Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~bradw/cv/papers/CXwinds.ps Äàòà èçìåíåíèÿ: Wed Jan 24 00:59:30 2001 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 06:58:36 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: annular

L57
The Astrophysical Journal, 546:L57--L60, 2001 January 1
# 2001. The American Astronomical Society. All rights reserved. Printed in U.S.A.
OBSERVABILITY OF STELLAR WINDS FROM LATE­TYPE DWARFS VIA CHARGE EXCHANGE X­RAY EMISSION
Bradford J. Wargelin and Jeremy J. Drake
Harvard­Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138; bwargelin@cfa.harvard.edu
Received 2000 September 16; accepted 2000 October 25; published 2000 December 29
ABSTRACT
Despite the fact that the overwhelming majority of stars are of late spectral type (F--M) and lie on the main
sequence, we know nothing about their stellar winds. Existing measurements of winds only apply to high­mass O
and B stars, red giants, and supergiants and only extend down to a few times 10 #10 yr #1 , as compared to the
M,
solar rate of # yr #1 . Attempts to detect winds from late­type dwarf stars have to date resulted only
#14
2 # 10 M,
in loose upper limits of order 10 #12 to 10 #11 yr #1 . We propose a novel method of studying stellar winds through
M,
observation of charge exchange--induced X­ray emission. Recent X­ray detections of comets suggest that charge
transfer between highly charged ions in the solar wind and neutral gases in cometary atmospheres is responsible
for much or all of the observed emission, a hypothesis that has been strengthened by Chandra observations of
comet C/1999 S4 (LINEAR). It has also been proposed that charge transfer between the solar wind and the local
interstellar medium (ISM) produces a substantial fraction of the soft X­ray background observed by ROSAT and
various rocket experiments. We show that the same process may be observable in nearby dwarf star systems using
Chandra and future large­area high­resolution observatories, which would provide hitherto unobtainable information
on wind geometry, ion composition, mass­loss rates, and the distribution of neutral gas in the ISM.
Subject headings: atomic processes --- ISM: structure --- stars: winds, outflows --- X­rays: stars
On­line material: color figures
1. INTRODUCTION
Although it is generally assumed that all stars lose mass via
stellar winds, our understanding of that process, even for the
Sun, is poor. Current estimates of stellar mass­loss rates rely on
measurements of P Cygni profiles, optical and molecular emis­
sion lines, IR and radio excesses, and absorption lines in bi­
nary systems and range from a few times 10 #10 to more than
10 #5 yr #1 (Lamers & Cassinelli 1999). These rates, however,
M,
apply only for massive OB and Wolf­Rayet stars or cool red
giants and supergiants. No data exist for the more modest stellar
winds from late­type main­sequence stars such as the Sun, a G2
dwarf that has a mass­loss rate of # yr #1 , more
#14
2 # 10 M,
than four decades lower than the current limit for nonsolar
measurements.
Measurements of stellar wind parameters such as mass loss,
wind velocity, and ion composition are needed to constrain mod­
els of stellar evolution, mass­loss mechanisms, and coronal cool­
ing as well as permit estimates of the rate of element dispersal,
kinetic heating, and ionization within the interstellar medium
(ISM). The mass­loss rate is of critical importance in models of
angular momentum loss in late­type stars, but it has proven very
difficult to estimate such properties from theoretical first prin­
ciples (see, for example, Krishnamurthi et al. 1997).
Another example of the importance of dwarf star winds is the
proposal that such winds are responsible for sweeping out the
gas and dust ejected by red giants in globular clusters (Coleman
& Worden 1977). Recently, Smith (1999) has shown that if all
M dwarfs have the same wind velocity and mass­loss rate per
unit mass as the Sun, then the resulting outflows are sufficient
to strip red giant ejecta from globular clusters, assuming plausible
values for the giants' mass­loss rates. The case for cluster gas
stripping by dwarfs is made even stronger if one assumes that
low­mass flare stars have significantly higher mass­loss rates, as
seems plausible (Badalyan & Livshits 1992).
Several authors have argued that main­sequence stars may
have mass­loss rates near or just below the current detection
threshold of a few times 10 #10 yr #1 . Willson, Bowen, &
M,
Struck­Marcell (1987) suggested that dwarf A and F stars that
lie in the d Scuti pulsational instability strip might have mass­
loss rates between 10 #9 and 10 #8 yr #1 , but Brown et al.
