Investigating Mercury’s Environment with the Two-Spacecraft BepiColombo Mission

Abstract

The ESA-JAXA BepiColombo mission will provide simultaneous measurements from two spacecraft, offering an unprecedented opportunity to investigate magnetospheric and exospheric dynamics at Mercury as well as their interactions with the solar wind, radiation, and interplanetary dust. Many scientific instruments onboard the two spacecraft will be completely, or partially devoted to study the near-space environment of Mercury as well as the complex processes that govern it. Many issues remain unsolved even after the MESSENGER mission that ended in 2015. The specific orbits of the two spacecraft, MPO and Mio, and the comprehensive scientific payload allow a wider range of scientific questions to be addressed than those that could be achieved by the individual instruments acting alone, or by previous missions. These joint observations are of key importance because many phenomena in Mercury’s environment are highly temporally and spatially variable. Examples of possible coordinated observations are described in this article, analysing the required geometrical conditions, pointing, resolutions and operation timing of different BepiColombo instruments sensors.

1 Introduction

Mercury’s environment is a complex system where the magnetosphere and exosphere are inherently coupled, and interact with the interplanetary medium and the surface (e.g.: Milillo et al. 2005; Killen et al. 2007). The planet’s close proximity to the Sun creates particularly strong external forcing conditions, involving extreme solar wind conditions and intense solar energetic particles and X-ray fluxes. Mercury possesses a weak, intrinsic, global magnetic field that supports a small magnetosphere which is populated by charged particles originating from the solar wind, from the planet’s exosphere and from the surface (a comparison of Mercury’s characteristics with Earth’s is summarised in Table 1). On the other hand, the exosphere is continually refilled and eroded through a variety of chemical and physical processes acting both on the surface and in the planetary environment and which are driven by external conditions, like the Sun’s irradiance and particles, and micrometeoroid precipitation toward the surface. These external conditions show a high variability along the eccentric orbit of Mercury (0.31–0.46 AU), so that, even though Mercury lacks seasons linked to its rotational axis inclination, it is generally assumed that the aphelion part of the orbit (True Anomaly Angle—TAA between 135° and 225°) is Winter, the perihelion (TAA between -45° and +45°) is Summer, while at TAA between 45° and 135° is Autumn and at TAA between 225° and 315° is Spring.

Table 1 Mercury’s and Earth’s parameters

The first direct encounters with Mercury’s environment comprised three flybys by the Mariner 10 spacecraft spanning 1974-75, allowing detection of Mercury’s magnetic field, and observations of its exosphere and surface (Russell et al. 1988). In 2011, NASA’s MESSENGER (Mercury Surface, Space Environment, Geochemistry and Ranging) mission (Solomon and Anderson 2018) was placed into a highly elliptical polar orbit around Mercury, carrying a suite of instruments designed to explore the fundamental characteristics of the planetary surface and environment. The mission concluded in 2015 with a low altitude campaign and finally impacted the planet.

Thanks to MESSENGER observations, we know that the Hermean magnetosphere is highly dynamic, with total reconfiguration taking place within a few minutes (Imber and Slavin 2017). The coupling of the interplanetary magnetic field and the solar wind with the planetary magnetosphere is much stronger than previously believed, owing to the almost-continuous dayside magnetic reconnection (e.g. Slavin et al. 2012), as proved by the frequent observations of flux transfer events (FTE) (Imber et al. 2014; Leyser et al. 2017) (Fig. 1 upper panels). Slavin et al. (2014, 2019a) examined several MESSENGER passes during which extreme solar wind conditions both compressed the dayside magnetosphere due to high dynamic pressure, and eroded it due to extreme reconnection. The solar wind-planet interaction is further complicated by the existence of Mercury’s large metallic core, within which induction currents are driven during these extreme events, acting in opposition to this compression/erosion (Jia et al. 2015, 2019; Dong et al. 2015). Only simultaneous observations of the external conditions of solar wind, plasma precipitation, micrometeoroid, and exosphere distributions will allow a full understanding of the Na exospheric behaviour and help to solve the mystery of the highly volatile component in this close-to-star planet. Moreover, there is an indication of a strong correlation between the distribution of energetic refractory elements in the exosphere and the crossing of micrometeoroid streams (Killen and Hahn 2015), but MESSENGER had no dust monitor on board that might have been able to confirm this. We know from ground-based and from MESSENGER observations that the composition of the constituent particles in the Mercury’s environment includes, besides H and He, Na and Na⁺, K, Mg, Ca and Ca⁺, Mn, Fe and Al (Broadfoot et al. 1974; Potter and Morgan 1985, 1986; Bida et al. 2000; McClintock et al. 2008; Bida and Killen 2017), while unexpectedly, no signature of oxygen atoms was detected by the Mercury Atmospheric and Surface Composition Spectrometer (MASCS; McClintock and Lankton 2007). However, the mass resolution of the Fast Imaging Plasma Spectrometer (FIPS; Raines et al. 2011) was too low to discriminate between individual ion species, while atom groups including oxygen could not be detected by MASCS. Mass spectrometers with high mass resolution would allow the detection and characterisation of the majority of the constituent of the exosphere and planetary ions including molecules and atom groups that would provide an important information for describing the surface release processes, for explaining the fate of oxygen, and ultimately for tracing the planet’s evolutionary history.

In summary, in the 1980s, the analysis of the data of the three Mariner-10 fly-bys revealed unexpected features of Mercury’s environment including the intrinsic magnetic field and the presence of high-energy electron burst events. With the new millennium, the MESSENGER mission, thanks also to the exosphere ground-based observations, have greatly improved our knowledge of the complex Hermean environment. However, both missions left the planet with new intriguing questions. In Sect. 2, we summarise the main findings about the Hermean environment and the still unsolved points.

The ESA-JAXA BepiColombo mission is expected to provide a tremendous improvement in the knowledge of the functioning of Mercury’s environment, and solve the numerous questions that are still open after previous space missions together with ground-based observations. In fact, this technologically advanced and optimally designed mission exhibits all the capabilities to accomplish the requirements mentioned above. It allows simultaneous two-point measurements thanks to the two spacecraft, MPO and Mio, with optimal orbits for exploring external, close-to-planet, far-tail and flank conditions (Milillo et al. 2010). BepiColombo, unlike MESSENGER, has a dedicated full plasma instruments package in the Mio spacectraft that offers a unique opportunity to study in details the magnetosphere of Mercury. Among the instruments devoted to the study of the Mercury’s environment, BepiColombo includes sensors and experiments that never have been operated at the innermost planet, like the plasma wave experiment Mio/PWI, the dust monitor Mio/MDM, the neutral mass spectrometer MPO/SERENA-STROFIO and Na imager Mio/MSASI, two Energetic Neutral Atom (ENA) imagers MPO/SERENA-ELENA and Mio/MPPE-ENA. Furthermore, many of the BepiColombo instrument types previously flown on Mariner-10 and MESSENGER (i.e.: magnetometers MPO/MAG and Mio/MGF, charged particle detectors Mio/MPPE and MPO/SERENA, UV spectrometer MPO/PHEBUS, X-rays spectrometers MPO/MIXS and SIXS) have significantly improved performances in spatial coverage, energy, mass and time resolutions. In other words, BepiColombo will offer an unprecedented opportunity to investigate magnetospheric and exospheric dynamics in the deepest level ever reached at Mercury as well as their interactions with solar wind and radiation, and interplanetary dust. In Sect. 3, the main characteristics of the BepiColombo composite mission and of instruments for the environment are briefly described.

In Sect. 4, some possible joint investigations performed by coordinated measurements of different instruments on board of the two spacecraft are suggested. These are intended as examples of the potentialities of the BepiColombo mission for the study of the coupled magnetosphere-exosphere-surface-interior system of Mercury. Summary and conclusions are given in Sect. 5.

2 Findings and Open Questions on the Hermean Environment

Thanks to Mariner-10 three fly-bys, the MESSENGER mission and the ground-based observations of the exosphere, we now have some grasp on the complexity of the Hermean environment. Starting from the space environment at Mercury’s orbit, proceeding with the magnetosphere, the exosphere and finally the surface relevant characteristics, in this section, we provide a summary of the current state of knowledge of Mercury’s environment, along with a discussion of the related unanswered key questions.

2.1 Solar Wind, Radiation Environment and Dust at Mercury Orbit

It is necessary to understand the background solar wind, radiation environment and dust conditions at Mercury to interpret future measurements and identify specific space weather characteristics. Observations taken by the Helios 1 and 2 and MESSENGER missions have characterised the interplanetary conditions at the orbit of Mercury (Marsch et al. 1982; Pilipp et al. 1987; Sarantos et al. 2007; James et al. 2017; Korth et al. 2011).

The key solar wind parameters that influence planetary space weather are the flow speed ($v$), density ($n$), proton and electron temperatures ($T_{p}$, $T_{e}$) and interplanetary magnetic field (IMF) strength and direction and parameters derived from them, like plasma and magnetic pressures, Alfvénic/sonic numbers and the plasma beta, $\beta$, (e.g., Pulkkinen 2007; Lilensten et al. 2014; Plainaki et al. 2016). The solar wind speed does not change significantly with radial distance from the Sun, its average value being 430 km/s, however it shows a significant variability (peaks of 800 km/s). The solar wind density and the strength of the IMF decrease with the square of heliocentric distance, so that on average the density and IMF strength at Mercury’s orbit are 5–10 times larger than that at the Earth orbit (see Table 2) (Burlaga 2001; Slavin and Holzer 1981). The Parker spiral at Mercury orbit forms an angle of about 20° with the solar wind flow direction, which implies a change of the relative magnitude of the IMF components with respect to the near-Earth conditions (the angle at the Earth’s orbit is ∼45°). The Alfvénic Mach number ($M_{\mathrm{A}} = V_{\mathrm{sw}}/V_{\mathrm{A}}$, where $V_{\mathrm{A}}$ is the Alfvén speed) at Mercury is about 2–5, while it is between 6–11 at the Earth (Winterhalter et al. 1985) and consequently also $\beta$, the ratio between the plasma and the magnetic pressures, is lower at Mercury, ranging between 0.5 and 0.9 against 1.7 at the Earth (Slavin and Holzer 1981). The average parameters are also slightly variable depending on the solar cycle phase (e.g.: Schwenn 1990; Korth et al. 2011). A summary of the typical solar wind parameters at Mercury and at the Earth is given in Table 2.

