Killen 2007

Space Sci Rev (2007) 132: 433–509
DOI 10.1007/s11214-007-9232-0
Processes that Promote and Deplete the Exosphere

of Mercury
Rosemary Killen · Gabrielle Cremonese · Helmut Lammer · Stefano Orsini ·

Andrew E. Potter · Ann L. Sprague · Peter Wurz · Maxim L. Khodachenko ·
Herbert I.M. Lichtenegger · Anna Milillo · Alessandro Mura
Received: 2 February 2007 / Accepted: 5 June 2007 / Published online: 30 October 2007
© Springer Science+Business Media B.V. 2007
Abstract It has been speculated that the composition of the exosphere is related to the com-
position of Mercury’s crustal materials. If this relationship is true, then inferences regarding
the bulk chemistry of the planet might be made from a thorough exospheric study. The most
vexing of all unsolved problems is the uncertainty in the source of each component. Histor-
ically, it has been believed that H and He come primarily from the solar wind (Goldstein,
B.E., et al. in J. Geophys. Res. 86:5485–5499, 1981), Na and K come from volatilized mate-
rials partitioned between Mercury’s crust and meteoritic impactors (Hunten, D.M., et al. in
Mercury, pp. 562–612, 1988; Morgan, T.H., et al. in Icarus 74:156–170, 1988; Killen, R.M.,
et al. in Icarus 171:1–19, 2004b). The processes that eject atoms and molecules into the ex-
osphere of Mercury are generally considered to be thermal vaporization, photon-stimulated
desorption (PSD), impact vaporization, and ion sputtering. Each of these processes has its
own temporal and spatial dependence. The exosphere is strongly influenced by Mercury’s
highly elliptical orbit and rapid orbital speed. As a consequence the surface undergoes large
R. Killen ()
Department of Astronomy, University of Maryland, College Park, USA
e-mail: rkillen@astro.umd.edu
G. Cremonese
Osservatorio Astronomico-INAF, Padova, Italy
H. Lammer · M.L. Khodachenko · H.I.M. Lichtenegger

Space Research Institute, Austrian Academy of Sciences, Graz, Austria
S. Orsini · A. Milillo · A. Mura

Istituto di Fisica dello Spazio Interplanetario-CNR, Rome, Italy
A.E. Potter
National Solar Observatory, Tucson, AZ, USA
A.L. Sprague
Lunar and Planetary Laboratory, University of Arizona, Tucson, USA
P. Wurz
Physics Institute, University of Bern, Bern, Switzerland
434 R. Killen et al.
fluctuations in temperature and experiences differences of insolation with longitude. Be-

cause there is no inclination of the orbital axis, there are regions at extreme northern and
southern latitudes that are never exposed to direct sunlight. These cold regions may serve
as traps for exospheric constituents or for material that is brought in by exogenic sources
such as comets, interplanetary dust, or solar wind, etc. The source rates are dependent not
only on temperature and composition of the surface, but also on such factors as porosity,
mineralogy, and space weathering. They are not independent of each other. For instance, ion
impact may create crystal defects which enhance diffusion of atoms through the grain, and
in turn enhance the efficiency of PSD. The impact flux and the size distribution of impactors
affects regolith turnover rates (gardening) and the depth dependence of vaporization rates.
Gardening serves both as a sink for material and as a source for fresh material. This is ex-
tremely important in bounding the rates of the other processes. Space weathering effects,
such as the creation of needle-like structures in the regolith, will limit the ejection of atoms
by such processes as PSD and ion-sputtering. Therefore, the use of laboratory rates in esti-
mates of exospheric source rates can be helpful but also are often inaccurate if not modified
appropriately. Porosity effects may reduce yields by a factor of three (Cassidy, T.A., and
Johnson, R.E. in Icarus 176:499–507, 2005). The loss of all atomic species from Mercury’s
exosphere other than H and He must be by non-thermal escape. The relative rates of photo-
ionization, loss of photo-ions to the solar wind, entrainment of ions in the magnetosphere
and direct impact of photo-ions to the surface are an area of active research. These source
and loss processes will be discussed in this chapter.
Keywords Mercury · Exosphere · Surface composition · Particle release processes
1 Observations of the Mercury Exosphere
Early ground-based efforts to detect an atmosphere on Mercury were unsuccessful, leading

only to successively smaller upper limits for the atmospheric density. The first real informa-
tion about the Mercury atmosphere came with the Mariner 10 flybys (Broadfoot et al. 1976).
Atomic hydrogen, helium and atomic oxygen were detected by the Mariner 10 ultraviolet
photometers from sunlight scattered from these atoms.
A comprehensive discussion of the Mariner 10 measurements is given in “The Mercury
Atmosphere” by Hunten et al. (1988), and the reader is referred to this source for details.
Subsolar point densities were estimated at 6,000 atoms/cm3 for helium, and 230 atoms/cm3
for the thermal component of atomic hydrogen. A nonthermal component of hydrogen with
a scale height of about 70 kilometers was observed near the limb above the subsolar point,
providing a total number density of 23 atoms/cm3 at the surface. Atomic oxygen was de-
tected in the third flyby at a level of 44,000 atoms/cm3 , the number being dependent on
the assumed scale height. This chapter is concerned with Mercury atmospheric observations
made after the Mariner 10 flybys, all from ground-based observations.
1.1 Sodium in Mercury’s Exosphere
In 1985, Drew Potter and Tom Morgan were at the McDonald Observatory of the University
of Texas measuring the spectrum of light reflected from the Moon. They were looking for
the “filling-in” of reflected solar lines caused by interactions of solar radiation with the
lunar surface. It occurred to them that Mercury was similar in many ways to the Moon, and
should show similar “filling-in” effects. The first solar lines they observed on Mercury were
Processes that Promote and Deplete the Exosphere of Mercury 435
Fig. 1 Sodium discovery

spectrum showing D1 and D2
sodium resonance emission lines
within the solar Fraunhofer
absorption lines reflected from
the Mercury surface. Measured at
the Harlan Smith telescope at the
McDonald Observatory of the
University of Texas
the pair of sodium Fraunhofer lines at 5,890 and 5,896 angstroms. They were astonished to
find that the spectra of these features showed bright emission lines that extended well above
the continuum. The wavelengths of the lines identified them as resonance scattering lines of
sodium in the exosphere of Mercury. The discovery spectrum is shown in Fig. 1 (Potter and
Morgan 1985).
The calculation of sodium column content from the spectra first requires that the emission
intensity be extracted from the combination of surface reflection and sodium atom emission.
The common practice has been to interpolate the solar Fraunhofer continuum from one side
of the emission to the other, and subtract it to obtain the sodium emission line. Procedures
that fit the Fraunhofer absorption line with a combination of Voigt profiles give the best
results for defining the Fraunhofer background to be subtracted. However, Hunten and Wal-
lace (1993) pointed out that this procedure underestimates the sodium emission, since about
a fifth of the sodium light is hidden in a depression in the continuum caused from extinction
by the sodium atoms.
Once the sodium emission intensity is determined, two methods have been used to cal-
culate the column content of sodium atoms. The ratio of D2 to D1 emissions is a function
of both temperature and the column content of sodium and the surface albedo (e.g. Killen
2006), so by assuming a value for temperature, the column content can be calculated from
the ratio. This procedure was used to estimate the column content of sodium atoms in the dis-
covery spectrum. However, this procedure is subject to large uncertainties. A better method
is to use the planet itself as a “standard candle” to calibrate the sodium emission, since the
solar continuum reflected from the planetary surface is available in each spectrum. The re-
flectance of Mercury can be calculated using the Hapke formulation for reflectivity with
the optical constants for the Mercury surface. (The most recent set of optical constants was
provided by Mallama et al. (2002).) It is then necessary to take into account the effect of
atmospheric blurring of the Mercury image, which reduces the surface brightness as seen by
ground-based observers.
Since the discovery of sodium emission, there have been a number of additional observa-
tions. Table 1 shows a summary of the observations that have been reported to date. These
observations show that average sodium column densities computed from emission spectra
seen above the planetary surface range between 1–10 × 1011 atoms/cm2 . Column densi-
ties measured from absorption spectra seen during solar transit are smaller, in the range
1–3 × 1010 . However, these densities are measured at the evening terminator, where sodium
densities are normally seen to be small.
Table 1 Column densities of sodium in the Mercury exosphere
Observer Column content, Remarks

atoms/cm2
8.1 ± 1.0 × 1011* Slit spectra

Potter and Morgan (1987) 1.9 ± 1.4 × 1011 Slit spectra
Potter and Morgan (1985) 2.8–3.8 × 1011 Slit spectra
Potter and Morgan (1997) 4.4–6.0 × 1011 5 × 5 image slicer,
peak values
Sprague et al. (1997) Min. <1 × 1010 Slit spectra
equatorial
Max. 15.0 × 1011 Slit spectra
Avg. 1.20 ± 1.49 × 1011 Slit spectra
Sprague et al. (1998) 8.6 × 1011 Northern spot,
slit spectra
8.0 × 1011 Southern spot,
slit spectra
Potter et al. (1999) 0.84 × 2.9 × 1011 10 × 10 image slicer,
average values
Barbieri et al. (2004) 0.43–1.97 × 1011 Slit spectra
Scanning Fabry Perot
Schleicher et al. (2004) 3.4 ± 0.1 × 1010 North pole
3.0 ± 0.1 × 1010 South pole
1.5 ± 0.1 × 1010 West equatorial
<0.2 × 1010 East equatorial
Leblanc et al. (2006) Slit spectra,
terminator low,
solar limb high
Potter et al. (2007) 1.5–3.5 × 1011 10 × 10 image slicer,
average values**
* Calibration from D /D ratio. Other observations, except Schleicher et al. (2004), were calibrated from the
2 1
Mercury reflectance signal
** Sodium seen in absorption. Column content derived from equivalent width
*** Average values from 10 × 10 arc second slicer were scaled to the approximate area of planet
1.2 Temperature of the Sodium Exosphere
The velocity distribution of sodium atoms in the exosphere is an important indicator of the
source of the sodium. Source processes such as photon-stimulated desorption and ion sput-
tering will produce non-Maxwellian velocity distributions (e.g. Madey et al. 1998). Sub-
sequent interaction of the atoms with the surface can relax the distribution back towards
Maxwellian. Thus, analysis of the velocity distribution can provide information on source
Fig. 2 The sodium D2 line

profile for the central disk of
Mercury observed with the
Anglo-Australian telescope on
January 6, 1998 (solid line).
A 1,500 K Maxwellian velocity
distribution is plotted as a dotted
line, an 1,100 K distribution is
plotted as a dot-dash line, and a
750 K distribution is plotted as a
long-dash line. The best fit to the
line profile is given by the
1,500 K distribution (Killen et al.
1999)
processes and surface interactions. The measurable quantity that is related to the velocity
distribution is the spectral profile of the emission line, which can be compared with line
profiles computed assuming combinations of Gaussian velocity profiles for the atomic tran-
sitions. An initial effort was reported by Potter and Morgan (1987), who obtained a some-
what noisy line profile of the sodium D2 emission at a dispersion of 7.2 mÅ per step. They
concluded that the half-width of the line was consistent with a thermalized gas at a temper-
ature of 500 K. A subsequent analysis that took into account the wings of the line emission
suggested that the line profile was better fit by a temperature of about 1,000 K (Shemansky
and Morgan 1991). Measurements at higher resolution and better signal-to-noise were re-
ported by Killen et al. (1999). Line profiles were measured at the McDonald Observatory
of the University of Texas at a dispersion of 1.98 mA per step, and at the Anglo-Australian
Telescope (AAT) at Coonabarabran, Australia at a dispersion of 3.14 mÅ per step. The
line profiles from the AAT were fit assuming a Chamberlain-type exosphere, summing the
Gaussian line profiles for each of the hyperfine levels of the sodium resonance transition. An
example showing three trial fits to the line profile of the central disk of Mercury observed is
shown in Fig. 2.
The best fit to the central disk profile was given by assuming a 1,500 K Maxwellian
velocity distribution. Line profiles from above the north and south pole were best fit by
a 750 K gas. In all cases, the gas temperature appeared to be everywhere hotter than the
surface temperature by 600–700 K. The importance of high signal-to-noise is evident in this
plot, where the differences between profiles for different temperatures are small, becoming
evident only in the wings of the emission profile.
Another approach to measuring the velocity profile is to measure the altitude distribu-
tion of the atoms. The observations by Schleicher et al. (2004) of Mercury in transit across
the Sun measured the altitude distribution of sodium above the terminator. These data pro-
vide direct measures of scale heights and consequently temperatures. They also measured
the width of the absorption line, finding that it corresponded to a Doppler temperature of
3,540 K. This apparent high temperature might be the consequence of sputtering over the
polar regions and/or radiation acceleration, since the view of sodium atoms during a transit
is such that velocity changes caused by radiation acceleration should be detectable (Potter
et al. 2007). Their results are compared with the line profile estimates of scale height and
temperature shown in Table 2. The temperatures from line profiles and estimated altitude
distributions are in fair agreement near the center of the planet, but differ considerably at
Table 2 Temperatures and scale heights for the Mercury sodium exosphere
Observer/author Location Temperature, K
Shemansky and Morgan (1991) Planetary average 1000
Killen et al. (1999) North polar region 750

Center 1500
South polar region 750
Schleicher et al. (2004) North polar region 1380 ± 307

South polar region 1330 ± 307
West edge 1540
East edge 1540
Planetary average 1540
the poles. The temperatures from altitude distribution are almost twice those derived from
line profile data. The difference might be the result of slight differences in location of the
observations near the poles.
The high temperature sodium absorptions in the polar regions reported by Schleicher et
al. (2004) were at a maximum in regions offset towards the dawn terminator by about 15
degrees. In polar regions, 15 degrees away from the dawn terminator, the altitude distribu-
tion is much shorter, and the temperature is obviously much lower. Consequently, the AAT
and transit observations may not have viewed the same regions of the exosphere near the
poles, and temporal variations are certainly important. Recalling that the subsolar surface
temperature of Mercury is about 700 K, and polar temperatures are much lower, the results
summarized in Table 2 show that sodium is produced by an energetic process, and is not
completely thermally accommodated. A planet-wide average temperature of 1,000 K for
sodium corresponds to a scale height of about 97 km. For this scale height, the surface con-
centrations of sodium derived from the column densities of Table 1 would be in the range
1–10 × 104 atoms/cm3 .
1.3 Spatial Variation of Sodium Emission
One of the first things that investigators of the sodium emission noticed was that it was not
uniform over the surface of Mercury and did not conform to the distribution expected for a
classical exosphere. This is illustrated in Fig. 3 by profiles of the sodium emission across
the planet taken with a slit spectrograph (Potter et al. 2006).
A similar set of profiles for the period April 3–6, 1988, were analyzed by Killen et al.
(1990) using radiative transfer theory to make profiles of simple smooth models of a uniform
exosphere. The observations could not be fit by these models. Compared to the models, the
observed sodium emissions were concentrated in the sunward direction with polar enhance-
ments. Several north–south distributions showed more sodium at the north polar regions
than could be explained by a simple model. Furthermore, the distribution changed notice-
ably from one day to the next.
1.4 North–south Emission Peaks
Two-dimensional images of sodium distribution over the surface were obtained by Pot-
ter and Morgan (1990), using a 5 × 5 (arcsec) image slicer. Images from two different
Fig. 3 Profiles of sodium emission (solid lines) and surface continua (dotted lines) across Mercury. The
upper left profile (a) from April 1, 1987, shows symmetric limb brightening at the north and south polar
regions. At the upper right (b), there is a north–south profile from December 3, 1986, that shows excess
emission in the north polar region. At the lower left (c), there is a profile that starts at the dusk terminator and
proceeds to the bright limb. Sodium emission is effectively nonexistent at the dusk terminator, in contrast to
the lower right profile (d), which shows a terminator-to-limb profile for the case that the dawn terminator is
in view. Here, the sodium emission decreases only about 30% as the terminator is approached
Fig. 4 Maps of the sodium D2

emission from Mercury on
July 7, and August 23, 1989
dates (July 7, 1989, and August 23, 1989) are shown in Fig. 4. These maps confirmed the
long slit observations that the sodium is not distributed uniformly over the surface. Excess
sodium emission was seen at both polar regions, plus some excess sodium near the subsolar
point. Almost all the maps showed north–south excess sodium, and a sequence of three im-
ages from July 16–18, 1989, showed variations from one day to the next. The combination
of north–south excess sodium emission and daily variability suggested to the authors that
sodium was produced, at least in part, by sputtering at high latitudes by solar wind particles
precipitated to the surface in the cusps of the magnetosphere.
The images shown in Fig. 4 view approximately opposite sides of the planet. Both images
show emission peaks at north and south high latitudes and a smaller peak in the vicinity of
the subsolar point. The peak values of sodium emission are 3.8 MR and 1.2 MR, respectively.
The difference in peak intensity between the two images is due in large part to the difference
of radial velocity. The July image was taken at a time when the radial velocity was about
−4.7 km/sec, while the August image was taken near aphelion, when the radial velocity was
about 0.69 km/sec. The peak values of atomic sodium density in the two images are similar.
The images were made using a 5 × 5 image slicer having 0.5 slices to yield 5 arc second
square images, each having 10 × 10 pixels of 0.5 . To cover the whole planet, two such
images offset from one another were combined, registered to one another by matching the
surface reflection signals. The directions shown in this figure are those seen by a ground-
based observer, and the terminator is the dusk terminator for the July image, and the dawn
terminator for the August image.
The July image displays longitudes in the range from 270° (terminator) to 360° and 0°
to 47° (limb). This range includes the high latitude radar- and optical-bright albedo spots
B (north) and A (south) (reported by Slade et al. 1992; Butler et al. 1993; Harmon 1997),
as well as the relatively freshly excavated Kuiper-Murasaki crater complex on the eastern
limb. There is excess sodium in the vicinity of these features. The August image displays
longitudes in the range 71° (limb) to 175° (terminator), approximately the opposite side of
the planet from the July image. The sodium-bright spot pattern is similar to that seen for the
July image, but there are no obvious anomalous geographic features in this case, other than
the far eastern edge of Caloris Basin.
Sprague et al. (1997) made an extensive series of Mercury sodium measurements using a
long-slit spectrograph. The slit provided three to five spatial resolution elements across the
planet and, by stepping the slit across the planet, they were able to obtain low-resolution
maps. They found that high-latitude enhancements of sodium emission were common, with
the added characteristic that there was usually more at one hemisphere than the other. They
observed notable sodium enhancements when the spectrograph slits were placed over the
radar-bright spots B and A which are centered near at 355°W longitude and 55°N and 25°S
latitude, respectively. In a repetition of the measurements on October 4, 1996, they again
found sodium enhancements at the approximate locations of the radar-bright spots (Sprague
et al. 1998). They also observed an exceptionally high emission value, corresponding to a
column content of 15 × 1011 atoms/cm2 at a location close to the Caloris Basin. Similar
enhancements close to these locations were also observed by Potter and Morgan (1990), and
the sodium map for July 7, 1989, shown in Fig. 4 includes the region of the radar-bright
spots. Sprague et al. (1994, 1993) proposed that these sodium-bright spots were the result
of a special quality of the radar-bright regions and Caloris Basin—most likely a surface
composition rich in sodium as the result of geologic processes.
Potter et al. (2006) used a 10 × 10 element image slicer having 1 slices to produce
10 × 10 images with 1 pixels. Many of the images showed high latitude emission en-
hancements in one hemisphere relative to the other, and many showed enhancements in
both hemispheres. The ratio of north to south hemispheric emission intensity was calculated
for each image and the results are plotted in Fig. 5 as a function of the longitude of the center
of the surface reflection image. The ratios appear to be random with respect to longitude,
and there is no concentration of excess sodium emission in the northern hemisphere near the
longitude of the Caloris Basin centered at latitude 22° north and longitude W 180°.
A correlation of high or low ratios shown in Fig. 5 with any specific longitude would
suggest that specific sodium-rich areas of the planet were responsible for excess sodium
at that longitude. However, there do not appear to be any strong correlations with specific
longitudes. Images in which bright spots appeared in both hemispheres were of special in-
terest in view of the possibility that the north–south radar-bright spots are sources of sodium
Fig. 5 Ratios of the northern

hemisphere emission to the
southern hemisphere emission
plotted against longitude of peak
brightness in the continuum
reflection image
Table 3 Sodium images that show bright symmetric north–south emission peaks (Potter et al. 2006)
Date Terminator True Radiation Terminator Bright limb

anomaly, acceleration, longitude, longitude,
degrees cm/sec2 degrees degrees
Aug. 31, 1999 Dusk 39.4 166.9 270.1 65.0

Feb. 6, 2000 Dawn 315.9 170.6 89.0 320.5
May 30, 2000 Dawn 81.5 157.8 281.1 181.0
Jun. 3, 2000 Dawn 98.4 135.1 289.9 200.7
Jan. 6, 2002 Dawn 297.7 172.2 85.3 325.0
Nov. 6, 2001 Dusk 68.8 173.1 96.2 225.7
Apr. 9, 2003 Dawn 113.4 113.4 269.3 162.6
emission. A selection of images from Potter et al. (2006) that showed bright spots in both
hemispheres is listed in Table 3. For three of the observations listed in Table 3 (Aug. 31,
1999, Feb. 6, 2000, and Jun. 6, 2000) the radar-bright spots at 355° longitude are in view.
However, they are not in view for the remaining four observations. These images shared
high values of solar radiation acceleration and, in fact, images measured at high values of ra-
diation acceleration almost always showed bright spots in both hemispheres. Ip (1990) mod-
eled the effect of radiation acceleration on the planetary sodium distribution, and showed
that high values of radiation acceleration would push sodium towards the terminator. At
the poles, the terminator and bright limb meet, so that at the poles, we expect to see limb
brightening enhanced by sodium pushed to the terminator by radiation acceleration.
North–south excess sodium has also been observed in absorption against the Sun. Schle-
icher et al. (2004) observed the sodium D2 line in absorption against the surface of the
Sun during the transit of 2003. As shown in Fig. 6, sodium absorptions were observed at
high northern and southern latitudes, in agreement with all the emission observations that
find high latitude concentrations of sodium. This was the case despite the fact that radiation
acceleration was not high, about 13.2% of surface gravity. Sodium absorption is observed
along the western equatorial region, but not along the eastern equatorial region. It is puz-
Fig. 6 Distribution of sodium