M,
(1990) argued that Very Large Array measurements placed
much lower limits on the ionized mass­loss rate from such stars
and that this mass­loss rate was representative of the total rate.
Mullan et al. (1992) analyzed data from the James Clerk
Maxwell Telescope and concluded that flaring M stars might
have mass­loss rates of a few times 10 #10 yr #1 , but other
M,
authors using more detailed models have concluded that the
loss rates cannot be higher than a few times 10 #12 yr #1
M,
(Lim & White 1996; van den Oord & Doyle 1997). As we will
show, however, even these modest mass­loss rates should be
detectable with current and planned X­ray observatories.
2. CHARGE EXCHANGE X­RAYS FROM THE SOLAR WIND
It has been known for decades that charge exchange (CX)
occurs between solar wind ions and neutral gas in the ISM,
but it is only very recently that the resulting X­ray emission
from highly charged ions in the wind has been considered. The
basic process in CX is the radiationless collisional transfer of
one or sometimes multiple electrons from a neutral atom or
molecule to an ion. Electron transfer can also occur between
two ions, but Coulombic repulsion greatly reduces the inter­
action cross section. The recipient ion is, for X­ray--emitting
highly charged ions, left in a high­n excited state, which then
decays via single or sequential radiative transitions.
The first astrophysical observation of such X­ray emission
was from comets (Lisse et al. 1996; Dennerl, Englhauser, &
Tru˜mper 1997). As first proposed by Cravens (1997), these
X­rays are produced via the charge exchange of highly charged
solar wind ions with neutral atoms and molecules in the comet's
atmosphere. Subsequent papers (Ha˜berli et al. 1997; Krasno­
polsky 1997; Wegmann et al. 1998; Lisse et al. 1999; Neu­
gebauer et al. 2000; Schwadron & Cravens 2000; Kharchenko

L58 X­RAY DETECTION OF STELLAR WINDS Vol. 546
& Dalgarno 2000) have supported and expanded that idea and
have included detailed models of the expected X­ray spectrum
(Kharchenko & Dalgarno 2000), which is dominated from 200
to 1000 eV by K shell emission from H­like and He­like ions
of C, O, N, and Ne. Recent Chandra observations of comet
C/1999 S4 (LINEAR) by Lisse et al. (2000) detected prominent
O vii emission at 570 eV along with several other weaker lines,
providing strong evidence that CX is indeed the dominant
X­ray emission mechanism.
Following the observation of cometary X­rays, Cox (1998)
suggested that X­rays should also be produced from CX of the
solar wind with neutral gas streaming into the heliosphere from
the ISM. Using a fairly conservative model, Cravens (2000)
estimated that this mechanism accounts for about 25%--50% of
the observed soft X­ray background below roughly 0.5 keV. This
is consistent with data from the Wisconsin soft X­ray background
sky survey (McCammon & Sanders 1990) and ROSAT obser­
vations (Snowden et al. 1994, 1995, 1998), which suggest that
roughly half of the keV background emission comes from a
1
4
``local hot plasma.'' Cravens also pointed out that temporal var­
iations seen in the ROSAT data are similar to what would be
expected from CX X­ray emission caused by variations in the
solar wind, further strengthening a heliospheric interpretation.
This same process must occur for all stars with highly
charged winds, and we show that large­area high­resolution
observatories such as Chandra should be capable of detecting
winds that are only moderately stronger than the Sun's, cor­
responding to mass­loss rates orders of magnitude smaller than
have been previously measurable.
3. EMISSION FROM STELLAR WINDS
As most recently noted by Kharchenko & Dalgarno (2000),
the rate of CX emission is higher at energies below 200 eV
than above 300 eV because only the most highly charged ions
emit photons at X­ray energies, while virtually all metal ions
in the solar wind emit at extreme ultraviolet energies and below.
From the viewpoint of detecting CX/wind emission, however,
the higher energy photons are easier to see because of their
higher contrast with emission from the central star.
In addition, useful spectral information can be obtained at
X­ray energies with nondispersive detectors, unlike at lower
energies. Nevertheless, extreme UV observations might be very
revealing, but no EUV missions to date have had the requisite
combination of large collecting area, high spatial resolution,
and low detector background.
