Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/~bradw/cv/papers/FeCX.ps.gz Äàòà èçìåíåíèÿ: Tue Nov 15 22:45:37 2005 Äàòà èíäåêñèðîâàíèÿ: Sat Dec 22 03:09:18 2007 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: comet tail

CHARGE­EXCHANGE SPECTRA OF HYDROGENIC AND He­LIKE IRON
B. J. Wargelin, 1
P. Beiersdorfer, 2
P. A. Neill, 3
R. E. Olson, 4
and J. H. Scofield 2
Received 2005 April 27; accepted 2005 July 31
ABSTRACT
We present H­like Fe xxvi and He­like Fe xxv charge­exchange spectra resulting from collisions of highly charged
iron with N 2 gas at an energy of #10 eVamu #1 in an electron beam ion trap. Although individual high­n emission
lines are not resolved in our measurements, we observe that the most likely level for Fe ×25 ! Fe ×24 electron capture
is n max # 9, in line with expectations, while the most likely value for Fe ×26 ! Fe ×25 charge exchange is significantly
higher. In the Fe xxv spectrum, the K# emission feature dominates, whether produced via charge exchange or
collisional excitation. The K# centroid is lower in energy for the former case than for the latter (6666 vs. 6685 eV,
respectively), as expected because of the strong enhancement of emission from the forbidden and intercombination
lines, relative to the resonance line, in charge­exchange spectra. In contrast, the Fe xxvi high­n Lyman lines have a
summed intensity greater than that of Ly# and are substantially stronger than predicted from theoretical calculations
of charge exchange with atomic H. We conclude that the angular momentum distribution resulting from electron
capture using a multielectron target gas is significantly different from that obtained with H, resulting in the observed
high­n enhancement. A discussion is presented of the relevance of our results to studies of diffuse Fe emission in the
Galactic center and Galactic ridge, particularly with Astro­E2/Suzaku.
Subject headinggs: atomic data --- atomic processes --- X­rays: diffuse background --- X­rays: general
1. INTRODUCTION
Within the past decade, astrophysical X­ray emission via
charge exchange (CX ) has been recognized to occur in comets,
in the atmospheres of planets including the Earth, throughout the
heliosphere, and around other stars (see review by Cravens 2002
and references therein). Recently, observations with moderate
spectral resolution by Chandra ( Wargelin et al. 2004; Smith
et al. 2005) and XMM­Newton (Snowden et al. 2004) have de­
tected clear signatures of geocoronal and heliospheric CX, most
prominently in time­variable oxygen line emission, which may
contribute a significant fraction of the soft X­ray background.
All the aforementioned CX emission is from moderately ionized
species such as He­like and H­like C, N, O, and Ne, which
originate in solar or stellar coronae. Those ions emit X rays when
they CX with neutral molecules such as H 2 O in comets, neutral
H in the Earth's outer atmosphere, and neutral interstellar H and
He within the heliosphere or astrospheres around other stars.
CX has also been proposed ( Tanaka et al. 1999) to explain
some and perhaps most of the line emission from more highly
ionized species such as He­like and H­like Si, S, Ar, Ca, and Fe
observed in diffuse emission from the Galactic ridge (GR) and
Galactic center (GC; Koyama et al. 1996; Kaneda et al. 1997;
Ebisawa et al. 2001; Muno et al. 2004). According to this hy­
pothesis, which is one of several competing explanations of GC/
GR line emission (see Muno et al. 2004 and references therein),
the highly charged ions are low­energy cosmic rays that CX with
neutral gas in the plane of the Galaxy. This CX mechanism
would naturally explain the remarkable similarity in the spectral
shapes of GC and GR diffuse emission from widely separated
regions of the Galaxy, since the emission arises from essentially
the same population of ions with the intensity level primarily
determined by the supply of neutral gas.
