. In 2010, two of the five THEMIS probes were re-directed to enter a lunar orbit, where they continue to explore the magnetotail when the Moon is on the night side, and function as near-Earth solar wind monitors when the Moon is on the dayside. In Fig. , Juice's position is shown in red, with ARTEMIS P1 and P2 in blue and purple, respectively, while the Moon is shown in grey. The lines show the trajectories over the period for which J-MAG was measuring, with the final positions of the bodies in this time period shown with markers.

Juice passed close to both the ARTEMIS probes, though significantly closer to P2. At the time of lunar closest approach (21:15), Juice was approximately 6600 km (∼1RE) from ARTEMIS P2, and 16 600 km (∼2.6RE) from P1, while the two ARTEMIS probes were separated by 12 000 km (∼1.9RE). At these distances, one can expect good agreement in the large-scale structures seen by all spacecraft, and we therefore find it interesting to compare J-MAG observations with those from the fluxgate magnetometers aboard ARTEMIS . This is shown in Fig.  where the magnetic field measurements from all three spacecraft are plotted in GSE. Data for all three spacecraft are shown at 16 Hz. Here we show OBS and IBS data separately, with OBS in red and IBS in yellow. Offsets of the J-MAG sensors have been adjusted to best match ARTEMIS data for this comparison. The shaded area marks the time for which Juice was in eclipse, and a vertical line marks the point of closest approach to the Moon.

We can make several observations from this short window of data. It is immediately noticeable that both J-MAG sensors show very clean signals, with the level of noise significantly lower than in the corresponding ARTEMIS measurements. This is yet another excellent demonstration of the instrument quality and magnetic cleanliness of the spacecraft . Over the two-hour period, the field is dominated by a strong X-component, which makes sense given the position in the magnetotail. Overall, the observed fields are relatively stable, with the maximum field variation around 2 nT. This makes for an excellent opportunity to probe any differences in the four signatures.

There are some discrepancies between the two ARTEMIS probes. This is most notable in the GSE z axis, where they report a difference of up to about 1 nT in the latter portion of the flyby. While curious, this is not totally unexpected given their spatial separation. More puzzling, however, are the differences seen between the two J-MAG sensors. The second half of the flyby, after closest approach, shows generally good agreement, however, significant differences are seen in the earlier portion of the flyby. We see that OBS measures a larger field magnitude, primarily in the GSE x axis, and is at points nearly 1 nT larger than IBS and around 0.6 nT larger than the two ARTEMIS probes. In this period IBS generally sits at a lower magnitude than ARTEMIS, and shows an amount of fluctuation.

We offer some possible explanations for these discrepancies. The lunar pass occurred almost immediately after the instrument was switched on. This means that there was a steep temperature gradient at the time as the sensors and electronics began to warm up. The period of eclipse is also likely to have had an effect on the instrument temperatures, both in terms of the shadow, and the restricted operation of other spacecraft processes while running from battery power. As discussed earlier, temperature is known to affect magnetometer offsets, which may explain the differing behaviour of IBS and OBS. It appears that these irregularities mostly settled after around an hour of operation. In future we would recommend that J-MAG be switched on several hours in advance of any period considered crucial for recording science or calibration data in order to allow for temperature equilibrium to be reached.

As well as relative drifts between the two J-MAG sensors, which we may attribute to temperature fluctuations, there were also some sharp disturbances of order ∼0.5 nT seen by IBS during the eclipse period, most notably at around 21:05, and primarily in the GSE x and y axes. When the data are viewed in the J-MAG OBS frame, the disturbance is seen in the x and z axes, which are perpendicular to the magnetometer boom by definition, with no detectable signature in the y axis. It is therefore believed that a current source along the boom harness may be responsible for the signal seen by IBS, where a field generated by current aligned with the J-MAG y axis would manifest in the other two axes. To create a field perturbation of order 0.5 nT, say, 10 cm away from the source, a current of 0.25 mA would be required, which may be plausible given the boom electronics.

This may also explain why it was not seen by OBS, which is mounted 3 m further along the boom, very close to the tip. Two other instruments are placed between the two fluxgates, the radio wave instrument antennae and search coil magnetometer – both of the Radio and Plasma Wave Investigation (RPWI) suite . Therefore, as well as the cabling from the OBS and SCA sensors, IBS also has harnesses from these other instruments running past it, making it more susceptible to electrical disturbance.

We note, however, that the fact we are able to distinguish disturbances at this scale is a testament to the magnetic cleanliness of the spacecraft and the low noise level of J-MAG. While curious, the magnitude of these fluctuations is small and does not provide much concern with regard to J-MAG EMC requirements (e.g. DC offset at OBS<2nT), provided we can ensure more stable temperatures during continuous operation at Ganymede .

It is likely that a combination of factors contribute to the OBS and IBS discrepancies here, and further investigation into sensor behaviour shortly after switch-on and during temperature changes will help to confirm the source of what we see here. Despite the initial calibration, unstable temperatures, and possible electrical disturbances, we consider the overall performance of J-MAG during the lunar phase to be very encouraging. It has provided several areas of investigation and calibration activities, and helps planning of future flyby operations.

The longest portion of continuous operation during the LEGA occurred after the spacecraft exited the magnetosphere. In this section we take a closer look at its transit through the magnetosheath and entry into the solar wind. Figure a shows J-MAG measurements in GSE across an 8-h period covering the magnetospheric exit, from the outer magnetosphere on the left, through the magnetopause and magnetosheath, across the bow shock, and into the solar wind. Figure b is a zoom of the red box drawn on the upper panel, which shows the shock crossing in detail over a period of 8 min.