Table 2 Solar wind parameters at Mercury’s and Earth’s orbit

Corotating Interaction Regions (CIRs), High Speed Streams (HSSs) and ICMEs are commonly observed at Mercury. During these events the plasma conditions in the solar wind are known to vary significantly from the average.

HSSs are a domain of solar wind plasma flowing at a higher speed than usual, typically reaching a speed of 700 to 800 km/s. They are characterised by relatively weak IMF, with rapidly changing direction due to Alfvenic fluctuations, and low density. They are considered to originate in the coronal holes on the solar surface in which the magnetic field forms an open-field structure. Very high $M_{\mathrm{A}}$ might be observed at Mercury’s orbit during HSSs at solar maximum (Baumjohann et al. 2006). CIRs (e.g., Pizzo 1991; Richardson 2018) are flow structures evolving in the background solar wind due to a velocity difference between adjacent plasma streams, e.g. slow solar wind and HSSs. A stream interface forms between the two different plasma regimes and develops to a well defined structure near the orbit of Earth. At the orbit of Mercury, CIRs are typically not yet evolved (Dósa and Erdős 2017) or are less pronounced (Schwenn 1990). As CIRs evolve radially outward, compression and shear between the two streams increases. These two factors give rise to fluctuations that are superposed upon Alfvénic fluctuations generated close to the Sun. This means that inner, “younger” regions of interplanetary plasma tend to carry signatures of solar origin and their investigation can provide insight into solar processes. Measurements of magnetic field fluctuations at low frequencies can help to constrain different models of solar wind heating mechanisms and acceleration by low frequency waves (e.g. Hollweg and Isenberg 2002; Dong and Paty 2011; Dong 2014; Suzuki 2002).

ICMEs (e.g., Sheeley et al. 1985; Gopalswamy 2006; Kilpua et al. 2017) are macro-scale interplanetary structures related to Coronal Mass Ejections (CMEs) characterised generally by a higher fraction of heavy multi-charged ions (Galvin 1997; Richardson and Cane 2004). Their integral part is a magnetic flux rope and if sufficiently faster than the preceding solar wind, ICMEs have leading shocks and turbulent sheath regions ahead. Winslow et al. (2015) studied 61 ICMEs detected by MESSENGER and found high magnetic field intensity and fast mean velocity (86.2 nT, and 706 km/s, respectively). Good et al. (2015) analysed the radial evolution of a magnetic cloud ICME, using data from MESSENGER and Solar Terrestrial Relations Observatory (STEREO)-B, and found evidence that the structure was clearly expanding, with a radius increasing by about a factor of two between Mercury’s and Earth’s orbits. Unlike HSSs, CIRs and the ICME sheaths, ICMEs generally present very low M_A at the orbit of Mercury. Study of ICME propagation has been carried out in the past by using Helios1, Helios 2 and IMP data (Burlaga et al. 1980). The in situ observations of the interplanetary conditions at Mercury’s orbit by BepiColombo/Mio instrumentation coupled with observations at 1 AU or at different distances from the Sun, performed by other space missions like Solar Orbiter (Müller et al. 2013) or Parker Solar Probe (Fox et al. 2016), could be compared to the results of propagation disturbance models at varying heliospheric distances or used to constrain them (e.g., Möstl et al. 2018) and for improving the knowledge of the acceleration mechanisms. In Sect. 4.1, coordinated measurements by BepiColombo and other missions coupled to possible models/tools for the interpretations are suggested.

Galactic Cosmic Rays (GCRs) are a homogeneous, nearly isotropic background of high-energy charged particles (mostly protons) with an energy reaching GeV to even $10^{24}$ eV, originating outside the Solar System, and constituting an important component of the particle radiation environment at Mercury. They continuously bombard Mercury’s surface, generating cascades of secondary particles, including neutrons and gamma rays, providing a diagnostic of the Mercury surface composition (e.g., Goldsten et al. 2007). MESSENGER observations suggested that GCR protons are a potential energy source to stimulate organic synthesis at Mercury’s poles, where wide water ice deposits are thought to be present in permanently shadowed regions, which may contain organics (e.g., Lawrence et al. 2013; Paige et al. 2013). To characterise the Hermean radiation environment and better understand this phenomenon, an accurate evaluation of the GCR flux at Mercury’s orbit is needed, which so far has been possible only through modelling of the GCR propagation in the heliosphere (e.g., Potgieter 2013). BepiColombo will be able to monitor the GCR radiation environment and to estimate its intensity and modulation features (see Sect. 4.1).

The interplanetary medium is also populated by dust grains. Three major populations of the interplanetary dust have been identified in the inner solar system (0.3 to 1.0 AU) by previous in-situ dust observations, the Pioneer 8 and 9 and Helios dust experiments (e.g., Grün et al. 2001). Particles of one population have low-eccentricity orbits about the Sun and are related to particles originating in the asteroid belt, while particles of the second population have highly eccentric orbits and are allegedly emitted from short-period comets (Dermott et al. 2001; Jackson and Zook 1992). In situ measurements of those particles revealed grains with size from 100 down to 1 micrometer at an impact speed of 10 km/s. For the interplanetary grains in that size range, the dynamics is primarily affected by the gravitational force of the Sun, $F_{gr}$, and the solar radiation pressure, $F_{\mathit{rad}}$, the ratio $\beta = F_{\mathit{rad}}/ F_{gr}$ being close to unity. Due to the component of the radiation force tangential to a grain’s orbit, called Poynting-Robertson Light drag (e.g., Dermott et al. 2001), micron-sized particles spiral down toward the Sun. The third population identified in the inner solar system, called “$\beta$ meteoroids”, composed of small particles in size range between tens of nanometer and 0.1 micrometer and detected to arrive from the solar direction (Iglseder et al. 1996). Due to such small size, those dust particles are accelerated radially outwards by the solar radiation force against solar gravity and finally they could reach escape velocity, therefore, having hyperbolic orbits they exit the solar system (Zook and Berg 1975; Wehry and Mann 1999). In addition to these dust populations, the presence of a circumsolar ring near Mercury’s orbit was recently recognized by remote observation (Stenborg et al. 2018).

Dust grains are charged due to UV radiation or collision with charged particles; hence, they are subject to the Lorentz force that for sub-micron dust grains in the inner solar system becomes more important than the other forces, such as gravity and radiation pressure (Leinert and Grün 1990). The electric potential depends on the density and temperature of the surrounding plasma as well as the photoelectron intensity due to solar radiation. Therefore, the charge number on the grain, proportional to the dust size and potential, can be dynamic.

Using in situ measurements Meyer-Vernet et al. (2009) revealed the presence of anti-sunward directed nanograins near the Earth’s orbit. Interestingly, the distribution of the nanograins at 1 AU is highly variable, where periods of high- and zero-impact rates alternate with a period of about 6 months (Zaslavsky et al. 2012). This dust structure could be due to the complex dynamics of the charged grains in a non-uniform solar wind structure (Juház and Horányi 2013). Consequently, the electromagnetic properties in the solar wind are important to understand the dynamics of the sub-micron and micrometer sized dust.

These dust grains/micro-meteoroids impact Mercury’s surface along its orbit. Mercury has an inclined orbit, and since Mercury is away from the ecliptic plane at aphelion, it is expected that close to the aphelion phase the flux of meteoroids im**ing on the surface of Mercury and the flux of ejecta particles will both decrease (Kameda et al. 2009). The impacting rate and the local time asymmetries are poorly characterised (Pokorný et al. 2017), but a clear relation with a comet stream crossing has been observed in the exosphere composition by MESSENGER/MASCS (Killen and Hahn 2015) (See Sects. 2.6 and 2.8). The interplanetary dust grains will be detected and characterised by BepiColombo Mio instrumentation, for the first time at the Mercury environment and related to the exospheric distribution and composition (see Sect. 4.9).

2.2 What Is the Magnetosphere Configuration, Its Relation with the Planet’s Interior Structure and Its Response to Solar Activity?