D2 absorption around the disk of
Mercury seen in transit across the
Sun. Concentrations of sodium
vapor exist at high north and
south latitudes. Sodium
absorption is seen along the west
equatorial limb (dawn), but not
along the east equatorial limb
(dusk). The true anomaly angle
was 149.2° and radiation
acceleration was 49.3 cm/sec2 , or
about 13.2% of surface gravity.
Figure from Schleicher et al.
(2004)
zling to note that almost no sodium is observed over most of the eastern half of the planet.
Observations of sodium emission do not show much, if any, deficit of emission on the dusk
hemisphere of the planet. The temperatures derived from the altitude distribution of the
sodium absorption are listed in Table 2.
Conclusions from this discussion are: (1) Excess sodium emissions often appear in either
the northern or southern hemisphere, usually both. Sometimes they appear over the bright-
albedo regions A and B. (2) Near the peak values for radiation acceleration, symmetric
north–south spots tend to appear, which may be the result of the pile-up of sodium near
the terminator pushed there by radiation acceleration. (3) North–south excess emissions
sometimes appear even when radiation acceleration is relatively low (see Fig. 4). In addition
to radiation acceleration effects, causes suggested for the appearance of sodium-bright spots
include sodium-rich areas and high-latitude sputtering effects.
1.5 Diurnal Variation of Sodium Distribution
Sprague et al. (1997) reported a diurnal variation in sodium emission, such that the sodium
abundance increased as the sunrise terminator was approached. This is illustrated in Fig. 7
(Fig. 4b from Sprague et al. 1997), where a bar chart shows the average sodium column
abundance as a function of Mercury’s local time. Bin 1 = early morning, 2 = mid-morning,
3 = mid day, 4 = mid afternoon, 5 = late afternoon. By mid- and late- afternoon the Na
abundance is down by close to a factor of 3. Sprague et al. (1997) proposed that this effect
results from evaporation of cold-trapped and ion implanted Na that is released as the sun
rises and heats the surface of Mercury.
The ratio of terminator to limb emission over a range of true anomaly angles was reported
by Potter et al. (2006). To calculate this ratio, sodium emission on the terminator side of
the image was summed, and divided by the sum of sodium emission on the limb side of
the image. For terminator-to-limb distributions where the emission falls to low values at
Fig. 7 The variation of sodium

column content is shown at
hourly intervals past sunrise
(Fig. 4b from Sprague et al.
1997). This bar chart shows that
the column content of sodium in
bins of Mercury’s local time: bin
1 = early morning, 2 =
mid-morning, 3 = mid day, 4 =
mid afternoon, 5 = late afternoon
Fig. 8 The terminator-to-limb ratio variation with true anomaly angle is shown for cases when the dawn
terminator is in view, and when the dusk terminator is in view Potter et al. (2006). The strength of solar
radiation acceleration is plotted for comparison
the terminator (as seen in Fig. 3c), the terminator-to-limb ratio will be less than unity. For
terminator-to-limb distributions where there is appreciable sodium at the terminator (as seen
in Fig. 3d), the ratio will be unity or larger. Figure 8 shows a plot of this ratio against true
anomaly angle. The ratio of radiation acceleration to surface gravity is also plotted.
For dawn-side observations on the “out” leg of the orbit, where true anomaly angles
(TAA) < 180°, the ratios appear to be consistent with the hypothesis that dawn enhance-
ment from evaporation of sodium occurs. The dawn-side ratio is larger than unity, indicating
enhancement of sodium at the terminator. The dawn-side ratios are largest near maximum
radiation acceleration. For values of TAA larger than 140°, the ratios for dawn-side observa-
tions drop down to near-unity values, coincident with the decrease of radiation acceleration.
This change is inconsistent with evaporative enhancement of sodium near the terminator. It is
reasonable to expect that the rate of evaporation of condensed sodium and hence the amount
of sodium vapor near the terminator should depend on the rate of exposure of new surface
at the terminator. In other words, the amount of excess sodium should depend on the rate
of terminator advancement. The rate of terminator advancement is about −0.2 degrees/day
at perihelion, rising to a maximum value of +3.38 degrees/day at aphelion. Consequently,

one expects that the maximum amount of dawn enhancement should be found near aphe-
lion, rather than near maximum radiation acceleration, but this does not occur. For the “in”
leg of the orbit (TAA > 180°), the effect is much less pronounced. There are only a few
dusk-side ratios. On average, the ratio values scatter around unity, with some below unity
near maximum radiation acceleration.
Although some of the observations are consistent with the explanation that dawn en-
hancement occurs from evaporation of condensed sodium, it seems reasonable to expect
that radiation acceleration could also play a role by pushing sodium towards the terminator,
thus producing an excess of sodium there (see Ip 1990). This would explain why the effect
is largest near maximum radiation acceleration.
The transit observations reported by Schleicher et al. (2004) (shown above in Fig. 6),
found sodium emission clearly evident along the western equatorial region, which is the
dawn terminator. Little or no sodium was visible along the dusk terminator on the eastern
equatorial region. This observation is consistent with the existence of a dawn terminator en-
hancement of sodium vapor. Dawn/dusk asymmetries can also be caused by an asymmetric
magnetosphere, with resulting asymmetric ion-sputtering and preferential re-implantation
of photoions on the dawnside (Killen et al. 2004a).
1.6 Variation of Sodium Emission with Time
Sodium emissions sometimes vary in an apparently random fashion. An example was pub-
lished by Potter and Morgan (1990), who observed the distribution of sodium over the plan-
etary disk to change over a three-day period, from February 16 to 18, 1989. A bright spot
appeared in the southern hemisphere on February 17, and disappeared the next day. The
peak emission intensities rose from 2.2 MR on February 16 to 3.4 MR on February 17, and
then dropped back to 2.2 MR on February 18.
Sprague et al. (1997) reported similar behavior. During the October 10–15, 1987 ob-
serving sequence, they saw a dramatic onset of north–south enhancement, followed by its
total disappearance. The sodium abundance at high northern and southern latitudes exceeded
equatorial values by a factor of 2 on October 14, while before and after this date, the abun-
dance was similar at all latitudes.
Potter et al. (1999) reported observations of Mercury sodium over six days during the
period November 13–20, 1997. Daily changes took place in both the total amount of sodium
and its distribution over the planet. The cause of these variations is unknown, but their
episodic nature suggests some connection with solar activity, or heating of the surface, both
of which are known to change over short periods of time. Sprague et al. (1997) were un-
able to find any correlation between changes in sodium emission and the F10.7 cm solar
radio flux. Likewise, Potter et al. (1999) found no correlation between the variations in their
November 1997 data and F10.7 cm solar flux. They did note that there were a number of
coronal mass ejection (CME) events at this time, some of which were directed towards the
general area of Mercury, and suggested a possible connection with the changes observed on
Mercury.
In addition to the random variations, there is a secular trend of sodium emission from one
day to the next. This is illustrated in Fig. 9 (from Potter et al. 2007) which shows a plot of the
planet-wide averaged emission intensity against true anomaly angle. The intensity is seen to
change from day to day, mostly in a regular fashion, and is closely correlated to emission
rate calculated for an individual sodium atom, caused by a variation in the continuum at the
rest frequency of a sodium atom in the exosphere.
Fig. 9 The planetary average sodium emission was measured using a 10 × 10 image slicer over the period
from 1997 to 2003, and is plotted here against the true anomaly angle. Depending on the true anomaly angle,
the average sodium emission can change up to 20% per day. The calculated rate of emission from a single
sodium atom is also shown in the plot (solid line)
The observed emission intensity shown in Fig. 9 tracks the calculated emission very well
except for regions near maximum radiation acceleration. On the “out” leg of the orbit (TAA
< 180°), there is a minimum just past the maximum acceleration at 50°. On the “in” leg of
the orbit (TAA > 180°) there are deviations from the calculated curve both before and after
the maximum radiation acceleration.
In large part, the secular changes of sodium emission intensity illustrated in Fig. 9 are
the result of changes in the intensity of solar radiation in the rest frame of the sodium atoms
as the radial velocity of Mercury changes.
However, changes in sodium column content also affect the emission intensity, and the
deviations noted from the calculated emission curve may be due to changes in the sodium
column content. By normalizing the emission intensities to a constant value of solar radiation
intensity it is possible to separate the changes due to solar radiation from those due to sodium
column content. Figure 10 shows the data from Fig. 9 normalized to the solar intensity seen
by sodium atoms on Mercury at a true anomaly angle of 150.26°.
The maximum values of normalized emission, and hence of sodium column densities,
occurs near perihelion and aphelion, where solar radiation acceleration effects are nearly
negligible. In between perihelion and aphelion, there is a decrease of nearly a factor of
three. The most likely cause for this decrease is the effect of radiation acceleration, which
reaches a maximum as the normalized emission reaches a minimum, most clearly evident
for the “out” leg of the orbit (TAA < 180°).
1.7 Effects of Radiation Acceleration on Sodium Emission
As noted in the previous paragraph, radiation acceleration appears to affect the amount of
sodium visible on Mercury. Previous observations of the effect of radiation acceleration were
inconsistent. Potter and Morgan (1987) found a decrease in column content as radiation
acceleration increased. The average column content of sodium decreased by about 30%
when the radiation acceleration relative to gravity increased from about 15% to 45%.
Fig. 10 Sodium emission intensities have been normalized to the values they would have at a constant
value of solar radiation intensity (that seen by Mercury at a true anomaly angle of 150.26°). The normalized
intensities are proportional to the average column content of sodium, in the absence of any secondary effects
of radiation acceleration. The normalized emission shows maxima near perihelion and aphelion. The ratio
of solar radiation acceleration to surface gravity (solid curve; axis RHS) reaches maximum values near true
anomaly angles of 60 and 300 degrees
Fig. 11 The normalized sodium

emission plotted against radiation
acceleration. Dawn terminator
observations are shown as open
circles and dusk terminator
observations as filled circles.
There is a wide scatter of the
data. A linear fit to the data yields
an R-square value of 0.2
However, Sprague et al. (1997) found no correlation at all between column content and
radiation acceleration. Recently, Potter et al. (2007) used the data shown in Fig. 10 to ex-
amine the effect of radiation acceleration. The data points from Fig. 10 are replotted against
radiation acceleration in Fig. 11. There is no consistent trend with radiation acceleration.
There are some sequences where the normalized emission intensity actually increases with
increasing radiation acceleration.
Potter et al. (2007) found that most of the data scatter was not random, but could be
explained by assuming that the sodium atoms on Mercury were exposed to the accelerating
effects of sunlight for about 1,700 seconds. The effect of radiation acceleration was different
whether the “out” leg or the “in” of the orbit was observed. There is a positive feedback loop
Fig. 12 The normalized sodium observations were corrected for the effect of solar radiation acceleration.
The wide scatter of data points seen in the uncorrected emission values shown in Fig. 11 has diminished.
A linear fit to the data yields an R-square value of 0.015 compared to 0.2 prior to the correction. An initial
decrease is followed by a wide region of near-constant corrected emission, with changes near the end of the
plot at maximum radiation acceleration. Near the end, there is a sharp increase of about 20% for observations
on the “in” leg of the orbit. This effect may be real, or it might be an artifact of the simplified concept used
for the correction
in the “out” leg of the orbit, such that radiation acceleration increases the solar continuum
intensity seen by the atoms, and a negative feedback loop in the “in” leg of the orbit, such
radiation acceleration decreases the continuum intensity. The result of this effect is that the
emission intensity per sodium atom is increased over the non-accelerated value on the “out”
leg of the orbit, and is decreased on the “in” leg of the orbit. The emission values corrected
for this effect showed much less scatter, with a general trend of about 30% to lower values
from minimum to maximum radiation as shown in Fig. 12.
After the correction for the secondary effects of radiation acceleration, the emission in-
tensities should be truly proportional to the column content. To obtain column content val-
ues, the corrected emission intensities were multiplied by 9.712, which is the g-factor at a
true anomaly angle of 150.26°. The results were plotted against true anomaly angle with the
result shown in Fig. 13.
The results are compared with the theoretical predictions of Smyth and Marconi (1995) in
Fig. 13. For true anomaly angles less than 180°, most of the data fall near the line predicted
for perfectly elastic collisions with the surface (β = 0), suggesting that the interaction of
sodium atoms with the surface is weak. The negative change in continuum with increasing
radiation pressure at true anomaly angles greater than 180° causes the intensity per atom to
decrease, so that the emission is actually under-corrected at true anomaly angles between
300 and 330 degrees of true anomaly angle. The curve should in fact be compressed, so that
the atoms have a similar emission rate per atom in the entire region between 330 and 360
degrees of true anomaly angle.
Comparison of column densities at aphelion with those at perihelion is important, be-
cause at these two points, the effect of radiation acceleration is negligible. The column
content at aphelion was larger than at perihelion by a factor of about 1.3, suggesting that the
source processes for sodium do not quite keep up with loss processes as Mercury approaches
the Sun, falling short by about 30%.
Fig. 13 Average column content data computed from the acceleration-corrected data of Fig. 12 is compared
with predictions of Smyth and Marconi (1995). (Note that this average is taken over the area of the entire
10 × 10 arc second image slicer. The angular area of Mercury in view is about a third of this value, so
densities referred to the Mercury surface would be about three times larger.) There are three overlays on this
plot, taken from Fig. 15 of the Smyth and Marconi (1995) paper, in which they took into account radiation
acceleration and the interaction of the sodium atoms with the surface
These results show that radiation acceleration can alter the emission intensity without any
change in the column content of sodium. Consequently, the effect of radiation acceleration
on emission intensities should be taken into account if column densities are to be calculated
from emission intensities.
1.8 The Sodium Tail of Mercury
Models of the Mercury sodium exosphere published by Ip (1986) and Smyth and Marconi
(1995) predicted that radiation acceleration could sweep sodium completely off the planet
into a down-sun tail, depending on the energy of the atoms. Potter et al. (2002b) were able
to observe the sodium tail of Mercury in twilight, mapping it downstream to a distance of
about 40,000 km, as shown in Fig. 14. At that point, the velocity of the sodium had increased
to about 11 km/sec as the result of solar radiation acceleration. The cross-section of the tail
at 17,500 km downstream had a half-width of about 20,000 km, which implied transverse
velocities of sodium in the tail of 2 to 4 km/sec. These velocities imply source velocities
from the planet of the order of 5 km/sec. The total integrated flux of sodium in the tail
was approximately 1023 atoms/sec, which corresponds to 1 to 10% of the estimated total
production rate of sodium on the planet.
Considering that Smyth and Marconi (1995) estimated that atoms with velocities in ex-
cess of 2.1 km/sec could escape the planet into the tail, it appears the solar radiation acceler-
ation can provide this velocity over a considerable range of the orbit. The radial velocity at
which the observed tail disappears would be a measure of the initial velocity of the sodium
atoms, but this would be different on the inbound and outbound legs of the orbit, on account
of the fact that the effect of radiation acceleration is different for the inbound and outbound
legs.
Fig. 14 The sodium tail of

Mercury as observed May 26,
2001. A 10 × 10 arc second
image slicer was used to capture
square sections of the tail
downstream from the planet
1.9 Potassium in the Exosphere of Mercury
Potassium in the exosphere of Mercury was reported by Potter and Morgan (1986). Sodium
and potassium are chemically and physically very similar, so the appearance of potassium
is not surprising.
The discovery spectrum is shown in Fig. 15, where it is seen that the emission line is
very much weaker than for sodium. The abundance of potassium is much less than that of
sodium. In the discovery spectra, the average column content of potassium was estimated to
be about 109 atoms/cm2 , about 1% of the column abundance of sodium. Potassium obser-
vations reported by Sprague et al. (1990) found typical potassium abundances of 5.4 × 108
atoms/cm2 .
When they placed the slit over the longitudes of the “hot poles” where the Sun is over-
head at perihelion, they found enhanced potassium abundance over the Caloris Basin and its
antipode, as shown in Fig. 16.
Sprague et al. (1990) suggested that the potassium enhancements are consistent with an
increased source of potassium from the well-fractured crust and regolith associated with the
fractures in the basin floor and the hummocky terrain at the antipode. Killen et al. (1991)
Fig. 15 Mercury reflectance

spectrum showing emission from
the potassium D1 line at
7698.05 Å, Doppler-shifted from
its rest wavelength of 7698.98 Å.
The stronger D2 emission line
can only be observed when the
Doppler shift is such as to move
it out from under an atmospheric
oxygen absorption line
Fig. 16 Potassium observations

(Sprague et al. 1990) showed an
enhanced column abundance
when the spectrograph slit was
over the longitude of Caloris
Basin and over the antipodal
terrain 180° away. These
longitudes are both under the Sun
during perihelion
Fig. 17 Potassium observations

of Sprague et al. (1990) show
higher abundance in morning
than in afternoon. This is
explained in Sprague (1992) as a
result of K ion implant into
regolith materials during the long
Hermean night with subsequent
release to the exosphere the
following morning. By afternoon,
the source of implanted K ions is
diminished
presented an alternate interpretation of the observation: because the localized sources of

sodium could not be consistently correlated with specific locations on Mercury’s, she noted
that potassium sources may also not be consistently correlated with hermeographic locations
either.
Sprague (1992) noted that potassium observations showed a marked morning/afternoon
asymmetry (Fig. 17), similar to that observed for sodium, and presented a model in which
potassium ions implant on the nightside of the planet where the footprint of the magne-
tosheath intersects the surface. In the morning the potassium is released by heating and
thermal desorption.
Hunten and Sprague (2002) summarized all their data concerning the diurnal variation of
potassium column content, and concluded that potassium would evaporate from the surface
Fig. 18 Same-day maps of the sodium and potassium emission over the surface of Mercury. The sodium
emission is on the left side of the image, potassium on the right. The resolution elements of the sodium image
are 0.5 square, while those of the potassium image are 1 square. The weaker signal from the potassium
exosphere required use of a lower resolution. Both sodium and potassium show excess emission in the south-
ern hemisphere. Sodium also shows emission in the northern hemisphere equivalent to that in the southern
hemisphere, while the potassium emission in the northern hemisphere is weak compared to that of the south-
ern hemisphere. The phase angle was about 90°. The terminator is located at about 85°W longitude, and the
limb at about 348°W longitude. The radar-bright spots are centered near 355°W longitude at about 55°N and
28°S, about 7° towards the center of the planet from the bright limb. True anomaly was about 284° and the
radial velocity was near maximum at −9.8 km/sec. The enhancements for sodium are in the region of the
radar- and albedo-bright spots A (south) and B (north). The potassium enhancement is over bright spot A
at temperatures of 400 K and higher. The chemical and physical similarity of sodium and
potassium suggests that they should behave similarly on Mercury.
A comparison of sodium and potassium images supports this view. Same-day images
of sodium and potassium emissions were measured by Potter and Morgan (1997) over a
five-day period from December 6–10, 1990. Both the sodium and potassium emission were
more intense at high northern and southern latitudes, and varied from one day to the next.
The distributions of sodium and potassium emission over the Mercury surface were similar
(but not identical), supporting the view that they are generated by similar processes. For this
series of observations, the ratio of peak sodium intensity to peak potassium intensity was
about 190. An example of the sodium and potassium images is shown in Fig. 18.
A series of sodium and potassium observations reported by Potter et al. (2002a) showed
that the ratio of sodium to potassium was highly variable. This is illustrated in Fig. 19, where
the ratios of planetary averages of sodium to potassium column densities are plotted against
the potassium column densities.
Table 4 compares the Na/K ratio in different solar system bodies.
The ratio is found to vary from a low of 2 in the Earth’s crust to a high of up to 190 in
Mercury’s exosphere. It is instructive to note that the highest values of this ratio are both
seen in atmospheres (Earth) or exospheres (Mercury), indicative of differential loss. In the
atmosphere of the Earth, the source of sodium is primarily sea salt, which is enriched in
sodium due to differential solubility of Na/K in water. Thus a factor of 3 enrichment in
the atmosphere must be attributed to removal of K, which is the heavier of the two. A low
value of Na/K in the Earth’s crust indicates dissolution of Na salt in the seawater, which
is ultimately subducted to the mantle. In Mercury’s case, we do not know the initial Na/K
Fig. 19 The ratio of sodium to

potassium emission plotted as a
function of potassium column
content. Although the average
ratio of sodium to potassium is
about 100, there are significantly
higher and lower values. The
ratio decreases with increasing
potassium density, suggesting
that sodium and potassium
densities can vary independently
of one another
Table 4 Sodium to potassium

ratios in the solar system Object Na/K Source
Mercury exosphere 40–140 Potter et al. (2002a)

Io exosphere 10 Brown (2001)
Europa exosphere 25 Brown (2001)
Moon exosphere 6 Potter and Morgan (1988)
Lunar crust 7–9 Lodders and Fegley (1998, p. 177)
Meteorites 7–14 Lodders and Fegley (1998, p. 311)
Earth atmosphere 20–150 Gault and Rundle (1969)
Earth seawater 27 Lodders and Fegley (1998, p. 164)
Earth crust 2 Lodders and Fegley (1998, p. 143)
Solar system 15 Lodders and Fegley (1998, p. 80)
ratio in the crust. However, radiation acceleration is higher for potassium than for sodium,
and the shape of the Fraunhofer lines are different. Consequently, the effect of radiation
acceleration on emission intensity will be different for the two species, and that may be the
cause of the variations of Na/K ratio observed. There also may also be ion/wave resonances
that increase the loss rate of potassium ions. At present this is speculation, but measurement
of wave modes by the upcoming missions should investigate these processes.
1.10 Calcium in the Exosphere of Mercury
Sprague et al. (1993) reported an unsuccessful search for calcium emissions at 4226.73 Å,
estimating an upper limit for calcium of 7.4 × 108 atoms/cm2 . A tenuous calcium exosphere
at Mercury at a density of 1–1.5 × 108 atoms/cm2 , principally seen in the polar regions, was
first observed in July 1998, using the High Resolution Echelle Spectrograph (HIRES) at the
W.M. Keck I telescope (Bida et al. 2000). Figure 20 shows results for calcium observations
Fig. 20 Variation of column

abundance and velocity for
calcium as a function of radial
distance from Mercury (Killen et
al. 2005). Data are shown for
three dates: July 19, 1998, June
30, 1999, and June 9, 2000
in 1998, 1999, and 2000. The emission was observed off the disk, but near the poles of
Mercury.
Killen et al. (2005) summarized four years of observations of the calcium exosphere of
Mercury. As seen in Fig. 20, the observations show a persistent but spatially variable blue
shift, indicating an excess velocity toward the observer of up to 3 km s−1 , with an average
excess velocity of 2.2 km s−1 above the south pole. In addition, the line profiles reveal a
hot corona at the equivalent of 12,000–20,000 K in a thermalized atmosphere, indicating a
large range of motion with respect to the observer. The calcium is not confined to the polar-
regions: rare and low Ca abundance is seen in the equatorial regions. Strong emission was
seen anti-sunward on May 3, 2002. Apparent weak emission on the sunward hemisphere
may be due to scattered light from the surface, or may indicate a high latitude source. Sput-
tering and impact vaporization could introduce calcium into the exosphere from Mercury’s
crust.
Killen et al. (2005) suggested that the likely source of the calcium is either impact va-
porization in the form of CaO and clusters, which are subsequently photo-dissociated, or
ion-sputtering of atoms, molecules and ions. The column abundance is somewhat, but not
strongly, correlated with solar activity but data are sparse. Koehn and Sprague (2007) ex-
plore the possibility that the Sun is the primary source of O and Ca in Mercury’s exosphere,
a result of highly ionized atoms of O+6 and Ca+11 delivered by the Sun to Mercury.
1.11 Unknown Exospheric Components and Surface Composition
Despite several unpublished searches for visible and near-infrared emission from other
elements in the Mercury exosphere, none have been found. Sprague et al. (1996) re-
ported a search for lithium in the Mercury exosphere, with negative results. They esti-
mated the upper limit for lithium abundance to be 8.4 × 107 atoms/cm2 . It is expected
that the observed species represent only a small fraction of Mercury’s exosphere, be-
cause at the surface the total pressure, derived from the sum of these known species,
is almost two orders of magnitude less than the exospheric pressure of approximately
10−10 mbar, obtained by the Mariner 10 occultation experiment (Fjeldbo et al. 1976;
Hunten et al. 1988). The surface composition plays a crucial role in the production and com-
position of Mercury’s exosphere. To try to understand the importance of various processes
that may be at work on Mercury one can initially use a suite of several possible surface
compositions to bound the exospheric models.
The properties of Mercury’s surface, especially its composition, age, origin and evolu-
tion, are not known precisely at the moment, due to the lack of sufficient remote and in situ
investigations. Therefore, the surface composition can only be estimated, based on model-
ing and comparison with the Moon. Several scientific efforts have been made to estimate
the surface composition of Mercury using spectral reflectance measurements in comparison
with spectra of analog materials in laboratory studies.
Since the 1960s considerable optical and near-infrared spectra have been obtained, as
discussed by Warell (2003). The measurements were improved with infrared detectors in
the 1980s. The infrared spectra of Mercury, combined with laboratory studies of terrestrial,
Lunar and meteoritic materials, indicate that the rock composition is dominated by feldspars
and low iron pyroxene (Warell and Blewett 2004). Possible Mercurian surface compositions
range from Lunar meteorites up to mixtures of Mercury analog materials such as labradorite
and enstatite.
Burbine et al. (2002) used synthetic Mercury analogs to compare low-FeO anorthositic
compositions with that of partial melts, derived from melting experiments on the EH4
chondrite Indarch (enstatite-rich chondrite). The goal of their work was to relate the
compositions of basaltic partial melts and their residual aubritic materials to that of the
Hermean crust and mantle, respectively. Blewett et al. (1997) in previous experiments
used lunar anorthositic breccia MAC 88105, which is related to lunar meteoroid mate-
rial, as analog to rocks of the Hermean crust. The synthetic Mercury composition used
by Burbine et al. (2002) is depleted in FeO relative to the lunar anorthosite MAC 88105.
However, to produce the observed spectral reddening of the surface by space weather-
ing, surface soils should contain at least a few wt% of FeO in the bulk (Hapke 2001;
Burbine et al. 2002). Mid-infrared spectral studies of Mercury’s surface indicate Na-rich
feldspars and pyroxene (Sprague and Roush 1998) and alkali basalt (Sprague et al. 1994).
In addition, clino-pyroxene was identified (Sprague et al. 2002).
Mercury soil analogs which mirror the spectroscopic observations and range from Lunar
meteorites up to mixtures of Mercury analog materials like labradorite and enstatite can be
summarized as the following:
Synthetic Mercury: Burbine et al. (2002) compared low-FeO anorthositic compositions
with that of partial melts, derived from melting experiments on the EH4 chondrite Indarch
which is an enstatite-rich chondrite. The aim of their work was to relate the compositions of
basaltic partial melts and their residual aubritic materials to that of the Hermean crust and
mantle, respectively.
Lunar anorthositic breccia MAC 88105: This soil composition relates Lunar meteoroid
material to rocks of the Hermean crust (Blewett et al. 1997). The Synthetic Mercury com-
position (Burbine et al. 2002) is depleted in FeO relative to Lunar anorthosite MAC 88105.
However, to produce the observed spectral reddening of the surface by space weathering,
surface soils should comprise at least a few wt% of FeO in the bulk (Hapke 2001).
by75 + en25, la75 + en25 (Bytownite–enstatite mixtures and labradorite–enstatite mix-
tures): Warell and Blewett (2004) investigated several mixtures of Mercury analog materials
by means of visible near infrared (VNIR) reflectance spectroscopy. The best spectral fit was
reached with a mixture of labradorite and enstatite (3 : 1) with about 0.1 wt% of submicro-
scopic metallic iron to simulate spectral reddening due to weathering. The USGS feldspar
sample, which was used in the Warell and Blewett (2004) study.
an75 + en25, ol75 + en25 (Andesine–enstatite mixtures and oligoclase–enstatite mix-
tures): VNIR reflectance spectra show clear spectroscopic features related to electronic ef-
fects among transition elements, but is insensitive for thorough discrimination among pla-
gioclase components. Therefore, the range was extended to andesine- and oligoclase-rich
compositions.
VNIR reflectance spectra show clear spectroscopic features related to electronic effects
among transition elements, but are insensitive for thorough discrimination among plagio-
clase components. Therefore, one cannot exclude andesine- and oligoclase-rich composi-
tions. However, a mixture of andesine and enstatite adjusted to ratios of observed surface
elements in the exosphere may be a good analog for Mercury’s geochemical surface com-
position which contains most likely Si, Ti, Al, Fe, Mg, Ca, Na, K, Mn and O.
The presently observed and various expected exospheric species are shown in Table 5.
Note that some of the reported species (e.g., N2 , O2 , CO2 and H2 O) are just estimated as
upper limits.
One should also note that the composition of Mercury’s exosphere is non-stoichiometric
with respect to the surface composition (Morgan and Killen 1997). The composition and
temporal variability of the exosphere in part help to explain the weathering rate of the sur-
face, and the volatile redistribution rate over both short and long time scales. What resides
on the surface at this epoch is not the pristine surface, but a highly space-weathered surface
that has been overturned by meteoroid bombardment, and desiccated of volatile content by
many processes including photon-stimulated processes, ion sputtering and vaporization. The
relative importance of these processes and their effectiveness at redistributing volatiles ei-
ther to space or to high-latitude cold traps or to the nightside can be addressed by studying
the exosphere (e.g., Leblanc and Johnson 2003).
2 Exospheric Sources and Their Models
2.1 Impact Vaporization
Impacting particles of small sizes (<100 µm) constantly rain onto Mercury’s surface at
a mean velocity of 20 km/s (Cintala 1992), churning the regolith and vaporizing the sur-
face. Larger meteors impact more sporadically, but with higher mean velocity (Marchi et al.
2005). Many minor species, which are refractory during vaporization of silicates in vacuum,
are highly volatile during hypervelocity impacts due to high temperatures and pressures dur-
ing impacts (Gerasimov et al. 1998). Therefore the vapor ejected from impact vaporization
will be the most representative of the surface composition.
Table 5 Expected abundances in