Galactic plane X­ray emission from cosmic­ray CX was in
fact first considered by Silk & Steigman (1969). Subsequent
studies of this idea, including those by Watson (1976), Bussard
et al. (1978), and especially Rule & Omidvar (1979), concluded
that the fraction of nearly fully ionized cosmic rays is negligible
below several MeVamu #1 , well above the energy at which CX
cross sections begin a precipitous decline (#25q 0.5 keVamu #1 ;
Ryufuku & Watanabe 1979). Most of the K­shell line flux was
therefore predicted to be emitted from cosmic rays with kinetic
energies of 1 MeVamu #1 or more (#10 MeVamu #1 for Fe). For
Fe lines this corresponds to Doppler widths of roughly 1 keV,
much larger than the #170 eV FWHM broadening measured by
Koyama et al. (1996) and Tanaka et al. (2000) in an Advanced
Satellite for Cosmology and Astrophysics (ASCA) spectrum of
the GC. More recently, Muno et al. (2004) observed several GC
fields with Chandra and deduced that Fe line broadening was
probably no more than #100 eV and could be consistent with
zero. For comparison, the Doppler broadening for Fe Ly# in a
plasma with kT ¼ 10 keV is #7 eV. More importantly, because
of the small CX cross sections at these high collision energies,
CX line emission would be several orders of magnitude too weak
to explain the Galactic plane emission.
Because of solar modulation, however, there are no mea­
surements of the interstellar cosmic ray flux or ionization state
below #1 GeVamu #1 ( Fulks 1975), so these predictions remain
theoretical. Likewise, experimental cross section data for ion--
interstellar medium ( ISM ) collisions in the #MeV amu #1 en­
ergy range are sparse, and theoretical calculations can have large
errors. In addition, none of the studies to date have modeled
more poorly understood processes such as multielectron CX or
multielectron ionization. Although these effects were believed to
be relatively minor, the last process in particular will extend the
approximately fully ionized regime to somewhat lower energies.
Rule & Omidvar (1979) noted that multielectron ionization may
be especially important when the target nuclear charge is greater
than that of the cosmic ray, and they also assumed for simplicity
1
Smithsonian Astrophysical Observatory, Harvard­Smithsonian Center for
Astrophysics, 60 Garden Street, MS­70, Cambridge, MA 02138; bwargelin@
cfa.harvard.edu.
2 Department of Physics, Lawrence Livermore National Laboratory, Liver­
more, CA 94550.
3 Department of Physics, University of Nevada, Reno, NV 89557.
4 Department of Physics, University of Missouri, Rolla, MO 65401.
687
The Astrophysical Journal, 634:687--697, 2005 November 20
# 2005. The American Astronomical Society. All rights reserved. Printed in U.S.A.

that the interstellar medium was entirely neutral. Heavy­element
abundance enhancements toward the center of the Galaxy, which
were not considered, will also increase some effective cross
sections, given that the importance of O, Fe, and other metals as
neutral targets outweighs that of atomic H at the relevant colli­
sion energies.
Given these uncertainties, a cosmic­ray CX explanation for
some of the Galactic center and Galactic ridge diffuse line
emission should not yet be completely discounted, particularly
for emission from lower charged cosmic­ray ions (e.g., #1 keV
L­shell photons from CX of He­like through F­like Fe), which
are more abundant at lower energies where CX cross sections are
larger. It is also worth noting that consideration of the cosmic­ray
ionization balance naturally leads to predictions of roughly 2--
3 times as much He­like emission as H­like emission for all
elements ( Rule & Omidvar 1979), in agreement with the ratios
observed in ASCA spectra of the GC and GR ( Tanaka 2002).
Under certain conditions, CX may also be a significant con­
tributor to X­ray emission from highly charged ions that are
thermally ionized. In addition to the many examples of solar
wind CX described above, such conditions may occur in the
Galactic center as highly ionized plasma interacts with neutral
gas on the boundaries of the dense molecular clouds that exist
there. Another possibility is the mixture of shocked gas and
evaporating clouds in supernova remnants, as has been consid­
ered by Wise & Sarazin (1989). The fraction of total emission in
such cases is likely to be small, but CX emission can be distin­
guished from electron impact excitation by its unique spectral
signatures, as discussed in x 2. The X­Ray Spectrometer ( XRS)
microcalorimeter detector on Astro­E2 ( Mitsuda et al. 2004),
which has 6 eV resolution, is particularly well suited to the study
of supernova remnants given its nondispersive spectral capabil­
ity. The XRS also should be able to provide definitive measure­
ments regarding the relevance of CX emission in the Galactic
center and Galactic ridge, although its modest collecting area and
small field of view will necessitate very long exposures to do so.