The Hermean dipole moment is relatively weak ($m_{M} =195\ \mbox{nT}\cdot R_{M}^{3}$; almost perfectly aligned with the rotational axis as derived by the average MESSENGER measurements of Anderson et al. 2012). The planet is engulfed by the inner heliosphere solar wind with relatively intense dynamic pressure ($p_{sw,M} = \frac{1}{2} \rho _{M} v_{sw}^{2} \approx 10$ nPa). In comparison, the terrestrial values are very different ($m_{E} \approx 31000~\mbox{nT}\cdot R_{E}^{3};\ p_{sw,E} = \frac{1}{2} \rho _{E} v_{sw}^{2} \approx 0.5$ nPa). Nonetheless, Mercury’s intrinsic magnetic field interaction with the solar wind results in formation of a proper planetary magnetosphere, which is unique in the Solar System, being the only one of the same length scale as the planet itself. The structure of the magnetosphere resembles the terrestrial one, but differs in details. On the dayside, the planetary magnetic field is compressed by the solar wind flow, while on the night-side the magnetic field lines become stretched and elongated away from the planet and form two lobe regions in the tail separated by the current sheet. The outer boundary of Mercury’s magnetosphere towards the magnetosheath is the magnetopause, whereas the inner boundary is the surface itself. Due to the weak magnetic field of the planet and the high dynamic pressure in the solar wind, only a small magnetosphere is created. The average sub-solar magnetopause distance is only 1.41 $R_{M}$ from the planet center (Korth et al. 2017), while at the Earth it is about 10 $R_{\mathrm{E}}$. The relatively strong interior quadrupole moment with respect to the dipole causes a northward shift of the equatorial magnetosphere by 0.196 $R_{\mathrm{M}}$ (Anderson et al. 2012; Johnson et al. 2012; Wicht and Heyner 2014). This dipole offset has been the result of an analysis of the MESSENGER magnetic equator crossing done in the range $3150\leq \rho _{z} \leq 3720$ km (with $\rho _{z}$ as distance to the planetary rotation axis). Other analysis methods yielded different values of the dipole offsets. Thébault et al. (2018) reports an offset of 0.27 $R_{M}$.

Mercury’s small magnetosphere, therefore, controls, guides, and accelerates the solar wind plasma and solar energetic particles such that charged particles (>keV) precipitation can occur with enhanced intensity focused at particular locations on the surface (see Sect. 2.4). This is in contrast to the Moon or asteroids where one side of the object is bathed by unfocused solar wind, which, apart from solar eruptions, usually has lower energies (about 1 keV/nucleon) (Kallio et al. 2008). In addition, the low $M_{\mathrm{A}}$ of the solar wind causes Mercury’s bow shock and magnetopause boundary to vary dynamically over short timescales.

As Mercury does not possess an ionosphere, the planet body is directly subject to magnetospheric variations. Changes in the external magnetic field (e.g. from the magnetospheric dynamics) drive currents within the electrical conducting interior of the planet (e.g.: Janhunen and Kallio 2004). As the electromagnetic skin depth $\delta$, i.e. the characteristic depth to which a changing magnetic field penetrates a conductor, depends on the frequency of variations and the conductivity $\sigma$ of the material as $\delta =\sqrt{ \frac{(2)}{\omega \mu _{0} \sigma }}$, one has to consider the frequency band of the variation as well as the conductivity structure of the planet. Using models for the closure of field-aligned currents as observed by MESSENGER through the planet, Anderson et al. (2018) estimated the planetary conductivity structure. There the conductivity exponentially rises with depth from the crust/mantle ($\sigma \approx 10^{-8}$ S/m) to the highly conducting core-mantle boundary ($\sigma \approx 10^{6}$ S/m) at $r\approx 2000$ km from the planet center (Hauck et al. 2013; Johnson et al. 2016). They estimated that up to 90% of the total current might close in this manner (Anderson et al. 2018). Short time variations penetrate only the upper planetary layers whereas long time variations may penetrate to the core causing induction currents (e.g., Hood and Schubert 1979; Suess and Goldstein 1979; Glassmeier et al. 2007a). The effects of the induction currents on the large-scale configuration of Mercury’s magnetosphere have been inferred from the MESSENGER data for cases of extreme (Slavin et al. 2019a), strong (Slavin et al. 2014; Jia et al. 2019) and modest (Zhong et al. 2015a, 2015b; Johnson et al. 2016) variations in solar wind pressure. Global simulations that self-consistently model the induction effects (Jia et al. 2015, 2019; Dong et al.

Important science questions that BepiColombo can answer are: how do the currents circulate inside the planetary crust? to what extent does the planetary field shield the planetary surface from direct impact of particles from the solar wind on the dayside and from the central plasma sheet on the night side? is the shielding effective only during the largest induction events, or always effective except during the most intense reconnection events, or at some intermediate point between these two extremes? The planned orbits for the two spacecraft will enable Mio to acquire direct measurements of the upstream solar wind while at the same time MPO will monitor the space environment close to the planet. Such a conjunction between the two spacecraft is ideal for studying Mercury’s planetary response to the external solar wind forcing (see Sect. 4.3)

Not only do the induction currents produce dayside magnetosphere reconfiguration but, simultaneously, the nightside current systems are also significantly altered (Fig. 4 upper panel) (see Sect. 2.4). This delicate interplay between induction and reconnection, proposed by Slavin and Holzer (1979), was estimated by Heyner et al. (2016). Johnson et al. (2016) showed that the 88-day-variation in the magnetosphere due to the planetary orbit around the Sun changes the dipole moment of the planet (by about 4%) by driving induction currents deep inside the planet. Thereby, measurements of the variation of Mercury’s magnetospheric structure can be used to constrain its core-mantle boundary independently from geodetic measurements. By studying correlated periodic temporal variations, of external and induced origins, Wardinski et al. (2019) estimated the size of the electrically conductive core to be 2060 km, slightly above previous geodetic estimates. Variations on geological time scales may actually penetrate inside the core and give rise to a negative magnetospheric feedback on the interior dynamo (Glassmeier et al. 2007b; Heyner et al. 2011). The dual probe BepiColombo mission is highly suited to further study the relation between day- and nightside processes in particular during configuration with one spacecraft at the dayside and the other in the magnetotail (see Sect. 4.7). Moreover, the north-south symmetry of the BepiColombo orbits will allow a characterisation of the southern hemisphere environment, which was not well covered by MESSENGER. Furthermore, any long-term variations in the magnetospheric field may be used to sound the electrical conductivity structure of the planet in a magnetotelluric fashion.

Last but not least, the final phase of the MESSENGER mission enabled the discovery of crustal magnetic anomalies in the northern hemisphere (Hood et al. 2018). Their analysis may provide important information about the temporal variation of the planetary magnetic field in the past (Oliveira et al. 2019). However, if the southern hemisphere magnetic anomalies are similar to those of the northern hemisphere, where the magnetic field arising from the known anomalies is at maximum 8 nT at 40 km (Hood 2016; Hood et al. 2018), their effect should be negligible for magnetospheric dynamics. Possible deviation of charged particles at the surface by local magnetic fields would be recognised by BepiColombo as ion back-scattering intensification at the interface between the micro-magnetosphere and its internal cavity, as it has been observed in the case on Mars (Hara et al. 2018) and the Moon (Saito et al. 2008; Deca et al. 2015; Poppe et al. 2017).

2.3 How Does the Solar Wind Mix with Mercury’s Magnetosphere?

The pristine solar wind does not directly interact with the Hermean magnetosphere. Instead, it is modified by processes in the upstream foreshock, at the bow shock, and in the magnetosheath before encountering the magnetopause. At the bow shock, the solar wind plasma is decelerated and heated from super-magnetosonic to sub-magnetosonic speeds, enabling it to flow around the obstacle that the Hermean magnetosphere constitutes (e.g., Anderson et al. 2010, 2011). As the IMF cone angle (the angle between the IMF and the Mercury-Sun-line) is typically ∼20° (Table 2), the subsolar bow shock is most often quasi-parallel (e.g., Slavin and Holzer 1981).

Upstream of the quasi-parallel shock, a foreshock region can be found that is magnetically connected to the shock, into which shock-reflected (so-called back-streaming) particles are able to travel along the IMF (Jarvinen et al. 2019). Those particles interact with the solar wind, generating waves and steepened magnetic structures (e.g., Burgess et al. 2005; Jarvinen et al. 2019). These waves and structures are convected with the solar wind stream back to the quasi-parallel shock. Hence, the regions upstream and downstream of that shock are generally more variable in comparison to the quasi-perpendicular shock and adjacent regions (e.g.: Le et al. 2013 Eastwood et al. 2005; Sundberg et al. 2013, 2015; Karlsson et al. 2016; Jarvinen et al. 2019).

The terrestrial quasi-parallel bow shock is highly structured and allows for high-speed jets of solar wind plasma to regularly form, penetrate the magnetosheath, and impact onto the magnetopause (e.g., Hietala et al. 2009, 2012; Plaschke et al. 2013a, 2013b, 2018). Signatures of high-speed jets have not yet been found in the Hermean magnetosheath (Karlsson et al. 2016), however, structures similar to hot flow anomalies have been identified near Mercury (Uritsky et al. 2014).

Differences are also apparent with respect to the quasi-perpendicular side of the bow shock and the corresponding magnetosheath, where ion cyclotron and mirror mode waves can originate from anisotropic particle distributions (e.g. Gary et al. 1993). At Earth, both modes exist, while at Mercury, only ion cyclotron waves have been observed (Sundberg et al. 2015). Mirror modes have only been predicted in simulations (Herčík et al. 2013). Their growth in the dayside magnetosheath may be inhibited by the limited size of the region and by the low plasma $\beta$ (Gershman et al. 2013).

Mio observations are expected to shed light on the existence and basic properties of several foreshock and magnetosheath phenomena, including foreshock cavities, bubbles, hot flow anomalies, jets, and mirror mode waves, due to the optimised orbit and advanced plasma instrumentation with respect to MESSENGER. In addition, Mio and MPO dayside conjunctions will allow, for the first time, simultaneous observations near Mercury and in the upstream foreshock, shock, or magnetosheath regions. This will make it possible to study the impact of transient phenomena emerging in these regions of the Hermean magnetosphere.