Mercury’s exosphere (Milillo et Species Surface abundance Total Zenith Column
al. 2005) (cm−3 ) (cm−2 )
a Hunten et al. (1988):

H 23; 230a 3 × 109h
He 6.0 × 103a <3 × 1011h
measurements or upper limits
b Potter and Morgan (1997) Li <8.4 × 107n
c Hodges (1974): model O 4.0 × 104 <3 × 1011h
20 Ne 6 × 103 dayc
abundance
d Morgan and Killen (1997): 7 × 105 nightc
model abundances Na 1.7–3.8 × 104a 2 × 1011i
e Morgan and Killen (1997): Mg 7.5 × 103d 3.9 × 1010d
model abundances Al 654c 3.0 × 109d
f Sprague et al. (1993): measured
Si 2.7 × 103d 1.2 × 1010d
upper limit
S 5 × 103d 2.0 × 1010d
g Sprague et al. (1996, 1995):
6 × 105g 2.0 × 1013g
prediction
h Shemansky (1988): Mariner 10 Ar <6.6 × 106a <9 × 1014b
measurements 1.3 × 109k
i Killen et al. (1990): measured K 3.3 × 102b 2 × 109b
abundance 5 × 102h
j Bida et al. (2000) Ca 387d <1.2 × 109d
k Killen (2002): model abundance <239f <7.4 × 108c
l Killen and Ip (1999) 1.1 × 108j
m Huebner et al. (1992) ionisation Fe 340d 7.5 × 108d
rates: experimental (e); H2 <1.4 × 107p <2.9 × 1015p
theoretical (t) for quiet and active O2 <2.5 × 107p <9 × 1014p
Sun N2 <2.3 × 107p <9 × 1014p
n Sprague et al. (1996): model
OH 1.4 × 103d,e 1 × 1010d,e
abundance
p Broadfoot et al. (1976)
CO2 <1.6 × 107p <4 × 1014p
H2 O <1.5 × 107p <1 × 1012c <8 × 1014p
q Cremonese et al. (1997)
In addition, macro-meteors impact Mercury but at an unknown rate. Marchi et al. (2005)
provided the distribution of impact probability as a function of impactor radius, up to objects
of 100 m in radius. In particular, meteoritic impactors coming from the Main Asteroid Belt
are expected to impact on Mercury. The contribution by these larger meteorites to the global
Hermean exosphere is negligible; nevertheless, their impact is expected to produce strong,
localized, but temporary increases in the exospheric density, enriched by material coming
from deeper layers (Mangano et al. 2007). The impact frequency of such objects (especially
in the lower size range) at Mercury is not negligible relative to the nominal duration of the
BepiColombo mission (one year nominal plus one year extension).
Regardless of the size of the impactor, the initial ejecta from an impact will be high-
temperature vapor (∼5,000 K). This will quickly be followed by the “liquid and vapor” at a
slightly lower temperature (2,500 K). However, only the vapor ejected from Comet Tempel
I in the first milliseconds was hot, followed quickly by thermalized vapor. This suggests that
impact vapor may be much cooler than previously supposed.
Impact events probe to a depth of several diameters of the impacting body. Because
meteorite impacts probe much deeper than any process other than venting, and because
the energy density of the process is very high, the exospheric products of this emission
Fig. 21 Density versus time for an impacting object of 0.1 m radius, at 400 km altitude for the species
whose mean density value does not change between day- and nighttime (left); separately, for Na and K
(right). Horizontal lines represent the exospheric background for each species (from Mangano et al. 2007)
Table 6 Vaporization rates for sodium at Mercury due to micrometeoritic bombardment
Reference Orbital Projectile Total Vaporization rate Na

position → target vaporization (atoms cm−2 s−1 ), with
rate f (Na) = 0.005
(g cm−2 s−1 )
Morgan et al. (1988) aphelion combination 1.74 × 10−15 2.27 × 105

perihelion combination 6.42 × 10−15 8.40 × 105
Cintala (1992) aphelion diabase → 8.18 × 10−15 1.07 × 105
regolith
regolith → 7.25 × 10−16 0.95 × 105
regolith
perihelion diabase → 2.75 × 10−15 3.60 × 105
regolith
regolith → 2.46 × 10−15 3.22 × 105
regolith
process most closely represent the surface composition as a whole. Simulations performed
by Mangano et al. (2007), analyze the effects in terms of the gaseous cloud produced by
impacts of objects in the range 1 cm – 1 m. Particularly noticeable is the case of 10-cm
meteor for which the enhancement, depending on the considered species, varies from 1 to
4 orders of magnitude higher than the mean exospheric background values (see Fig. 21).
Figure 21 shows density versus time for an impacting object of 0.1 m radius, at 400 km
altitude.
Durations are generally larger than 2,000 s, and their extension larger than 50° (calculated
with respect to the center of the planet). Estimated vaporization rates from various groups are
summarized in Table 6. A considerable variation in estimated rates results from uncertainties
in the physical state of the surface as well as the impact flux as discussed in the following.
2.2 Interplanetary Space at Mercury’s Orbit
The impact flux onto Mercury (as well as on other planets) is the consequence of several
different physical processes which produce bodies on planetary crossing orbits. Detailed
studies for the Earth have shown that the range of sizes impacting with our planet span over
more than eight orders of magnitude: from µm up to hundreds of meters. There is no reason
to doubt that the same is true also for the other terrestrial planets. Such a flux of material onto
Mercury has several effects, like the formation of craters and the well-known “maturation”
of the soils, for example. The role of the meteoroid flux on Mercury’s exosphere is not well
known.
A small fraction of volatiles released to the exosphere is thought to be produced by impact
vaporization of meteoritic material. The composition of the Hermean exosphere thus reflects
the chemical composition of the surface, and of meteorites impacting Mercury, mixed with
traces of solar wind. Unfortunately, the meteoritic gardening and the impact history of the
Mercury surface is presently unknown because it depends on variables related to the com-
position of the surface and the flux of meteoroids. The meteoroid flux used in literature for
Mercury studies are roughly derived from estimate at the Earth’s heliocentric distance, then
extrapolated to the inner Solar System. It means we may not have a good estimate of the
statistics on the number of impacts and the velocity distribution of the meteoroids. Cintala
(1992) published very nice calculations on meteoroid impacts, but his work was restricted to
sizes less than 1 cm, which are subject to Poynting–Robertson drag. His calculations cannot
be directly extrapolated to larger bodies. Particles having a larger size follow a completely
different dynamical evolution (Marchi et al. 2005).
The radiation pressure force deflects small particles directly antisunward. They may leave
the Solar System after they are ejected from a comet or formed by collision, the exact con-
dition depending on the initial orbital parameters. Particles for which solar gravity amounts
to more than twice the radiation pressure may stay in bound orbits, and they form the main
content of the interplanetary dust cloud. The radiation pressure influences the orbital evolu-
tion of this latter component mainly through the Poynting–Robertson effect. The momentum
transfer caused by radiation falling onto a moving particle includes, when seen in the refer-
ence frame of the Sun, a small component anti-parallel to the particle’s velocity that stems
from the Lorentz transformation of radiation pressure force in the frame of the particle. This
is the case for particles that move in orbital motion about the Sun and are exposed to the
photon flux that is directed radially outward. The small deceleration that is induced by the
anti-parallel component is denoted as the Poynting–Robertson effect. Thus a drift toward
the Sun is superimposed on the motion in Keplerian orbits which limits the lifetime of the
dust particles, just as collisions do.
Although the Poynting–Robertson effect may vary strongly with the size, composition,
and structure of particles, the radial drift of the particles that it causes is small compared
to the orbital velocities. The deceleration of particles by the Poynting–Robertson effect re-
duces the eccentricities and semi-major axes of their orbits. This leads to an increase of dust
number density with decreasing solar distance.
The particles having size larger than 1 cm are not dominated by the Poynting–Robertson
effect and they have to be studied following a different dynamical approach. Most of the
meteoroids arriving on the terrestrial planets come from the asteroid main belt and the main
delivery routes are the well-known resonances. Among them, the most efficient in ejecting
material toward the inner Solar System are the 3 : 1 and ν6 (Morbidelli and Gladman 1998;
Bottke et al. 2002).
2.3 Impacts of Meteoroids
Following the description of the meteoroid fluxes at the Mercury orbit we infer that we
have to take into account two different populations of meteoroids based on their size and
consequently different dynamical evolution.
In the case of the smaller particles Cintala (1992) provided a very good model of mete-
oroid impact vaporization considering objects with radii in the range 10−8 –10−2 m, that has
been used by Cremonese et al. (2005, 2006) to calculate the differential number of impacts
and the mass distribution to obtain the mass of the vapor produced by the impacts. Although
the larger particles have been studied in a specific dynamical model realized for Mercury
(Marchi et al. 2005), the size distribution of impactors on Mercury has been calibrated with
the flux observed on the Earth, for which reliable data are available (Brown et al. 2002).
Indeed, in their numerical simulations, they estimated the ratio of impacts on Mercury
versus the Earth for each projectile size, and they used this ratio to scale the impact rate with
Mercury relative to that observed for the Earth. For this reason the dynamical model has
been obtained for meteoroid in the size range of 10−2 –102 m, but we have to consider here
only those impacts that are relevant for the daily production of the exosphere’s elements that
can affect the ground-based observations, limiting the size range to 10−2 –10−1 m.
The most important result concerning the velocity distribution of large particles is the
wide range of impact velocities on Mercury: the mean impact velocity is about 30 km s−1 ,
but the tails span from about 15 to 80 km s−1 . For comparison, the Moon’s impact distribu-
tions are much narrower with a maximum impact velocity of about 50 km s−1 (see Fig. 22).
Moreover, Mercury’s impact distributions depend on the impactor sizes—that is, on the
simplification f (v, r) = f (v)—used in some works (e.g., Cintala 1992), which does not
hold in this case. To quantify the effects of the impactor sizes, we note that the percentage
of high velocity impactors (defined as those having v > 50 km s−1 ) are 25% and 19% for
r = 1 and 10−2 m, respectively. Note that indeed on Earth it is possible to have impacts
with velocities up to 80 km s−1 , but they are sporadic events related to retrograde swarm of
fragments, presumably of cometary origin. Figure 23 shows f (v, r) onto Mercury’s surface
as a function of the projectile velocity, averaged in the size range 10−2 and 10−1 m, in the
average and perihelion cases.
Fig. 22 Left panel: Impact probability f (v, r), as a function of the projectile velocity, averaged over the large
meteoroids range (10−2 –0.15 and 10−2 –0.10 m for Moon and Mercury, respectively). Right panel: Impact
probability f (v) as a function of the projectile velocity for small meteoroids (10−8 –10−2 m for Moon and
Mercury, respectively)
Fig. 23 Impact probability f (v, r) onto Mercury’s surface as a function of the projectile velocity, averaged
in the size range 10−2 and 10−1 m, in the average and perihelion cases
Since the orbit of Mercury is quite eccentric (e = 0.2) there is some variation from
the mean impact rate along its orbit. According to the model of Marchi et al. (2005)
the distribution obtained for aphelion is almost the same as in the average case. On the
contrary, for the perihelion case, the impact distribution is quite different and the rel-
ative number of high-velocity impacts is about 43% and 33%, respectively for r = 1,
10−2 m. Thus, impacts at perihelion happen at considerably greater velocity than the av-
erage case. Moreover, asymmetries in the rate of impacts onto planets or satellites have
been widely studied for synchronous rotating bodies (e.g., see Horedt and Neukum 1984;
Marchi et al. 2005) and for non-synchronous rotating bodies, like Mercury, the same con-
siderations hold, but the asymmetry is related to the morning-evening (am/pm) hemispheres
instead of to leading-trailing ones.
Following the model of Marchi et al. (2005) in the average case the ratio am/pm is greater
than 1 except for r < 13 cm, while at perihelion the ratio is always am/pm > 1. The increase
of the am/pm ratio with particle size is normal, as already pointed out by Morbidelli and
Gladman (1998). It is due to the more numerous meteoroids having small semi-major axis,
increasing the density of particles inside the Mercury’s orbit, which typically tend to fall on
the morning hemisphere. Also, it is normal that the am/pm ratio is larger for Mercury at
perihelion, because the orbital velocity of the planet is higher, and the planet tends to catch
up the meteoroids, rather than being caught up by them.
In the size range of 10−2 –10−1 m, the dynamical model by Marchi et al. (2005) estimates
at perihelion am/pm = 1.2, while at aphelion am/pm = 0.8. Figure 23 shows the impact
probability f (v, r) in the average and perihelion cases.
The precise calculation of the amount of neutral atoms refilling the exosphere due to the
impacts is not possible. Recent models that calculated the amount of vapor produced during
the impact possibly represent an upper limit. The evaluation of the contribution to the ex-
osphere requires the knowledge of the partitioning of the kinetic energy of the impact in the
elements composing the vapor, allowing inference of the velocity distribution of the neutral
atoms. The models considered by Cintala (1992) and Cremonese et al. (2005, 2006) treat
vertical impacts of spherical projectiles into a regolith, with the shock behavior depending
on the composition of the target and the meteoroid.
The quantity of melted and vaporized regolith produced by each impactor type is a func-
tion of the impact velocity and target temperature. An impact event of sufficiently high
velocity creates what can be visualized ideally as a spheroidal volume, centered below the
impact point, that grades from vapor through a liquid and vapor field into a completely liquid
phase and on until a solid–liquid region merges into highly shocked, unfused target material.
In this context it is clear that a smaller increase in internal energy will be required to initiate
fusion in a hot target than a cold one, simply because the former is closer to its melting
point. According to Cintala (1992) the effects of target temperature are not trivial, but they
are secondary to the role played by impact velocity.
The above-mentioned models have been used to estimate the production rate of sodium,
the main element observed with ground-based telescopes. That of potassium can be inferred
from the ratio between the two atoms, in the composition assumed. It must be borne in mind
that the cause of the high variability of Na/K observed in the exosphere is unknown. In the
following we will report this calculation for the sodium.
Given the assumed velocity distribution of micro-meteors at Mercury, Cintala (1992)
concluded that 1.2 projectile masses of vapor and 8.6 projectile masses of melt would be
produced by a micro-meteor impact onto Mercury, whereas Morgan et al. (1988) calculated
5.4 projectile masses of vapor. As pointed out by Cintala Morgan et al. (1988) used a spatial
density which was a factor or four greater than that used by Cintala, 1.8 g cm−3 , a velocity
distribution with a higher mean impact velocity, and a lower sound speed than recommended
by O’Keefe and Ahrens (1986). Without these differences, Morgan et al. (1988) would have
obtained 0.63 projectile masses of vapor, only half that calculated by Cintala. Given these
unknowns, the exact vaporization rate must be considered to be very uncertain. An addi-
tional factor that must be kept in mind when discussing the results of impact vaporization
calculations is that, for micrometeoritic impact onto a regolith, the impactors are of the order
of µm and the regolith particles are of the order of 100 µm in diameter. One might imagine
that a certain fraction of the energy would go into angular velocity of the regolith particles,
and in addition, that the vapor produced would be largely directed downward, and would
fill voids in the regolith. Therefore our knowledge of the rate and temperature of the vapor
ejected into the exosphere is rudimentary at best.
2.4 Physics of the Evaporation and Production Rate
The volume of target (regolith) material vaporized by a spherical projectile of radius, r, and
impacting velocity, v, can be estimated using the relation of Cintala (1992)
4
Vvap (v, r) = πr 3 c + dv + ev 2 . (1)
3
The constants c, d (km−1 s) and e (km−2 s2 ) depend on target temperature and projectile
composition (Cintala 1992, Table 3, page 952). In these computations the constant values
have been obtained for a diabase projectile and the target at 400 K. Equation (1) was de-
rived by Cintala (1992) for meteoroids smaller than 10−2 m. Cremonese et al. (2005, 2006)
assumed that (1) is valid up to 10−1 m, and for a regolith target with a modal composi-
tion similar to those of the basalt-derived soils from the Taurus Littrow floor (Ahrens and
Cole 1974): pyroxene ∼60%, plagioclase ∼30%, olivine ∼5% and ilmenite ∼0.2%. In do-
ing so the model is based on a regolith target with a modal composition having a higher
plagioclase/pyroxene ratio. It follows that (1) should underestimate the amount of material
vaporized with respect to the model of Cremonese et al. (2005, 2006) using a different com-
position. In fact, the energy needed to vaporize the plagioclase is lower than that needed to
vaporize the pyroxene (i.e., Ahrens and O’Keefe 1972). Therefore, under the same condi-
tions of stress, plagioclase-rich rocks produce more vapor than pyroxene-rich rocks.
The total mass of the vapor produced from the infalling of meteoroids on Mercury’s
surface can be obtained by using (1) given the flux of bodies impacting the planet, Φ, and
their velocity distribution and size distribution

Φ= φ(v, r) · dr · dv, (2)
where φ(v, r) is the differential number of impacts as a function of the meteoroid velocity
and radius. Several authors (Cintala 1992; Cremonese et al. 2005; Marchi et al. 2005) used
the following relation
φ(v, r) = f (v)h(r), (3)
where f (v, r) is the differential velocity distribution of meteoroids (km−1 s), and h(r) is
the differential number of impacts per year and per unit of impactor radius on the entire
surface of the planet (year−1 m−1 ). These functions can be taken from Cintala (1992), for
small meteoroids (10−8 –10−2 m), as follows
3
v √ 2 )+v 2
f (v) = κd 0.2
e−γ d(v 2 −vMEe Ee , (4)
d(v 2 − vMEe
2
) + vEe
2
where κ = 3.81, γ = 0.247 (km−1 s), d is Mercury’s orbital distance in AU, vMEe = 4.25
km s−1 is the escape velocity at the surface of Mercury and vEe = 11.1 km s−1 is the escape
velocity for the Earth at 100 km altitude; and
11
3SM T 4 (i−1)
11
4 i
h(r) = − 3
ici ln ρP πr 3 exp ci ln ρP πr 3 , (5)
4ρP πr F1 i=1 3 i=0
3
where F1 = 0.373 and the constants ci (i = 0, 1, . . . , 11) were reported by (Cintala 1992,
Table A1, page 968), SM is the area of the planet surface, and T is the number of seconds in
1 year. ρP is the meteoroid density, which is assumed to be 2.5 g cm−3 consistent with the
measurements of the densities of stratospheric cosmic dust particles (Rietmeijer 1998) and
with densities data of S-type igneous asteroids (e.g., Krasinsky et al. 2002), which are the
main constituents of the inner part of the Main Belt. Equation (4) simply assumes that the
flux is governed by the gravity field.
In the case of the large meteoroids (10−2 –10−1 m), h(r) was given by (Marchi et al. 2005;
Cremonese et al. 2005) as
a1
h(r) = 1 − a3 exp −a4 r 0.5 , (6)
r a2
where a1 = 1.22, a2 = 3.7, a3 = 0.511, a4 = 0.85. Equation (6) is valid for a large size
range, but the calculations have been limited to an upper limit of 0.1 m, because meteoroids
with radius larger than 0.1 m are not relevant to the daily production of the exosphere.
Figure 24 shows the differential distribution h(r) of number of impacts per year and per unit
of projectile radius in the case of large meteoroids.
The vapor composition is determined by the target compositions impact velocity (i.e.,
Flynn and Stern 1996), then the production rate (atoms cm−2 s−1 ) of the neutral Na, SNa , is
Fig. 24 The differential distribution h(r) of number of impacts per year and per unit of projectile radius (on
the entire surface of the planet) in the case of large meteoroids. The horizontal line corresponds to impacts
that occur with a daily time scale
calculated by the following relation (Morgan and Killen 1997)
fNa
SNa = Mvap NA , (7)
mNa
where Mvap is the vapor production rate (g cm−2 s−1 ), mNa is the atomic weight of Na, NA
is the Avogadro’s number and fNa is the mass fraction of Na in the regolith. Mvap has been
calculated by solving the following equation
vmax rmax
ρ
Mvap = φ(v, r)Vvap (v, r) dv dr, (8)
SM T vmin rmin
where ρ is the target density (1.8 g cm−3 ), vmin = 4.25 km s−1 and vmax = 114 km s−1 ,
rmin = 10−8 m and rmax = 0.1 m.
Assuming that fNa = 0.038, the Na production rate at mean orbit due to impact of me-
teoroids in the entire range considered (10−8 –0.1 m), is 1.82 × 106 atoms cm−2 s−1 , corre-
sponding to 1.36 × 1024 s−1 reported in Fig. 25 as a function of the minimum meteoroid size
(Cremonese et al. 2005).
The contribution of sodium to the exosphere due to large meteoroids, 10−2 –0.1 m, is less
than 1%, and only the 7% of the Na comes from the impacts of meteoroids larger than 10−3
m. The number of impacts on the surface of Mercury are 6.69×10−7 impacts cm−2 s−1 in the
size range of 10−8 –10−2 m and 2.59 × 10−21 impacts cm−2 s−1 in the range of 10−2 –10−1 m
(Cremonese et al. 2005).
To estimate the contribution of the meteoroids to the production of Na the mass of me-
teoroids impacting the surface, for the entire size range, per unit of time and unit of surface
Fig. 25 Cumulative production

rate of the neutral sodium atoms
released in the vapor for the
complete meteoroid size range,
as a function of the minimum
meteoroid radius. Each
production rate value is due to all
the meteoroids having a size
larger than the corresponding
radius in the x-axis. The rapid
decline in the production rate at
larger radius (at about 10−3 m)
underlines the fact that the main
contribution to the sodium
production comes from this
radius range
has been calculated as follows (Bruno et al. 2006)
vmax
F= ψ(v) dv, (9)
vmin
F is meteoritic flux in g cm−2 s−1 , and ψ(v) is the differential meteoritic flux in g cm−2 s−1
(km/s)−1
rmax
4πρp
ψ(v) = r 3 φ(v, r) dr, (10)
3SM T rmin
Figure 25 shows the cumulative production rate of the neutral sodium atoms released in the
vapor for the complete meteoroid size range, as a function of the minimum meteoroid ra-
dius. Cremonese assumed that the atomic fraction of sodium in the regolith is fNa = 0.038
(Cremonese et al. 2005) which yields a source rate of 2.91 × 10−16 g cm−2 s−1 . By assum-
ing that the meteoroids are completely vaporized and have the same regolith composition
(fNa = 0.038), the Na derived from the meteoroid material is 5.3 × 105 atoms cm−2 s−1 . By
adding this quantity to the previous one, they obtained a Na production of ∼2.3 × 106 atoms
cm−2 s−1 if fNa = 0.038, or 3.0 × 105 if fNa = 0.005. This value of sodium production rate
calculated by Cremonese et al. (2005, 2006) is higher than those reported by Hunten et al.
(1988), 1.34 × 104 atoms cm−2 s−1 , but lower than that reported by Leblanc and Johnson
(2003), 6.7 × 105 atoms cm−2 s−1 . It is in fair agreement with the upper limits given by
Morgan et al. (1988), 1.9 × 106 atoms cm−2 s−1 . Killen et al. (2001) used fNa = 0.005, and
obtained a sodium source rate of 2.2 × 106 atoms cm−2 s−1 at perihelion and 1.15 × 105 at
aphelion. Normalizing to the same composition, Cremonese et al. (2005) would have ob-
tained a source rate for Na of 3.0 × 105 at mean orbit, in good agreement with the Killen et
al. (2001) results.
3 Exospheric Sources and Their Models
3.1 Photon and Electron Stimulated Desorption of Surface Elements
In addition to the solar wind, plasma related to Coronal Mass Ejections (CMEs), cosmic rays
and solar energetic particles, Mercury’s surface environment is continuously bombarded by
solar radiation and electrons. Solar radiation in the form of infrared and visible and ener-
getic photons is absorbed by the surface and causes heating and alteration of the dayside
surface area. This heating can reach dayside temperatures on Mercury of up to about 700 K
at the planet’s equator, which is relevant for thermal desorption. Moreover, solar photons
with energies ≥4 eV can induce bond-breaking, which is important for photon-stimulated
desorption (PSD) of absorbed elements. Lower energy electrons in the order of tens of eV
are also important in addition to surface charging because they can cause electronic exci-
tations that also lead to electron stimulated desorption (ESD) of adsorbed elements from
the planetary surface. In the following subsections we describe these processes and their
expected role in refilling Mercury’s exosphere in more detail.
3.2 Photon-Stimulated Desorption
Photon-stimulated desorption (PSD) corresponds to the desorption of surface elements as a

result of electronic excitation by a photon of a surface atom. Madey and Yakshinskiy (1998)
showed that this process efficiently ejects alkalis from surfaces under laboratory conditions.
Yakshinskiy and Madey (2004, 1999) found from their laboratory experiments with Na-
covered lunar basalt samples that UV photons with energies of about 3–5 eV or greater than
5 eV cause desorption of “hot” Na atoms. This process acts through electronic transitions
such as band gap excitation, valence electron excitation, or core excitation. Near-UV pho-
tons with energies ≤4 eV caused little or no detectable desorption of Na. Yakshinskiy and
Madey (2004) estimated the PSD cross-section σ at photon energies of ≈5 eV to be about
10−20 cm2 , which is about seven times larger than that used by Killen et al. (2001). They
found from their experiments that the desorbed Na atoms are suprathermal with a velocity
peak in the PSD distribution of about 900 m s−1 . It was also discovered that desorption of Na
varies with surface temperature and increased by a factor of 10 after the sample was heated
from about 100 K to about 470 K (Yakshinskiy and Madey 2004). Cassidy and Johnson
(2005) estimated that desorption from a regolith is reduced by about a factor of about three
from that on a flat surface.
Bombardment of lunar silicates by UV photons (λ < 300 nm) was found to produce effi-
cient desorption of Na atoms (Yakshinskiy and Madey 1999). The flux of atoms of species
X desorbed by PSD can be given by

