Auroral emissions from Uranus and Neptune

Uranus and Neptune possess highly tilted/offset magnetic fields whose interaction with the solar wind shapes unique twin asymmetric, highly dynamical, magnetospheres. These radiate complex auroral emissions, both reminiscent of those observed at the other planets and unique to the ice giants, which have been detected at radio and ultraviolet (UV) wavelengths to date. Our current knowledge of these radiations, which probe fundamental planetary properties (magnetic field, rotation period, magnetospheric processes, etc.), still mostly relies on Voyager 2 radio, UV and in situ measurements, when the spacecraft flew by each planet in the 1980s. These pioneering observations were, however, limited in time and sampled specific solar wind/magnetosphere configurations, which significantly vary at various timescales down to a fraction of a planetary rotation. Since then, despite repeated Earth-based observations at similar and other wavelengths, only the Uranian UV aurorae have been re-observed at scarce occasions by the Hubble Space Telescope. These observations revealed auroral features radically different from those seen by Voyager 2, diagnosing yet another solar wind/magnetosphere configuration. Perspectives for the in-depth study of the Uranian and Neptunian auroral processes, with implications for exoplanets, include follow-up remote Earth-based observations and future orbital exploration of one or both ice giant planetary systems. This article is part of a discussion meeting issue ‘Future exploration of ice giant systems’.

LL, 0000-0002-8428-1369 Uranus and Neptune possess highly tilted/offset magnetic fields whose interaction with the solar wind shapes unique twin asymmetric, highly dynamical, magnetospheres. These radiate complex auroral emissions, both reminiscent of those observed at the other planets and unique to the ice giants, which have been detected at radio and ultraviolet (UV) wavelengths to date. Our current knowledge of these radiations, which probe fundamental planetary properties (magnetic field, rotation period, magnetospheric processes, etc.), still mostly relies on Voyager 2 radio, UV and in situ measurements, when the spacecraft flew by each planet in the 1980s. These pioneering observations were, however, limited in time and sampled specific solar wind/magnetosphere configurations, which significantly vary at various timescales down to a fraction of a planetary rotation. Since then, despite repeated Earth-based observations at similar and other wavelengths, only the Uranian UV aurorae have been re-observed at scarce occasions by the Hubble Space Telescope. These observations revealed auroral features radically different from those seen by Voyager 2, diagnosing yet another solar wind/magnetosphere configuration. Perspectives for the in-depth study of the Uranian and Neptunian auroral processes, with implications for exoplanets, include follow-up remote Earth-based observations and future orbital exploration of one or both ice giant planetary systems.
This article is part of a discussion meeting issue 'Future exploration of ice giant systems'.  Figure 1. Auroral radio spectrum of solar system magnetized planets derived from Voyager/PRA low frequency observations [8], updated for Jupiter from Cassini/RPWS observations [9]. The vertical dashed line plots the terrestrial ionospheric cutoff, below which the atmosphere is opaque. The kilometric radiations produced by the Earth, Saturn, Uranus and Neptune display very similar spectra, although UKR/NKR are typically weaker than TKR/SKR by an order of magnitude. Only Jupiter can be observed from the ground. (Online version in colour.) Radiation (TKR), the Jovian broadband kilometric, hectometric and decametric radiations (bKOM, HOM and DAM) and Saturn's kilometric radiation (SKR). They are all non-thermal, strongly beamed and highly circularly (and sometimes elliptically) polarized emissions produced above the magnetic poles along high latitude magnetic field lines associated with the atmospheric visible or UV aurorae. They are mainly produced on the right-handed extraordinary (R-X) and marginally on the left-handed ordinary (L-O) free-space modes at frequencies near the local electron cyclotron frequency f ce , itself proportional to the local magnetic field strength [8]. These auroral planetary radio emissions share a common generation mechanism known as the Cyclotron Master Instability (CMI), first identified at Earth [10] and recently confirmed in situ at Saturn and Jupiter [11][12][13], which amplifies radio waves from non-Maxwellian mildly relativistic electrons of a few keV in magnetized and sufficiently tenuous regions, namely when the electron plasma frequency f pe is significantly lower than f ce . When flying by Uranus and Neptune, Voyager 2/PRA revealed two novel, strikingly similar, radio emitters promptly referred to as 'radio twins' [14,15]. The diversity of ice giant radio emissions is reviewed at length in [16][17][18]. Hereafter, we focus on their prominent auroral component, observed within 15-900 kHz and 10-1300 kHz (figure 2) and thus termed Uranian and Neptunian Kilometric Radiations (UKR and NKR), respectively. With large scale properties reminiscent of those of the terrestrial, jovian and kronian cases, they were rapidly attributed to the CMI [8]. Their remote observation therefore provided the first evidence of a magnetic field at both planets.