Two-point measurements in the magnetosheath will also give information on how the turbulence develops downstream of the bow shock, which will give an interesting comparison to the situation at Earth. Turbulence is probably the best example in plasma physics of multi-scale, nonlinear dynamics connecting fluid and kinetic plasma regimes and involving the development of many different phenomena spreading the energy all over many decades of wave numbers. To date, the near-Earth environment and the solar wind represented the best laboratory for the study of plasma turbulence (Bruno and Carbone 2013 and references therein) providing access to measurements that would not be possible in laboratories. Turbulent processes were observed in MESSENGER magnetic field data (Uritsky et al. 2011), especially at kinetic scales. Thanks to the Mio full plasma suite and to the MPO instruments for the space plasma observations, BepiColombo will offer the opportunity not only to conduct thorough turbulent studies, but also the great opportunity, from a physical point of view, to access the physical parameters and, therefore, the plasma regimes that are not available in the terrestrial magnetosphere and nearby solar wind. In particular, we will have access to low beta regimes inside the Mercury magnetosphere and to a fully kinetic turbulence lacking the large-scale MHD component typical of the Earth’s magnetosheath. Last but not least, during operations of BepiColombo at Mercury and in the nearby solar wind, combined analysis with Solar Orbiter and Parker Solar Probe are expected to be a great opportunity to build a more complete view on the inner solar wind turbulence properties at different distances from the Sun.

The solar wind flowing around the Hermean magnetosphere, producing turbulence, is a driver of both magnetic and plasma fluid instabilities eventually producing an efficient mixing of the two plasmas. In this context, magnetic reconnection plays a key role by ultimately allowing for the entering of solar wind plasma into the magnetosphere, and thus a net momentum transport across the magnetopause. Nevertheless, it is not yet fully understood how dayside reconnection is triggered at the sub-solar point of Mercury’s magnetopause. Based on observations at the Earth, magnetic reconnection between the southward oriented IMF and the planetary magnetic field is the most effective plasma mixing process. However, analysis of MESSENGER data demonstrated that reconnection at Mercury is significantly more intense than at the Earth (e.g. Slavin et al. 2009, 2012, 2014). Di Braccio et al. (2013) reported that the reconnection rate in the subsolar region of the magnetopause is independent of the IMF orientation, attributed to the influence of low-$\beta$ plasma depletion layers (Gershman et al. 2013), however a larger statistical study found that reconnection-related signatures were observed at a significantly higher rate during southward IMF intervals, and concluded that the relationship between clock angle and reconnection rate is akin to that observed at the Earth (Leyser et al. 2017).

BepiColombo will consistently provide very good estimates of the plasma pressure values, enabling more comprehensive studies of this phenomenon. A much better understanding of the dayside reconnection processes at Mercury is crucial, being the dominant process allowing for the solar wind plasma to enter the magnetosphere. Frequent reconnection made the measurement of large amplitude Flux Transfer Events (FTEs) (Slavin et al. 2012; Imber et al. 2014; Leyser et al. 2017) by MESSENGER a common occurrence. Solar wind particles reaching the cusps are eventually partially mirrored in the strengthening field or impacting the surface there, as observed by MESSENGER (Winslow et al. 2014; Raines et al. 2014). Poh et al. (2016) observed isolated, small-scale magnetic field depressions in the dayside magnetosphere, known as cusp filaments, thought to be the low latitude extent of FTEs (Fig. 5). Since this particle bombardment at the cusp regions is on going over geological time scales, the surface material may actually be darkened in certain spectral bands (see also Sect. 2.7 and Rothery et al. 2020, this issue). The cusp location depends on the Hermean heliocentric distance as well as the IMF direction. The northern cusp region has been readily identified by analysing the magnetic field fluctuations and its anisotropy related to the reconnection (He et al. 2017). The BepiColombo two-spacecraft configuration in the cusp region will offer an optimal opportunity for a detailed analysis of FTEs, filaments, and plasma entering the magnetosphere in both hemispheres (see Sect. 4.5).

Fig. 5

(a) Example of a MESSENGER orbit (black solid line) on 26 August 2011 projected onto the X-Z plane in aberrated Mercury solar magnetic (MSM’) coordinates during a period without filamentary activities in the cusp. The model bow shock ($BS$) and magnetopause ($MP$) from Winslow et al. (2013) (marked by the two dots at the dayside magnetosphere) are shown in dotted lines; the Sun is to the right. The thick portion of the orbit represents the cusp region, and the dot at the nightside magnetosphere represents the magnetotail current sheet ($CS$) crossing. The arrow denotes the spacecraft trajectory. (b) Full-resolution magnetic field measurements (top to bottom, $X$, $Y$, and $Z$ components and field magnitude) acquired along the orbit shown in Fig. 1a. The vertical dashed lines mark the boundary crossings shown in panel (a). CA denotes the closest approach, and all times are in UTC. (Poh et al. 2016)

Magnetopause reconnection is not limited to the dayside. Müller et al. (2012) showed in a simulation how reconnection at the equatorial dawn flank allows magnetosheath plasma to enter the magnetosphere and contribute to a partial ring current plasma. Di Braccio et al. (2015a) provided the first observations of the plasma mantle, a region in the near-tail where plasma is able to cross the magnetopause along open field lines. A subsequent statistical analysis by Jasinski et al. (2017) demonstrated that the mantle was more likely to be observed during southward IMF, and (due to the observations being entirely in the southern hemisphere), during periods of negative B_X. The BepiColombo orbit will allow an in-situ analysis of such phenomena in both hemispheres (see Sect. 4.2).

Apart from reconnection processes, instabilities (e.g. KH, mirror and firehose instabilities) also play an important role in the mixing process and can be associated with different kinds of waves. At Mercury, KH instabilities were predicted (Glassmeier and Espley 2006) and observed by MESSENGER (Sundberg et al. 2012a, Liljeblad et al. 2014). By using MESSENGER data, Liljeblad et al. (2015) showed that the local reconnection rate was very low at the magnetopause crossing associated with the presence of a low-latitude boundary layer (LLBL), ruling out direct entry by local reconnection as a layer formation mechanism. In fact, at Mercury KH waves (which have been suggested to provide particle entry into the LLBL at Earth; e.g. Nakamura et al. 2006) predominantly occur at the dusk side of the magnetopause where, due to kinetic effects resulting from the large gyroradii, ions counter-rotate with the waves (Sundberg et al. 2012a, Liljeblad et al. 2014). On the opposite flank ions co-rotate with the waves resulting in reduced growth rates and larger LLBL (Liljeblad et al. 2015). Mid-latitude reconnection associated with KH instabilities was also discussed by Faganello and Califano (2017), and Fadanelli et al. (2018). In fact, Gershman et al. (2015b) showed that a subset of nightside KH vortices actually has wave frequencies close to the Na⁺ ion gyrofrequency, indicating that those ions can alter KH dynamics, probably through kinetic effects. James et al. (2016) showed evidence that the KH instability is driving Hermean ULF wave activity that could be better identified by BepiColombo instrumentation (magnetometers and charged-particle detectors). ULF waves are also used to estimate plasma mass density profiles along field lines and, generally, within the magnetosphere (James et al. 2019) complementing observations made by particle instrumentation.

Another peculiarity of the Hermean magnetosphere compared to the terrestrial one is that the plasma density gradients are observed to have much smaller spatial scales and are much more pronounced than that on Earth, due to the interaction of a strongly choked plasma (the solar wind) with a nearly empty cavity constituted by the small-scale magnetosphere. The role of the instability of the density gradient in the planet-Sun interaction is an interesting topic to investigate also as an example of other comparable exoplanetary environments. BepiColombo will enable analysis of the smaller scales instabilities or “secondary” instabilities that may be much more efficient than fluid-scale instabilities in plasma mixing processes (Henri et al. 2012, 2013). One of those mechanisms could be the lower-hybrid induced, non-adiabatic ion motion across the magnetopause. Since the lower hybrid waves are almost electrostatic, it was not possible to test this hypothesis with MESSENGER. Some coordinated observations between the two BepiColombo spacecraft are suggested in Sect. 4.2.

A further important aspect of solar wind/magnetospheric plasma mixing is the phenomena associated with violent solar wind events. The proximity of Mercury to the Sun compared to the Earth and the small scale of its magnetosphere make it even more responsive to unusually strong events (see next Sect. 2.4).

Another key question concerns whether the mechanism of impulsive penetration observed at Earth (e.g. Echim and Lemaire 2002) is operating at Mercury. We know that at least the small-scale variation in the momentum and density necessary for this mechanism exists in the form of magnetic holes (Karlsson et al. 2016). These findings should be further investigated with linked observations from MPO and Mio, because these will bring a better understanding of the space weathering at Mercury and its contribution to the generation of Mercury’s exosphere, as detailed in Sect. 2.7.

2.4 How do the Solar Wind and Planetary Ions Gain Energy, Circulate Inside the Magnetosphere and Eventually Impact the Planetary Surface? What Is the Current System in Mercury’s Magnetosphere?