φX = fX NS φph (λ)QX (λ) dλ, (11)
1
PSD
φNa = φph QNa fNa NS , (12)
4
where φph is the solar UV flux at Mercury, QX (λ) is the PSD cross-section for species X at
wavelength, λ, fX is the fraction of species X in the regolith, and Ns is the total regolith sur-
face density in number of atoms cm−2 /mean free path. The experimental PSD cross section
for Na has been given as QNa = 1–3 × 10−20 cm2 , integrated over the effective wavelength
range, 250–400 nm (Yakshinskiy and Madey 1999). However, the actual yield in a regolith
Table 7 Solar UV flux at Earth’s orbit compared with periherm (0.29 AU) and apoherm (0.44 AU)
Solar photons 1 AU [cm−2 s−1 ] 0.44 AU [cm−2 s−1 ] 0.29 AU [cm−2 s−1 ]
UV-A (3.1–3.9 eV) 2 × 1016 1 × 1017 2.4 × 1017

UV-B (3.9–4.4 eV) 2.5 × 1015 1.3 × 1016 3 × 1016
UV-C (4.4–12.4 eV) 1 × 1014 5.15 × 1014 1.2 × 1014
is reduced, perhaps by a factor of three (Cassidy and Johnson 2005) due to the possibility
that ejected atoms will stick to grain surfaces before they can emerge from the regolith. The
regolith surface number density is often assumed to be Ns = 7.5 × 1014 cm−2 /MFP, where
MFP is the photon mean free path. The solar UV flux at Mercury integrated from 100–318
nm is φph = 3.31 × 1015 cm−2 s−1 /Rm2 , where Rm is Mercury’s orbital distance from the
Sun in AU. The sodium fraction in the lunar regolith is 0.0053, and this value has been
extensively assumed for the sodium fraction in Mercury’s regolith.
Photon-stimulated desorption is induced by electronic excitations rather than by thermal
processes or momentum transfer. However, a temperature dependence in the yield was found
and attributed to diffusion rates to the extreme surface (Yakshinskiy and Madey 2004). The
velocity distribution of emitted atoms can be described by a Weibull distribution, which has
a high-velocity tail. The energy distribution of the emitted atoms from electron stimulated
desorption (ESD) has been given by Johnson et al. (2002) as
EU β
fPSD (E) = β(1 + β) , (13)
(E + U )2+β
where E is the energy of the emitted particle, U is the characteristic energy for PSD of a
given species, and β is the shape parameter of the distribution. The shape parameters for Na
and K were given by Johnson et al. (2002) as βNa = 0.7 and βK = 0.25. The peak of the Na
velocity distribution from a PSD source is similar to that for a 1,100 K gas (Yakshinskiy and
Madey 1999), corresponding to a U = 0.0098 eV in (13).
The temperature dependence for PSD, as given by the Yakshinskiy and Madey data,
was fit by the following expression (1.1448E-5T 2 − 0.00163T + 0.02128)/1.8 (Killen, in
preparation) where the maximum surface temperature used in this correction is 475 K. The
desorption rate from a plane surface probably differs from that of a regolith due to the
probability that a desorbed atom in a regolith will collide with a surface element before
reaching the extreme surface (Cassidy and Johnson 2005). PSD is not effective in ejecting
refractory species.
To investigate the PSD-induced release of atoms from the surface of Mercury, the varia-
tion of solar UV photons incident on the planetary surface over the planet’s orbit have to be
considered. The solar flux at Mercury’s eccentric orbit differs substantially from the average
condition present at 1 AU. Table 7 shows the solar UV-A, UV-B and UV-C flux at periherm
of about 0.29 AU and apoherm at about 0.44 AU. One can see from Table 7 that the solar UV
flux is about a factor 11 higher at the periherm and more than five times higher at apoherm
than that at the Earth’s orbit in 1 AU.
Lammer et al. (2003), studied the PSD-induced release of Na and K atoms along Mer-
cury’s orbit. The flux of desorbed Na atoms by incoming solar UV photons could be cal-
culated by using (12) (Yakshinskiy and Madey 1999; Lammer et al. 2003). The study of
Lammer et al. (2003) showed clearly that the largest PSD fluxes of released Na occur near
equatorial latitudes at periherm, where the flux could reach values depending on used PSD
Fig. 26 The left panel shows the solar UV irradiance at Earth’s orbit during low solar activity, while the right
panel shows the solar UV irradiance at Earth’s orbit during high solar activity where the irradiance could
be up to 100 times higher than during quiet solar periods (M. Schoeberl and H. Mitchell, UARS/SUSIM,
NASA-GSFC/SVS)
cross-section of about 4.5 × 106 − 3.15 × 107 cm−2 s−1 . At apoherm the Na flux values at
the equatorial regions are about three times lower.
The PSD-induced Na fluxes at latitudes higher than 75° are lower. Lammer et al. (2003)
concluded that there should not be any noticeable PSD sources at Mercury’s polar areas.
The PSD-induced Na fluxes at the equatorial regions at apoherm and periherm of Lammer
et al. (2003) are lower than the estimated fluxes of McGrath et al. (1986) of about 2.0 ×
107 –2.0 × 108 cm−2 s−1 but larger than the estimated average flux value of about 2.0 × 107
cm−2 s−1 by Killen et al. (2001) and Killen and Ip (1999). A reason for the larger fluxes
estimated by McGrath et al. (1986) could be that the fluxes estimated by these authors are
overly optimistic since they were based on data for alkali halides (Killen and Morgan 1993).
By estimating the PSD-induced Na source rates one obtains between periherm and apo-
herm values in the order of about 1.0 × 1024 s−1 , which is in agreement with the observations
by Killen et al. (2001) who found 7.6 × 1023 s−1 for November 13 and 1.4 × 1024 s−1 for
November 20, 1997.
However, as one can see in Fig. 26 the solar UV flux increases up to about 100 times
from low solar activity to active solar periods or during flare events. Therefore, one may
expect that the Na PSD flux could reach values during high solar activity periods or flare
interaction with Mercury of the order of about ≥108 cm−2 s−1 .
A second element that can be desorbed from Mercury’s surface due to solar UV radiation
is K. Laboratory experiments by Madey et al. (1998) regarding the desorption of alkalis on
oxide surfaces yield PSD cross-sections for K atoms which vary between 1.4 ± 0.6 × 10−20
cm2 and 1.9 ± 0.8 × 10−21 cm2 for wavelengths between 247.5 nm (5.0 eV) and 365 nm
(3.5 eV). The most efficient cross-section in their experiments is about 1.8 × 10−20 cm2 at
253.7 nm (4.9 eV). The exospheric observation by Potter and Morgan (1997) give an upper
limit to Na/K ratio of about 200, which corresponds to PSD-induced K fluxes in the order
of about 104 cm−2 s−1 for K atoms corresponding to the observed Na/K fractionation along
a latitude strip that is directly facing the Sun. Because the Na/K ratio is extremely variable
in the exosphere (Potter et al. 2002a) it is likely that loss rates play a role as well as source
rates.
However, note that real regolith on Mercury’s surface is different from the material stud-
ied by Madey et al. (1998). Since it has been irradiated, the alkali binding could be altered,
the porosity of the material is unknown and sticking could be an efficient process. These
factors can cut down the cross-section by more than an order of magnitude. Further, the
surface layer can be depleted in alkali (Hapke 2001; Madey et al. 1998).
3.3 Electron-Stimulated Desorption
In addition to PSD experiments with adsorbed Na and K atoms on lunar silicates, Yak-
shinskiy and Madey (1999, 2004) also studied electron-stimulated desorption (ESD). They
found that exposure of Na covered surfaces by low energetic electrons with energies from
3–50 eV causes also desorption of “hot” Na atoms. Generally there are intimate connections
between ESD and PSD because similar electronic processes cause desorption of the atoms
via electron or photon excitation. The release of Na via ESD is strongly temperature depen-
dent. The observed average ESD cross-section for Na atoms and 10–50 eV electrons is about
1–2×10−19 cm2 (Yakshinskiy and Madey 1999, 2004). The experimentally determined ESD
cross-section for atomic Na has its initial threshold at about 4 eV, which is comparable with
the PSD threshold. Furthermore, the desorption cross-sections have a similar magnitude for
electron energies of about 5 eV and there is a resonance like feature at about 11 eV (Yak-
shinskiy and Madey 1999). If one assumes quasi-neutral solar wind plasma, then electron
fluxes for electrons with energies of about 12 eV are about 5 × 109 – 2 × 109 cm−2 s−1 at
periherm and apoherm, respectively. The flux of adsorbed elements like Na or K due to ESD
can be calculated by
ΦESD = ΩΦe σESD f, (14)
where Ω is the area where the electrons can reach the planetary surface divided by the
whole planetary surface, Φe is the electron flux at Mercury’s environment, σESD is the ESD
photon cross-section and f is the composition of the regolith abundance for instance of Na
in Mercury’s surface. Because Φe is several orders of magnitude smaller than the photon
flux Φv , ESD during ordinary solar wind conditions will not be an efficient release process
for adsorbed surface elements like Na or K compared to PSD, sputtering or micrometeoroid
evaporization.
Leblanc et al. (2003b) considered a particular SEP event with particle energies larger
than 10 keV. The event was observed in detail at the Earth’s orbit (Reames et al. 1997)
and rescaled to Mercury’s orbit. Generally SEPs can reach Mercury before or few hours
after the arrival of the shock and a magnetic cloud usually associated with a Coronal Mass
Ejection (CME) that is for an unperturbed Mercury’s magnetosphere for quiet solar wind
conditions. Leblanc et al. (2003b) studied test particles representative of the energy flux
distribution for each SEP ion species and for the electrons. They launched the particles from
the magnetopause and followed them inside Mercury’s magnetosphere and surface by using
a magnetospheric model of Luhmann et al. (1998). They found that these particles can cause
ESD of Na atoms and, because they penetrate more than the solar wind ions or UV photons,
they can enhance the supply of Na atoms to the surface where it can be desorbed by thermal
or PSD.
3.4 Particle Surface Sputtering
The impact of energetic ions on a solid surface (e.g., Mercury’s surface) will cause the
release of particles via momentum transfer, which is called sputtering, or more precisely
physical sputtering. Particle sputtering will release all species from Mercury’s surface into
space, reproducing more or less the local surface composition on an atomic level. Prefer-
ential sputtering of the different elements of a compound will lead to a surface enrichment
of those elements with low sputtering yields in the top-most atomic layers. However, the
steady-state composition of the flux of sputtered atoms will reflect the average bulk com-
position. Thus, particle sputtering, when operative, will give us compositional information
about the refractory elements of the bulk surface.
The normalized energy distribution for particles sputtered from a solid, f (Ee ), with the
energy Ee of the sputtered particle, has been given as (Sigmund 1969)

6Eb Ee Ee + Eb
f (Ee ) = √ 1− , (15)
3 − 8 Eb /Ec (Ee + Eb )3 Ec
where Eb is the surface-binding energy of the sputtered particle and Ec the cut-off energy
for sputtered atoms. The cut-off Ec , which is the maximum energy that can be imparted to a
sputtered particle by a projectile particle with energy Ei , is given by the limit imposed by a
binary collision between a projectile atom, m1 , and the target atom, m2 , (to be sputtered) as
4m1 m2
Ec = Ei . (16)
(m1 + m2 )2
Figure 27 shows the normalized energy distribution for several elements. Note that the max-
imum of the energy distribution is at Emax = Eb /2, with Eb ranging from fractions of an eV
to several eV depending on species and mineral/matrix. At higher energies the distribution
falls off with Ee2 until the energy Ee approaches the cut-off energy Ec .
The polar angle distribution of sputtered atoms, f (α), for polycrystalline surfaces is
best described by a quadratic angular dependence,f (α) ∝ cos2 α for laboratory experiments
(Hofer 1991). By modeling the details of sputtering in loosely packed regolith grains, Cas-
sidy and Johnson (2005) found that for a fine-grained and porous regolith a better choice is
f (α) = cos α. For the azimuth angle a uniform distribution over 2π is a suitable description.
Having the energy, the azimuth, and elevation angle one can calculate all three components
of the initial particle velocity, v, and the trajectory of each sputtered particle in the exosphere.
Using many such trajectories a vertical density profile Ni (h) can be calculated (Wurz and
Lammer 2003; Wurz et al. 2007). The density profile can easily be integrated to obtain the
column content, which is the typical measurement obtained from telescopic observations
of the exosphere. Either the exospheric density at the surface or the column content can be
used to compare with observational data. The flux Φi of atoms sputtered from the planetary
Fig. 27 Normalized energy

distributions for sputtered O, Si,
Ca, and Fe atoms by impact of
protons of 1 keV energy using
(1). The symbols indicate the
energy corresponding to the
escape velocity of each sputtered
atom
surface can be calculated as
Φi = Φion Yitot = Φion Yirel Ci , (17)
where Φion is the energetic ion flux onto the surface and Yitot the total sputter yield of
species i; that is, the number of released surface atoms per incoming ion, from the sur-
face with a given elemental composition. The total sputter yield Yitot can be broken up into a
relative sputter yield Yirel and Ci the atomic abundance of species i on the surface. The total
sputter flux of species i can also be written as
Φi = Ni (0) vi , (18)
with Ni (0) the exospheric particle density at the surface (h = 0), and vi the average ve-
locity of sputtered particles. Combining (17) and (18), the exospheric density at the surface
resulting from the sputter process for species i is
1
Ni (0) = Φion Yitot , (19)
vi
Ni (0) can be used in a calculation as a starting point to derive a quantitative density profile
from the sputtering process for a given surface composition. The average release velocity is
derived from the sputter distribution, (1), as

3π 2Eb,i
vi = , (20)
4 m2
where Eb,i is the binding energy of species i in the particular chemical/mineralogical mix
of the surface (Wurz et al. 2007). Note that the most probable velocity is vmp = Eb,i /m2 ,
which is lower than the average release speed by a factor of about 3.3. These velocities
have to be compared to the Hermean escape speed of 4.250 km s−1 . If we take oxygen as
an example, with a binding energy of Eb = 2.0 eV, we get vi = 11.57 km s−1 , which
exceeds the escape velocity considerably. The same is true for other elements. Thus, many
sputtered atoms escape Mercury’s gravity field. This can been seen in Fig. 27 where the
energy corresponding to the escape speed is indicated in the energy distribution of sputtered
atoms.
Therefore, if the flux of ions impinging the planetary surface, Φion , is known one can
calculate ab initio, with the sputter yields Yitot for a particular surface composition, the sput-
tered flux, the surface density, the density profile, and the column content and compare these
numbers with the observations. This has been done for the Moon recently (Wurz et al. 2007).
In addition to atoms, clusters of two and more atoms can be released from the solid
surface via sputtering. For metallic surfaces the release of metallic dimers, trimers, etc.
with yields of about 10−1 , 10−2 , . . . with respect to the atomic sputter yield, has been
observed in the laboratory (Gnaser and Hofer 1989; Wurz et al. 1990, 1991; Hansen et
al. 1998). For oxide surfaces, which are more representative to the mineralogical surface
of Mercury, one observes monoxides like SiO or CaO and larger oxide molecules, with
yield ratios up to [Me/MeO] ∼ 1 (Oechsner et al. 1978; Wucher and Oechsner 1986;
Wurz et al. 1990). The sputter yield of such dimers (metallic and oxides) depends strongly
on the chemical environment on the surface, mostly on the surface being oxidized or
metallic. This has been investigated by performing sputter experiments on clean metallic
surfaces that were oxidized in a controlled way (Wurz et al. 1990; Hansen et al. 1999).
It was found that the oxide molecule yields correlated strongly with the oxygen cov-
erage, with the sputtered monoxide yield as large as the atom yield at maximum oxy-
gen coverage, and the metal clusters anticorrelated with the oxygen coverage. The en-
ergy distribution of sputtered clusters is similar to the energy distribution of sputtered
atoms, as given in (1) (Wucher and Oechsner 1986; Hansen et al. 1998), but the fall-
off at higher energy of sputtered particles is f (Ee ) ∝ Ee−n with n > 2 (Coon et al. 1993;
Betz and Husinsky 2004).
Since sputtering is a quite energetic process sputtered clusters have temperatures of sev-
eral 1,000 K when released, which limits their stability against falling apart. So far clusters
of sputtered atoms have not been observed directly in Mercury’s exosphere, but it has been
proposed that a significant contribution to the Ca exosphere arises from sputtered CaO mole-
cules that fall apart at high altitudes (Killen et al. 2005).
A fraction of the sputtering atoms are positive or negative ions. The ion fraction of the
sputtered atoms can be in the range between 10−4 and a few 10−1 , depending on element,
matrix, and primary ion (Benninghoven et al. 1987). Sputtered ions were used for compo-
sitional analysis of ground-up sample simulants of mare and highland soils to mimic the
sputtering behavior of lunar regolith (Elphic et al. 1993). Because of the lack of an at-
mosphere and ionosphere for Mercury, sputtered ions immediately are picked up by the
electro-magnetic fields of Mercury’s magnetosphere or escape with the solar wind.
Large areas on Mercury’s surface are exposed to solar wind even during regular solar
wind conditions, and solar wind ions are the most important ion population causing sput-
tering from Mercury’s surface. Solar wind velocities are in the range of 300 to 800 km/s
(slow and fast solar wind), which translates to energies of 0.5 to 3.3 keV/amu, and with a
typical value of 1 keV/amu. Note that the sputter yield has a maximum around ion energies
of 1 keV/amu. In the solar wind, protons and alpha particles make up more than 99% of the
ions, and heavy ions (from carbon to iron and up) together are about 0.1% of the solar wind
ions in the number flux (Wurz 2005, and references therein).
For the lunar surface total sputter yields are about 0.07 surface atoms per impinging ion
for the typical H/He mix of solar wind ions (Wurz et al. 2007). Heavier ions have sputter
yields even larger than 1, but because their abundance in the solar wind is very low their
contribution to the total sputter yield of the solar wind is negligible. The He/H ratio in the
normal solar wind is about 0.04 (Aellig et al. 2001), with He abundance varying from about
0.02 at solar minimum to 0.06 at solar maximum. For solar wind speed below 350 km/sec
the average He/H is 1.8%. Increasing solar wind speed implies increasing He abundance in
the solar wind, where He/H = a + bvsw , where a and b depend on the solar cycle (Aellig et
al. 2001).
During coronal mass ejections (CMEs) the He/H ratio can be enhanced to about 0.2 in
a CME (Sarantos et al. 2007), which will increase the sputter yield accordingly. Moreover,
the abundance of heavy elements can also be significantly increased in CMEs with respect
to undisturbed solar wind (Wurz et al. 2001, 2003), which may become important for the
sputter yields. Heavy ions are also highly enriched in SEP events (e.g., Wiedenbeck et al.
2005; Cohen et al. 2005); however, at these energies the sputter yields are very low since the
particles penetrate far into the solid.
It is expected that the sputter yield on Mercury’s surface will be about the same as that
on the moon. The sputter yield has to be reduced by the porosity of the surface (Cassidy
and Johnson 2005), thus we estimate that the actual sputter yield for Mercury’s surface is
between 0.02 and 0.025 surface atoms per impinging ion for the typical mix of solar wind
ions.
The heavy ions in the solar wind are highly charged because of the million-degree hot
solar corona. Oxygen, for example, is present in the solar wind with charge states of typi-
cally +6 and +7; iron is present with charge states in the range from +8 to +12. These high
charge states mean that the ions have high internal energies (potential energies), for exam-
ple, 295 eV for O6+ and 1055 eV for Fe10+ , as compared to singly or doubly charged ions.
The charge state affects the sputtering yield due to the potential (ionization) energy avail-
able, hence the term “potential sputtering”. However, these high internal energies (potential
energies) have to be compared to their kinetic energies in the solar wind of typically 16 keV
for oxygen and 56 keV for iron. It has been argued that the sputter yield for highly charged
ions impacting on a planetary surface is increased by a factor of 10 to 1,000 as a result
of their high internal energy (Shemansky 2003). The laboratory data on sputter yields for
highly charged ions have been reviewed by Aumayr and Winter (2004), and we will briefly
summarize their findings here. For metallic surfaces and semiconductors (Si and GaAs) no
deviation of the sputter yield for highly charged ions from the sputter yield of singly charged
ions was found, with the highest charge states investigated being Ar9+ and Xe25+ . Moreover,
all the measured sputter yields agree with the TRIM calculations, a software package which
considers only the kinetic energy of the impacting ion (Ziegler and Biersack 1985).
For ionic crystals (NaCl and LiF) a pronounced increase with ionic charge state was
observed; for NaCl the sputter yield increased by a factor of 4 for Ar8+ ions compared to
Ar+ ions, for LiF the sputter yield increased by a factor of 25 for Ar14+ ions compared to
Ar+ ions. Note that Ar charge states in the solar wind range from +8 to +11.
For oxides, which are the best analog for Mercury’s surface, a clear signature of a sput-
ter yield increase for highly charged ions was observed for SiO2 and Al2 O3 . For SiO2 this
increase was about 3 for Ar8+ ions compared to Ar+ ions, and about 65 for Xe25+ ions
compared to Ar+ ions. Similar enhancements were found for the Al2 O3 surface. Measured
sputter yields of 1.5 keV Xeq+ onto Al2 O3 show an approximately 40-fold increase in the
sputter yield for Xe28+ over that of Xe9+ . Both of these materials appears to have a finite
sputter yield at zero kinetic energy of the projectile. On the other hand, for a highly ionic ox-
ide such as MgO, even though potential energy greatly increases the sputter yield, potential
energy does not induce sputtering in the absence of kinetic energy of the projectile.
However, this enhancement is strongly depending on the ion dose the surface has been
exposed to. After a removal of about a monolayer from the oxide surface the sputter yield
for highly charged ions drops to about the values for singly charged ions. Removal of a
monolayer of surface material corresponds to a heavy ion flux of a few 1013 ions cm−2 s−1
at solar wind energies, which takes about two weeks at Mercury’s orbit. This reduction in
sputter yield is attributed to the very surface becoming reasonably conductive (by preferen-
tial loss of oxygen and the creation crystal defects) and thus the highly charged ions become
decharged; that is, they lose their internal energy when they approach the surface.
To model ion sputtering it is important not only to model Mercury’s magnetosphere but
also to understand the composition of the solar wind at Mercury and its variability. These
effects are being considered to correctly characterize the ion sputter source (Sarantos et al.
2007).
3.5 Surface Maps of Particle Fluxes and Energies
Sputtering of Na by solar wind ions impinging onto the surface of Mercury through the cusps
of the magnetosphere was first suggested by Potter and Morgan (1990) to explain rapid vari-
ations in the observed Na exosphere, with high- to mid-latitude enhancements appearing
and disappearing on intervals of less than a day. Such variations cannot be attributed to PSD
which varies slowly with true anomaly angle of the planet, and which would be character-
ized by a sub-solar maximum in the distribution. Variations in the sodium exosphere during
a week-long sequence in November 1997 were shown to correlate with possible variations
in Mercury’s magnetosphere such that increased ion sputtering correlated with opening of
the cusp regions and an increased ion flux to the surface (Killen et al. 2001). The role of the
precipitation of the solar wind plasma has received a noticeable interest because ion sputter-
ing is proposed as a potential candidate to explain rapid temporal variations in Mercury’s Na
exosphere observed in Earth-based remote sensing measurements (e.g., Potter et al. 1999;
Wurz and Lammer 2003; Leblanc et al. 2003b).
Sputtering occurs mostly due to solar wind precipitation, even if planetary ions may
contribute as well (Delcourt et al. 2003). Solar wind particles are expected to precipitate in
the dayside cusps (Massetti et al. 2003; Kallio and Janhunen 2003); during this motion, a
large fraction (∼90%) of the protons are bounced by the increasing magnetic field, the others
(10%) reach the surface of Mercury and lead to ion-sputtering, or (1%) are neutralized due
to charge-exchange effect (Mura et al. 2006a). During this motion, protons do not exactly
follow magnetic field lines: particles are drifted northward by the E × B drift and westward
by the grad-B drift (Mura et al. 2005).
The former is energy-independent; the latter is more efficient for the highest energies.
Moreover, non-adiabatic effects may become of crucial importance for most of the ion mag-
netospheric transport (Delcourt et al. 2003; Massetti et al. 2007). Figure 28 shows an exam-
ple of maps of surface precipitation for high- (1–10 keV) and low-energy (100 eV–1 keV)
protons; high-energy protons (up to several keV), accelerated by the reconnection mech-
anism, precipitate at lower latitudes with respect to low-energy ones (see Massetti et al.
2003).
The intensity and the shape of the H+ flux depend on the magnetospheric configuration
which, in turn, depends on both the intrinsic magnetic field of Mercury and variable external
parameters, such as the interplanetary magnetic field (IMF), and solar wind velocity and
density (Kabin et al. 2000; Kallio and Janhunen 2003; Massetti et al. 2003).
Even if Mariner 10 measurements revealed the existence of an intrinsic magnetic field,
this estimation has a considerable uncertainty, because it is difficult to separate the internal
and external magnetic field components (Connerney and Ness 1988). Nevertheless, the di-
Fig. 28 Example of solar wind proton precipitating flux onto the north hemisphere, in an open magne-
tospheric configuration (BIMF = [0, 0, −20] nT). Left panel: low energy (<1 keV) protons; right panel: high
energy (>1 keV) protons. Adapted from Mura et al. (2006c)
3
pole moment is probably between 284 and 358 nT RM (Ness et al. 1975); for comparison,
the Earth’s dipole is approximately 3 × 10 nT RE . Peculiarities in Mercury’s magnetosphere
4 3
arise also from the specific conditions of the solar wind at Mercury’s orbit (0.31–0.47 AU),
which differ substantially from the average conditions present at 1 AU.
Parker spiral forms an angle of about 20° with the solar wind radial direction, while it is
approximately 45° at the Earth. This implies a change of the relative ratio of the IMF com-
ponents with respect to the near Earth conditions, and a modified solar wind–magnetosphere
relationship. The average solar wind density is about ten times higher than at the Earth, and
this value varies considerably due to the high eccentricity of the orbit of the planet, with
average densities from 34 cm−3 at the aphelion, to 83 cm−3 at the perihelion. The dynamic
pressure is, on average, approximately 16 nPa (Massetti et al. 2003). By applying all the
above parameters, it has been estimated that the sub-solar point, where the internal and ex-
ternal pressures balance, is about 1.5 RM out from Mercury’s center (Siscoe and Cristofer,
1975; Goldstein et al. 1981), while at the Earth this value is 11 RE . In this respect, the
scale-length of Mercury’s magnetosphere is 1/7 compared to the Earth’s.
Ion precipitation at Mercury has been a subject of several analyzes motivated by Mariner
10 magnetic field and electron measurements made in 1974–1975. Unfortunately, there are
no direct ion measurements available from the Mariner 10 flybys. Many studies of the solar
wind-Mercury interaction have focused primarily on analyzing the role of the interplanetary
magnetic field (Luhmann et al. 1998; Kabin et al. 2000; Killen et al. 2001; Sarantos et
al. 2001; Ip and Kopp 2002; Kallio and Janhunen 2003; Massetti et al. 2003), and on the
solar wind dynamic pressure (Sarantos et al. 2007), the motion of ions in the Hermean
magnetosphere produced from the Hermean exosphere and emitted the surface (Ip 1987;
Delcourt et al. 2003, 2002; Killen et al. 2003), and the motion of the solar wind protons
injected from the tail (Lukyanov et al. 2001).
Both uncertain quantities and the large number of possible configurations make it very
difficult to include all of them in a model of proton precipitation. Evaluation of H+ flux on
the surface may be obtained from MHD simulations (Leblanc et al. 2003a), quasi-neutral
hybrid MHD simulations (Kallio et al. 2003), single particle models (Mura et al. 2005) and
loss-cone estimations (Massetti et al. 2003); in the first two cases, the magnetic field is ob-
tained from the model; in the other cases, it must be provided separately (for example, it may
be reconstructed by adapting a magnetic field model of the Earth to Mercury’s case (Mas-
setti et al. 2003), assuming proper values for the IMF, the solar wind density and velocity).
These precipitation models generally prescribe two areas of intense precipitation, roughly
corresponding to the cusps regions, or a big area of precipitation located in the planetary
dayside, depending on external conditions or simulation assumptions.
Table 8 summarizes some relevant quantities related to solar wind proton and planetary
ion precipitation.
The H+ flux onto the surface of Mercury may exceed values of 109 cm−2 s−1 , and the
total integrated H+ flux onto the surface of Mercury can be estimated as about 1025 s−1 .
During solar energetic particle (SEP) events, high-energy integrated proton flux may be up to
1026 –1027 s−1 (Leblanc et al. 2003a). Alpha particles exhibit a smaller flux (approximately
one tenth) but the yield for sputtering is considerably higher, so that they are expected to
contribute to this process as well. The flux of the exospheric ions, like sodium, is much
lower, and can reach 106 –107 cm−2 s−1 (Delcourt et al. 2003; Leblanc et al. 2003b); those
ions can precipitate and generate sputtering also in the nightside. The maximum sputtering
flux due to this process has been estimated by Delcourt et al. (2003) and is around 104
cm−2 s−1 . The flux of ions to the Hermean surface depends strongly on solar wind dynamic
pressure and on IMF (e.g., Kallio and Janhunen 2003; Delcourt et al. 2003), and strong
disturbances such as SEP or CME events (e.g., Cohen et al. 2005).
Table 8 Flux related to solar wind proton and planetary ion precipitation
Quantity Value References, notes
Flux (cm−2 s−1 ) 1.5 × 108 McGrath et al. (1986) (H+ )