(b) A rich diversity of components
The UKR and NKR average spectra displayed in figure 1 are strikingly similar, as expected from comparable magnetic environments at both planets. They also compare to the TKR and SKR spectra but with flux densities weaker by an order of magnitude or more. This explains why SKR was detected 10 months before the encounter with Saturn while UKR and NKR were only observed from about 5 and 8 days, respectively, before closest approach (CA) up to 1 month after. In the specific case of Uranus, figure 2a additionally shows that UKR was much brighter after the  CA, as the result of prominent nightside radiosources around the strong-field southern magnetic pole, beamed away from the solar direction. Two reasons have been proposed to account for this property. The large southward magnetic field offset implies a larger spatial region above the southern magnetic pole where f pe /f ce 1 and thus a larger CMI-driven wave growth region [17]. Also, a much high power radiated from the nightside pole evokes the terrestrial case and suggests that the solar wind may be the ultimate driver of Uranian auroral processes [16].
Contrary to the Earth or Saturn, which are relatively simple radiosources, Uranus and Neptune host a complex zoo of radio emissions, likely related to the tilt of their magnetic dipole, with a total of nine different components listed in table 1 and labelled in figure 2. These components, both broad-and narrow-banded, were divided in two categories based on their duration: (i) bursty and (ii) smooth (or continuous) emissions. The bursty adjective refers to intermittent emissions reminiscent of those usually observed at the Earth, Jupiter and Saturn, typically lasting up to a few minutes and associated with impulsive acceleration processes. At Neptune, however, the main bursty component was identified as the most intense, impulsive (approx. 30 ms) and narrowbanded (1-14 kHz) emission ever observed at a radio planet [18]. Conversely, smooth components refer to emissions radiating constant flux for hours or more and have only been observed at Uranus and Neptune to date. The broadband smooth components dominate the time-frequency spectrogram of both planets. Their apparent modulation at the planetary period evident in figure 2, with characteristic features (such as the high-frequency dropout at the center of each UKR b-smooth episode), results from beaming effects which depend on the viewing geometry. Few scenarios have been proposed to account for the free energy source of the (prominent) smooth emissions, which remains unknown.
There have been considerable efforts to accurately locate the source of each radio component and to identify the underlying active magnetospheric regions, as exhaustively reviewed in [16][17][18]. A key parameter in such studies is the assumption or determination of the beaming angle associated with the CMI-driven wave amplification, generally found to lie at large angles from the local magnetic field. Investigation of the beaming of bursty emissions at both planets royalsocietypublishing.org/journal/rsta Phil. Trans. R. Soc. A 378:   revealed the existence of Earth-like auroral cavities at their source with f pe /f ce as low as 0.01 [19]. Figure 3 shows the magnetic footprint of most of the UKR and NKR components listed in table 1 on top of cylindrical maps and magnetic iso-contours. At Uranus (figure 3a), the b-bursty and smooth prominent components are colocated near the nightside southern magnetic pole. The narrowband bursty and the b-smooth dayside emission are instead produced near the dayside northern magnetic pole. The former was thought to be associated with the northern cusp, while the latter is magnetically connected with the b-smooth emission in the nightside hemisphere. All these components lie along field lines approximately mapping to the brightest regions of patchy UV aurorae (white regions). At Neptune (figure 3b), sources of the main smooth emission were located above both magnetic poles, while the bursty source lies near the southern magnetic pole, where the magnetic field strength is high enough to support CMI-driven emission up to 1. 