The following discussion focuses on CX emission from He­
like O vii, which is the brightest component of solar­cometary
X­ray spectra and likely to be so in other stellar systems. The
population of bare and H­like C ions in the solar wind is ac­
tually larger than for O (von Steiger et al. 1992), but X­ray
emission from C lies between about 300 and 400 eV, in an
energy range where detection efficiency is very poor because
of the nearly inescapable use of thin plastic windows and filters
in X­ray detectors. The carbon in these filters is strongly ab­
sorbing at energies just above 284 eV, with the result that
detected O emission is much stronger than that from C.
Another reason to focus on O vii CX emission is that it is
dominated by Ka photons, allowing a tight energy­filtering
range to be applied to observational data and significantly re­
ducing the effective detector background in what will generally
be photon­starved observations. As explained by Beiersdorfer
et al. (2000), CX populates high­n levels in He­like ions such
as O vii, but the coupling of the two electrons' spins means
that an (triplet) state occurs about 3 times as often as
S p 1
an (singlet) state. The selection rule prohibits
S p 0 DS p 0
dipole decay of the triplet states to the singlet 2 1
1snl 1s S 0
ground, resulting in radiative cascades to triplet levels,
1s2l
which then decay via higher multipole radiation to ground and
emit Ka emission. Likewise, a significant fraction of the orig­
inal singlet states cannot decay directly to ground because of
other selection rules (such as ), so that nearly all the
Dl p#1
He­like X­ray emission is from transitions. CX
n p 2 r 1
emission from H­like ions, in contrast, has a much higher frac­
tion of decays from high­n levels directly to ground because
there are no spin­change radiative constraints, thus spreading
the emission over a broader energy range.
3.1. Modeling the CX Emission Distribution
The CX emissivity (in units of photons s #1 cm #3 ) of any ion
is given by
e p n n v j , (1)
H ion CX
ion
where n H is the neutral gas density, n ion is the density of the
``parent'' ion, is the collision velocity, and j CX is the CX
v ion
cross section. We have set the center­of­mass collision velocity
equal to the velocity of the ion since the neutral gas atoms are
less massive and generally much slower. Note that every CX
collision with a highly charged ion produces an X­ray and
lowers the ion charge.
We begin a quantitative assessment of stellar wind/CX emis­
sion by considering the solar wind, approximating it as being
spherically symmetric and purely radial, with no attempt to
include the effects of shocks and asymmetries within and be­
yond the heliosphere. Since, as will be shown, the bulk of the
CX emission occurs within the roughly 100 AU radius he­
liosphere and is concentrated toward the center, this simplifi­
cation has no significant impact on our main conclusions.
We model the density of neutral gas as n p n #
H H 0
, where r is the distance from the Sun, is the
exp (#l /r) n
H H 0
neutral gas density at large values of r (0.15 cm #3 ), and l H is
the scale length for depletion of neutral H near the Sun, primarily
by CX with solar wind protons but also by photoionization, etc.
In that simple model used by Cravens (2000), l H is roughly
20 AU on the ``downstream'' side of the Sun's motion through
the ISM, but we use the upstream value of 5 AU here. The
significance of l H will be discussed later, but its exact value is
of minor importance in our calculations for the Sun.
The product of and is equal to the particle flux
n (r) v (r)
ion ion
(particles per second) divided by . At 1 AU the wind
2
4pr
density is approximately 7 cm #3 with an average velocity of
400 km s #1 . The flux of solar wind O ions is equal to the solar
mass­loss rate, roughly yr #1 , times the fractional
#14
2 # 10 M,
abundance of the ion. For O ion abundances we use the values
recommended by Kharchenko & Dalgarno (2000) after their
search of the literature: for oxygen abundance rel­
#4
5.3 # 10
ative to hydrogen, with 0.09 and 0.24 for the fractions of bare
and H­like oxygen ions, respectively. This yields a rate of
O ix ions and O viii ions per second
31 31
3.6 # 10 9.6 # 10
injected into the solar wind.
No measurements of cross sections for CX between highly
charged O ions and H, He, or H 2 are available, but Greenwood
et al. (2000) have recently measured cross sections for CX with
H 2 O and CO 2 at energies somewhat higher than those found
in the solar wind and obtained typical values of cm 2
#15
6 # 10
for single­electron transfer. Phaneuf et al. (1982) measured

No. 1, 2001 WARGELIN & DRAKE L59
Fig. 1.---Model solar wind CX emission from He­like O vii, as would be
observed from afar, in 1 AU annular bins. Choppiness in the curves is an
artifact of the grid used in the simulation. [See the electronic edition of the
Journal for a color version of this figure.]