In x 2 we briefly review the CX mechanism and discuss key
diagnostics of CX emission that can be used in the analysis of
Astro­E2 spectra. In x 3 we describe our experiment, followed
by an explanation of data analysis procedures in x 4, discussion
of results in x 5, and conclusions in x 6.
2. CHARGE­EXCHANGE THEORY
CX is the radiationless collisional transfer of one or more elec­
trons from a neutral atom or molecule to an ion. If the recipient ion
is highly charged, it is left in an excited state, which then decays
via radiative cascades or, if the neutral species donates more than
one electron, by some combination of radiative decay and auto­
ionization. Single­electron capture generally dominates for highly
charged ions; multiple­electron CX is discussed in x 5.4.
Since no photons are emitted during the electron transfer, the sum
of the internal energies of the ion and atom/molecule is conserved,
and the donated electron(s) can be transferred only to specific levels
in the ion. The resonant character of the electron transfer is softened
somewhat by distortion of the energy levels of the ion and atom
during the collision, so that a range of atomic states is accessible.
For low collision energies (up to #100 keV amu #1 ), the n level
with the largest capture probability for single­electron transfer is
given approximately by Janev & Winter (1985; rewriting to ex­
plicitly include the neutral species ionization potential) as
nmax # q
I H
I n
# # 1=2
1 ×
q # 1
##### 2q
p
# # #1=2
; Ï1÷
where q is the ion charge, I n is the ionization potential of the
neutral species, and I H is the ionization potential of atomic H
(13.6 eV ). For Fe +26 and Fe +25 colliding with H at low ener­
gies, n max is therefore expected to be #12. Molecular nitrogen
has an ionization potential of 15.6 eV (14.5 eV for atomic N ),
so n max for CX with N 2 is nearly the same as with H. At low
collision energies, the n distribution has a fairly sharp maximum
but gradually broadens to its widest at #25q 0.5 keV amu #1 . At
even higher energies, n max slowly decreases, and the distribution
narrows again (Ryufuku & Watanabe 1979).
The angular momentum (l ) distribution varies more strongly
with collision energy. The details of this energy dependence are
important, because they affect how the excited ion can radia­
tively decay, e.g., directly to ground if l initial # l ground ¼ #1, or
via cascades for large values of initial l. The l distribution is
especially important in the CX of fully stripped ions, which
yields excited hydrogenic ions. For example, if the initial excited
level is an 11p state, it can decay directly to the 1s ground state,
yielding a Ly# photon. If the ion starts from an s, d, f, g, or other
state, however, it cannot decay to ground because of the #l ¼
#1 selection rule. Instead, the ion is likely to end up decay­
ing along the ``yrast chain'' in sequential #l ¼ #n ¼ #1 steps
with l ¼ n # 1 (: : : 4f ! 3d ! 2p ! 1s), ultimately resulting
in Ly# emission.
At low collision energies, low­l states are most likely to be
populated ( Ryufuku & Watanabe 1979), and the combined in­
tensity of high­n lines (n # 3 ! 1) may exceed that of Ly#
( Beiersdorfer et al. 2000). As energy increases, however, the
l distribution becomes more statistical in nature (in proportion
to 2l × 1), and fewer of the initial states can decay directly to
ground, resulting in a higher fraction of Ly# emission. The
hardness ratio of high­n versus Ly# emission can thus be used
as a diagnostic of collision energy, as illustrated for O viii and
Ne x by Beiersdorfer et al. (2001).
At the higher energies of relevance for cosmic­ray CX
(k100 keVamu #1 ), only a few percent of the X­ray emission is
from high­n states. The absence of significant high­n Fe lines in
observations of diffuse emission from the Galactic ridge and
Galactic center therefore does not necessarily indicate the ab­
sence of cosmic­ray CX emission (cf. Masai et al. 2002). En­
hanced high­n emission is expected, however, when collision
energies are low, e.g., in the highly ionized plasma/molecular
cloud scenarios described in x 1.