3 MHz. An unexpected result came with the localisation of the n-smooth UKR and high-frequency smooth NKR sources near the magnetic equator instead, the former being associated with the ring. The generation of auroral radio waves near the equator was tentatively related to low plasma densities at Uranus [20] and to a close-in magnetic anomaly at Neptune [21]. Overall, the conditions in which the CMI develops in, and especially outside, these auroral regions are poorly understood. While knowing their hemisphere of origin, measuring the polarization of the observed radio waves was used to identify their (magneto-ionic) mode of propagation. All the components but one listed in table 1 correspond to R-X mode. Only the so-called dayside component corresponds to L-O mode, which is weaker than the prominent nightside R-X mode emission by approximately 15 dB [22]. While bright R-X and faint L-O mode emissions are regularly observed from the same source regions for TKR at Earth or SKR at Saturn, only Uranus displays L-O mode emission with no simultaneous R-X counterpart. This peculiarity was attributed to the high plasma density prevailing on the Uranian dayside, making the ratio f pe /f ce closer to unity and in turn corresponding to an unusual CMI regime which favours L-O mode with respect to R-X mode emission. Elliptical polarization was also routinely observed despite the limited PRA polarization capabilities [23]. Exhaustive full polarization measurements are needed to understand the CMI framework at the source and probe the plasma environment encountered along the raypath.

(c) Magnetospheric dynamics
The quasi-continuous remote observations of Voyager/PRA for tens of days revealed fundamental aspects of the ice giant magnetospheric dynamics.

(i) Rotation period
The strong rotational modulation of UKR/NKR was first of all used to determine the intrinsic rotation period of both planets with a fair accuracy. At Uranus, the value of 17.24 ± 0.01 h (or 17 h 14 m 24 ± 36 s) was obtained from the spectral analysis of 30 days of b-smooth postencounter observations (least-squares fitting of its stable repeatable dropouts) weighted by an independent spectral analysis of 6 h of in situ magnetic measurements acquired around the CA [24]. At Neptune, the value of 16.108 ± 0.006 h (or 16 h 06 m 30 ± 24 s) was achieved by combining the values obtained from independent spectral analyses of two NKR components [25]. The study of the main smooth component, which displays highly repeatable polarization reversals with a source fixed in sub-spacecraft longitude, used 15 days of observations across the CA. The analysis of NKR bursty episodes used a longer interval of 60 days, while testing both possible scenarios of a source fixed either in longitude (searchlight) or in local time (flickering). To date, the above values remain the best available estimates of the ice giant internal sidereal rotation rates.
(ii) Solar wind and possible satellite control Both magnetospheres are controlled by the solar wind to some extent. On the one hand, the occurrence/intensity of some radio components are linked to interplanetary conditions. At Uranus, the long-term activity of n-bursts produced within/near the cusp correlates with the solar wind density [26]. At Neptune, the flux density of NKR bursts was instead found to best correlate with the merging electric field [27]. On the other hand, the total power radiated by UKR and NKR, reaching approximately 3 × 10 7 W and 10 7 W, confirmed the relationship previously suspected between the radio power radiated by a planet and the incident magnetic power convected by the solar wind on the magnetospheric obstacle [8]. This 'radio magnetic' Bode's Law supports the solar wind as the primary driver of auroral processes and accounts for a lower energy input than at Earth or Saturn, whose magnetospheres are strongly controlled by the solar wind [8, and references therein]. This limited evidence of a weak solar wind control needs to be confirmed and quantified, though.
In terms of internal drivers, it has been argued that an Alfvenic interaction between the moon Ariel and the Uranian magnetic field powered a single, pulsing, UKR high-frequency arc [28] (labelled in grey in figure 2). No evidence of satellite-driven auroral radio emissions has been reported yet at Neptune. Although Triton has a suitable atmosphere, it is orbiting too far from the planet to sustain a powerful current system capable of producing radio arcs, such as in the Jupiter-Io case (see, for instance, the comparative study of [29]).