As explained in Sect. 2.3, solar wind plasma enters the magnetosphere through dayside magnetopause reconnection, as at Earth, but this process takes place at Mercury even when the magnetic shear angle, the angle between the IMF and planetary magnetic field (Di Braccio et al. 2013) is low. The time resolution (10 s) and the angular coverage (1.15$\pi$) of MESSENGER particle measurements was insufficient to study the resulting acceleration, while BepiColombo will provide resolutions up to 4 s and full angular coverage (see Sect. 4.5). Newly-reconnected magnetic field lines containing solar wind plasma are convected through the magnetospheric cusps to form the plasma mantle, where the competition between down-tail motion and E x B drift toward the central plasma sheet determines which solar wind ions end up in the lobes of the magnetotail (Di Braccio et al. 2015b; Jasinski et al. 2017). Solar wind H⁺ and He²⁺ that make it to the central plasma sheet retain their mass-proportional heating signatures when observed there. Reconnection between lobe magnetic field lines in the tail sends plasma sheet ions and electrons both tail-ward and planet-ward, where they escape downtail, impact the surface or are lost across the magnetopause. Precipitating ions observed on the nightside are mainly at mid- to low- latitudes where the magnetic field lines are closed, that is, with both ends connected to the planet, (Korth et al. 2014), providing evidence of a relatively large loss cone (Winslow et al. 2014). The magnetic flux carried in this process is returned to the dayside completing the Dungey cycle (Dungey 1961) in a few minutes (Slavin et al. 2010, 2019b).

At Earth, energetic charged particles trapped inside the planetary magnetic field azimuthally drift around the planet, because of gradient and curvature drifts. The drift paths are along the iso-contours of the magnetic field. A consequence of the small Hermean magnetosphere with a relatively big portion occupied by the planet is the lack of a significant ring current (Mura et al. 2005; Baumjohann et al. 2010). In fact, closed drift paths around the planet are not allowed.

MESSENGER provided evidence of planetary ions in various regions of the Mercury’s magnetosphere (e.g., Zurbuchen et al. 2011; Raines et al. 2013), primarily in the northern magnetospheric cusp and central plasma sheet. Sodium ions originating from the sodium exosphere are one of the main contributors of planetary ions to the magnetospheric plasma at Mercury. Such processes have been explored with statistical trajectory tracing in the electric and magnetic field models of the Mercury’s magnetosphere either using empirical (e.g., Delcourt et al. 2007, 2012) or MagnetoHydroDymanic (MHD) simulations (e.g., Seki et al. 2013; Yagi et al. 2010, 2017). Even with steady magnetospheric conditions, the dynamics of sodium ions can change dramatically with conditions of the surface conductivity (Seki et al. 2013) or solar wind parameters (Yagi et al. 2017). Various mechanisms can contribute to the energisation of sodium ions, including $i$) acceleration by convective electric field around the equatorial magnetopause, resulting in the partial sodium ring current (Yagi et al. 2010), $\mathit{ii}$) the centrifugal effect due to curvature of the electric field drift paths (Delcourt et al. 2007), and iii) induction electric field during substorms (Delcourt et al. 2012). Some evidence consistent with centrifugal acceleration has been observed, such as sodium ions being predominantly observed in the pre-midnight sector of the magnetotail (Raines et al. 2013; Delcourt 2013) but observations of ions undergoing such acceleration have not been reported. The much more comprehensive instrument complement on BepiColombo, including mass spectrometers on both spacecraft, should enable these concrete connections between models and observations to be established (see Sect. 4.5).

Waves should play a substantial role in particle acceleration at Mercury, due to the highly dynamic nature of Mercury’s magnetosphere. MESSENGER observations of wave activity at different frequencies (Boardsen et al. 2009, 2012, 2015; Li et al. 2017; Sundberg et al. 2015; Huang et al. 2020) indicates different physical processes at work and has shown the expected turbulent cascade of energy from MHD scales down to ion-kinetic scales. Much of this power spectrum lies within the ion-kinetic regime, and so wave-particle interactions like ion cyclotron dam** should play a role in the acceleration of both solar wind and planetary ions within the system. However, understanding of particle acceleration through such mechanisms remains limited, requiring further treatment by both theory and numerical modelling. Ultimately, BepiColombo measurements will resolve magnetospheric wave activity in considerably more detail and enable significant progress towards answering the question of how both ions and electrons are accelerated (see Sect. 4.5).

While much has been learned about ions >100 eV from MESSENGER observations, low-energy planetary ions are a complete mystery. Born around 1 eV, these ions have never been observed as MESSENGER’s lower energy bound was 46 eV for most of the mission (Raines et al. 2014). If present, these ions could have substantial effects. At Earth, such low-energy ions have been shown to alter the kinetic physics of magnetic reconnection on the dayside magnetopause (Borovsky and Denton 2006; Li et al. 2017). Studies of field line resonances show that the total plasma mass density on the dayside may be >200 AMU/cm³, in contrast to the very low (sometimes undetected) densities measured by MESSENGER (James et al. 2019). In the magnetotail, a substantial cold planetary ion population in the central plasma sheet would substantially change the mass density and may be one of the unseen factors causing asymmetries observed there, in reconnection signatures (Sun et al. 2016), current sheet thickness (Poh et al. 2017), and field line curvature (Rong et al. 2018). Thus far, none of these asymmetries have been tied to the 0.1–10 keV planetary ions observed in this region by MESSENGER. BepiColombo/Mio will be able to measure the spacecraft potential, thus estimation of the density of the lower-energy ions could be obtained.

At the Earth, there are two primary large-scale current systems which flow into/out of the high latitude ionosphere, known as region 1 and region 2 currents. These currents specifically couple the magnetopause and the inner magnetosphere, closing through Pedersen currents in the ionosphere (Fig. 6a), and are enhanced during periods of high magnetospheric activity. The region 1 currents map to higher latitudes in the ionosphere and are upward on the dusk side and downward on the dawn side. The region 2 currents map to locations equatorward of this and have opposite polarity. Mercury’s magnetosphere differs from that of the Earth for the smaller size relative to the planetary radius, the higher amplitude and smaller timescales for magnetospheric dynamics at Mercury, and the lack of a conducting ionosphere for current closure. Glassmeier (2000) suggested that current closure is not required in any ionosphere and it is possible in the magnetospheric plasma proper, while early simulations predicted that in this small magnetosphere region 1 currents could develop and close inside the highly conductive planetary interior while the region 2 currents could not fully develop (Janhunen and Kallio 2004). Field-aligned currents are typically observed at the Earth using magnetic field measurements taken by spacecraft passing over the high latitude regions. A model of the internal magnetic field is subtracted from each pass, and the residual magnetic field is analysed for perturbations indicative of a local current. Anderson et al. (2014) performed this analysis on Mercury’s magnetic field using MESSENGER data and concluded that region 1 field-aligned current signatures were identifiable, particularly during geomagnetically active times (Fig. 6b). These signatures suggested typical total currents of 20–40 kA (up to 200 kA during active times), which may be compared with current strengths ∼MA at the Earth. The signatures were relatively smooth and occurred on every orbit passing through the current regions, which implied that the current systems were stable. This raises an important open question of current stability, given the short timescales for dynamics at Mercury (e.g. Slavin et al. 2012; Imber and Slavin 2017).

Fig. 6

(a) A schematic of the large-scale field-aligned currents observed in the northern hemisphere at the Earth, from Carter et al. 2016. (b) An example of northern hemisphere field-aligned currents accumulated over many MESSENGER orbits during 2012, from Anderson et al. 2018

While Mercury does not have a conducting ionosphere, it does have a large metallic core with a radius of ${\sim} 0.8~R_{\mathrm{M}}$, above which is a lower conductivity silicate mantle. This unique topology allows field-aligned currents to close across the outer surface of the core (see also Sect. 2.2). One of the key open questions for BepiColombo in the realm of magnetospheric dynamics is the closure mechanism for region 1 field-aligned currents and the extent to which induction currents are driven at Mercury.

Finally, Anderson et al. (2014), in agreement with previous modelling results (Janhunen and Kallio 2004), did not find evidence for region 2 currents at Mercury, suggesting that plasma returning to the dayside from the magnetotail may impact the surface. This suggestion is strongly supported by observations of MESSENGER/X-Ray spectrometer (XRS) which show evidence of X-ray emission from Mercury’s nightside surface, mainly located between 0 and 6 h local time (Lindsay et al. 2016), caused by fluorescence attributed to precipitation of electrons originating in the magnetotail (Starr et al. 2012) (Fig. 7a). Electrons around 10 keV were observed in association with magnetic field dipolarisations in the magnetotail (Dewey et al. 2018). The offset dipole magnetic field at Mercury is expected to cause an asymmetry in the north-south precipitation intensity and location, however MESSENGER’s elliptical orbit did not allow all regions of the surface to be equally accurately characterised.

Fig. 7

(a) Maps of XRS footprint locations associated with XRS records containing magnetospheric electron-induced surface fluorescence in latitude-local time coordinates centered at midnight. The southern hemisphere data have a lower spatial resolution. (Lindsay et al. 2016). (b) The distribution of suprathermal electron events by latitude and local time centered at noon. The events spanned all local times but had the highest concentration in the dawn and dusk sectors (Ho et al. 2016)

Observations of the loading and unloading of open magnetic flux in Mercury’s magnetotail (Imber and Slavin 2017), combined with in situ measurements of reconnection-related phenomena such as dipolarisation fronts (Sundberg et al. 2012a, 2012b), flux ropes (e.g. Smith et al. 2017) and accelerated particles (Dewey et al. 2017), reproduced by ten-moment multifluid model (Dong et al. 1995; Baumgardner et al. 2008; Schmidt et al. 2012) (Fig. 8). Other species can be quickly photo-ionised and begin circulating in the magnetosphere as planetary ions. The released atomic groups (mainly after MIV) can be further dissociated, gaining energy. For a full characterisation, multiple instruments and systematic observations of the exosphere at different conditions are required, as well as a combination of simultaneous measurements of possible drivers and the resultant final particles, such as photo-ionised ions. BepiColombo will simultaneously observe the exospheric composition, solar wind, planetary ions and dust (see Sects. 4.6, 4.8, 4.9 and 4.11).