Before leaving the magnetosphere, J-MAG saw significant field oscillations from around 00:15 until 02:30. These oscillations take place in a background field of around 40 nT (though gradually decreasing throughout), with amplitudes of order 10 nT and periods between 1–9 min, increasing with proximity to the magnetopause. Given these scales, it seems likely that these are MHD-scale ultra low frequency (ULF) oscillations (Pc5 band), perhaps associated with field line resonances e.g., global cavity modes e.g., or Kelvin Helmholtz instabilities e.g.. The magnetopause crossing itself is poorly defined, with several possible partial crossings in quick succession (seen by field rotations, predominantly BZ, at the current sheet boundary), and a possible pile-up layer. After this, Juice entered the magnetosheath. That is, a region of shocked, slowed, compressed solar wind. The field magnitude in this period is around 18 nT, though a degree of high-frequency fluctuations and occasional field rotations are seen. This is consistent with known properties of the turbulent magnetosheath e.g..

The bow shock crossing shown in Fig. b is much more abrupt than the magnetopause crossing, and shows the expected structure of a quasi-perpendicular shock with a foot, ramp, and overshoot e.g.. The shock-normal to interplanetary magnetic field angle is around 70°, calculated using the magnetic coplanarity theorem. Upstream of the shock, there is a short foreshock with high-frequency fluctuations, perhaps kinetic whistler waves e.g., lasting approximately 2 min for Juice's passage (from 06:33:30 to 06:35:20). This corresponds to around 440 km in space. Downstream of the shock (left side of plot) we see some evidence for mirror mode waves, which often can be seen in high β plasma downstream of perpendicular shocks and exhibit proportionally large fluctuations in |B|, though further investigation (including plasma moments) would be needed to confirm this.

While there is clearly a wealth of physical interest in this short interval, we do not explore it further in this paper, but note once again that this is an excellent demonstration of J-MAG's ability to observe such phenomena, which is very promising for future endeavours at Jupiter and Ganymede, where we expect to see comparable field magnitudes to some of the outer magnetospheric regions explored here.

Figure  shows J-MAG measurements (red) at a 1 min cadence over several days, after Juice exited the magnetosphere and was travelling duskward through the solar wind. Overlaid are corresponding measurements from the OMNI database (grey), sampled at L1 and propagated to Earth's bow shock, also at 1 min cadence. Note that these are the same data shown in the latter half of Fig. a.

No timing adjustments have been made to either dataset to align the observations, though the OMNI data are inherently time-shifted and evolved from their L1 observation point to represent the conditions at Earth's bow shock nose. As discussed earlier (alongside Fig. ), the bow shock X-position varied throughout this time interval according to the solar wind dynamic pressure. This variation (of order ∼3RE) is less than the movement of Juice over the same period, whose GSE X-position varied from 5RE to 14RE (from 9.5RE behind bow shock to 1RE ahead). At the average solar wind speed in this interval, this corresponds to a maximum difference in propagation time of approximately 3 min. We therefore conclude it is reasonable to expect solar wind evolution to be minimal between the bow shock and Juice X-positions. We note, however, that the propagation of Wind data to the bow shock covers around 200RE (∼60 min).

On large scales, the two datasets agree remarkably well, both in terms of absolute magnitudes and in vector fluctuations. This is another excellent validation of J-MAG's performance. We do see some notable differences on smaller scales, however, particularly towards the second half of the dataset, where there are several instances of rotational discontinuities seen by Juice but not captured by OMNI, and we notice that the alignment of observed structures also worsens. While the X-alignment of Juice and the bow shock generally improves throughout the interval (and so should improve temporal alignment), the displacement in YGSE increases significantly as Juice moves duskward of the Sun-Earth line (see Fig. ). In addition, the Wind spacecraft, from which the OMNI data originate, was actually located dawnward of the Sun-Earth line. This means that the lateral displacement from Wind to Juice extended from 130RE to 250RE over this period. It is well-known that spatial structures in the solar wind can vary considerably on this scale e.g., and so is unsurprising that the finer scale field structure differs more towards the end of the interval.

Despite these differences, there is a generally very good correspondence between the two datasets. This also demonstrates the potential utility of Juice (and other planetary missions) during cruise phase as solar wind monitors for the purposes of space weather forecasting, if the science payload is active and data can be transmitted to ground with low latency with a minimal (and fully automated) calibration applied. Though this calibration level would not be suitable for science studies, these low latency data may be useful for identifying and parametrising large events and geoeffective conditions, such as strongly negative BZ. This has been successfully demonstrated with Solar Orbiter, where low-latency magnetometer data have been used for real-time predictions of CME arrival times and geoffectiveness when sunward of L1 see, e.g., giving greater advance warning than current operational systems. We would argue that planetary missions should be used to the same end when in cruise phase, though this would require the science payload to be operational more often and for more extended periods than is typical at present.

Even if near-real time downlink is not possible, continuous operation during cruise would aid the reconstruction of solar wind transients and post-event analysis of space weather. The often unusual orbits of these missions during cruise can allow for sampling of regions of the heliosphere for which we do not have regular coverage. For example, observations at 1 AU but significantly away from the Sun-Earth line may provide useful training data for the development of the upcoming Vigil mission to the L5 Lagrange point e.g.. This extended coverage may also improve understanding of potential space weather hazards at other planets and spacecraft and, as is becoming increasingly relevant, conditions conducive for human spaceflight.