(d) Implications for exoplanets
Uranus and Neptune proved to be auroral radio emitters analogous in many ways to the Earth, Jupiter and Saturn, supporting a 'universal' emission mechanism, together with unique properties attributed to their large magnetic tilt. Their detailed characterization has important implications for understanding which energy sources sustain the CMI, and how it operates in asymmetric magnetospheres devoid of plasma, in the solar system and beyond. Oblique magnetic/radio rotators are common in astrophysics with the most famous examples being pulsars. The generalized 'radio magnetic' Bode's confirmed by the radio power radiated by Uranus and Neptune has been widely used to predict the expected radio flux from exoplanets, depending on the nature of their interaction with the host star [30, and references therein]. Intensive observations of exoplanetary systems have, in turn, been carried out in the past decades and currently form a major objective of large ground-based radiotelescopes [31]. While still waiting for a confirmed detection, a few tens of stellar objects have already been detected at radio wavelengths, with bright, highly circularly polarized, bursts assigned to CMI [32]. The coolest of these (so-called ultracool dwarves, with spectral type later than M7) display bursty and quiescent emissions. Their strong 'planet-like' magnetism makes them a potential missing-link between stars and planets. Beyond the total radiated radio power, several other key parameters such as the source locus and beaming are relevant to build updated figures of merit to identify the most relevant exoplanets to track at radio wavelengths. Overall, Uranus and Neptune radio emitters thus hold a very special place in the solar system reference framework.

Ultraviolet aurorae
The extreme-to far-UV (EUV to FUV) spectra of Uranus and Neptune have been sampled by UV spectro-imagers from space telescopes in orbit around Earth such as the International Ultraviolet Explorer (IUE) in the 1980s and HST since the late 1990s as well as from the UltraViolet Spectrometer (UVS) onboard Voyager 2. A variety of emissions have been detected from their atmospheres [33,34] (figure 3 of the latter article compares the dayside UV spectra of both planets, as measured by UVS) and reviewed at length in [35,36]. Here, we focus on the prominent auroral emissions resulting from particle impact onto neutral hydrogenic species in the upper atmosphere, including the H-Lyα line at 121.6 nm and the H 2 bands ranging from 80 to 160 nm. Unlike in the radio domain, the UV aurorae of Uranus and Neptune significantly differ from each other and will be dealt with separately.  auroral contribution, better seen around the nightside southern pole. The bright dayside H-Lyα sporadic signal observed by IUE was thus a posteriori attributed to variable dayglow [35]. The spatial distribution of H-Lyα additionally revealed an extended non-thermal H corona, possibly powered by dayglow and/or dayside aurora [40, and references therein].

(a) Uranus
The H 2 -band aurorae were analysed from a few tens of hours of observations of both the sunlit and darkside hemispheres, best observed by UVS between 95 and 111 nm [41]. Their spatial distribution is displayed in figure 4. Overall, the aurorae did not form complete ovals and rather appeared as patchy emissions clustered near the magnetic poles, at the footprint of tailwarddirected flux tubes, within 135-225 • ULS longitude and 60-65 • ULS latitude, mapping to 5-10 R U in the magnetosphere. The aurorae were more spatially extended and two to four times more intense around the weak-field northern magnetic pole than around the strong-field southern magnetic pole (as opposed to the UKR components), and radiated a total power of 3-7 GW at H-Lyα and shorter wavelengths. Assuming a 10% efficiency, this required 10 times more precipitated power. Averaged over the disc, this power corresponds to 0.008 erg cm −2 s −1 precipitation fluxes, 10 times less than required to account for the observed temperature of 850 K [43].
The spatial distribution of the UV aurorae directly probes the magnetic field topology. Assuming a spatial magnetic conjugacy of southern and northern aurora, [42] built up a fifthorder magnetic field model termed AH 5 , updating the Q 3 model derived from magnetic data [44] with larger high-order multipole components.