Fig. 8

A composite of four images of sodium at Mercury showing spatial scales ranging from the diameter of the planet, to approximately 1000 times that size. Image obtained using (a) the 3.7 m AEOS telescope on Maui on 8 June, 2006, and (b) the 0.4 m telescope at the Tohoku Observatory on Maui on 10 June, 2006. (c, d) Obtained using the 0.4 m and 0.1 m telescopes at the Boston University Observing Station at the McDonald Observatory on the night of 30 May, 2007. The tail brightness levels at distances larger than ${\sim} 10~R_{M}$ are higher in Figs. 1c than 1b, a manifestation of exospheric variability. (Baumgardner et al. 2008)

The exosphere of Mercury was discovered by the Mariner 10 Ultraviolet Spectrometer (Broadfoot et al. 1974), which measured H and He and obtained an upper limit for O (Shemansky 1988). Hydrogen was also measured by the Ultraviolet and Visible Spectrometer (UVVS), a subsystem of MASCS, on board of MESSENGER (McClintock et al. 2008; Vervack et al. 2018). Although the scale heights of H measured by Mariner 10 and MESSENGER agree, consistent with a temperature of ∼450 K, the intensities measured by MASCS were a factor of about 3–4 greater than those measured by Mariner 10. The surface number density, $n_{0}$, of H inferred from the first MESSENGER flyby was $70 < n_{0}< 250$ cm⁻³, while that inferred from the second flyby was $65 < n_{0}< 95$ cm⁻³.

MESSENGER did not observe He since the wavelength range of MASCS did not extend to the 58.4 nm He emission line. Mariner 10 obtained a maximum He column density of $2.5\times 10^{12}$ cm⁻², and a single scale height consistent with $T = 450$ K. One intriguing point of the Mariner 10 observations is the mismatch between the models and the altitude profile closer to the terminator, while altitude profiles closer to the subsolar point were reproduced accurately by the models (Broadfoot et al. 1976). This mismatch was interpreted by Shemansky and Broadfoot (1977) and Smith et al. (1978) to be due to a poorly understood thermal accommodation, which makes the exospheric density more or less dependent on the surface temperature (hence the mismatch closer to the terminator, where surface temperature was less constrained). Some insights have come from the Moon, where helium observations by the orbiters LADEE and LRO can be explained by helium being fully accommodated to the lunar surface (Hurley et al. 2016; Grava et al. 2016). But for Mercury, the lack of measurements has significantly impeded further progress in understanding the gas-surface interaction, a fundamental parameter in the study of temporal evolution of exospheres. The UV and mass spectrometers on board BepiColombo will be able to detect helium, filling this decade-long gap.

Another interesting aspect of ⁴He is that some unknown fraction of it can come from outgassing from the Hermean interior, since ⁴He is the radioactive decay product of ²³²Th, ²³⁵U, and ²³⁸U within the crust. Current data cannot constrain that, but again some insights come from the Moon, where ∼15% of the exospheric helium is unrelated to the solar wind alpha particle (He⁺⁺) influx, the main source of helium, and presumed to be endogenic (Benna et al. 2015). BepiColombo will measure the exospheric ⁴He and simultaneously the solar wind alpha particle fluxes, hence it will be able to constrain the Hermean endogenic ⁴He source rate (see Sect. 4.6).

Concerning diffusion from the crust of exospheric species, ⁴⁰Ar is another radiogenic gas, being the result of radioactive decay of ⁴⁰K in the crust that ultimately finds its way to the exosphere through cracks or fissures (Killen 2002). The Mariner 10 UV spectrometer could only place a generous upper limit of subsolar density of $6.6\times10^{6}$ cm⁻³ (Shemansky 1988), based on the sensitivity of the instrument. This density would make ⁴⁰Ar one of the most abundant species in the Mercurian exosphere. Interestingly, on the Moon this is indeed the case: ⁴⁰Ar and ⁴He are the most abundant exospheric elements identified so far, peaking at a few $10^{4}$ cm⁻³ (Hoffman et al. 1973). MESSENGER MASCS bandpass did not include the emission line doublet of ⁴⁰Ar at 104.8 and 106.7 nm. A measurement by BepiColombo of the column density of ⁴⁰Ar (and hence of its source rate) would constrain the abundance of ⁴⁰K within the crust. This measurement coupled with the measurement of the ionised component ⁴⁰Ar⁺, providing the loss rate for this element (photo-ionization and electron impact ionization being the major loss processes), with important implications for Mercury formation (see Sect. 4.11).

Although searches for O were regularly conducted by the MASCS instrument on board MESSENGER, there was no clear detection of O in the spectrum. The upper limit for O is ∼2 R (Rayleighs) at 130.4 nm (Vervack et al. 2016). The value reported by Mariner 10 (60–200 R; Broadfoot et al. 1976), well above this number, would have been easily detected during the UVVS observations. It is possible that depletion of oxygen in the Hermean exosphere occurs by condensation of metal oxides and by formation of slowly photolyzed oxides (Berezhnoy 2018) rather than being ejected as neutral atoms. Where is the expected Mercurian oxygen? The identification of atom groups by mass spectrometers on BepiColombo is a unique way for looking for oxides (see Sect. 4.9).

The sodium exosphere of Mercury was first observed from the ground by Potter and Morgan (1985) using the high-resolution echelle spectrograph at the McDonald Observatory. Since that time sodium has been the most observed species in the Hermean exosphere, thanks to its intense intrinsic brightness. North/south asymmetries and variable high latitude enhancements of sodium have been observed with ground-based instruments (e.g. Potter et al. 1999; Killen et al. 1999; Mangano et al. 2009, 2015) (Fig. 9), including transit observations showing northern or southern enhancements at the limb (Schleicher et al. 2004; Potter et al. 2013; Schmidt et al. 2018). K has been observed by ground-based observations too, showing behaviour similar to that of Na (Potter and Morgan 1986; Killen et al. 2010). An extended sodium tail was first observed by Potter et al. (2002), and subsequently studied by Potter et al. (2007), Baumgardner et al. (2008), Potter and Killen (2008), and Schmidt (2013). These studies demonstrated that the extent of the tail strongly depends on the TAA. Mouawad et al. (2011) showed that the simulated Na exosphere strongly depends on the assumed velocity distributions of the source processes, the composition of the regolith, and sticking and thermal accommodation factors assumed in the simulation. The low to medium-energy source processes such as PSD and MIV are more likely to provide Na to the tail (Schmidt et al. 2012). A fairly repeatable seasonally varying equatorial sodium exosphere was reported by Cassidy et al. (2015). Cassidy also reported a repeatable pattern of East/West sodium asymmetries tied to the Mercury TAA of the planet explained as due to higher Na condensation in the surface regions where the average temperature is colder (cold poles) (Cassidy et al. 2016) (Fig. 10). Leblanc and Johnson (2010) suggested that thermal desorption and photon-stimulated desorption are the dominant source processes for the Na exosphere of Mercury. Recently, Gamborino et al. (2019), analysing the MESSENGER vertical profile of the equatorial subsolar Na exosphere concluded that the main process responsible for Na release in this region seems to be the thermal desorption. In contrast, Orsini et al. (2018), by analysing the Na images obtained by the THEMIS telescope coupled with the magnetic and ion measurements of MESSENGER, reported a variation of Na shape related to an ICME arrival at Mercury, thus linking the ion precipitation to the sha** of the Na distribution (see Sect. 2.4). A multi-process mechanism, involving ion sputtering, chemical sputtering and PSD have been invoked to explain the Na relationship with the precipitation of ions (Mura et al. 2009).

Fig. 9

Examples of recurrent Na emission patterns identified in the Hermean exosphere. Equatorial Peak South, and Wide Peak, Bottom, from left to right: 2 symmetric peaks, 2 peaks with northern spot dominant, 2 peaks with southern spot dominant and 2 peaks connected. (Mangano et al. 2015)

Fig. 10

Observed sodium column density projected onto Mercury’s equatorial plane over the course of one Mercury year. These observations show a sodium enhancement that rotates with the surface and peaks near Mercury’s cold-pole longitudes (white dashed lines) when they are sunlit. The enhancement grows over the course of the morning, reaches a peak near noon, and then fades in the afternoon. (Cassidy et al. 2016)

Currently, the scientific community is divided between those favouring an interpretation stating that the two variable peaks are linked to solar wind precipitation and that the Sun’s activity is the major driver of the Na exosphere configuration at Mercury (e.g: Killen et al. 2001; Mura et al. 2009; Mangano et al. 2013; Massetti et al. 2017; Orsini et al. 2018), and those favouring the variations being due only to the surface temperatures considered over long time scales or according to position along the orbit (e.g: Leblanc and Johnson 2010; Schmidt et al. 2012; Cassidy et al. 2016) or due to the crossing of the interplanetary dust disk (Kameda et al. 2009). The question is even more tricky considering the question of how this volatile element survived throughout Mercury’s evolution history. Multi-point and multi-instrument observations by BepiColombo will provide full characterisation of Na together with possible drivers of its surface release (see Sect. 4.10).

Calcium was discovered in Mercury’s exosphere by Bida et al. (2000) using the échelle spectrograph HIRES at the Keck I telescope. It was determined to have a very large-scale height consistent with high temperature (Killen and Hahn 2015); in fact, Killen et al. (2005) suggested that the hot calcium atoms are most likely produced by a non-thermal process. This was verified by the MESSENGER MASCS observations which determined that the calcium is ejected from the dawnside with a vertical density profile that has been interpreted resulting from a characteristic energy of about 6.4 eV, which Burger et al. (2014) converted to a temperature of 70’000 K.