The hardness ratio of emission from He­like ions is much less
sensitive to collision energy, because the n ¼ 2 ! 1 line ( K# )
always dominates. From simple spin statistics, following elec­
tron transfer a He­like ion will have a total spin S ¼ 1 about 3/4
of the time, and S ¼ 0 only 1/4 of the time. Since only #S ¼
0 transitions are allowed, none of the high­n S ¼ 1 (triplet) states
can decay to the S ¼ 0 (singlet) 1 S 0 ground state, and instead the
excited electron cascades to one of the n ¼ 2 triplet states, from
which it ultimately decays via a forbidden or semipermitted
transition.
Within the n ¼ 2 level, the triplet 3 P 2,1 and 3 S 1 states that give
rise to the ``intercombination'' and ``forbidden'' lines, respec­
tively, receive much more of the cascade­derived population
than the singlet 1 P 1 state that yields the ``resonance'' line. The
triplet lines are therefore much stronger relative to the resonance
line in CX spectra than they are in thermal plasmas. (See recent
measurements by Beiersdorfer et al. [2003] and theoretical
predictions by Kharchenko et al. [2003].) Given adequate energy
resolution, this is an excellent indicator of CX emission, re­
gardless of the ion­neutral collision energy. As we illustrate in
x 5.1, even if the K# lines are instrumentally blended, one may
WARGELIN ET AL.
688 Vol. 634

still be able to use the energy centroid of the blend to distinguish
between CX and thermal emission.
Intriguingly, Muno et al. (2004) report that in the two re­
gions near the GC that yielded the most precise spectral fitting
results, the Fe xxv K# energy was measured to be 6670 ×6 #8 and
6671 ×4 #5 eV, significantly less than the 6700 eV energy of the
resonance line that dominates collisional spectra and very close
to the centroid energy we measure in our CX spectrum (see x 5).
The energy calibration for these measurements, however, had
to be shifted by 33 eV for reasons that are not well understood;
the absolute uncertainty in energies is probably something like
# 15 eV, so drawing firm conclusions about forbidden­line en­
hancement in the GC is risky.
3. EXPERIMENTAL METHOD
Our experiment used the Lawrence Livermore National
Laboratory ( LLNL) EBIT­II electron beam ion trap to collect
Fe xxvi and Fe xxv CX spectra using N 2 as the neutral gas. The
operation of EBITs has been described extensively elsewhere
( Levine et al. 1988), as has the magnetic trapping mode
( Beiersdorfer et al. 1996b) used for these measurements. To
briefly summarize, singly or doubly charged Fe ions are injected
into the EBIT­II trap region from a metal vapor vacuum arc
( MeVVA), where they are longitudinally confined by an elec­
trostatic potential and radially confined by a 3 T magnetic field,
as well as by electrostatic attraction of the narrow (60 #m)
electron beam. The beam, with a current of 140 mA for all our
measurements, passes vertically through the short trap region
(2 cm tall ), where it collisionally ionizes and excites the rela­
tively stationary ions. A neutral gas injector and several spec­
trometers are arrayed azimuthally around the ion/electron/gas
interaction region.
The neutral target gas is injected directly into the trap, where
some N 2 molecules CX with the trapped Fe ions before being
ionized and dissociated themselves. Although CX cross sections
are much larger than those for electron impact excitation and
ionization (of the order of 10 #14 vs. 10 #21 cm 2 in this case), the
neutral gas has a much lower density than the electron beam
(10 6 --10 7 vs. #10 12 cm #3 ), and the CX collision velocity (es­
sentially equal to the ion velocity, 10 6 --10 7 cm s #1 ) is much
smaller than the electron beam velocity (#10 10 cm s #1 ). The
product of ion density and the effective emission volume (de­
termined by the electron beam diameter when the beam is on and
the ion cloud size when the beam is off ) is roughly the same
whether the beam is on or off, so the rate of CX interactions is
only #1% of that for electron­ion collisions.
CX spectra are therefore collected in the magnetic trapping
mode with the electron beam turned off (response time #60 #s)
once the desired ion charge balance has been attained, which
takes roughly 1 s. The ions are still confined (although less
densely) within the trap region, where they collide with neutral
nitrogen molecules, undergo CX, and emit photons. As illus­
trated in Figure 2 of Beiersdorfer et al. (2000), the magnetic
trapping mode allows a weak CX spectrum, which would other­
wise be swamped by the much stronger electron­impact colli­
sional spectrum, to be revealed. Our measurements record the
net result of all relevant CX processes, whether from single
or multielectron transfer, radiative decays, or autoionization, as
manifested by their spectra.