The limited spatial extension of the aurorae, mapping to small distances in the magnetotail, brings a strong constraint on the underlying driving mechanism [35]. It suggests a prominent role for sunward plasma convection in the inner magnetosphere, as opposed to an elongated terrestrial-type aurora at the boundary between open and closed magnetic field lines. The UV southern and northern aurorae were additionally found to be co-located with the footprint of UKR (southern) b-smooth/b-bursty and (northern) dayside smooth components, respectively (figure 3), indicating a common electron population at the origin of both emission types. The UV aurorae also coincided with the maximal intensities of whistler-mode emissions [45] and 22-35 keV electron fluxes [46] sampled in situ by Voyager 2. Overall, this suggests that the Uranian aurorae are primarily excited by the precipitation of magnetotail-side energetic electrons, plausibly originating from the plasma sheet and scattered into their phase space loss cone by whistler waves, with a possible contribution from field-aligned potentials. While the limited UVS dataset prevented from the tracking of time-variable emissions, the dual nature of smooth and bursty auroral radio emissions implies multiple drivers, either steady-state or impulsive. In the latter case, it is worth noting that Earth-like quasi-periodic substorms have been separately   In addition, when observed several weeks apart, the auroral spots observed in 2011 and in 2014 were strikingly spatially coincident, revealing a single active magnetospheric region. The detected emissions were variable over a few minutes, with roughly periodic pulses every 2.5 min in 2014. The observed spots radiated a few GW of unabsorbed H2 emission over 70-180 nm, also rising with time from approximately 2 GW in 2011 to approximately 9 GW in 2014 [51]. These numbers are comparable to the Voyager/UVS auroral radiated powers given above (which corresponded to emission at H-Lyα and shorter wavelengths) and, likewise assuming a 10% efficiency, imply an input power 10 times larger. A separate analysis of STIS long-slit spectra acquired in parallel to images in 2011 and 2012 quantified the contribution of airglow and auroral emissions in the H 2 bands, with precipitating fluxes estimated from 0.01 up to 0.05 erg cm −2 s −1 , respectively [53].
Overall, these near-equinoctial aurorae strongly differ from those observed at solstice and from the usual terrestrial or kronian large-scale auroral storms driven by interplanetary shocks. They were instead interpreted as cusp aurorae driven by dayside magnetic reconnection between planetary and interplanetary magnetic fields, likely triggered by sudden magnetospheric compressions [51], a phenomenon regularly observed at Earth. At equinox, the formation of a stable magnetotail is inhibited by the fast rotation and the plasma sheet is unlikely to survive more than half a rotation period [54], so that it cannot sustain significant magnetic reconnection in the tail driving powerful auroral precipitations [55]. Additionally, magnetic reconnection between planetary and interplanetary magnetic fields is expected to occur periodically with highly dynamical reconnection sites [56,57].

(iii) Intermediate configuration between past equinox and solstice
The reanalysis of past HST data revealed one additional auroral detection in a single STIS image of 1998 (figure 5a), sampling yet another solar wind/magnetosphere configuration, intermediate between northern solstice and equinox [50]. The emissions appeared as partial rings in both hemispheres, brighter around the weak-field northern pole, and recorded under quiet solar wind conditions. Such features strikingly resemble the auroral ovals of the Earth, Jupiter or Saturn and differ from the spots observed at northern summer solstice and past equinox. They likely require another acceleration mechanism, yet unknown, capable of sustaining longitudinally extended auroral precipitation.
When Uranus reached the reverse configuration between equinox and southern summer solstice, in late 2017, HST tracked the planetary auroral response to unusually powerful solar wind disturbances, sampling the peak of three interplanetary shocks with four separate observations. For the first time, auroral emissions were seen in all the obtained images and, whenever visible, in both hemispheres simultaneously (figure 5g-j). Interestingly, these signatures display spot-like features similar to those of 2011-2014, although brighter, together with faint elongated arcs comparable to those of 1998, suggesting that different auroral processes may actually co-exist under this intermediate configuration. The detailed analysis of these images, and of the new constraints that they provide onto the rotation rate, is in progress.