Magnesium was discovered by the MASCS spectrometer on-board the MESSENGER spacecraft during the second flyby (McClintock et al. 2009). The flybys observations, analysed by an exospheric model (Sarantos et al. 2011), are consistent with a source located in the post-dawn equatorial region producing a dual temperature distribution (determined by fitting the vertical profile with a Chamberlain model): hot energetic distribution (up to 20000 K) and cool distribution (less than 5000 K). Retrieved temperatures from the MESSENGER MASCS data along the orbit evidenced periods of a single source in the dayside at betwen 4000–6000 K and a double source near the dawn terminator, as registered during the flyby, for 15% of the time (Merkel et al. 2017).

Observations showed evidence of a dawn enhancement also correlated to the Mg-rich surface region (Merkel et al. 2018). Both Ca and Mg are consistent with impact vaporization in the form of molecules, which are subsequently dissociated by a high-energy process (Killen 2016; Berezhnoy and Klumov 2008; Berezhnoy 2018). The location and timing of the enhanced Ca emission near TAA = 30° are suggestive of a connection with the comet 2P/Encke dust stream (Killen and Hahn 2015; Christou et al. 2015; Plainaki et al. 2017) (Fig. 11), which is also suggested to be the primary driver of Mg, Al, Mn and Ca⁺ observed at these particular TAA (Vervack et al. 2016).

Fig. 11

(a) Ca vaporization rate at Mercury due to the interplanetary dust-disk (magenta line) plus a cometary stream whose peak density occurs at TAA 25^∘. The red line is the summed contributions from the cometary dust stream plus that due to an interplanetary dust-disk that is inclined $10^{\circ}$ from Mercury’s orbital plane, and whose ascending node is $290^{\circ}$ when measured from Mercury’s longitude of perihelion, with the dust density varying as $R^{2}$, where $R$ is the heliocentric distance. The MASCS observations are plotted in black. (b) Sketch of the Mercury orbit crossing the 2P/Encke dust stream. (Killen and Hahn 2015)

Ca⁺ was first detected in Mercury’s exosphere during the third MESSENGER flyby (Vervack et al. 2010). Although Ca⁺ was not regularly detected by UVVS, it was detected on several occasions during the last year of the mission. The FIPS team was unable to unambiguously confirm the detection of Ca⁺ due to limited mass resolution of the instrument and possible overlap with K⁺ (Zurbuchen et al. 2008). From modelling the ion measurements of FIPS, it was found that the Ca⁺ abundance is about two decades larger than the K⁺ (Wurz et al. 2019).

Bida and Killen (2011) reported measurements of Al at line-of-sight abundances of $(2.5\mbox{--}5.1) \times 10^{7}$ cm⁻² from 860 to 2100 km altitude from observing runs at the Keck 1 telescope during 2008 and 2011 (Bida and Killen 2017). Al was also detected by MASCS late in the MESSENGER mission, at a line-of-sight column abundance of $7.7\times10^{7}$ cm⁻². The UVVS value pertains to lower altitudes (250–650 km) than those measured by Keck and are thus considered consistent.

The UVVS data revealed the unexpected presence of Mn at an estimated column of $4.9\times 10^{7}$ cm⁻² (Vervack et al. 2016), but it is estimated to be highly variable. Because the geometry of the observation was complicated, the column abundance is considered as an order-of-magnitude.

In conclusion, Mariner-10, MESSENGER, and the numerous ground-based observations proved the presence of H, He, Na, K, Ca, Mg, Al, Mn in Mercury’s exosphere, but other species and atom groups are expected to be present. The BepiColombo multi-type of instrument approach to the exosphere identification (combining remote sensing and in-situ measurement) warrants that new elements will be added to the list (see Sect. 4.11).

2.7 What Are the Relationships Between the Solar Wind or the Planetary Ions and the Exosphere?

As anticipated in the previous sections, Mercury’s weak magnetic field, the high reconnection rate, and its exosphere allow the solar wind to reach a large portion of the dayside surface, primarily focusing at the base of the cusp regions. However, as solar wind is conflated inside the magnetosphere, it can impact the surface in other regions as well, such as the dawn polar regions (Raines et al. 2014, 2015). The planetary ions directly released from the surface or resulting from the exosphere photo-ionization, circulate and are accelerated in the magnetosphere, and can be convected back onto the surface mainly on the night side at middle latitudes, but also in the dusk flanks and the dayside (Raines et al. 2013, 2015; Wurz et al. 2019) (see Sect. 2.4). Therefore, these charged particles of solar wind or of planetary origin can impact the surface over a wide range of local times. The impact of an energetic ion onto a surface can have a number of different, inter-related consequences, including: the reflection of the neutralized impacting particle, the ejection of neutrals and charged particles from the surface (ion-sputtering process), the production of X-rays in the case of higher energy and high charge-state ions as in the SEP events, and the alteration of the chemical properties of the surface, causing the so-called “space weathering” (Domingue et al. 2014; Strazzulla and Brunetto 2017; see also Rothery et al. 2020, this issue). The ion-sputtered neutral particles contribute to filling the exosphere, but the contribution of this process with respect to other surface-release processes is still unclear and remains a matter of debate within the science community.

While MESSENGER observed ions directed toward the surface and was able to provide estimates of precipitation rates in the northern magnetospheric cusp by analysing deep magnetic field depressions (Poh et al. 2017) and average pitch angle distributions (Winslow et al. 2014). Despite an estimation of the proton precipitation flux in the range $10^{6}~10^{7}$ cm⁻² s⁻¹ (Raines private communication), MESSENGER was not able to provide a direct proof of the impact onto the surface or the connection with the exosphere generation. The ENA instruments on board BepiColombo will be able to detect the back-scattered ions, thus providing evidence of the impacts (Milillo et al. 2011).

In spite of the rapid changes in the precipitating proton flux due to the fast magnetospheric activity (time scales of 10s of seconds) and magnetic reconnection processes, the exosphere would show a much smoother response, because of the time delay of the exosphere transport. In fact, ballistic time scale is about 10 minutes after MIV (Mangano et al. 2007) while the exosphere requires some hours to recover after a major impulsive ion precipitation event (Mangano et al. 2013; Mura 2012). However, the predicted close connections linking the ion precipitation with clear changes in the exosphere have been elusive. In fact, ground-based observations by THEMIS telescope suggest that the Na exospheric double peaks can show a variability on a time scale smaller than 1-hour, and possibly shorter-term fluctuations of about 10 minutes (Massetti et al. 2017). On the other hand, MESSENGER observations show an equatorial Na exospheric density almost repeatable from year to year, even if a time variability of some 10s of % not related to orbit or planetographic position, but probably linked to transient phenomena, is clearly registered (Cassidy et al. 2015). The analysis of THEMIS observations recorded during the transit of an ICME at Mercury, registered by MESSENGER (Winslow et al. 2015; Slavin et al. 2014) shows that this event could be put in relation with a variation of the global shape of the Na exosphere, i.e.: at ICME arrival time the two peaks seem to extend toward the equator becoming an exosphere almost uniformly distributed throughout the whole dayside (Orsini et al. 2018). This observation suggests that the magnetospheric compression, could notably affect the whole Na exospheric emission by modifying the cusps extension and driving ion impacts at low latitudes. This agrees with the thick and low-$\beta$ plasma depletion layers observed by MESSENGER during extreme conditions (Slavin et al. 2014; Zhong et al. 2015b). Mercury’s magnetopause reaches the planet’s surface ∼30% of the time during ICMEs (Winslow et al. 2017), but even if the magnetopause was not compressed to the planet surface, the ICME, being rich of heavy ions (Galvin 1997; Richardson and Cane 2004) at large gyroradii, could allow the access of ions to the closed-field line regions (Kallio et al. 2008). Since the heavy-ion sputtering yield is higher than the proton one (Johnson and Baragiola 1991; Milillo et al. 2011), a significant enhancement of surface release could be seen during ICMEs or SEPs (Killen et al. 2012).

Even if the influence of plasma precipitation seems the only way to explain the double peak shape and variability of the Na and K exospheres observed from ground-based telescopes (e.g.: Mangano et al. 2009, 2013, 2015, Massetti et al. 2017, Potter et al. 2006), the ultimate mechanisms responsible for such surface release is not unambiguously identified. In fact, the yields (number of particles released after the impact of a single ion) measured in laboratory simulations in the case of few-keV protons onto a rocky regolith surface are too low to explain the observed Na exosphere (Johnson and Baragiola 1991; Seki et al. 2015). A multi-process action was suggested by Mura et al. (2009) for Mercury’s exosphere generation and by Sarantos et al. (2008) in the Moon’s case, that includes the alteration of the surface properties induced by the impact so that the subsequent action of photons (PSD) is more efficient, but up to now there is no unambiguous evidence supporting it. BepiColombo, having a full suite of particle detectors for the close-to-surface environment and a Na global imager, will allow a unique comprehensive set of measurements of the exosphere and of the ion precipitation in dayside as well as in the nightside (see Sects. 4.6, 4.8 and 4.10).

The nightside surface is also subject to energetic electron precipitation, as suggested by the MESSENGER X-ray observations (Starr et al. 2012). Possible consequences of these impacts, besides the X-ray emission, would be the release of volatile material due to ESD. The observation of the nightside exosphere is particularly difficult by remote sensing UV spectrometers since there is no solar radiation able to excite the exospheric atoms. BepiColombo’s X-ray imager, together with its mass spectrometer, would provide a new important set of measurements (see Sect. 4.8).