Two high­purity Ge detectors with energy resolutions of
#250 and #370 eV (FWHM at 7 keV) were used to collect spec­
tra. The signal­processing lower level discriminators ( LLDs)
were set at 5 and 4 keV, respectively, to exclude unnecessary
events and prevent event pile­up. All the results we present were
obtained with the higher resolution detector, but the second de­
tector with its lower energy threshold was helpful in identifying
trap contaminants.
Because the detector resolution was insufficient to directly
separate the spectra of Fe xxvi and Fe xxv, data were collected in
two measurements using different electron beam energies. The
low­energy run ( L) used E beam ¼ 9:2 keV, and the two high­
energy runs ( I and J ) were at 17.2 keV. For comparison, the
ionization potentials of Li­like Fe +23 , He­like Fe +24 , and H­like
Fe +25 are 2.046, 8.828, and 9.278 keV, respectively (see Table 1).
During run L most of the trapped ions were He­like, with a small
fraction that were H­like. The observed CX spectrum was there­
fore a nearly pure He­like spectrum. ( The Li­like CX spectrum
lies below 2 keV, well below the lower level discriminator set­
ting.) In runs I and J (17.2 keV ), the trap contained significant
fractions of He­like, H­like, and bare ions, with a roughly 2 :1
ratio of Fe +25 and Fe +26 , resulting in a mixed CX spectrum of
He­like Fe xxv and H­like Fe xxvi lines.
In the 31 hr L run, ions were electrostatically trapped and
ionized for 3.5 s (the beam­on phase), followed by 2.5 s of
magnetic trapping (the beam­off CX phase). In the I and J runs
(18 and 20 hr), the beam­on phase lasted 4.5 s. The trap electric
potential was 300 V for run L and 100 V for runs I and J; the
difference in trap potentials was inadvertent and results in only a
small difference in effective ion­neutral collision energies. Based
on past measurements of ion energies as a function of trapping
parameters ( Beiersdorfer et al. 1996a), we estimate the average
ion energy in both cases to be roughly 10 eV amu #1 : between
5 and 20 eV amu #1 for runs I and J and approximately double
that for run L. The ions have a nonthermal energy distribution,
so one cannot strictly speak of an equivalent temperature, but
setting (56 amu)(10 eV amu #1 ) ¼ kT yields T ¼ 6:5 ; 10 6 K.
Although the Fe xxv CX spectrum was collected under two
different trap conditions, its weak dependence on collision en­
ergy, as explained in x 2, means that the results from run L can be
applied to runs I and J with little error.
4. DATA ANALYSIS
4.1. Energy Calibration
Because all the lines of interest are at least partially blended,
precise knowledge of the line energies and the detector energy
scale is essential for proper spectral fitting. Published line en­
ergies were weighted by theoretical cross sections to predict the
centroid energies of all relevant emission­line blends. Detector
energy scales were calibrated by fitting the beam­on spectra (see
Fig. 1), which have far more counts than the beam­off (CX ) data.
The CX spectra were then fitted using the derived energy cali­
brations and detailed spectral models.
To simplify analysis and increase the signal­to­noise ratio, and
because their spectra were virtually identical, data from runs I
and J were combined by scaling the energy/channel relationship
of run J by 0.9980 with an offset of +0.077 channels (2.7 eV )
and rebinning to match run I. As explained below, the He­like
Fe xxv K# line was used as an absolute energy reference, while
radiative recombination ( RR) ``lines'' in the beam­on spec­
tra, which are widely separated with well known energy differ­
ences, were used to deduce the scaling factors (eV per detector
channel ).
4.1.1. Polarization and Effective Line Energies
During the beam­on phase, the unidirectional nature of the
electron­ion collisions leads to polarization effects and non­
isotropic emission. During the beam­off phase, however, there is
Fe CHARGE­EXCHANGE SPECTRA 689
No. 1, 2005

no preferred direction for ion­neutral collisions, and the resulting
CX emission is unpolarized. Each emission­line and RR feature
is usually a blend of several transitions, so the strength of each
transition and its polarization must be known in order to predict
the energy centroid of the observed feature.