(b) Neptune
Voyager 2/UVS detected two kinds of faint H 2 band emissions from the darkside hemisphere of Neptune, as displayed in figure 6. UVS scanned only a portion of the disc (figure 6a) with successive observations lasting for approximately 20 h. A weak emission clustered around 60 • longitude and constant at all latitudes was attributed to airglow. It was proposed to be powered by photoelectrons from the dayside, magnetically conjugate, hemisphere during poleon configurations occurring once a rotation. A brighter signal, peaking at approximately 240 • longitude around the southern magnetic pole, was tentatively identified as a faint southern  aurora. This emission, radiating power as low as 5 × 10 7 W (roughly 100 times less than Uranus), was thought to be magnetically conjugate to a northern more intense aurora which remained outside the UVS field-of-view and thus escaped detection [34,58]. The authors argued that auroral emissions could be powered by intense enough field-aligned current systems driven by the Neptune-Triton electrodynamic interaction (such as in the Io-Jupiter case) along isolated 'plasma arcs' located at the intersection between the centrifugal equator and Triton's orbit (figure 6d). Although the Neptunian southern aurora was sampled only once by UVS, on the nightside, during specific solar wind conditions and in a particular planetary configuration, no new UV observation of Neptune has been obtained in the following 28 years. This is not unsurprising considering the low intensity auroral signal and large distance from Earth. However, late in 2017, powerful interplanetary shocks propagated in the outer heliosphere and successively reached Uranus and Neptune. Each planet was imaged through four HST/STIS orbits. The Uranian observations sampled three magnetospheric compressions and revealed powerful aurora around both magnetic poles, as already discussed above. The Neptunian observations sampled a single magnetospheric compression, as displayed in figure 7. While providing the first ever instantaneous FUV images of the planet, the observations did not reveal evidence of localized

Search at other wavelengths
Voyager 2 did not include any NIR instrument. Long-term NIR observations of Uranus and Neptune thus started with ground-based facilities in the 1990s. The NIR range encompasses thermalized ro-vibrational lines of the ionospheric H + 3 ion and quadrupole lines of H 2 in the 2-5 µm range. While both emissions have been early discovered at Uranus [59,60], such as for Jupiter and Saturn, no detection has yet been achieved at Neptune yet [61, and references therein]. An exhaustive review on ice giant ionospheres and upper thermospheres is addressed in a companion paper [6], and we hereafter restrict ourselves to a brief discussion of H + 3 emissions and their suspected association with auroral processes at Uranus.
Uranus displays H + 3 emission from the whole disc, mainly attributed to EUV solar-driven ionization of the upper atmosphere and witnessing a hot thermosphere. Long-term spectroscopic observations revealed that the H + 3 total emission, column density and temperature display a long-term variability, together with a few significant excursions on a shorter time-scale [62, and references therein]. The latter have been tentatively attributed to transient auroral contributions. The long-term decrease of the thermospheric temperature (cooling) has been originally suspected to be associated with the solar cycle or Uranus seasons, until the most recent observations of [62] confirmed a monotonically long-term cooling (from approx. 700 K in 1992 to approx. 500 K in 2018) inconsistent with a cyclical seasonal behaviour. The authors instead proposed that field-aligned currents (driving Joule heating) gradually decreasing in strength with the transition of Uranus from northern summer to southern summer solstice may be responsible for this variation. In the light of the previous section, it however appears that (i) the UV auroral power (or precipitated input) was similar at solstice (prominent emission on the nightside) and near equinox (dayside aurorae) and (ii) the past-equinox UV aurorae were brighter and more frequently observed from 2011 to 2017. These results thus question any direct association between the observed auroral precipitations and the H + 3 long-term cooling. A few H + 3 resolved maps of Uranus have provided additional insights to this intriguing question. In 1993, several IRTF instantaneous images taken with the ProtoCAM instrument interestingly revealed a bright area, apparently co-rotating across the disc [63]. The authors could not confirm their auroral nature but set up an upper limit of 20% for any auroral contribution to the total H + 3 signal (a few 10 −26 W m −2 once disc-averaged). Late in 2017, coordinated multispectral observations of Uranus (involving the HST, VLT, Gemini and Chandra observatories) sampled a powerful solar wind shock interaction, providing the opportunity to directly test any direct relationship between UV, NIR and X-rays emissions [52]. The six NIR images were spread across a 9 day long interval, sampling the full interplanetary shock. They revealed disc-wide H + 3 emission (most clearly seen with Gemini/NIRI, and brighter around the southern polar cap) but no auroral features consistent with the bright southern UV aurora successfully tracked with a single HST image. As mentioned previously (and shown in figure 5), three additional HST images, sampling two earlier interplanetary shocks, also revealed aurorae around both magnetic poles. Although HST/Gemini data were not strictly simultaneous, this campaign shows that while auroral emissions are routinely detected in the FUV, they have no detectable H + 3 counterpart, suggesting that auroral precipitations do not trigger bright enough H + 3 emission. As a result, the origin of H + 3 long-term variability and the lack of any detectable H + 3 emission of auroral nature (unlike the jovian and kronian cases) remains not understood.