Fig. 2.---Simulation of O vii CX emission from 100 times solar stellar wind
at a distance of 3 pc, observed with Chandra ACIS­S detector for 100 ks,
with 5# annular binning (corresponding to 15 AU). Results are shown for two
different values of l H , the length scale for neutral H depletion near the star.
Wings from the central coronal emission are shown assuming a detected rate
of 0.1 counts s #1 in the O Ka energy band.[See the electronic edition of the
Journal for a color version of this figure.]
cross sections for CX of various O (and C) ions with H and
H 2 , but only up to Li­like O vi. We adopt the estimated value
of # cm 2 used by Cravens (2000) and others for CX
#15
3 # 10
of O vii and O viii with neutral H atoms. With a neutral H
density of 0.15 cm #3 for the local ISM, the path length for CX
collisions, , is therefore roughly 150 AU. Ra­
l p 1/(n j )
CX H CX
diative recombination is of negligible importance because of
its much smaller cross section and the low electron density.
Emissivity is computed numerically using equation (1) as a
function of radial distance from the Sun, starting at 1 AU and
extending out to 1000 AU in 1 AU steps. Note that because
fully ionized O ix ions contribute to He­like X­ray emission
via sequential CX to H­like O viii and then He­like O vii, the
densities of both O ix and O viii ions must be computed. The
relevant equations are
dn (r) 2 1
ix p # # n (r), (2)
ix
[ ]
dr r l (r)
CX
dn (r) 2 1 1
viii p # # n (r) # n (r). (3)
viii ix
[ ]
dr r l (r) l (r)
CX CX
The #2/r terms reflect the dependence of the wind density,
2
1/r
while the terms represent the depletion or addition of
1/l CX
specific ions via CX collisions. Note that l CX is a function of
r because of its dependence on n H , which is depleted near the
Sun.
The results computed above are interpolated to determine
the emissivity at every point in a 1000 AU radius octant of
three­dimensional space with 1 AU grid spacing. The emission
is then collapsed into two dimensions, as it would appear on
the sky when observed far from the Sun, and finally divided
into radial (annular) bins. Figure 1 shows the resulting distri­
bution of X­ray emission from He­like O vii and the individual
contributions from the parent O ix and O viii ions.
3.2. Detectability of Nearby Stellar Winds
Now consider a star with a wind 100 times stronger than the
Sun at a distance d of 3 pc observed with the Advanced CCD
Imaging Spectrometer­S (ACIS­S) detector on the Chandra
X­Ray Observatory for ks. Since essentially every bare
t p 100
and H­like O ion will eventually emit a He­like X­ray, the
O vii emission rate R phot will be 31
100 # (3.6 # 10 # 9.6 #
photons s #1 . With an effective area A of
31 34
10 ) p 1.32 # 10
roughly 400 cm 2 at 570 eV (Chandra Proposers' Observatory
Guide [POG]), the number of detected Ka counts, given by
R At
phot
N p , (4)
det 2
4pd
will be close to 500, nearly all of which will be detected at
radii greater than 5# from the star. At smaller radii, coronal
emission from the star would likely swamp the CX signal.
Figure 2 shows the distribution of Ka counts with 5# radial
binning, using two values of l H (5 and 50 AU). The larger value
is roughly appropriate for our example 100 times solar wind;
proton­hydrogen CX will tend to sweep out a larger volume of
neutral gas, with l H scaling as roughly . Higher fluxes of
1/2
—
M
ionizing radiation will also increase the extent of neutral gas
depletion. The histogram using the smaller value of 5 AU is
shown to illustrate the effect of different values of l H .
The background signal is based on a rate of counts
#7
1.5 # 10
s #1 arcsec #2 within a 200 eV--wide energy band centered on
570 eV (Chandra POG; M. Markevitch 2000, private commu­
nication). Coronal emission from the central star is shown as­
suming a pessimistically high rate of 0.1 counts s #1 within the
same energy band. Chandra's spatial resolution is extremely high
( ), with a surface brightness distribution that falls
FWHM # 0#.5
off as (Chandra POG), so that only about 130 of the 10,000
#2.5
r
coronal counts fall more than 5# from the center. The net signal­
to­noise ratio (S/N), , is larger than
1/2
N / (N # N # N )
CX CX BG wings
3 out to radii of more than 50#.