Table 1 lists theoretical energies of individual levels. Un­
certainties are based on comparisons among different sources
where possible. Polarization corrections were made according
to the prescriptions described by Wong et al. (1995). These
corrections have a negligible effect on emission­line centroids in
nearly all cases, but polarization is more important for RR fea­
tures, as described below.
In H­like Fe xxvi, the simple 2 : 1 intensity ratio of transitions
from p 3/2 and p 1/2 levels is slightly modified by polarization
adjustments to 2.1 :1. For Ly#, this shifts the centroid from
6966.0 to 6966.2 eV. Energy shifts for higher n lines are even
smaller.
For He­like emission from n # 3, selection rules dictate that
transitions from 1snp 1 P 1 to ground dominate, but there are also
significant contributions from 1snp 3 P 1 levels because of level
mixing between states with the same L and J. The K# 2 / K# 1
( 3 P 1 / 1 P 1 ) ratio has been measured to be #1/3 at 8 and 10 keV
(Smith et al. 2000). At higher energies, the excitation cross
section for K# 2 decreases rapidly, while that for K# 1 remains
roughly constant for energies up to several times the threshold.
We therefore assume that K# 2 / K# 1 is 0:1 # 0:1 at 17 keV. For
higher n, the ratio is smaller than for n ¼ 3; the actual values for
n > 3 are of little importance, so we conservatively assume
0:2 # 0:2 for both 9.2 and 17 keV.
For K# , emission from other triplet levels ( 3 S 1 and 3 P 2 ) is also
important, and the relative intensities of all the lines, which have
a relatively large energy spread, cannot be predicted accurately.
We therefore leave the K# energy free in spectral fits.
Line energy centroids, appropriate for both the beam­on and
beam­off phases, are listed in Table 1. Average energies for RR
into n ¼ 1 and 2 are also listed and were derived from the RR
cross sections listed in Table 2. The RR cross sections include
polarization effects appropriate for our instrumental geometry
(with observations perpendicular to the electron beam axis) and
were calculated (Scofield 1989) using matrix elements obtained
TABLE 1
Fe Ion Energy Levels and Lines
Level
Energy
(eV ) Reference Line Energy Centroid
E RR bound
a
(eV)
Fe xxvi
1s ......................................................................... 0 . . . . . . 0
2p 1/2 ..................................................................... 6951.9 # 0.2 1 ( Ly# ) 6966.2 # 0.3 6956.8 # 2, b 6955.3 # 2 c
2s ..................................................................... 6952.4 # 0.2 1
2p 3/2 ................................................................. 6973.1 # 0.2 1
3p 1/2 ..................................................................... 8246.3 # 0.2 1 ( Ly# ) 8250.5 # 0.3 8247.5 # 2
3s ..................................................................... 8246.5 # 0.2 1
3p 3/2 ................................................................. 8252.6 # 0.2 1
4p 1/2 ..................................................................... 8698.5 # 0.2 1 ( Ly#) 8700.2 # 0.3 8699.0 # 1
4p 3/2 ................................................................. 8701.1 # 0.2 1
5p 1/2 ..................................................................... 8907.4 # 0.2 1 ( Ly# ) 8908.3 # 0.3 8908.0 # 1
5p 3/2 ................................................................. 8908.8 # 0.2 1
6p 1/2 ..................................................................... 9020.7 # 0.2 1 ( Ly#) 9021.3 # 0.3 9021.0 # 1
6p 3/2 ................................................................. 9021.5 # 0.2 1
7p 1/2 ..................................................................... 9089.0 # 0.2 1 ( Ly# ) 9089.6 # 0.3 9089.5 # 1
7p 3/2 ................................................................. 9089.9 # 0.2 1
Ion. Pot................................................................ 9277.6 # 0.2 1 . . . . . .