Uranus was also imaged by the Chandra X-ray observatory with the High Resolution Camera (HRC) on two occasions in 2017 (green vertical lines in figure 8), with approximately 25 ks exposures, to search for any auroral signature reminiscent of those known at the Earth or Jupiter, driven by charge exchange processes or electron bremstrahlung [64]. Despite the second observation revealing a weak excess of photons from the Uranian disc (1.6 standard deviations above the background level), both images were essentially dominated by photon noise (a similar and Chandra (X-rays) observations. One HST/STIS image (43-min exposure, broadband filter including H-Lyα) taken near the peak of dynamic pressure revealed a powerful southern aurora fitted by a model oval (pair of solid red lines). Six NIR images obtained with VLT/NACO (two exposures of 56 and 72 min, using adaptative optics and the L' broadband filter) and Gemini/NIRI (four exposures ranging from 33 to 74 min, using both the L' broadband and the Brα narrowband filter) covered a 9-days long interval sampling the full interplanetary shock. The observed H + 3 emission (most clearly seen in the Gemini/NIRI Brα images) extends all over the disc, brightening around the southern polar cap, but does not display any localized emission associated with either the northern blue or the southern red model auroral oval. Adapted from [52]. (Online version in colour.) negative result was previously obtained at Saturn using the same approach [65]). The authors also retrospectively analysed the only other Chandra image of Uranus obtained to date, an approximately 30 ks exposure acquired in 2002 with the Advanced CCD Imaging Spectrograph (ACIS). This observation revealed a convincing excess of photons from the planet within the 0.5-1.1 keV range (reaching 10.2 standard deviations above the background level). While this level of emission is consistent with fluorescent scattering of solar photons by the disc, it does not discard transient auroral signatures and shows that Uranus X-rays can be tracked with sufficiently sensitive telescopes. No X-ray emission has been detected from Neptune.

Conclusion and perspectives
Uranus and Neptune produce weak but complex, and often unique, auroral emissions, only detected at radio and UV wavelengths to date. Along the numerous open questions detailed above, their understanding remains at its infancy. Re-observing such radiations remotely, on a long-term and regular basis, is therefore essential to sample their morphology, energetics and dynamics, and unravel how energy is transferred and dissipated in distant asymmetrical magnetospheres, with implications for the search of exoplanets.
UV and NIR Earth-based facilities proved to be capable of detecting emissions from Uranus, and are worthy to pursue with current (HST, NIR ground-based telescopes) and future (JWST, LUVOIR) observatories, while X-ray observations (with Chandra/XMM-Newton and soon Athena) remain exploratory so far. The early IMP-6 radio observations also reminds us that ice giant radio emissions are in principle accessible from near-Earth space-based sensitive radio instruments, ideally located on the far side of the moon to be shielded from solar/terrestrial radio emissions and/or anthropogenic radio frequency interferences [66].
Nonetheless, understanding ice giant auroral processes unavoidably requires in situ observations from dedicated orbital exploration probes. Modern radio and UV instruments observing the aurorae from a large variety of positions are primarily needed to update the rotation period, track the solar wind magnetosphere interaction and investigate possible electrodynamic coupling with moons. The highly tilted magnetic dipole will also facilitate easy sampling of polar regions with equatorial orbiters, especially in the Uranian case, and repeated crossings of auroral radiosources and acceleration regions such as Cassini and Juno at Saturn and Jupiter. In situ radio, magnetic and plasma measurements from these key regions are crucial to assess the microphysics of auroral processes, and establish their specific and/or universal character.
Data accessibility. The Voyager/PRA low band data plotted in figure 2 were retrieved from the MASER repository