Coarser radial binning will increase the S/N somewhat. All
other things being equal, the number of detected CX counts
goes as the inverse square of the distance to a star, but the
angular size of the emission region scales as so that the
1/d
number of detected photons per radial bin (counts per arcse­
cond) is roughly proportional to . For our example of a 100
1/d

L60 X­RAY DETECTION OF STELLAR WINDS Vol. 546
times solar wind at 3 pc, the S/N is mostly determined by the
number of CX counts out to several tens of arcseconds and
therefore scales as roughly , so that the S/N is greater
1/2
(N )
CX
than 2 even for CX emission that is several times weaker.
Likewise, the example wind can be seen at distances well be­
yond 10 pc.
So far we have considered only He­like O vii emission.
Given a typical detector CCD energy resolution of roughly
150 eV (FWHM), much of the signal from H­like O viii (at
eV) and N vii (at eV) will be detected within
E # 654 E # 500
our example 470--670 eV energy band. In more highly ionized
winds there will be a higher fraction of these X­ray--emitting
ions, and L shell emission from intermediate­charge Fe ions
will also contribute. In most cases the S/N will be further
improved by using a wider energy band to include emission
from more ions. As Cravens (2000) suggests (for the solar
case), there may also be enhanced emission at large radii be­
cause of the higher neutral gas density near the astropause and
the higher wind density in the shocked flow.
4. APPLICATIONS
There are many candidates for stellar wind/CX observations
with Chandra or future large­area high­resolution X­ray mis­
sions. The dMe stars are of particular interest because of their
ubiquity and because one might expect flaring M stars to have
stronger winds than their nonactive counterparts. Their winds
are also likely to be more highly ionized than the Sun's because
of higher coronal temperatures, which are up to an order of
magnitude higher than the solar value of K
6
T # 2 # 10
(Schmitt et al. 1990). Within 5 pc there are 19 X­ray--detected
dMe stars, out of 42 M stars in the ROSAT survey (Wood et
al. 1994), plus over a dozen A, F, G, and K stars, of which
most are K type. Within 3 pc, six of the seven M stars are
classified as flaring.
Detection of such stars' wind­driven CX X­ray emission
would provide sorely needed insight into dwarf star winds and
their interaction with the ISM. In addition to basic information
on mass­loss rates, other things that could be studied are stellar
wind geometry, providing information on polar versus equa­
torial flows, coronal mass ejections, and other spatial and tem­
poral variability; astrosphere geometry, from observations of
emission on the leading and trailing edges of stellar motion
through the ISM (Wood & Linsky 1998); and correlations be­
tween wind characteristics and metallicity, coronal emission,
flaring activity, and stellar type.
The distribution of neutral gas in the local ISM could also
be investigated. The Sun is believed to lie within and near the
boundary of a high­density local interstellar cloud (LIC; Holzer
1989); for stars lying in low­density regions of the ISM, the
path length for CX would be at least several hundred parsecs,
and so their CX/wind emission would be spread out over a
much larger volume, rendering it undetectable. Likewise, the
detection of a stellar wind would almost certainly imply that
the star lies within the LIC or other region of enhanced density.
The heavy­ion composition of stellar winds could also be
inferred from spectra, particularly on future missions employ­
ing microcalorimeter technology. Detectors such as the X­Ray
Spectrometer (Kelley et al. 1999) built for the Astro­E mission
have energy resolution of better than 10 eV FWHM at soft
X­ray energies, which is more than adequate to distinguish
between the emission from different ions.
One could also determine wind velocities by measuring hard­
ness ratios in H­like spectra (high­n transitions vs. n p 2 r 1
lines). As explained by Beiersdorfer et al. (2000) and experi­
mentally demonstrated for elements including Ar and Xe, the
absence of spin­coupling effects in H­like ions means that elec­
trons captured into high­n states with can decay directly
l p 1
to the ground state. The hardness ratio therefore depends
l p 0
on the angular momentum of the electron captured by CX, which
has a probability distribution that is a function of collision ve­
locity at energies similar to those in the solar wind.