4 × 108 Massetti et al. (2003) (H+ , IMF = [0, 0, −10] nT)
2 × 109 Massetti et al. (2003) (H+ , upper limit)
2 × 109 Mura et al. (2005) (H+ , upper limit)
105 –106 Delcourt et al. (2003) (Na+ , perihelion-aphelion)
Integrated flux 8 × 1024 Leblanc et al. (2003b) (SEP, H+ , 10 keV – 10 MeV)

(total), 3 × 1023 Leblanc et al. (2003b) (SEP, He2+ )
(s−1 ) 1.1 × 1025 Massetti et al. (2003) (H+ , IMF = [0, 0, −10] nT)
4 × 1025 Mura et al. (2005) (H+ IMF = [0, 0, −20] nT)
3.9 × 1025 Kallio and Janhunen (2003) (H+ , IMF = [0, 0, 10] nT)
3.4 × 1025 Kallio and Janhunen (2003) (H+ , IMF = [0, 0, −10] nT)
2.7 × 1025 Kallio and Janhunen (2003) (H+ , IMF = [32, 10, 0] nT)
30 × 1025 Kallio and Janhunen (2003) (H+ , High USW)
Fraction of 8% Leblanc et al. (2003b) (SEP, H+ , 10 keV – 10 MeV)

precipitating 11% Leblanc et al. (2003b) (SEP, other species)
ions 10% Mura et al. (2005) (H+ )
Open field area 7.3 × 1016 Killen et al. (2001) (H+ )

(cm) 2.8 × 1016 Massetti et al. (2003) (H+ , IMF = [0, 0, −10] nT)
1.8 × 1017 Mura et al. (2005) (H+ IMF = [0, 0, −20] nT)
One criticism of this source process is that the sputtering efficiency of protons, the domi-
nant ion in the solar wind, is quite small and cannot account for significant sputtering (Koehn
et al. 2003). When the IMF has either a negative Bz or a strong Bx component, magnetic re-
connection occurs between IMF and the Hermean magnetic field. In this case, H+ particles
from the magnetosheath can cross the magnetopause, enter the magnetosphere and precipi-
tate following open field lines. The shape of proton flux onto the surface (and, hence, of the
ion-sputtering flux) approximately mimics the shape of the reconnection at the MP.
In a purely IMF-Bz case, reconnection occurs only if Bz is negative; however, the Bx
component plays a substantial role in the magnetosphere solar wind coupling (Kabin et al.
2000; Sarantos et al. 2001; Kallio and Janhunen 2003; Massetti et al. 2007). In fact, the
different angle of the Parker spiral at Mercury compared to the Earth suggests that the Bx
component could be dominant. Non-zero Bx values introduce a north–south asymmetry in
the H+ surface precipitation; this feature is very important because it can easily explain the
north–south asymmetries in the earth-based observation of some exospheric components
(e.g., sodium).
An asymmetry in impacting ion flux translates into an asymmetry in the neutral at-
mosphere only when ion sputtering is a strong comparative source of atmospheric neutrals.
This generally happens when the IMF is southward (Bz < 0). Since southward fields were
observed roughly only half the time in Helios I and II data, a north–south asymmetry in
impacting proton flux should be expected for only 16–28% of Mercury’s lifetime. Still, vis-
ible asymmetries could sometimes result from a northward IMF when photon-stimulated
desorption is weak—namely at aphelion. At a first look, this result implies that north–south
asymmetries in impacting ion flux might be quite frequent at Mercury.
When the radial component of the IMF is dominant, most of the precipitating solar wind
ions impact on one hemisphere: for a sunward-pointing component (Bx > 0) most ions reach
the surface of Mercury’s southern hemisphere; the opposite is true for an anti-sunward-
pointing field (Bx < 0). This north–south asymmetry was first pointed out by Sarantos et al.
(2001) who qualitatively showed that, for a negative Bx , solar ions have a velocity compo-
nent parallel to Mercury’s (open) magnetic field configuration in the northern hemisphere
but antiparallel in the southern hemisphere, and vice versa. Sarantos et al. (2007) analyzed
the Helios I and II magnetic field data collected while the spacecrafts were within Mercury’s
orbital range. A dominant Bx was defined as one being at least twice that of either of the
other two components. Results show that the Bx was “dominant” according to the above
definition between 32–57% of all time. (Each set of data is defined as one spacecraft pass
between 0.31 AU to 0.47 AU. The lowest occurrence of a dominant Bx in a set was 32%,
while the highest was 57%.)
To quantify this, Kallio and Janhunen (2003) tested one case of a Parker IMF (Bx = 32
nT, By = 10 nT, Bz = 0 nT). They found that in that case only 36% of the total proton
flux impacted the northern hemisphere. This result implies that when there is a north–south
asymmetry of the sputtering emissions in the atmosphere, an upper limit of a 2 : 1 ratio of
total content in each hemisphere should be expected.
Figure 29 shows a numerical, single-particle simulation of the H+ flux onto the dayside
surface, as a function of IMF-Bx : in the figure, the Bx > 0 case is shown, and in this config-
uration the proton flux on the surface is expected to be higher in the southern hemisphere. To
find how often the radial component of the IMF is dominant, Sarantos et al. (2007) imple-
mented an analysis of the Helios data collected while the spacecrafts were within Mercury’s
orbital range. These authors used 40-second averages of the IMF and computed the prob-
ability density (normalized occurrence rate) estimate (see Fig. 30) At first look, this result
implies that north–south asymmetries in impacting ion flux might be quite frequent at Mer-
cury.
However, an asymmetry in impacting ion flux translates into an asymmetry in the neutral
atmosphere only when ion sputtering is a strong comparative source of atmospheric neutrals.
This generally happens when the IMF is southward (Bz < 0). Since southward fields were
observed roughly only half the time in Helios I and II data, a north–south asymmetry in
impacting proton flux should be expected for only 16–28% of Mercury’s lifetime. Still,
visible asymmetries could sometimes result from a northward IMF when photon-stimulated
desorption is weak—namely at aphelion. Figures 31 and 32 show the precipitating flux of
solar wind ions impacting Mercury’s surface for likely aphelion and perihelion conditions.
The relative recycling and escape rates of ions of Hermean origin has been a long-standing
question (e.g., Ip 1987).
Global tracings of magnetospheric ions (Na+ and K+ ) in static magnetic field configura-
tions given by the modified TH93 model (Killen et al. 2004a; Sarantos 2005) indicate that
impacts dominate, while the escape rate of these species to the solar wind responds very
slightly to external conditions: between aphelion and perihelion the escape ratio was seen to
range from 30% to almost 40% (Sarantos 2005, PhD thesis). While long-term recycling is
very strong, it is reduced by about a factor of 1.5 at perihelion. This prediction could help
explain why the sodium atmosphere is observed to be denser at aphelion. In contrast, Del-
court et al. (2003) and Leblanc et al. (2003a) found significantly reduced relative recycling
rates of ∼10%.
The sodium ion precipitation in Delcourt et al. was found to be similar to auroral pre-
cipitation at the Earth. We can only speculate about the reasons for these differences. The
Processes that Promote and Deplete the Exosphere of Mercury
Fig. 29 Solar wind H+ fluxes on the surface of Mercury, with different IMF-Bx external condition, from 5 nT (left) to 30 nT (right). Top panels: northern hemisphere; bottom
panels: southern hemisphere. As a positive IMF-Bx component increases, higher fluxes are expected in the southern hemisphere. From Mura et al. (2006c)
477
Fig. 30 Probability density estimates for important solar wind parameters from Helios 40-second data
around the Hermean aphelion and perihelion (Sarantos et al. 2007)
models were run for different input conditions, with different resolutions and, most impor-
tantly, using different exosphere models which provided the initial ion distribution to be
tracked. Possible explanations for these differences include pitch-angle and IMF Bx inclu-
sion effects.
A pitch-angle effect may be related to the location of photoionization events by different
exosphere models: the loss cone angle for particles launched below or at scale-height (e.g.,
Killen et al. 2004a; Sarantos 2005) is wider than for particles launched in the exoionosphere,
so the latter may mirror instead of impact the surface.
On the other hand, the inclusion of IMF Bx in the TH93 model widens the open area,
that is, the area of auroral precipitation, and makes it asymmetric. In conclusion, both in-
vestigations were limited in scope as they tested one case at aphelion and one at perihelion.
These differences point to the need for more comparative simulations, with single input and
boundary conditions.
Figure 33 (upper panel) illustrates the effect of likely solar wind conditions at perihelion
versus aphelion on the fraction of surface area open to the solar wind.
At perihelion the solar wind is more dense than at aphelion, and the IMF is more radial,
both of which act to create a more open magnetosphere. For the same Bz , the perihelion
magnetosphere is more likely to have a larger surface area exposed to the solar wind. The
likely precipitating solar wind flux (lower panel) is proportional to the open area and the
Fig. 31 Precipitating flux (log scale) of solar wind ions impacting Mercury’s surface for likely aphelion
conditions. These maps were produced with the modified TH93 model of the Hermean magnetosphere. In-
put conditions were selected by analyzing probability density estimates that are consistent with the Helios
40-second data in the 0.31–0.47 AU range. Vertical columns address the effects of increasing pressure on the
cusp location and precipitation flux for the same IMF, while horizontal rows illustrate the effects of a more
southward IMF for given pressure. The open-closed boundary exhibits a strong dawn–dusk asymmetry for
the Bz = −5 nT cases as a result of the dominant Bx . In turn, the cusp becomes more symmetric as Bz grows
comparable to Bx (cases with Bz = −10 nT). (Adapted from Sarantos et al. 2007)
dynamic pressure. For a precipitating integrated flux of 1.4 × 1026 , the upper and lower
limits of ion sputter yield averaged over the solar wind species (0.05 and 0.15, respectively,
give a yield averaged over the entire surface of the planet of 9 × 106 and 2.6 × 107 cm−2 s−1 ,
respectively. These yields are comparable to or greater than the PSD yields.
Heavy ions of planetary origin may be relatively energetic due to the centrifugal acceler-
ation during E × B transport over the polar cap (Delcourt et al. 2003). Most of the Na+ ions
are lost into the dusk flank, but a localized region of energetic Na+ precipitation develops at
the planet’s surface, and extends over a large range in longitude at mid-latitudes (30◦ –40◦ ).
These ions may sputter additional material. Characteristics of precipitating Na photoions are
shown below in Fig. 34.
3.6 Extreme Solar Events
Because dense or fast solar plasma features could compress Mercury’s magnetosphere so
that more surface elements can be released due to sputtering into the exosphere, it is impor-
Fig. 32 Maps of the solar wind flux (log scale) that precipitates onto the Hermean surface for likely con-
ditions at aphelion (a, c) and perihelion (b, d). Shown in parentheses are the density (cm−3 ) and velocity
(km/s) tested in each case. The open-closed boundary moves equatorwards by about 10 degrees for perihelion
conditions. Responding to the denser plasma and stronger field magnitude, the precipitating flux at perihe-
lion clearly increases both in the dayside and in the tail. The modeled integrated precipitating source (s−1 )
increases fourfold, while the open area available to the solar wind doubles. (Adapted from Sarantos et al.
2007)
tant to study collisions with extreme solar particle events on Mercury’s magnetospheric and
surface environment.
Leblanc et al. (2003b) studied the interaction of a solar energetic particle (SEP) event
reported by Reames et al. (1997) of protons with energy larger than 10 keV with Mer-
cury’s magnetospheric-surface environment. They expected that if a SEP encounters Mer-
cury, a significant f1ux of energetic particles will reach Mercury’s surface which may refill
the exosphere. The simulations indicate that after the arrival of a SEP at Mercury, a popu-
lation of quasi-trapped energetic ions and electrons is expected close to Mercury, which is
stable for hours after their arrival. A significant dawn/dusk charge separation is observed
and a fraction of about 10% of the initial energetic particles may impact the surface with
a spatial distribution that exhibits north/south and dawn/dusk asymmetries. Furthermore,
Leblanc et al. (2003a, 2003b) found that the flux of particles impacting Mercury’s surface
and the ability of a quasi-trapped population to be maintained near Mercury are highly de-
pendent on the Bz sign of the interplanetary magnetic field. Leblanc et al. concluded that
impacting SEPs can eject a non-uniform distribution of Na atoms into Mercury’s exosphere,
Fig. 33 Effective open area