The magnetospheric ions of solar wind origin circulating close to the surface may experience another possible interaction with the exosphere that has never been investigated through observations in Mercury’s environment: where the exospheric density is higher, an ion can charge exchange with a local neutral atom. The product of the interaction is a low-energy ion (the previously neutral atom) and an ENA (the neutralised energetic ion) having almost the same energy and direction of the parent ion (Hasted 1964; Stebbings et al. 1964). The collection of the charge-exchange ENA generated along a line-of-sight will provide a threefold information: a) remote sensing of the plasma population circulating close to the planet; b) the signature of exospheric loss; and c) a source of planetary ions (Orsini and Milillo 1999; Mura et al. 2005, 2006). BepiColombo, having two ENA sensors, will provide observations of ENA from different vantage points, allowing a kind of reconstruction of the 3D ENA distribution (see Sects. 4.5 and 4.11).

2.8 What Effects Does Micrometeoroid Bombardment onto the Surface Have on the Exosphere?

The MIV process results in release of solid, melt and vapor from a volume where the meteoroid hits the surface (Cintala 1992). The released vapour leaves the surface with a Maxwellian energy distribution (corresponding to temperature between 1500 and 5000 K), thus the exosphere is refilled with a cloud contributed by the surface material depending on the energy of vaporization, and gravitationally differentiated (Berezhnoy 2018). Models of surface-bounded exospheres have extremely varied estimates of the importance of impact vaporization, and also vary by degree of volatility of the species. Although impact vaporization is an established field of study (Melosh 1989; Pierazzo et al. 2008; Hermalyn and Schultz 2010) uncertainties regarding the importance of impact vaporization on extra-terrestrial bodies include the uncertainty in impact rates for both interplanetary dust and larger meteoroids and comets (Borin et al. 2010, 2016, Pokorný et al. 2018), the relative amounts of melt and vapour produced in an impact (Pierazzo et al. 1995), the temperature of the vapour—which affects escape rates (e.g. Cintala 1992; Rivkin and Pierazzo, 2005), the relative amount of neutral versus ionized ejecta (Hornung et al. 2000), and the gas-surface interaction of the downwelling ejecta (Yakshinskiy and Madey 2005). Finally, the enhanced volatile content of the Mercurian regolith measured by the MESSENGER instruments requires a reanalysis of previous models.

At Mercury, observations of Ca and Mg exospheres seem to indicate clearly that MIV is the primary responsible process of their generation (see Sect. 2.6). In fact, the identified source is located in the dawn hemisphere where higher micrometeoroid impact flux is expected. The Ca column density increases where the 2P/Encke comet meteoroid stream is expected to cross the Mercury’s orbit (Killen and Hahn 2015). The importance of impact vaporization as a source of exospheric neutrals has been constrained in part by observation of the esca** component of the exospheres—the Mercurian tail (Schmidt et al. 2012), the Ca exosphere of Mercury (Burger et al. 2014) and the lunar extended exosphere and tail (Wilson et al. 1999; Colaprete et al. 2016). These results all depend critically on the assumed temperature or velocity distribution of the initial vapour plume, on the assumed photoionization rate and on the interaction of the material with radiation (which is species-dependent). In fact, after release, the material can be subjected to other processes like dissociation or photoionization. Estimates of the Na photoionization rate have varied by a factor of three, and values of the Ca photoionization rate have varied by a factor of about four (e.g. see Killen et al. 2018). This obviously introduces a huge uncertainty in the escape rate, and hence the source process. The quenching temperature of the cloud defines the final constituents of the vapour cloud (Berezhnoy and Klumov 2008; Berezhnoy 2018). The hypothesis of energetic dissociation of atom groups (Killen 2016) considered for explaining the high temperature of the two refractories, Ca and Mg, observed at Mercury exosphere, is unable to fully explain the observed intensity mainly due to uncertainty in the physics of dissociation processes (Christou et al. 2015; Plainaki et al. 2017).

The evaluation of the residence time of the material in the exosphere is an open issue. Colaprete et al. (2016) conclude, based on observations of the UV spectrometer onboard the LADEE mission to the Moon, that released MIV material persists in the exosphere-surface system for much longer than the ionization lifetime. Residence times in the lunar environment of 45 to 90 days (mainly on the lunar surface) can be expected before escape to the solar wind, which would explain the long-term smooth increase and decrease in the Na column density observed as the result of meteoroid streams. This is the result of each particle residing in the regolith for approximately an ionization lifetime (i.e., several days) between bounces, combined with the many bounces that it has to take before being lost from the exosphere. Given the long residence time of Na on the surface deduced by both Leblanc et al. (2003) and Colaprete et al. (2016), it has been suggested that ion or micrometeoroid impacts are the primary source of atoms migrating from the regolith to the extreme surface, and these atoms feed the subsequent release by photons or thermal processes (Mura et al. 2009; Killen et al. 2018). If this is the case, then micrometeoroid impacts play the dominant role in maintaining the exospheres while the less energetic processes such as PSD and thermal desorption serve to keep the atoms in play until they are destroyed by photoionization.

Schmidt et al. (2012) studied the extended sodium tail of Mercury and concluded that both photon-stimulated desorption and micrometeoroid impacts are required to simulate the ∼20% loss of Mercury’s sodium atmosphere, depending on orbital phase, and that the two mechanisms are jointly responsible for the observed comet-like tail as driven by solar radiation pressure. Furthermore, the Na escape rate derived by the observation of the Moon’s distant tail increased by a factor of 2 to 3 during the most intense period of the 1998 Leonid meteor shower (Wilson et al. 1999), evidencing a strong influence of meteor impacts on the lunar sodium exosphere and its escape rate. It was found theoretically that NaO and KO photolysis lifetimes are significantly shorter than its ballistic flight times (Valiev et al. 2020). Planned BepiColombo studies of Na and K content as a function of altitude will be able to estimate properties of photolysis-generated Na and K atoms in the Hermean exosphere (see Sect. 4.9). BepiColombo will offer an unprecedented opportunity to observe simultaneously the exosphere in refractories and volatile components, the Na tail, and the dust environment around Mercury.

Apart from the average exospheric condition, the possibility to register a bigger (1, 10 cm and 1 m) meteoroid impact during the BepiColombo mission has been investigated by Mangano et al. (2007). They concluded that the noticeable increase of some species over the average exospheric density, the amplitude of the resulting cloud, the duration, and the favourable detection probability of the MIV event all work in the direction of positive detections of a cloud from a 10-cm vaporized meteoroid. The vaporized surface hemispherical volume could reach a dimension of meters, depending also on the density and porosity of the regolith; thereby, allowing a kind of remote sensing of the planet endogenous material.

2.9 How do the Surface Composition, Mineralogy and Physical Condition Affect the Surface Release Processes?

Surface mineralogy, composition and grain size determine the thermal capacity and the surface release efficiency. MESSENGER observations revealed many structures specific to Mercury. BepiColombo observations will investigate correlations between the surface composition and any discernible patterns in the ratios of species released from the surface thus, having an effect on the exosphere (see also Rothery et al. 2020, this issue).

A new target, not anticipated until they were discovered by MESSENGER, is hollows (Blewett et al. 2018 and references therein; Lucchetti et al. 2018). These are steep-sided, flat-bottomed depressions where the upper 10–20 m of surface material has been somehow lost. They are clearly young, and probably still active today losing volatile components of the surface material at their edge (Blewett et al. 2018). Any enhancement in exospheric species that can be traced back to fields of hollows would help to illuminate the nature and rate of volatile loss. Other possible sites of enhanced volatile loss that might still be ongoing and which should be checked by BepiColombo include conical ejecta blocks surrounding the Caloris basin (Wright et al. 2019), down-slope streaks on the inner walls of a few impact craters and the Nathair Facula vent (Malliband et al. 2019). High spatial and mass resolution measurements of the exosphere and planetary low energy ions above these regions will provide information on the volatile release and on possible changes still active.

Radar-bright units inside permanently-shadowed polar craters have been identified by ground-based radar observations (Harmon and Slade 1992; Harmon 2007; Harmon et al. 2011; Chabot et al. 2018). They were consistent with water-ice or sulfur or supercooled silicates, the MESSENGER neutron spectrometer observations in the northern regions identified hydrogen-rich substance in favor of a water ice interpretation (Lawrence et al. 2013). Reflectance measurements performed by the MESSENGER laser altimeter in the same regions showed surfaces with albedos distinctly different from that surrounding terrain. A few locations have very high albedo consistent with water-ice, but numerous locations have very low albedo, interpreted to be complex carbon-bearing organic compounds (Chabot et al. 2016; Rothery et al. 2020, this issue). In these regions, the surface properties could be quite different from the rest of the planet. In particular, the backscattering rate of solar wind could be higher thus the space weathering effect could be enhanced. The release efficiency could be higher than on the rocky surface (e.g.: Seki et al. 2015) and presence of organic material synthetized by GCR (e.g., Lawrence et al. 2013; Paige et al. 2013; see Sect. 2.1) could affect also the yield of the release processes. Finally, an effect on the polar exosphere could be expected, in particular the thermal H component is probably higher at the poles.

MESSENGER could not compare the surface features to the exosphere composition measurements since the MASCS observations did not cover the whole planet at high resolution. Nevertheless, the MASCS data analysis proved a connection between the Mg-rich region (Weider et al. 2015) and the observed local exosphere (Merkel et al. 2018). Eventually, the characterisation of the relationship between the surface properties and the exosphere is a task for BepiColombo, which will provide map** of visual, NIR, IR, X and gamma spectra (see Sect. 4.12 and Rothery et al. 2020, this issue).