Even if wind/CX emission is not detected, much firmer limits
can be placed on stellar wind properties than has been possible
up to now. As noted earlier, a mass­loss rate of several times
10 #13 yr #1 , which is roughly 1000 times lower than has
M,
been previously detected, should be observable out to several
parsecs for stellar winds having solar­type wind composition,
and more highly ionized winds will be even brighter.
The authors were supported by NASA contract NAS8­39073
to the Chandra X­Ray Center during the course of this research.
REFERENCES
Badalyan, O. G., & Livshits, M. A. 1992, Soviet Astron., 36, 70
Beiersdorfer, P., et al. 2000, Phys. Rev. Lett., 85, 5090
Brown, A., Veale’, A., Judge, P., Bookbinder, J. A., & Hubeny, I. 1990, ApJ,
361, 220
Coleman, G. D., & Worden, S. P. 1977, ApJ, 218, 792
Cox, D. P. 1998, in The Local Bubble and Beyond, ed. D. Breitschwerdt,
M. J. Freyberg, & J. Tru˜mper (Berlin: Springer), 121
Cravens, T. E. 1997, Geophys. Res. Lett., 24, 105
---------. 2000, ApJ, 532, L153
Dennerl, K., Englhauser, J., & Tru˜mper, J. 1997, Science, 277, 1625
Greenwood, J. B., Williams, I. D., Smith, S. J., & Chutjian, A. 2000, ApJ,
533, L175
Ha˜berli, R. M., Gombosi, T. I., de Zeeuw, D. L., Combi, M. R., & Powell,
K. G. 1997, Science, 276, 939
Holzer, T. E. 1989, ARA&A, 27, 199
Kelley, R., et al. 1999, Proc. SPIE, 3765, 114
Kharchenko, V., & Dalgarno, A. 2000, J. Geophys. Res., 105, 18,351
Krasnopolsky, V. 1997, Icarus, 128, 368
Krishnamurthi, A., Pinsonneault, M. H., Barnes, S., & Sofia, S. 1997, ApJ,
480, 303
Lamers, H. J. G. L. M., & Cassinelli, J. P. 1999, Introduction to Stellar Winds
(Cambridge: Cambridge Univ. Press)
Lim, J., & White, S. M. 1996, ApJ, 462, L91
Lisse, C. M., et al. 1996, Science, 274, 205
---------. 1999, Icarus, 141, 316
Lisse, C. M., et al. 2000, IAU Circ. 7464
McCammon, D., & Sanders, W. T. 1990, ARA&A, 28, 657
Mullan, D. J., Doyle, J. G., Redman, R. O., & Mathioudakis, M. 1992, ApJ,
397, 225
Neugebauer, N., Cravens, T. E., Lisse, C. M., Ipavich, F. M., von Steiger, R.,
Shah, P. D., & Armstrong, T. P. 2000, J. Geophys. Res., 105, 2315
Phaneuf, R. A., Alvarez, I., Meyer, F. W., & Crandall, D. H. 1982, Phys. Rev.
A, 26, 1892
Schmitt, J. H. M. M., Collura, A., Sciortino, S., Vaiana, G. S., Harnden,
F. R., Jr., & Rosner, R. 1990, ApJ, 365, 704
Schwadron, N., & Cravens, T. E. 2000, ApJ, 544, 558
Smith, G. H. 1999, PASP, 111, 980
Snowden, S. L., et al. 1994, ApJ, 424, 714
---------. 1995, ApJ, 454, 643
---------. 1998, ApJ, 493, 715
van den Oord, G. H. J., & Doyle, J. G. 1997, A&A, 319, 578
von Steiger, R., Christon, S., Gloeckler, G., & Ipavich, F. M. 1992, ApJ, 389,
791
Wegmann, R., Schmidt, H. U., Lisse, C. M., Dennerl, K., & Englhauser, J.
1998, Planet. Space Sci., 46, 603
Willson, L. A., Bowen, G. H., & Struck­Marcell, C. 1987, Comments Astro­
phys., 12, 17
Wood, B. E., Brown, A., Linsky, J. L., Kellet, B. J., Bromage, G. E., Hodgkin,
S. T., & Pye, J. P. 1994, ApJS, 93, 287
Wood, B. E., & Linsky, J. L. 1998, ApJ, 492, 788