(upper panel) and precipitating
flux (lower panel) for likely
aphelion and perihelion
conditions, respectively (Sarantos
et al. 2007)
which may be the origin of several unexplained observed exospheric features, although they
could not explain the total amount of Na atoms needed to reproduce the observations of
Potter et al. (1999). However, much stronger SEP events than the one used in this work
have been observed at the Earth (Mason et al. 1999) during the same month as the Potter
et al. (1999) observations, and may have sufficient intensities. These authors proposed that
the encounter of Mercury with CME or magnetic cloud (MC) events could cause such an
enhancement.
Significant advances in the study of CMEs have been made by the Large Angle and
Spectrometric Coronagraph (LASCO) on board of ESA’s Solar and Heliospheric Observa-
tory (SoHO), which observed more than 8,000 CMEs since January 1996. The observational
data on CMEs are related to two spatial domains: the near-Sun region (up to 30 RSun ≈ 0.14
AU) remotely sensed by coronagraphs; and the outer region, including the geospace and
beyond, where in situ observations are made by spacecraft. At the larger distances such as
at Mercury’s orbit or beyond CMEs are traditionally called as Interplanetary CMEs or mag-
Fig. 34 Characteristics of precipitating Na+ ions (left) at perihelion and (right) at aphelion. The panels from
top to bottom show ion flux, average energy, and residence time in the magnetosphere (Delcourt et al. 2003)
netic clouds. CMEs are associated with flares and prominence eruptions and their sources
are usually located in active regions and prominence sites. The basic characteristic of CME
producing regions is closed magnetic structure. Recent studies on temporal correspondence
between CMEs and flares provide arguments in favor of the so-called common-cause sce-
nario, according to which flares and CMEs are different manifestations of the same large-
scale magnetic process (Zhang et al. 2001). Although the details of this process still remain
unclear, it definitely can be stated that an intensive flaring activity of a star should be accom-
panied by an increased rate of CME production. The probability of CME-flare association
increases with the duration of a flare (Sheeley et al. 1983): ≈26% for flare duration <1 h;
and ≈100% for flare duration >6 h. Multi-thermal structure of CMEs includes 1) coronal
material in the front region (≈2 MK); and 2) possibly a core from solar prominence material
(≈104 K), or hot flare plasma (≈10 MK).
The basic and widely considered characteristic of CMEs is their velocity, determined
by tracking a CME feature in coronagraph image frames taken with a certain time ca-
dence. According to the data from SoHO/LASCO, the velocity of solar CMEs ranges
from tens of km s−1 to >2,500 km s−1 near the Sun, with an average value of about
490 km s−1 (e.g., Gopalswamy 2004). Due to the relatively large statistics (>8,000) of
the considered SoHO/LASCO CMEs, the average ejecta velocity value can be consid-
ered as a representative quantity. It is also consistent with the results of other veloc-
ity measurement techniques applied to separate CME phenomena (Lindsay et al. 1999;
Gopalswamy et al. 2001a, 2001b, 2004; Lara et al. 2004). Halo CMEs, a subclass of CMEs
propagating toward the Earth, are currently believed to be the main drivers of space weather
disturbances at the Earth. Compared to the general population of CMEs, halos have much
higher average speeds of 1,004 km/s (Yashiro et al. 2004).
Besides the velocity of CMEs an additional important parameter for studying the plasma
interaction with Mercury’s magnetospheric-surface environment is the plasma density. Esti-
mates of CME plasma density from white light (Vourlidas et al. 2002), radio (Gopalswamy
and Kundu 1993), and UV observations (Ciaravella et al. 2003) give similar values of about
106 cm−3 at the distances ≈3–5RSun which are consistent with the assumption of the coro-
nal value of density of a CME material at the moment of ejection. At larger distances >30
RSun (i.e., >0.14 AU), the density and duration of ICMEs and associated MCs are measured
in situ by spacecraft (Henke et al. 1998; Lepri et al. 2001). For the plasma density in MCs
observed between 0.3–1 AU by the Helios satellites, Bothmer and Schwenn (1998) found a
power law
−2.4±0.3
d
nMC = n0MC , (21)
d0
where d is the radial distance to the Sun in units of AU, quantity n0MC = 6.47 ± 0.95 cm−3
is the MC plasma density at the near-Earth orbit, and d0 is at 1 AU (Bothmer and Schwenn
1998). By using this power law one gets CME number density values at periherm of about
80–260 cm−3 and about 45–130 cm−3 at about 0.38 AU.
Finally, another important characteristic of CME activity and its influence of Mercury’s
environment is the CME occurrence rate. The data from Skylab, SMM, Helios, Solwind,
and SoHO indicate a correlation between sunspot numbers (SSN) and the CME occurrence
rate (Hildner et al. 1976; Howard et al. 1986; Webb and Howard 1994; Cliver et al. 1994;
Cyr et al. 2000; Gopalswamy et al. 2003). At the same time, SoHO/LASCO observations
found that although there is an overall similarity between the SSN and the CMEs occurrence
rates, there are some differences in details. The most recent SoHO/LASCO observations
give a CME occurrence rate ≈0.8 CMEs/day for solar minimum and ≥6 CMEs/day for
solar maximum.
These numbers are consistent, but a bit higher as compared to previous estimations (Hild-
ner et al. 1976; Webb and Howard 1994; Cliver et al. 1994), which is attributed to the better
sensitivity and the high dynamic range of the LASCO coronagraphs. Focusing on halo and
fast-and-wide CMEs may help estimate the occurrence rate of ICMEs at Mercury. First, full
halo CMEs account for about 3.5% of all CMEs detected by SOHO. Second, the fraction of
fast-and-wide CMEs ranges from 2% (1996) to 6% (2003, past solar maximum) of the total
SOHO count (Gopalswamy 2004). Combining the observed occurrence rates of all CMEs
originating at the Sun with the aforementioned fraction of fast-and-wide CMEs, and assum-
ing that a typical halo has angular width of 120 degrees, we should expect 2–3 ICMEs/year
at solar minimum and about 44 ICMEs/year (one every nine days) at solar maximum to
impact Mercury.
Kallio and Janhunen (2003) studied the precipitation of the protons related to a MC or
CME event with a proton density of about 75 cm−3 and a velocity of about 800 km s−1
with a self-consistent quasi-neutral hybrid model. In their simulation the particle flux of the
plasma protons was calculated self-consistently with the kinetic model. By increasing solar
wind dynamic pressure to about 10−5 Pa it was possible to push the magnetopause toward
Mercury’s surface, in agreement with MHD model runs of Kabin et al. (2000). In such a
Fig. 35 3D exosphere
calculation by using the test
particle model for sputtered Ca
atoms as a function of planetary
distance in units of km for
ordinary solar wind conditions
(left panel) and during a
simulated MC collision with
Mercury with a plasma density of
75 cm−3 and a velocity of about
800 km s−1
case an intense H+ ion precipitation with fluxes up to about 109 cm−2 s−1 was found near
the subsolar point.
Figure 35 shows an equatorial cut of sputtered exospheric Ca atoms during ordinary solar
wind conditions where larger fluxes of solar wind protons only precipitate to the surface
around Mercury’s cusp regions (Massetti et al. 2003; Kallio and Janhunen 2003) and the
MC event discussed earlier and studied by Kallio and Janhunen (2003). One can see that
due to the large affected surface area the exosphere should be refilled by Ca atoms during
the MC collision.
One can see from Fig. 35 that the results which give much denser and more distributed
exospheric number densities of sputtered surface elements during expected collisions with
CMEs or MCs are pertinent to future measurements on the Messenger and BepiColombo
missions, which will be instrumented to observe the released exospheric particles.
Kinetic energies of solar wind ions are, on average, 1 keV/amu, where the sputtering ef-
ficiency peaks. Sputtering by H+ , which accounts for 85% of the total kinetic energy carried
by the solar wind, is relatively inefficient. On average, He2+ accounts for about 13% of the
kinetic energy carried by the solar wind, and is generally assumed to account for most of
the space weathering effects. However, although heavy ions (Z > 6) account for only about
2% of the kinetic energy carried by the solar wind, they also carry ∼1 keV each in potential
energy due to ionization into a high charge state. The charge state of the impinging ion has
little effect on the sputter efficiency of highly conducting targets (conductors and semicon-
ductors), but has considerable effect on some insulators as discussed earlier (Aumayr and
Winter 2004).
3.7 Sputter Effects on Surface Chemistry
Sputtering of the surface by any of the mechanisms discussed earlier may affect the top-
most layer of the regolith. Ion implantation of solar wind ions occurring for billions of years
may change the predominant apparent chemistry to more reflect the solar wind than the
pristine composition from Mercury’s origin. In addition, new chemical compounds may be
made. Production of sodium and water by proton sputtering of sodium-bearing silicates was
considered by the following mechanism (Potter 1995)
2H + Na2 SiO3 → 2Na + SiO2 + H2 O. (22)
The supply rate of water molecules is half the supply rate of Na by this process. Since the
free energy of this reaction is −4.7 kcal/mole, it will proceed spontaneously; however, the
activation energy is unknown. Chemical reactions, either on the surface or in the atmosphere,
can enhance the loss rates of the reaction products if the reaction products are created with
enough energy to escape. Very little work has been done to quantify rates for chemistry as a
source or loss process in the context of planetary sciences.
In laboratory research such processes are investigated and are referred to as plasma-
assisted etching, or chemically enhanced physical sputtering (e.g. Winters et al. 1983;
Winters and Coburn 1992). These processes are used for a large variety of industrial and
laboratory applications. Energetic ions and electrons of the plasma induce strong changes
in the surface chemistry. Plasma-assisted etching involves the interaction of a plasma dis-
charge with a solid surface to produce a volatile product. Laboratory research aims to pro-
duce product molecules at the surface that are weakly bound to the surface to enhance the
sputter yield for micro-fabrication. The reaction can be divided into three steps: 1) the ad-
sorption/implantation of external particles, 2) the product formation, and 3) the removal of
the product from the surface.
Physical sputter theory predicts that the sputter yield is inversely proportional to the
binding energy of the species to the surface (Sigmund 1969). Energy distributions under
the conditions of chemical sputtering have been measured in the laboratory (Haring et al.
1982) and two components in the energy distribution have been identified (Winters and
Coburn 1992). The first component is very well described by the energy distribution for
physical sputtering (see (15)). The second component, however, peaks at energies between
0.1 and 0.5 eV (indicative of a low surface binding energy) and can be described by a
Maxwellian distribution, and shows higher yield. Thus, the chemical alteration increased
the total sputter yield from the surface. In laboratory experiments the interaction of halo-
gens is studied mostly, but there are a few investigations where the increased sputter yield
of silicon by hydrogen ion impact is reported (Winters and Coburn 1992, and references
therein).
Such a low binding energy means that the created species can also be released via thermal
desorption, as is the case for the example of Na and water given above. Thus the chemical
alteration of Mercury’s surface by precipitating ions and the later release of volatiles (e.g.,
Na and water) can also occur with considerable delay. A possible scenario is ion implan-
tation when the surface is cold (e.g., at the nightside). At a later time, when the surfaces
warms up because of solar insolation, the thermal release of the such created volatiles may
occur.
Chemical reactions induced by solar wind ions impinging on the surface undoubtedly
occur on Mercury. Earlier, we discussed the example of a source for water molecules and
atomic sodium. The study of chemistry on the surfaces of bodies exposed to the solar wind
is in its infancy, and more so the effects this has on the sputter yields.
4 Thermal Vaporization of Alkali Atoms
The vapor pressure of a gas in thermal equilibrium with a liquid is given by
l
P = P0 exp − , (23)
RT
where l is the latent heat per mole, R is the gas constant and T is the temperature in Kelvins.
In the case of vaporization of an adsorbed volatile from a solid surface, the rate constant for
thermal desorption is described as
Edes
kdes = Ades exp − , (24)
kT
where A is a pre-exponential vibrational frequency, and Edes is the desorption energy. The
value of the pre-exponential, the vibrational frequency, most often used is 1013 s−1 (e.g.,
Hunten et al. 1988) but in fact it can vary from 104 to 1023 s−1 (Holmlid 1998). The
extremely large range for the preexponential term is due to the large number of physical
processes actually involved in thermal desorption: including diffusion to and from the bulk
rock or grains, surface diffusion between sites with different desorption energies, electronic
excitation and de-excitation, jumps during the near-desorption in excited states.
Spatially resolved studies of iron-rich minerals show the complexity of real metal oxide
surfaces: for instance, the desorption energy of alkali atoms is changed radically by the
addition of other atoms in low concentrations. The thermal barrier for K desorption from
the same type of mineral can range from 0.83 eV to 2.35 eV by adding 2 wt% Mn to the
material (Kotarba et al. 2004). Values quoted in the literature for desorption energy of Na
atoms range from 1.1 eV (Hunten and Sprague 1997), 1.8 eV (Madey et al. 1998) and 2.7
eV for metal and metal oxide surfaces (Holmlid and Olsson 1977). In the Monte Carlo study
of Leblanc and Johnson (2003) an average desorption energy of 1.85 eV was used with a
preexponential of 1013 s−1 . Holmlid (2006) suggested that this pre-exponential is much too
large, and that the desorption energy of alkali atoms from any real oxide mineral is unlikely
to be much smaller than 2 eV.
The rate of thermal desorption from a surface is in fact rate limited by the slowest process
acting in the chain of events leading to desorption (see Fig. 36). For the surface of Mercury,
Killen et al. (2004b) concluded that the rate-limiting process for thermal desorption from
the surface of Mercury is diffusion of atoms from the bulk of the grains. This conclusion
is consistent with the conclusions of Leblanc and Johnson (2003) that thermal desorption
rapidly depletes most of the sunlit surface of Mercury of adsorbed atoms.
The measured temperature for the sodium atmosphere is high, about 1,200 K, whereas a
high adsorption energy would imply efficient sticking at the surface, and hence rapid thermal
accommodation to the surface temperature (Holmlid 2006). However, observations of the
variation of sodium D2 intensities with true anomaly angle imply that the sticking coefficient
is quite small, on the order of 0.15 (Potter et al. 2006). This seems to be inconsistent with a
high adsorption energy. It is not inconsistent with a lower adsorption energy, and an efficient
Fig. 36 (a) Shows the time

required for the flux of solute
from the grain interior to the
grain surface to fall below a
constant value of 107 cm−2 s−1
(the maximal required at the
subsolar point to maintain the
exosphere) as a function of
temperature (diffusion
coefficient) and grain radius. We
show the limiting size grains
(radius in cm, color coded) that
can maintain the thermal fluxes
until half the solute is depleted
for (b) glass and (c) crystalline
minerals: 100 cm (glass) and
1 µm (mineral) at perihelion, and
103 cm (glass) and <1 µm
(mineral) at aphelion (Killen et
al. 2004b)
loss of adsorbed states, as described by Killen et al. (2004b) and by Leblanc and Johnson
(2003).
In this case the source process for Na and K to the exosphere would be dominated by
PSD, ion-sputtering and impact vaporization, which are all capable of ejecting the atoms
from the bound state. The source rates for all processes other than impact vaporization,
which accesses a depth equal to several impactor diameters, depends on the availability of
atoms at the extreme surface. To maintain a long-term supply to the exosphere by PSD or
ion-sputtering, atoms must diffuse from the bulk of the rock or grain to the extreme surface.
The long-term source rates are therefore limited by the diffusion rates. These are depen-
dent not only on temperature, but grain size and lifetime of the grain on the surface. Thermal
vaporization at the subsolar point on Mercury is very efficient, given a supply of adsorbed
atoms at the extreme surface (e.g. Leblanc and Johnson 2003). Fluxes are calculated using
(Crank 1975)
1/2
Dt 2 2 πF F
= − 1− − , (25)
a2 π π 3 3
where F is the fraction of solute lost at time t from a sphere of radius a. D is the diffusion
coefficient and the solute is assumed to be initially uniformly distributed.
To obtain the flux per unit surface area of the planet at time, t , we solved (25) for F , took
the time derivative of F and multiplied by the initial amount of solute and divided by the
cross-sectional area of the grain. Thus the loss rate per unit cross-sectional area is given by
4
dn πa 3 fNa ρ dF
= 3 dt
. (26)
dt πa 2
Once the adsorbed atoms have evaporated, the source rates drop to the values at which
atoms can be supplied to the surface by diffusion. Results show that large thermal fluxes
could be maintained by diffusion from hot glassy spheres as long as a mechanism exists for
desorption of the atoms from the surface.
Any glass sphere smaller than the limiting size would be able to maintain the flux. For
high fluxes, unreasonably small grains or high diffusion coefficients are required. Progres-
sively smaller grains supply atoms to the surface at faster rates but for shorter periods of
time, as the grains become depleted of volatiles (Killen et al. 2004b). The upper few µm
of the Moon is turned over about once in 103 years (Heiken et al. 1991). We expect the
turnover rate at Mercury to be about ten times that at the moon, since the impact flux is
about ten times that at Earth orbit.
Thus a µm-sized grain will sit on the surface of Mercury for about 100 years. If the
diffusion rate of a µm-sized grain drops to 107 cm−2 s−1 in one month, then a “thermal”
vaporization rate of 1011 cm−2 s−2 is sustainable for less than 0.1% of the µm-sized grains
on the surface, and none of the larger grains (Figs. 36a,b). It is concluded that the exospheric
rate required to maintain the observed exosphere 107 cm−2 s−11 , is therefore the limiting
value governed by diffusion of atoms to the surface of grains and by the gardening rate.
Radar bright spots near the poles of Mercury discovered in 1992 (Harmon and Slade
1992) were attributed to water ice. Numerous studies concluded that water ice is stable in
permanently shadowed craters near the poles of Mercury (Paige et al. 1992; Ingersoll et al.
1992). It is intriguing that 13 radar bright features were found between 70° and 80° latitude
(Harmon et al. 2001). All of these features are small, consistent with the limited amount of
permanent shading expected at these latitudes. It has been suggested that the volatile in the
cold traps is not water ice but in fact may be sulfur (Sprague et al. 1995) or cold silicate
(Starukhina 2000).
We should ask ourselves the question: Is thermal vaporization a loss process for water? In
other words, is evaporation a reversible or irreversible process? One must in fact show that
vaporization is an irreversible process to rule out the possibility of its presence at a given lat-
itude since atoms cannot be carried away by winds or other processes on an atmosphereless
body. Unless the atoms are lost by an irreversible process such as ionization and entrainment
into the solar wind, or chemical reactions, they must eventually return to the cold traps. Fur-
ther, as pointed out by Holmlid (2006), the adsorption energy and preexponential terms are
highly influenced by the local composition and charge state of the surface, which changes
rapidly near the terminator for instance.
5 Loss Processes
Loss of atoms from the exosphere must be defined into categories of loss: reversible loss
processes (sticking to the surface or ionization followed by impact with the surface and neu-
tralization) and irreversible loss processes (Jeans escape and ionization followed by crossing
the magnetopause boundary and entrainment into the solar wind). Reversible loss processes
must be further characterized as long-term loss, such as burial or chemical reaction into a
bound state with a large binding energy, or short-term loss such as adsorption on the surface
of grains. Jeans escape is an irreversible loss process, but photoionization may or may not
result in permanent loss. Sticking to the surface results in loss from the exosphere but not
from the regolith. These different types of loss processes will be discussed in the following.
5.1 Jeans Escape
Only H and He have significant rates of Jeans escape at the ambient surface temperature
on Mercury (Hunten et al. 1988). However, with the exception of thermal vaporization, the
source processes populating the Hermean exosphere create non-thermal populations and the
mean energy is species dependent. The mean velocities of atoms ejected by ion-sputtering
may be above escape velocity, 4.25 km/s, and will depend on the binding energy of the
atom in the substrate, and its mass (Table 9). As pointed out by Holmlid (2006) the binding
energy of a given atom will depend not only on the particular mineral in which it is found,
but also on the surrounding conditions. This is particularly important for a surface that is
highly subject to space weathering.
The mean velocity of Ca atoms observed in the Hermean exosphere is about 3.5 km/s
(Killen et al. 2005), consistent with the expected mean velocity for an ion-sputter source.
If the main source of Mg and Al to the exosphere is ion-sputtering, then the bulk of the
source would be lost by Jeans escape. Killen et al. (2005) suggested that dissociation of a
calcium oxide could also account for the high velocity observed for calcium in the Hermean
exosphere.
The temperature of meteoritic vapor is often described as being quite hot, perhaps as
high as 6,500 K. However, only the initial vapor plume seen by the Deep Impact space-
craft in the first few seconds was hot (1,000–3,000 K), implying that a rapid cooling takes
place in the dense fireball and possibly in the regolith (A’Hearn et al. 2005). The bulk of
Table 9 Mean velocity of an

ion-sputtered source for selected Species vmean [km/s]
atomic species
Mg 4.5
Al 4.3
Ca 3.5
Fe 3.0
the vapor—that in the ejecta curtain—which is composed of excavated material, was at am-
bient temperature. The vapor resulting from micrometeoritic bombardment may cool by
multiple collisions with the regolith as the micrometeoroids deposit their energy below the
surface.
5.2 Photoionozation
Photoionization is an irreversible loss process if the ions cross the magnetopause and be-
come entrained in the solar wind. Killen et al. (2004a) suggested that half of the photoions
produced near the surface return to the surface where they are neutralized. An east-west
electric field produces an asymmetric escape pattern such that those ions created on the
dusk side reimpact on the dawnside, and those ions created on the dawnside escape across
the magnetopause boundary or impact the nightside. However, the escape rate is higher as
the ions are produced at successively higher altitudes.
Photoionization rates of many likely exospheric species are known. Rates for He, C, N, O,
F, Na, S, Cl, Ar, Xe, OH were published by Huebner et al. (1992) and for Ca by Killen et al.
(2005). Photoionization rates for most refractory species (e.g., Mg) have not been published.
In any case, the energy with which refractory species are ejected by the two mechanisms
capable of ejecting refractory species, ion-sputtering and meteoritic vaporization, will be
energetic enough to eject most of the atoms at velocities exceeding escape velocity. Those
refractories that do not directly escape will probably stick to the surface on impact. Thus
their photoionization rates are not pertinent to their lifetimes in the exosphere.
The sodium ionization rate is controversial because of the discrepancy between the exper-
imental and theoretical cross-sections, and further because there are theoretical calculations
that agree with the experimental value. The Combi et al. (1997) value is quoted because
it is commonly used in the cometary community, and the observed cometary abundances
agree with calculations when this value is used for the ionization rate. However, the discov-
ery of the sodium tail on comet Hale-Bopp (Cremonese et al. 1997) and the high-resolution
spectroscopy performed along the tail was used to calculate a new value of photoionization
lifetime for the neutral sodium atom that confirmed the value suggested by Huebner et al.
(1992). The model used for the comet works also on the sodium tail of Mercury showing
that it does not depend on the object observed.
The rates for sulfur are theoretical (Huebner et al. 1992). Rates from Kumar (1982) are
50% higher. The calcium ionization rate was calculated by Walter Huebner and communi-
cated to the author. All other rates are from (Huebner et al. 1992). The photoionization rate
is not the limiting rate if atoms stick to the surface.
5.3 Charge Exchange
A charge-exchange process may occur when an energetic ion collides with a exospheric neu-
tral particle (target). In the interaction, an electron and a small amount of kinetic energy are
exchanged between the neutral and the ion; the net result is an energetic neutral atom (ENA)
and a thermal ion. The target (ionized) is scattered at an approximately perpendicular angle
with respect to the projectile path; the newly created ENA retains approximately both the
energy and the direction of the colliding energetic ion. The energy defect of the process is
equal to the difference of the two atomic ionization potentials (for complete discussions see,
e.g., Hasted 1964). Charge-exchange is a resonance process, which can be symmetrical (if
the species of the ion and the neutral are the same) or accidental (Stebbings et al. 1964;
Hasted 1964). The cross-section varies with species and energy and is of the order of
10−14 –10−17 cm−2 .
Table 10 Solar photo ionization

rates at 1 AU Species Rate coefficient Rate coefficient (active
(quiet sun) [s−1 ] sun) [s−1 ]
H 7.26 × 10−8 7.1 × 10−8
He 5.25 × 10−8 1.51 × 10−7
O(3 P) 2.12 × 10−7 5.88 × 10−7

O(1 D) 1.82 × 10−7 5.04 × 10−7
O(1 S) 1.96 × 10−7 5.28 × 10−7
Na 1.62 × 10−5(e) 1.72 × 10−5(e)

5.92 × 10−6(t) 6.42 × 10−6(t)
5.40 × 10−6(a)
S (3 P) 1.07 × 10−6 2.44 × 10−6

S(1 D) 1.08 × 10−6 2.46 × 10−6
S(1 S) 1.05 × 10−6 2.31 × 10−6
Ar 3.05 × 10−7 6.90 × 10−7
K 2.22 × 10−5 2.36 × 10−5
Ca 7.0 × 10−5(b) 7.8 × 10−5

(e) experimental
(t) theoretical OH → O(3 P) +H 6.54 × 10−6 7.17 × 10−6

OH → O(1 D) +H 6.35 × 10−7 1.51 × 10−6
(a) Combi et al. (1997)
OH → O(1 S) +H 6.71 × 10−8 1.64 × 10−7
(b) Huebner (personal
OH → OH+ 2.47 × 10−7 6.52 × 10−7
communication, unpublished)
Charge exchange at Mercury may occur due to solar-wind plasma (Mura et al. 2005) as
well as due to planetary ions. Generated neutrals have typical energies of 1 keV or more;
hence, such neutrals are no more trapped in ballistic orbits and the result is a net loss from
the planet. This process mostly occurs in the dayside and dawnside regions close to the
planetary surface (Mura et al. 2005); in general, the H-ENA production rate for unit length
reaches values up to 10 (cm−2 s−1 sr−1 m−1 ), close to the dayside planetary surface. It has
been estimated that, approximately, less than 1% of the solar wind plasma circulating inside
the magnetosphere of Mercury experience charge-exchange (Mura et al. 2005); the related
CE loss rate, cumulated on all neutral species, is between 1022 and 1024 s−1 , which is, on
average, small if compared to other loss mechanisms.
Figure 37 shows simulated fluxes of ENA coming from different directions (ENA im-
ages), as “seen” from a vantage point in the nightside, in a “fish-eye” projection. The fluxes
are integrated between 1–10 keV. The intense fluxes coming from the dawnside of the planet
(right side of the picture) are generated from westward drifting solar wind protons.
Fig. 37 Simulated ENA images, from a vantage point in the nightside (1.8, 0, 0.8 RM ). Color is coded
according to ENA flux, integrated between 1 and 10 keV. From Mura et al. (2005)
5.4 Gardening
Gardening, fragmentation and burial of the regolith, has been studied extensively for the
Moon (Heiken et al. 1991). Repetitive impacts agitate the surface by fragmenting, tum-
bling, burying and exhuming individual grains. The layer called the regolith is continuously
churned. Although numerous small impacts homogenize the upper part of the regolith, a
major role is played by the larger impacts in excavating previous sedimentary layers, frag-
menting rock layers and depositing fresh material onto the surface. The number of times the
lunar regolith has been turned over versus depth was calculated using results of laboratory
cratering experiments in fine grained unconsolidated targets (Gault et al. 1974), and is given
as a turnover versus depth as a function of time.
On the Moon, a depth of almost 1 cm is overturned once in 106 years with 50% probabil-
ity (Heiken et al. 1991, p. 87). If the meteoritic influx scales as 1/R 2 with distance from the
Sun, then the rates at Mercury are on average 6.7 times those at the Moon, and the turnover
times would scale as 0.15. Thus a layer of regolith 1 cm deep on Mercury should be over-
turned in 1.5 × 105 years with 50% probability. Turnover rates are important for long-term
renewal of material at the extreme surface. Survival times versus grain size were estimated
for the Hermean surface by Killen et al. 2004b, and compared with the time at which the
diffusive flux of sodium from the grain interior to the surface would drop to 50% of its
initial rate at a given prescribed rate (Fig. 38). This figure shows that the observed sodium
exosphere would be sustainable, even if all of the volatile loss were irreversible, because
fresh material will be supplied by regolith turnover before the volatiles are removed.
Fig. 38 Time it takes for a grain

of a given radius to lose 50% of
its solute at a given rate between
104 –107 cm−2 s−1 . The open
circles are lifetimes of grains on
the Moon, and the crosses are the
estimated lifetimes on the surface
of Mercury, with the assumption
that rates are approximately an
order of magnitude faster on
Mercury. This plot is a
two-dimensional cut through a
three-dimensional therefore
temperature may vary across the
plot
6 Space-Based Observations and Expected Results
The ESA cornerstone BepiColombo/MPO is planned to fly around Mercury in a polar orbit,
with 400 km periherm and 1,500 km apoherm; the orbital period is about 2.3 hours. Two
instruments in the MPO payload are mainly devoted to observing the exosphere: the FUV-
EUV spectrometer, PHEBUS, and the comprehensive suite for particle detection, SERENA.
The combination of these two experiments on BepiColombo will be an unprecedented op-
portunity to perform a detailed analysis of the exosphere composition and vertical profiles.
6.1 The BepiColombo/MPO/PHEUBUS UV Spectrometer
The BepiColombo/MPO/PHEBUS UV spectrometer PHEBUS is a dual FUV-EUV spec-

trometer (see the ESA BepiColombo webpage: http://www.rssd.esa.int/index.php?project=
BEPICOLOMBO) working in the wavelength range from 55 to 315 nm plus a small two-
channel detector used to measure potassium and calcium at 404 nm and 422 nm, respec-
tively. This instrument is devoted mainly to the characterization of Mercury’s exospheric
composition and dynamics. In addition, some ionized species have emission lines in the
spectral window; hence, this instrument will potentially contribute to the identification and
characterization of the exo-ionosphere.
Thanks to the remote sensing of the exosphere, PHEBUS will provide measurements of
density of many species (see Fig. 39) and their profiles at altitudes below the spacecraft
periherm. Observations of the low altitude exosphere are particularly important for the char-
acterization of the heavier atoms, such as N, C, Ne, Si, Fe, Mg and molecules like CO, that
are unlikely to arrive at the BC/MPO orbit and that can be observed, most probably, solely
by a remote-sensing instrument. PHEBUS will observe Na at 268 and 285.3 nm.
Since the observable emission lines are generated by the interaction between the so-
lar photons and the exospheric atoms, the PHEBUS can only observe by looking toward
the dayside or the terminator. In this sense, the two instruments, PHEBUS and the mass
spectrometer, SERENA/STROFIO (see next subsection), are completely complementary. In
fact, the combination of the PHEBUS remote sensing and the SERENA/STROFIO in situ
measurements will provide a complete view of the day–night exosphere for almost all the
foreseen components in a wide range of altitudes and Hermean conditions.
Fig. 39 PHEBUS EUV spectrum convolved with a Gaussian (σ = 1 nm)