Fe xxv
1s 2 ........................................................................ 0 . . . . . . 0
1s2s 3 S 1 ............................................................... 6636.7 # 0.3 2 ( K# ) measured 6656.4 # 3, b 6652.8 # 3 c
1s2p 3 P 0 .......................................................... 6665.6 # 0.3 2
1s2p 3 P 1 .......................................................... 6667.6 # 0.3 2
1s2s 1 S 0 ........................................................... 6668.1 # 0.3 2
1s2p 3 P 2 .......................................................... 6682.4 # 0.3 2
1s2p 1 P 1 .......................................................... 6700.5 # 0.3 2
1s3s 3 S 1 ............................................................... 7863.1 # 0.3 3 ( K# ) 7878.5 # 1.5, b
7880.0 # 1.5 c
7868 # 2, b
7867 # 2 c
1s3p 3 P 1 .......................................................... 7871.1 # 0.3 3
1s3d 3 D 2 .......................................................... 7880.3 # 0.3 3
1s3p 1 P 1 .......................................................... 7880.9 # 0.3 3
1s3d 1 D 2 .......................................................... 7882.3 # 0.3 3
1s4p 3 P 1 .............................................................. 8291.1 # 0.3 3 ( K#) 8294.6 # 1.0 8290 # 2
1s4p 1 P 1 .......................................................... 8295.3 # 0.3 3
1s5p 3 P 1 .............................................................. 8485.0 # 0.3 3 ( K# ) 8486.8 # 1.0 8485 # 2
1s5p 1 P 1 .......................................................... 8487.1 # 0.3 3
1s6p 3 P 1 .............................................................. 8590.0 # 0.3 4 ( K#) 8590.9 # 1.0 8590 # 2
1s6p 1 P 1 .......................................................... 8591.1 # 0.3 4
1s7p 1 P 1 .............................................................. 8653.9 # 0.3 4 ( K# ) 8653.9 # 1.0 8653.5 # 2
Ion. Pot................................................................ 8828.3 # 0.3 2 . . . . . .
WARGELIN ET AL.
690

TABLE 1---Continued
Level
Energy
(eV) Reference Line Energy Centroid
E RR bound
a
(eV)
Fe xxiv
2s ...................................................................................................... 0 . . . Below LLD 17.3 # 3, b 11.5 # 3 c
2p 1/2 .............................................................................................. 48.7 # 0.2 3, 5
2p 3/2 .............................................................................................. 64.65 # 0.2 3, 5
3s ...................................................................................................... 1149.2 # 0.3 3, 5 Below LLD 1154 # 2, b 1153 # 2 c
3p 1/2 .............................................................................................. 1162.7 # 0.3 3, 5
3p 3/2 .............................................................................................. 1167.4 # 0.3 3, 5
4s ...................................................................................................... 1544.9 # 0.4 3, 5 Below LLD 1547 # 2, b 1546 # 2 c
4p 1/2 .............................................................................................. 1550.5 # 0.4 3, 5
4p 3/2 .............................................................................................. 1552.5 # 0.4 3, 5
5s ...................................................................................................... 1726.6 # 0.5 3, 5 Below LLD 1727.5 # 2
5p 1/2 .............................................................................................. 1729.4 # 0.5 3, 5
5p 3/2 .............................................................................................. 1730.4 # 0.5 3, 5
6s ...................................................................................................... 1824.7 # 0.5 3, 5 Below LLD 1825 # 2
6p 1/2 .............................................................................................. 1826.3 # 0.5 3, 5
6p 3/2 .............................................................................................. 1826.9 # 0.5 3, 5
7s ...................................................................................................... 1883.7 # 0.5 5 Below LLD 1884 # 2
7p 1/2 .............................................................................................. 1884.7 # 0.5 5
7p 3/2 .............................................................................................. 1885.1 # 0.5 5
Ion. Pot............................................................................................. 2046.5 # 1.0 4 . . . . . .
Notes.---Line energy centroids are the same (within uncertainties) for both EIE ( beam­on) and CX (beam­off ) spectra, with the exception of the energy for
He­like K# , which is left free during spectral fitting. Energy level weightings for RR into n ¼ 2 are based on cross sections listed in Table 2; weightings for higher n
RR were extrapolated, as described in the text.
a RR spectral peaks appear at energy E beam × Ion: Pot: # E RR bound .
b With beam energy of 9.2 kV.
c With beam energy of 17 kV.