6.2 The BepiColombo/MPO/SERENA Neutral Particle Detectors
A comprehensive suite for particle detection in the Mercury’s environment, the SERENA in-
strument (see the ESA BepiColombo webpage: http://www.rssd.esa.int/ index.php?project=
BEPICOLOMBO), is included in the MPO payload of the mission. This package consists
of four units: STROFIO and ELENA will detect the neutral particles and measure their en-
ergies in the range from fractions of eV to a few keVs; MIPA and PICAM will measure and
analyze ionized particles of planetary and solar wind origin from tens eV to tens of keV.
ELENA (Emitted Low-Energy Neutral Atoms) is a neutral particle camera that investigates
neutral gases escaping from the surface of Mercury, their dynamics and the properties of
the related source processes. The ELENA sensor is a Time-of-Flight (TOF) detector, based
on state-of-the-art choppers and mechanical gratings. The new development in this field al-
lows unprecedented performances in timing and angular discrimination of low-energy neu-
tral particles. STROFIO is a mass analyzer able to measure the neutral composition of the
non-directional, thermal component of the exosphere.
The ELENA FOV (see Fig. 40, left panel) is always nadir-pointing, and most sectors
look towards the planetary surface. In this way, the ELENA sensor will be able to detect the
neutral flux coming from two different sources: charge-exchange ENAs, resulting from the
interaction of solar wind and planetary plasma with the neutral exosphere, and ion-sputtering
ENAs resulting from the precipitation of solar wind and planetary plasma onto the surface
of Mercury. In general, it is possible to discriminate the neutral flux coming from those
sources, since they have different typical energies and different generation regions.
In fact, instrumental simulations show that ELENA has both energy and angular reso-
lution able to discriminate between those sources. The STROFIO FOV (see Fig. 40, right
panel) points in the ram direction, so that low-energy exospheric particles will enter the
sensor head due to the spacecraft motion. Regardless of the source generation process (pro-
vided that particles will have enough energy to reach the instrument location, see Fig. 41),
STROFIO will count and identify the local exospheric components along the BepiColombo
MPO orbit.
Fig. 40 ELENA and STROFIO field-of-views along the BC-MPO trajectories. Left panel: MPO orbit (red
dashed curve); example of ELENA FOV (blue lines); boundaries of the projection of ELENA FOV onto the
planetary surface (green curves); sample of the sputtering flux, due to s/w precipitation inside cusps, emitted
from the surface (color-coded map). Right panel: MPO orbit (green curve; example of the STROFIO FOV
(red lines) and source location of observable particle with the related energy are indicated
Fig. 41 MPO height function of anomaly angle (red scale, LHS). Black scales (RHS): minimum release
energy necessary for a particle to get from the surface up to MPO orbit, for different species (O, Na, K, Fe),
function of the anomaly angle
The major processes causing surface emission at Mercury are listed in Sect. 2. Among
these processes, ion sputtering is the most effective in transporting exospheric gas up to
BC-MPO orbit, and hence it is worth simulating the signal measured by SERENA. The
ion-sputtering energy distribution function (fs ) (15) of the ejection energy usually peaks at
few eV (Sigmund 1969; Sieveka and Johnson 1984), with a high-energy tail, up to approx-
imately one hundred of eV. The value of the binding energy Eb determines the position of
the spectrum peak, which is located at Eb /2; the projectile impact energy Ei determines the
spectrum cut-off energy.
The spectrum differs from species to species but, in general, most of the ejected particles
are able to reach the BC-MPO altitude. For sodium, as an example, the energy needed to
reach the satellite altitude goes from 0.3 eV at S/C periherm, to 0.8 eV at S/C apoherm. If
we assume the distribution given by (15), then a fraction above 90% of the ejected sodium
is able to reach such altitudes. The detection of ion-sputtering ENAs deals with several
scientific objectives of BC-MPO/SERENA (Milillo et al. 2005):
(1) particle loss-rate from Mercury’s environment and surface emissivity;
(2) analysis of ion-sputtering process, since its yield, at high energy, is not well known; and
(3) analysis of proton precipitation, as a function of external conditions, performable in
addition to other ion measurements.
The SERENA-STROFIO unit, in principle, is able to reveal sputtered neutral particles,
depending on the relative composition of the S/C and particle velocity (see Fig. 42). Since
the process is, with some approximation, stoichiometric, the mass analysis of the flux com-
ing from a certain region on the surface gives information about the composition of the soil
in that region.
Fig. 42 Left: sample of ion sputtering density contours on the x–z plane (with trace of the MPO orbit). Right:
actual STROFIO simulated data, according to instrument constraints. From Mura (2005)
Fig. 43 2-D image of surface oxygen emission, obtained by the superimposition of 31 ELENA simulated
measurements. Color is coded according to ELENA 30 s count rates, integrated over all energies above
Oxygen escape energy. From Mura et al. (2006a)
On average, approximately 50% of the particles have more energy than the escape energy
(few eV); those particles travel along quasi-linear paths, allowing a remote-sensing of the
process via the SERENA-ELENA unit. The measurement of the total neutral flux, compared
to the precipitating ion flux (observed by SERENA-MIPA), gives information about the
effectiveness of the process (i.e., the process yield Y) and on the ion flux onto the surface,
including its spatial and temporal distribution (see Fig. 43).
The 2D image of surface oxygen emission is obtained with the help of S/C motion, as
a superimposition of 31 ELENA simulated measurements as the S/C is orbiting from 30°N
to 60°N at 12:00 MLT (the periherm is at 0°, 12:00 MLT). The intense proton precipitation
in the dayside, northern hemisphere results in an intense O-ENA production in the center of
the global 2D image. In general, it is possible to have a global view of proton precipitation
by means of ion-sputtered directional ENAs, with surface spatial resolution between 15 and
Fig. 44 Simulation of ELENA measurement, according to ENA flux in Fig. 37. FOV is shown by the black
bold grid; each square represents an instrument sector (2◦ × 2◦ ). From Mura et al. (2006b)
50 km, depending on S/C altitude. The color is coded according to ELENA 30 s count rates,
integrated over all energies above Oxygen escape energy.
Under different IMF conditions, the configuration of the Hermean magnetosphere
changes, so that the area of Solar Wind proton precipitation (and subsequent neutral release)
consequently drifts and changes in size (see, e.g., Massetti et al. 2003; Kallio and Janhunen
2003). According to the simulations shown, the ion-sputtering ENA signal from the dayside
is high enough to be detected by ELENA. In fact, the spatial and time resolution is good
enough to monitor instantaneous changes of the magnetospheric configuration; the spatial
resolution permits discrimination of surface emissivity variations. The intensity of the di-
rectional ENA signal originating from ion-sputtering depends on both proton precipitation
flux and surface properties (composition and yield).
However, our simulations show that it is possible to discriminate between these two fac-
tors. In fact, proton surface precipitation has a typical spatial scale factor of about 100 km
(Mura et al. 2006b). Since ELENA spatial resolution is lower (between 15 and 50 km), any
small-scale spatial change in ion-sputtering ENA signal would be probably due to surface
property variations. Moreover, temporal variability in the ENA signal should be ascribed
only to modifications in the proton circulation properties.
Charge exchange neutrals have typical energies of 1 keV or more (see Sect. 5.4); this
process mostly occurs in the dayside and dawnside regions (Mura et al. 2005), up to altitudes
of hundreds of km (see Fig. 37). The H-ENA production rate for unit length reaches values
up to 10 (cm−2 s−1 sr−1 m−1 ), close to the dayside planetary surface. This value, in some
cases (depending also on the line-of-sight length), could lead to high values of H-ENA flux
per steradian (up to 106 cm−2 s−1 sr−1 ). To facilitate the detection of such H-ENAs, the
ELENA central axis is tilted, with respect to the S/C nadir axis, by 8°. In this way, at least
three sectors point away from the planet, if the S/C is in a ∼15° orbital arc centered at the
apoherm (see Fig. 44). In fact, the H-ENA signal is lower within sectors looking towards
the planet, because the integration path is shorter. The S/C apoherm position will move, in
longitude, during the MPO mission, thus allowing different, optimal vantage points.
As an example, from a nightside vantage point it is possible to detect the H-ENA signal
generated from two different H+ populations: the first one originates from protons that are
precipitating into the cusp regions or circulating over the North pole. The second population
originates from protons that have been drifted westward by the grad-B drift, from dayside to
nightside through the dawn region. They reach low altitudes at their mirroring points, where
exospheric densities are higher: they may produce an intense ENA signal that can be seen in
the right side of the pictures. Charge-exchange ENA will be observed also by the other ENA
sensor onboard the BC MMO satellite, with a more extended view due to the more distant
apocenter (about 15,000 km).
The occurrence of impacts on Mercury by projectiles of the radius of 1 m seems to be
not infrequent (2 events/year); such large projectiles would vaporize the regolith and reach
a depth of meters, depending on the density and porosity of the regolith itself (Mangano
et al. 2005). For the surface deeper layers are believed to be less contaminated by space
weathering (Hapke 2001), the detection of the vaporized soil due to such impulsive events
could be the only way to remote sense the real Hermean endogenous material.
One of the most interesting goals of the SERENA-STROFIO observations is the identi-
fication and detection of meteoroid impact vaporization process. In fact, as stated in Sect. 2,
among the release processes active on Mercury, refractory species are released most effi-
ciently by impact events (Gerasimov et al. 1998); hence, MIV could be a valid mechanism
by which species like Mg, Al, and Si will be detected. Furthermore, larger projectiles (for
instance, 1 m radius meteoroids that have a probability of two impacts/year) would vaporize
the regolith reaching a depth of meters, hence layers less contaminated by space weathering
(Hapke 2001), the detection of the vaporized soil due to such impulsive events could be the
only way to remote sense the real Hermean endogenous material.
The frequent impact of meteoroids 10 cm in radius (more than two events/day, (Marchi
et al. 2005)) makes them particularly interesting. In this case, the enhancement, depending
from the considered species, varies from 1 to 4 orders of magnitude higher than the mean
exospheric background values (Mangano et al. 2007). Durations are generally larger than
2,000 s, and their extension larger than 50° (calculated with respect to the center of the
planet).
Such event will be observable by SERENA-STROFIO when the spacecraft flies over
the impact zone during the event, and also by the UV spectrometer PHEBUS when the
event happens in the daylight and along the instrument look direction. Figure 45 shows
the simulated Na density observed by STROFIO as a function of time after impact and
spacecraft position with respect to impact point. The signal is stronger when the impact is
toward the ram direction (i.e., the look direction of the instrument); when the spacecraft pass
the impact point fewer particles reach the instrument FOV.
In Fig. 46 the estimated counts rate of STROFIO for O, Mg and Na versus time in the
case of a meteoritic impact under the spacecraft position are shown. These simulations seem
to assure very high detectability, reaching almost the 100% of probability after only one
month of monitoring from a probe in polar orbit and periherm at 400 km, apoherm at 1,500,
as planned to be for the BepiColombo/MPO mission (Mangano et al. 2007).
Note that exospheric spatial inhomogeneities of refractory species with similar time
scales could be generated by ion-sputtering. In this case the shape and intensity of the cloud
would be significantly different as previously mentioned. Furthermore, the sputter-generated
cloud would be characterized by an energy spectrum reaching higher energies (Wiens et
al. 1997). The energy resolution of SERENA will allow discrimination between these two
processes.
Fig. 45 Simulated Na density observed by STROFIO as a function of time after impact and spacecraft
position with respect to impact point. The field of view of the instrument (20◦ × 20◦ in the ram direction)
and the particle trajectories have been considered
6.3 3D Exospheric Modeling and Measurement Feedback
Mercury’s exosphere consists of particles in ballistic orbits which originate from various
release processes like thermal release, particle/photon sputtering and micrometeorite impact.
The spatial distribution of a neutral exospheric component can be obtained by using a Monte
Carlo single-particle model. In these models, it is assumed that the trajectory of a relatively
small, but statistically representative, amount of test particles can reproduce the trajectories
of all the real exospheric particles. A weight w is associated to each test particle, which
takes into account the number of real particles that it represents.
For a given source process, the surface S where the process occurs is defined, then some
number (Ntp ) of test particles (usually some millions) are launched from randomly chosen
starting points P0 within S : A = S/Ntp , w = j0 A, with j0 being the initial flux through
the surface element A. The starting velocity v0 must be randomly chosen according to the
velocity distribution function of the source; for an arbitrary velocity distribution function
this can be done by using a Von Neumann (1951) algorithm. Alternatively, the surface of
the planet can be divided into a number of elements and the initial flux of the sputtered
particles can be prescribed for each element according to the composition of the soil and
the sputter agents considered. In this case, several thousand particles are launched from
each element corresponding to the initial conditions associated with the surface element
and the contribution of each particle to the total number density is weighted according to
the distribution function conforming to the release process considered. The initial elevation
angle is determined via a given distribution while the azimuth angle is chosen randomly
within the interval of 2π . Although the number of particles launched from each surface
element is rather limited, their statistics are expected to approximately represent that of the
many more particles forming the real exosphere of Mercury.
Processes that Promote and Deplete the Exosphere of Mercury
Fig. 46 Schematics of a Monte Carlo model simulation at Mercury. From the left column: Energy spectra of the source process; examples of trajectories, and simulated
exospheres for different release processes (based on Mura et al. 2007). Top row: thermal desorption; middle row: photon-stimulated desorption; bottom row: ion sputtering. The
501
planetary surface (in brown) represents a cut from −90° to 90° of latitude; the curvature is not to scale
The equations of motion are solved in spherical coordinates (r, , φ) and are given by:
GM
r̈ − r θ̇ 2 − r sin2 θ φ̇ 2 = − ,
r2
2
θ̈ + ṙ θ̇ − sin θ cos θ φ̇ 2 = 0, (27)
r
2
φ̈ + ṙ φ̇ + 2 cot θ θ̇ φ̇ = 0,
r
where M is the mass of Mercury and G the gravitational constant. The trajectories of the
particles are followed until they either leave the simulation box or hit the surface of the
planet. Particles falling back onto the surface are assumed to be trapped in the soil and
are therefore excluded from further calculations. If necessary, additional forces can easily
be incorporated into the model, for example, radiation pressure is expected to significantly
affect the trajectories of Na and K and should therefore be taken into account. On the other
hand, due to the slow rotation period of Mercury (P ∼ 59 days) the acceleration caused by
the Coriolis force might be neglected. Finally, exospheric species may eventually become
ionized and removed from the exosphere. Such loss processes can be taken into account both
by removing test-particles, or by decreasing w:
dw 1
= w(t) , (28)
dt i
τi
where τi is the lifetime of process i. To obtain the particle density, it is necessary to define a
spatial grid Q. Each time a test particle crosses a grid cell, a quantity q is deposited there:
q = w(t)t, (29)
where t is the time elapsed inside the cell. After all trajectories have been simulated, the
spatial dependent number density n can then be determined by dividing Q by the volume of
the cells:
Q(r, θ, φ)
n(r, θ, φ) = , (30)
V
further dimensions of the spatial grid can be used, as an example, to store the information
about the energy of the particles.
At Mercury, several Monte Carlo models (e.g. Smith et al. 1978) have been proposed
to reproduce the exospheric density due to relevant source processes such as thermal and
photon-stimulated desorption, ion sputtering or micro meteoritic impact vaporization; as
far as it concerns particle losses, it is necessary to include at least photo-ionization, being
charge-exchange process negligible as a neutral loss mechanism Milillo et al. 2005; Mura et
al. 2006b. Figure 46 shows an example of Monte Carlo modeling of the Hermean exosphere
(based on Mura et al. 2007).
In Fig. 46, samples of exospheric vertical profiles related to various release processes
are shown. TD produces a very dense exosphere, but at very low altitudes and only in the
dayside. This process is not effective in global loss from the planet, since the fraction of
particles that escape is well below 1% on average. In the case of PSD, the fraction of particles
that are lost varies between 1% to 20% (Mura et al. 2007), while for IS it is found to be
between 50% and 80%.
Fig. 47 Examples of Monte Carlo simulations of Na vertical profiles in the exosphere of Mercury. From
the left: Thermal desorption (over sub-solar point (SSP) and over the terminator (Term)); Photon-stimulated
desorption simulated using a Maxwellian-flux distribution with T = 1,500 K (over the sub-solar point, the
terminator and over the anti-sun point (Night)); Ion-sputtering (over the point of maximum ion precipitation
(Max) and averaged (Aver.) over all the dayside surface). Mercury–Sun distance is 0.38 AU, radiation pressure
as estimated by Smyth and Marconi (1995) included. Adapted from Mura et al. (2007)
Figure 47 shows examples of Monte Carlo simulations of Na vertical profiles released

from the surface due to thermal desorption, PSD and ion sputtering.
The Mercury exosphere simulations do provide significant background information, use-
ful for the understanding of the “real” measurements which will be taken by the forthcoming
in situ missions, like Messenger and BepiColombo. To deduce the physical meaning of these
future observations, it could be worth deriving some functional forms—based on a best-fit
approach of the available simulations—and link them to the basic physical exospheric para-
meters. In the future, such parameters could be properly tuned to best-fit the real data, and
immediately provide a reliable signature of the planet’s exospheric global characteristics.
In the simplified case of a surface with uniform concentration of a given component, the
exospheric density generated from both PSD and TD has a cylindrical symmetry around the
x axis (planet to Sun direction), and peaks at the sub-solar point. In this case, the vertical
profile over this point can be reproduced by the simplified exospheric r-profile model (Rair-
den et al. 1986) by Chamberlain (1963) (first three terms of (28)). The angular dependence
expressed by Rairden et al. (1986) for the Earth cannot be applied at Mercury since the
Hermean exosphere is generated directly from the planet surface, and not from the exobase.
Moreover, the radiation pressure acceleration can be very effective at Mercury (Smyth and
Marconi 1995; Potter et al. 2002a), ranging from 0.2 to 2 m/s for Na, and from 0.3 to 3 m/s
for K The remaining terms in (28) take into account the strong variation between day and
night conditions, and describe the shape of the tail:

r −1 F /r − 1 α−π 2
log10 (n) = A + Be−C(r−1) − − + G 1 − e−I (r−1) e 2 H , (31)
D α−π/2
1 + e− E
radiation pressure effect
Raydern
day/night modulation
where n is the density, α is the solar zenith angle (from the sub-solar point), r is the plane-
tocentric distance. A through I are free parameters depending on the process involved (PSD
or TD) and on boundary conditions (Mura et al. 2007); their values have been derived by
best-fitting the outcomes from Monte Carlo numerical model previously described.
The shape of the exosphere generated from ion sputtering strongly depends on the ion
precipitation pattern onto the surface. Recently, numerical models have taken into account
the contribution of protons (Sarantos et al. 2001; Massetti et al. 2003; Kallio and Janhunen
2003; Mura et al. 2007; Leblanc and Johnson 2003), minor s/w components such as alpha
particles (Leblanc and Johnson 2003) and planetary ions (Delcourt et al. 2003). According
to most authors, in the case of s/w protons, we expect a proton precipitation flux up to 109
cm−2 s−1 in the cusp regions, located in the dayside at mid latitudes. The flux, the shape
and the relative size of these regions depend on the magnetospheric configuration. As stated
before, the fraction of particles that escape from the planet is found between 50% and 80%.
Photoionization reduces, in general, the dayside exospheric density. Since the ionized
particles are accelerated by the electromagnetic fields, this process produces a net loss from
the planet. The estimated effect in exospheric profile, however, can be neglected.
Radiation pressure produces an increase of density in the nightside and a reduction in
the dayside. This effect can be easily seen, for example, in the PSD exosphere, since par-
ticles generated by this release process have long residence time in ballistic trajectories
(103 –104 s); hence, they are efficiently accelerated in the anti-sunward direction.
7 Conclusions
The best-studied constituent in Mercury’s exosphere, sodium, displays a rich variety in its
spatial and temporal variability. Observations of this constituent have revealed a rapid vari-
ation in the ion sputtering component due to a variability in the magnetosphere, possible
latitudinal and/or longitudinal variations in composition, long-term variations in photon-
stimulated desorption and radiation pressure acceleration, long-term and short-term varia-
tions in meteoritic vaporization, and possible sequestration of volatiles on the nightside and
at high latitudes. Observations of other species, both by Mariner 10 and by ground-based
telescopes, are more sparse. A great wealth of information, both about the surface and about
the interaction of high-energy radiation and particles with the surface, is expected to be
gained with the advent of the two planned spacecraft, Messenger and BepiColombo, to the
Hermean system. We expect to discover many more species in the exosphere in addition to
the six currently known, but their expected relative abundances is widely debated. A com-
parison of simulated data with actual data from these spacecraft will allow us to test our
current theories and revise them as appropriate. The abundances of the noble gases will tell
us about processes from implantation of solar wind to the abundances in the deep interior.
Asymmetries in the abundances of exospheric species will tell us not only how volatiles
are sequestered and lost, but about the refractory species as well. We will undoubtedly be
surprised by these discoveries, as all explorers have met with surprises in the past.
Acknowledgements A. Mura and H.I.M. Lichtenegger acknowledge support by the Europlanet project
(http://www.europlanet-eu.org) for supporting working visits to the Space Science Institute of the Austrian
Academy of Sciences in Graz, Austria and to the Istituto di Fisica dello Spazio Interplanetario-CNR, in Rome,
Italy.
References
M.R. Aellig, A.J. Lazarus, J.T. Steinberg, in Solar and Galactic Composition, ed. by R.F. Wimmer-
Schweingruber (2001)
M.A. A’Hearn, M.J.S. Belton, W.A. Delamere, J. Kissel, The Deep Impact Team, Science 310, 258–264
(2005)
T.J. Ahrens, D.M. Cole, Proc. Lunar Sci. Conf. 5th, (1974), pp. 2333–2345
T.J. Ahrens, J.D. O’Keefe, Moon 4, 214–249 (1972)
F. Aumayr, H. Winter, Phil. Trans. Roy. Soc. Lond. A 362, 77–102 (2004)
C. Barbieri, S. Verani, G. Cremonese, A. Sprague, M. Mendillo, R. Cosentino, D. Hunten, Planet. Space Sci.
52, 1169–1175 (2004)
A. Benninghoven, F.G. Rüdenauer, H.W. Werner, Secondary Ion Mass Spectrometry: Basic Concepts, Instru-
mental Aspects, Applications and Trends (Wiley, New York, 1987), 1227 pp
G. Betz, W. Husinsky, Phil. Trans. Roy. Soc. Lond. A 362, 177–194 (2004)
T.A. Bida, R.M. Killen, T.H. Morgan, Nature 404, 159–161 (2000)
D.T. Blewett, P.G. Lucey, B.R. Hawke, G.G. Ling, M.S. Robinson, Icarus 129, 217–231 (1997)
V. Bothmer, R. Schwenn, Ann. Geophys. 16, 1–24 (1998)
W.F. Bottke, A. Morbidelli, R. Jedicke, J.M. Petit, P. Levison, H.F. Michel, T.S. Metcalfe, Icarus 156, 399–
433 (2002)
A.L. Broadfoot, D.E. Shemansky, S. Kumar, Geophys. Res. Lett. 3, 577–580 (1976)
M.E. Brown, Icarus 151, 190–195 (2001)
P. Brown, R.E. Spalding, D.O. ReVelle, E. Tagliaferri, S.P. Worden, Nature 420, 294–296 (2002)
T.H. Burbine, T.J. McCoy, L.R. Nittler, G.K. Benedix, E.A. Cloutis, T.L. Dickinson, Science 37, 1233–1244
(2002)
B. Butler, D. Muhleman, M. Slade, J. Geophys. Res. 98, 15,003–15,023 (1993)
M. Bruno, G. Cremonese, S. Marchi, Mon. Not. Roy. Astron. Soc. 1, 1067–1071 (2006)
T.A. Cassidy, R.E. Johnson, Icarus 176, 499–507 (2005)
J.W. Chamberlain, Planet. Space Sci. 11, 901–960 (1963)
A. Ciaravella, J.C. Raymond, A. van Ballegooijen, L. Strachan, A. Vourlidas, J. Li, J. Chen, A. Panasyuk,
Astrophys. J. 597, 1118–1134 (2003)
M. Cintala, J. Geophys. Res. 97, 947–973 (1992)
E.W. Cliver, O.C. St. Cyr, R.A. Howard, P.S. Mc Intosh, in Solar Coronal Structures, ed. by V. Rusin,
J. Heinzel, C. Vial (VEDA Publishing House of the Slovak Academy of Sciences, 1994), pp. 83–89
C.M.S. Cohen, E.C. Stone, R.A. Mewaldt, R.A. Leske, G.M. Cummings, A.C. Mason, M.I. Desai, T.T. von
Rosenvinge, M.E. Wiedenbeck, J. Geophys. Res. 110, A09S16 (2005)
M.R. Combi, M.A. Disanti, U. Fink, Icarus 130, 336–354 (1997)
J.E.P. Connerney, N.F. Ness, in Mercury, ed. by F. Vilas, C.R. Chapman, M.S. Matthews (University of
Arizona Press, Tucson, 1988), pp. 494–513
S.R. Coon, W.F. Calaway, M.J. Pellin, J.M. White, Surf. Sci. 298, 161–172 (1993)
J. Crank, The Mathematics of Diffusion, 2nd edn. (Oxford Univ. Press, Oxford, 1975)
G. Cremonese, H. Boehnhardt, J. Crovisier, H. Rauer, A. Fitzsimmons, M. Fulle, J. Licandro, D. Pollacco,
G.P. Tozzi, R.M. West, Astrophys. J. Lett. 490, L199–L202 (1997)
G. Cremonese, M. Bruno, V. Mangano, S. Marchi, A. Milillo, Icarus 177, 122–128 (2005)
G. Cremonese, M. Bruno, V. Mangano, S. Marchi, A. Milillo, Icarus 182, 297–298 (2006)
D.C. Delcourt, T.E. Moore, S. Orsini, A. Millilo, J.A. Sauvaud, Geophys. Res. Lett. 29 (2002). doi:10.1029/
2001GL013829
D.C. Delcourt, S. Grimald, F. Leblanc, J.J. Berthelier, A. Millilo, A. Mura, S. Orsini, T.E. Moore, Ann.
Geophys. 21, 1723–1736 (2003)
R.C. Elphic, H.O. Funsten III, R.L. Hervig, Lunar Planet. Sci. Conf. Abst. 24, 439 (1993)
G. Fjeldbo, A. Kliore, D. Sweetnam, P. Esposito, B. Seidel, T. Howard, Icarus 29, 439–444 (1976)
B.C. Flynn, S.A. Stern, Icarus 124, 530–536 (1996)
W.A. Gault, H.N. Rundle, Can. J. Phys. 47, 85–98 (1969)
D.E. Gault, E.F. Horz, D.E. Brownlee, J.B. Hartung, Lunar. Planet Sci. Conf. 5, 260 (1974)
M.V. Gerasimov, B.A. Ivanov, O.I. Yakovlev, Earth Moon Planet. 80, 209–259 (1998)
H. Gnaser, W.O. Hofer, Appl. Phys. A 48, 261–271 (1989)

B.E. Goldstein, S.T. Suess, R.J. Walker, J. Geophys. Res. 86, 5485–5499 (1981)
N. Gopalswamy, in The Sun and the Heliosphere as an Integrated system, ed. by G. Poletto, S. Suess. ASSL
Series (Kluwer, 2004), pp. 201–240
N. Gopalswamy, M.R. Kundu, Sol. Phys. 143, 327–343 (1993)
N. Gopalswamy, S. Yashiro, M.L. Kaiser, R.A. Howard, J.L. Bougeret, J. Geophys. Res. 106, 29,219–29,230
(2001a)
N. Gopalswamy, A. Lara, S. Yashiro, M.L. Kaiser, R.A. Howard, J. Geophys. Res. 106, 29,207–29,218
(2001b)
N. Gopalswamy, A. Lara, S. Yashiro, R.A. Howard, Astrophys. J. 598, L63–L66 (2003)
N. Gopalswamy, S. Nunes, S. Yashiro, R.A. Howard, Adv. Space Res. 34(2), 391–396 (2004). doi:10.1016/
j.asr.2003.10.054
C.S. Hansen, W.F. Calaway, B.V. King, M.J. Pellin, Surf. Sci. 398, 211–220 (1998)
C.S. Hansen, W.F. Calaway, M.J. Pellin, B.V. King, A. Wucher, Surf. Sci. 432, 199–210 (1999)
B. Hapke, J. Geophys. Res. 106, 10,039–10,073 (2001)
R.A. Haring, A. Haring, F.W. Saris, A.A. de Vries, Appl. Phys. Lett. 41, 174–176 (1982)
J.K. Harmon, Adv. Space Res. 19, 1487–1496 (1997)
J.K. Harmon, M.A. Slade, Science 258, 640–642 (1992)
J.K. Harmon, P.J. Perilat, M.A. Slade, Icarus 149, 1–15 (2001)
G.B. Hasted, Physics of Atomic Collisions (Butterworths, London, 1964), p. 416
G. Heiken, D. Vaniman, B.M. French, Lunar Sourcebook: A User’s Guide to the Moon (Cambridge Univ.
Press, Cambridge, 1991)
T. Henke, J. Woch, U. Mall, S. Livi, B. Wilken, R. Schwenn, G. Gloeckler, R. von Steiger, R.J. Forsyth, A.
Balogh, Geophys. Res. Lett. 25, 3465–3468 (1998)
E. Hildner, J.T. Gosling, R.M. MacQueen, R.H. Munro, A.I. Poland, C.L. Ross, Sol. Phys. 48, 127–135
(1976)
R.R. Hodges Jr., J. Geophys. Res. 79, 2881–2885 (1974)
W.O. Hofer, in Sputtering by Particle Bombardment, ed. by R. Behrisch, R.K. Wittmaack (1991), pp. 15–90
L. Holmlid, J. Phys. Chem. 102, 10,636–10,646 (1998)
L. Holmlid, Planet. Space Sci. 54, 101–112 (2006)
L. Holmlid, J.O. Olsson, Surf. Sci. 67, 61–76 (1977)
G.P. Horedt, G. Neukum, Icarus 60, 710–717 (1984)
R.A. Howard, D. Michels, N.R. Sheeley, M.J. Koomen, in The Sun and the Heliosphere in Three Dimensions,
ed. by R. Marsden, D. Reidel. ASSL, vol. 123 (Norwell, 1986), pp. 107–111
W.F. Huebner, J.J. Keady, S.P. Lyon, Astrophys. Space Phys. 195, 1–294 (1992)
D.M. Hunten, A.L. Sprague, Adv. Space Res. 19, 1551–1560 (1997)
D.M. Hunten, A.L. Sprague, Meteorit. Planet. Sci. 37, 1191–1195 (2002)
D.M. Hunten, L.V. Wallace, Astrophys. J. 417, 757–761 (1993)
D.M. Hunten, T.H. Morgan, D. Shemansky, in Mercury, ed. by F. Vilas, C.R. Chapman, M.S. Matthews (Univ.
of Arizona Press, Tucson, 1988), pp. 562–612
A.P. Ingersoll, T. Svitek, B.C. Murray, Icarus 100, 40–47 (1992)
W.H. Ip, Geophys. Res. Lett. 13, 423–426 (1986)
W.H. Ip, Icarus 71, 441–447 (1987)
W.H. Ip, Astrophys. J. 356, 675–681 (1990)
W.H. Ip, A. Kopp, J. Geophys. Res. 07 (2002). doi:10.1029/2001JA009171
R.E. Johnson, Geophys. Monogr. 130, 203–219 (2002)
K. Kabin, T.I. Gombosi, D.L. DeZeeuw, K.G. Powell, Icarus 143, 397–406 (2000)
E. Kallio, P. Janhunen, Geophys. Res. Lett. 30, (2003). doi:10.1029/2003GL017842
R.M. Killen, Meteorit. Planet. Sci. 37, 1223–1231 (2002)
R.M. Killen, Publ. Astron. Soc. Pac. 118, 1347–1353 (2006)
R.M. Killen, W.H. Ip, Rev. Geophys. 37, 361–406 (1999)
R.M. Killen, T.H. Morgan, Icarus 101, 294–312 (1993)
R.M. Killen, A.E. Potter, T.H. Morgan, Icarus 85, 145–167 (1990)
R.M. Killen, A.E. Potter, T.H. Morgan, Science 252, 474–475 (1991)
R.M. Killen, A.E. Potter, A. Fitzsimmons, T.H. Morgan, Planet. Space Sci. 47, 1449–1458 (1999)
R.M. Killen, A.E. Potter, P. Reiff, M. Sarantos, B.V. Jackson, P. Hick, B. Giles, J. Geophys. Res. 106, 20,509–
20,526 (2001)
R.M. Killen, A.E. Potter, M. Sarantos, P. Reiff, Mercury, 25th Meeting of the IAU, Joint Discussion 2, Sydney,
Australia, 2003. Meeting abstract
R.M. Killen, M. Sarantos, P.H. Reiff, Adv. Space Res. 33, 1899–1904 (2004a)
R.M. Killen, M. Sarantos, A.E. Potter, P. Reiff, Icarus 171, 1–19 (2004b)
R.M. Killen, T.A. Bida, T.H. Morgan, Icarus 173, 300–311 (2005)
P.L. Koehn, A.L. Sprague, Planet. Space Sci. (2007). doi:10.1016/j.pss.2006.10.009
P.L. Koehn, T.H. Zurbuchen, K. Kabin, DPS Abstract, 35.2308, 2003
A. Kotarba, I. Kruk, Z. Sojka, J. Catal. 221(2), 650–652 (2004)
G.A. Krasinsky, E.V. Pitjeva, M.V. Vasilyev, E.I. Yagudina, Icarus 158, 98–105 (2002)
H. Lammer, P. Wurz, M.R. Patel, R. Killen, C. Kolb, S. Massetti, S. Orsini, A. Milillo, Icarus 166, 238–247
(2003)
A. Lara, J.A. González-Esparza, N. Gopalswamy, Geofísica Internacional 43, 75–82 (2004)
F. Leblanc, R.E. Johnson, Icarus 164, 261–281 (2003)
F. Leblanc, D. Delcourt, R.E. Johnson, J. Geophys. Res. 108, (2003a). doi:10.1029/2003JE002151
F. Leblanc, J.G. Luhmann, R.E. Johnson, M. Liu, Planet. Space Sci. 51, 339–352 (2003b)
F. Leblanc, C. Barbieri, G. Cremonese, S. Verani, R. Cosentino, M. Mendillo, A. Sprague, D. Hunten, Icarus
185, 395–402 (2006)
S.T. Lepri, T.H. Zurbuchen, L.A. Fisk, I.G. Richardson, H.V. Cane, G. Gloeckler, J. Geophys. Res. 106,
29,231–29,238 (2001)
G.M. Lindsay, J.G. Luhmann, C.T. Russell, J.T. Gosling, J. Geophys. Res. 104, 12515–12524 (1999)
K. Lodders, B. Fegley, The Planetary Scientists Companion (Oxford University Press, 1998)
J.G. Luhmann, C.T. Russell, N.A. Tsyganenko, J. Geophys. Res. 103, 9113–9119 (1998)
A.V. Lukyanov, S. Barabash, R. Lundin, P. C:son Brandt, Planet. Space Sci. 49, 1677–1684 (2001)
T.E. Madey, B.V. Yakshinskiy, V.N. Ageev, R.E. Johnson, J. Geophys. Res. 103, 5873 (1998)
A. Mallama, D. Wang, R.A. Howard, Icarus 155, 253–264 (2002)
V. Mangano, A. Milillo, S. Orsini, A. Mura, H. Lammer, P. Wurz, EGU Abstract, EGU05-A-01247, 2005
V. Mangano, A. Milillo, A. Mura, S. Orsini, E. De Angelis, A.M. Di Lellis, P. Wurz, Planet. Space Sci. (2007).
doi:10.1016/j.pss.2006.10.008
S. Marchi, A. Morbidelli, G. Cremonese, Astron. Astrophys. 431, 1123–1127 (2005)
G.M. Mason, J.E. Mazur, J.R. Dwyer, Astrophys. J. 525, L133–L136 (1999)
S. Massetti, S. Orsini, A. Milillo, A. Mura, E. De Angelis, H. Lammer, P. Wurz, Icarus 166, 229–237 (2003)
S. Massetti, S. Orsini, A. Milillo, A. Mura, Planet. Space Sci. (2007). doi:10.1016/j.pss.2006.12.008
M.A. McGrath, R.E. Johnson, L.J. Lanzerotti, Nature 323, 696–696 (1986)
A.S. Milillo, S. Orsini, P. Wurz, D. Delcourt, E. Kallio, H. Rillen, R.M. Lammer, S. Massetti, A. Mura, S.
Barabash, G. Cremonese, I.A. Daglis, E. De Angelis, A.M. Di Lellis, S. Livi, V. Mangano, K. Torka,
Space Sci. Rev. 117, 397–444 (2005)
A. Morbidelli, B. Gladman, Meteorit. Planet. Sci. 33, 999–1016 (1998)
T.H. Morgan, R.M. Killen, Planet. Space Sci. 45, 81–94 (1997)
T.H. Morgan, H.A. Zook, A.E. Potter, Icarus 74, 156–170 (1988)
A. Mura, Neutral Atom Emission from Mercury Magnetosphere. AOG, Conference. Singapore, June 20–24,
2005
A. Mura, S. Orsini, A. Milillo, D. Delcourt, S. Massetti, E. De Angelis, Icarus 175, 305–319 (2005)
A. Mura, S. Orsini, A. Milillo, A.M. Di Lellis, E. De Angelis, Planet. Space Sci. 54, 144–152 (2006a)
A. Mura, S. Orsini, A. Milillo, D. Delcourt, A.M. Di Lellis, E. De Angelis, S. Massetti, Adv. Geosci. 3
(2006b). ISBN 981-256-983-8
A. Mura, D. Delcourt, S. Massetti, A. Milillo, S. Orsini, A. Di Lellis, E. De Angelis, Geophysical Research
Abstracts, vol. 8, 06958, EGU General Assembly, Vienna (Austria), 2–7 April (2006c)
A. Mura, A. Milillo, S. Orsini, S. Massetti, Planet. Space Sci. (2007). doi:10.1016/j.pss.2006.11.028
N.F. Ness, K.W. Behannon, R.P. Lepping, Y.C. Wang, Nature 255, 204–205 (1975)
H. Oechsner, H. Schoof, E. Stumpe, Surf. Sci. 76, 343–354 (1978)
J.D. O’Keefe, T.J. Ahrens, Science 234, 346–349 (1986)
D.A. Paige, S.E. Wood, A.R. Vasavada, Science 258, 643–646 (1992)
A.E. Potter, Geophys. Res. Lett. 22, 3289–3292 (1995)
A.E. Potter, T.H. Morgan, Science 229, 651–653 (1985)
A.E. Potter, T.H. Morgan, Icarus 67, 336–340 (1986)
A.E. Potter, T.H. Morgan, Icarus 71, 472–477 (1987)
A.E. Potter, T.H. Morgan, Planet. Space Sci. 45, 95–100 (1997)
A.E. Potter, R.M. Killen, T.H. Morgan, Space Sci. 47, 1141–1148 (1999)
A.E. Potter, C.M. Anderson, R.M. Killen, T.H. Morgan, J. Geophys. Res. Planets 107 (2002a). doi:10.1029/
2000JE0014937
A.E. Potter, R.M. Killen, T.H. Morgan, Meteorit. Planet. Sci. 37, 1165–1172 (2002b)
A.E. Potter, R.M. Killen, M. Sarantos, Icarus 181, 1–12 (2006). doi:10.1016/j.icarus.2005.10.026
A.E. Potter, R.M. Killen, T.H. Morgan, Icarus 186(2), 571–580 (2007)
R.L. Rairden, L.A. Frank, J.D. Craven, J. Geophys. Res. 91(A12), 13613–13630 (1986)
D.V. Reames, S.W. Kahler, C.K. Ng, Astrophys. J. 491, 414–420 (1997)
F.J.M. Rietmeijer, Advanced Mineralogy (Springer, Berlin, 1998), pp. 22–28
M. Sarantos, Ion Trajectories in Mercury’s Magnetosphere, PhD thesis (Rice University, Houston, 2005)
M. Sarantos, P.H. Reiff, T.H. Hill, R.M. Killen, A.L. Urquhart, Planet. Space Sci. 49, 1629–1635 (2001)
M. Sarantos, R.M. Killen, D. Kim, Planet. Space Sci. (2007). doi:10.1016/j.pss.2006.10.011
H. Schleicher, G. Wiedemann, H. Wohl, T. Berkefeld, D. Soltau, Astron. Astrophys. 425, 1119–1124 (2004)
N. Sheeley, R.A. Howard, M.J. Koomen, D.J. Michels, Astrophys. J. 272, 349–354 (1983)
D.E. Shemansky, Mercury Messenger 2, 1 (1988)
D.E. Shemansky, AIP Conf. Proc. 63, Rarefied Gas Dynamics 23rd Intl. Symposium, 2003, p. 687
D.E. Shemansky, T.H. Morgan, Geophys. Res. Lett. 18, 1659–1662 (1991)
E.M. Sieveka, R.E. Johnson, Astrophys. J. 287, 418–426 (1984)
P. Sigmund, Phys. Rev. 184, 383–416 (1969)
G. Siscoe, L. Christopher, Geophys. Res. Lett. 2, 158–160 (1975)
M. Slade, B. Butler, D. Muhleman, Science 258, 635–640 (1992)
G.R. Smith, D.E. Shemansky, A.L. Broadfoot, L. Wallace, J. Geophys. Res. 83, 3783–3790 (1978)
W.H. Smyth, M.L. Marconi, Astrophys. J. 441, 839–864 (1995)
A.L. Sprague, J. Geophys. Res. 97, 18257–18264 (1992)
A.L. Sprague, T.L. Roush, Icarus 133, 174–183 (1998)
A.L. Sprague, R.W.H. Kozlowski, D.M. Hunten, Science 249, 1140–1142 (1990)
A.L. Sprague, R.W.H. Kozlowski, D.M. Hunten, F.A. Grosse, Icarus 104, 33–37 (1993)
A.L. Sprague, R.W.H. Kozlowski, F.C. Witteborn, D.P. Cruikshank, D.H. Wooden, Icarus 109, 156–167
(1994)
A.L. Sprague, D.M. Hunten, K. Lodders, Icarus 118, 211–215 (1995)
A.L. Sprague, D.M. Hunten, F.A. Grosse, Icarus 123, 345–349 (1996)
A.L. Sprague, R.W.H. Kozlowski, D.M. Hunten, N.M. Schneider, D.L. Domingue, W.K. Wells, W. Schmitt,
U. Fink, Icarus 129, 506–527 (1997)
A.L. Sprague, W.J. Schmitt, R.E. Hill, Icarus 136, 60–68 (1998)
A.L. Sprague, J.P. Emery, K.L. Donaldson, R.W. Russell, D.K. Lynch, A.L. Mazuk, Meteorit. Planet. Sci. 37,
1255–1268 (2002)
L.V. Starukhina, Proc. Lunar Planet. Sci. Conf. 31, 1301 (2000)
R.F. Stebbings, C.H. Smith, H. Ehrahardt, J. Geophys. Res. 69, 2349 (1964)
O.C. St. Cyr, R.A. Howard, N.R. Sheeley Jr., S.P. Plunkett, D.J. Michels, S.E. Paswaters, M.J. Koomen, G.M.
Simnett, B.J. Thompson, J.B. Gurman, R. Schwenn, D.F. Webb, E. Hildner, P.L. Lamy, J. Geophys. Res.
105, 18169–18185 (2000)
J. Von Neumann, Various Techniques Used in Connection with Random Digits. National Bureau of Standard
Applied Mathematics Series, vol. 12. (1951), pp. 36–38
A. Vourlidas, D. Buzasi, R.A. Howard, E. Esfandiari, in Solar Variability: From Core to Outer Frontiers, ed.
by A. Wilson, ESA SP-506 (ESA Publication, Noordwijk, 2002), pp. 91–94
J. Warell, Icarus 161, 199–222 (2003)
J. Warell, D.T. Blewett, Icarus 168, 257–276 (2004)
D.F. Webb, R.A. Howard, J. Geophys. Res. 99, 4201–4220 (1994)
M.E. Wiedenbeck et al., AGU Fall Mtg 2005, abstract SH11B-0267
R.C. Wiens, D.S. Burnett, W.F. Calaway, C.S. Hansen, K.R. Kykkem, M.L. Pellin, Icarus 128, 386–397
(1997)
H.F. Winters, J.W. Coburn, Surf. Sci. Rep. 14(3), 161–269 (1992)
H.F. Winters, J.W. Coburn, T.J. Chuang, J. Vac. Sci. Technol. B 1(2), 469–480 (1983)
A. Wucher, H. Oechsner, Nucl. Instr. Methods B18, 458–463 (1986)
P. Wurz, in The Dynamic Sun: Challenges for Theory and Observations. ESA SP-600 (2005), pp. 5.2, 1-9
P. Wurz, H. Lammer, Icarus 164, 1–13 (2003)
P. Wurz, W. Husinsky, G. Betz, in Symposium on Surface Science, ed. by J.J. Launois, B. Mutaftschiev, M.R.
Tempère (La Plagne, France, 1990), pp. 181–185
P. Wurz, W. Husinsky, G. Betz, Appl. Phys. A 52, 213–217 (1991)
P. Wurz, R.F. Wimmer-Schweingruber, K. Issautier, P. Bochsler, A.B. Galvin, F.M. Ipavich, Composition
of Magnetic Cloud Plasmas During 1997 and 1998. CP-598 (American Institute Physics on Solar and
Galactic Composition, 2001), pp. 145–151
P. Wurz, R. Wimmer-Schweingruber, P. Bochsler, A. Galvin, J.A. Paquette, F. Ipavich, in Solar Wind X
(American Institute Physics, 2003), pp. 679, 685–690
P. Wurz, U. Rohner, J.A. Whitby, C. Kolb, H. Lammer, P. Dobnikar, J.A. Martín-Fernández, Icarus (2007).
doi:10.1016/j.icarus.2007.04.034
B.V. Yakshinskiy, T.E. Madey, Nature 400, 642–644 (1999)
B.V. Yakshinskiy, T.E. Madey, Icarus 168, 53–59 (2004)

S. Yashiro, N. Gopalswamy, G. Michalek, O.C. St. Cyr, S.P. Plunkett, N.B. Rich, R.A. Howard, J. Geophys.
Res. 109(A7) (2004). doi:10.1029/2003JA010282
J. Zhang, K.P. Dere, R.A. Howard, M.R. Kundu, S.M. White, Astrophys. J. 559, 452–642 (2001)
J.F. Ziegler, J.P. Biersack, TRIM and SRIM Program Version SRIM-2003.26 (Pergamon, New York, 1985).
http://www.srim.org/

Killen 2007

Uploaded by

Copyright:

Available Formats

Killen 2007

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Killen 2007

Uploaded by

Copyright:

Available Formats

Space Sci Rev (2007) 132: 433–509

Processes that Promote and Deplete the Exosphere

Rosemary Killen · Gabrielle Cremonese · Helmut Lammer · Stefano Orsini ·

H. Lammer · M.L. Khodachenko · H.I.M. Lichtenegger

S. Orsini · A. Milillo · A. Mura

fluctuations in temperature and experiences differences of insolation with longitude. Be-

Keywords Mercury · Exosphere · Surface composition · Particle release processes

1 Observations of the Mercury Exosphere

Early ground-based efforts to detect an atmosphere on Mercury were unsuccessful, leading

1.1 Sodium in Mercury’s Exosphere

Fig. 1 Sodium discovery

Table 1 Column densities of sodium in the Mercury exosphere

Observer Column content, Remarks

8.1 ± 1.0 × 1011* Slit spectra

1.2 Temperature of the Sodium Exosphere

Fig. 2 The sodium D2 line

Observer/author Location Temperature, K

Shemansky and Morgan (1991) Planetary average 1000

Killen et al. (1999) North polar region 750

Schleicher et al. (2004) North polar region 1380 ± 307

1.3 Spatial Variation of Sodium Emission

1.4 North–south Emission Peaks

Fig. 4 Maps of the sodium D2

Fig. 5 Ratios of the northern

Date Terminator True Radiation Terminator Bright limb

Aug. 31, 1999 Dusk 39.4 166.9 270.1 65.0

Fig. 6 Distribution of sodium

1.5 Diurnal Variation of Sodium Distribution

Fig. 7 The variation of sodium

at perihelion, rising to a maximum value of +3.38 degrees/day at aphelion. Consequently,

1.6 Variation of Sodium Emission with Time

1.7 Effects of Radiation Acceleration on Sodium Emission

Fig. 11 The normalized sodium

1.8 The Sodium Tail of Mercury

Fig. 14 The sodium tail of

1.9 Potassium in the Exosphere of Mercury

Fig. 15 Mercury reflectance

Fig. 16 Potassium observations

Fig. 17 Potassium observations

presented an alternate interpretation of the observation: because the localized sources of

Fig. 19 The ratio of sodium to

Table 4 Sodium to potassium

Mercury exosphere 40–140 Potter et al. (2002a)

1.10 Calcium in the Exosphere of Mercury

Fig. 20 Variation of column

1.11 Unknown Exospheric Components and Surface Composition

2 Exospheric Sources and Their Models

2.1 Impact Vaporization

Table 5 Expected abundances in

a Hunten et al. (1988):

Table 6 Vaporization rates for sodium at Mercury due to micrometeoritic bombardment

Reference Orbital Projectile Total Vaporization rate Na

Morgan et al. (1988) aphelion combination 1.74 × 10−15 2.27 × 105

2.2 Interplanetary Space at Mercury’s Orbit

2.3 Impacts of Meteoroids

2.4 Physics of the Evaporation and Production Rate

calculated by the following relation (Morgan and Killen 1997)

Fig. 25 Cumulative production

has been calculated as follows (Bruno et al. 2006)

3 Exospheric Sources and Their Models