References.---(1) Erickson 1977; (2) Plante et al. 1994; (3) Vainshein & Safronova 1985; (4) this work; (5) D. Liedahl 1998 (private communication) using
HULLAC.
Fig. 1.---Spectra from beam­on phase, with linear (top) and logarithmic (bottom) vertical scales. Plots on the left are for run L (E beam ¼ 9:22 keV), and plots on
the right are for run IJ (E beam ¼ 17:21 keV ). Top panels are close­ups of the electron impact excitation Fe spectra (5--10 keV ). These spectra were used to precisely
calibrate the energy scales (34.27 eV channel #1 for L, 34.62 eV channel #1 for IJ ) by measuring separations between RR peaks; the K# lines were used to fix absolute
energies. Cr emission was scaled from the corresponding Fe lines, normalized by the fitted ratio of Cr and Fe n ¼ 2 ! 1 lines.

from a version of the GRASP Code ( Parpia et al. 1996) that was
modified to calculate the wave functions of the free electrons
along with their phase shifts. The extension beyond a central
potential model was needed to treat the recombination onto open
subshells. Results were extrapolated from n ¼ 2 to higher n
using the same s­ and p­state weightings and assuming that RR
into l levels other than s and p is small. Errors in those assump­
tions become less important as n increases because the energy
spread within a given n level decreases, and the quoted uncer­
tainties are conservative.
For both emission lines and RR, uncertainties in the weighted
energies are driven largely by uncertainties in the energies of in­
dividual levels. In the end, errors in the calibration of the energy
scale have an insignificant effect on the spectral fitting results.
4.1.2. Beam­on Spectral Fitting
The beam­on spectra were fitted using the Chandra Interac­
tive Analysis of Observations (CIAO) Sherpa fitting package
( Freeman et al. 2001). All lines were fitted using Gaussians, in
three energy ranges encompassing electron impact excitation
( EIE ) lines, RR into n # 2, and RR into n ¼ 1 (see Fig. 1).
Within the EIE group, line energies and widths were linked to
those for Fe xxv K#, except for the Fe xxv K# blend and the
unresolved K and Lyman series limit blends. Continuum emis­
sion (from two­photon radiation, bremsstrahlung, and low­
energy tails and other instrumental effects in the Ge detectors and
signal­processing electronics) was fitted with power laws. EIE
lines from n ¼ 2 7 and n # n limit (8--1) were included in the
fits, although for n # 5 the intensities of individual lines could
not be reliably constrained.
Contaminants are always present in the trap, usually at an
insignificant level and/or with emission at energies that do not
interfere with Fe. One minor exception here was Cr, the presence
of which was deduced from its Cr xxiii K# emission (and Cr xxiv
Ly# in run IJ ). In run L the He­like Cr K# intensity is #1%
of that for Fe K# ; in run IJ the He­like Cr/Fe ratio is 3% with
H­like Cr Ly# at 13% of the Fe Ly# intensity. Higher n Cr EIE
and RR lines were included in the fits, with intensities scaled to
the corresponding Fe lines. A few other contaminant lines ( Ba
from the electron gun filament, Ar from other spectrometers' pro­
portional counters, and Ti from impurities in the Fe MeVVA
wire) were included to slightly improve the quality of the RR fits.
In run L, the EIE spectrum was dominated by He­like Fe xxv
lines, but there was also a small contribution from Fe xxvi. Fit
results for Ly# (with 2.8% of the strength of K# ) were used to
scale the H­like ! He­like n ¼ 2 and 3 RR lines relative to their
He­like ! Li­like counterparts.
Fit sensitivities were studied by varying continuum levels, the
number and strength of contaminant lines, and links between line
energies and widths. In all cases, uncertainties in Fe line positions
are dominated by counting statistics, but to be conservative we
set the overall line errors equal to double the statistical errors.
TABLE 2
Radiative Recombination Cross Sections
E beam ¼ 5 keV E beam ¼ 8 keV E beam ¼ 11 keV
Ion Level # #
(# , #) P # #
(# , #) P # #
(# , #) P
Fe xxvi .............................. 1s 226.37 26.422 99.9 120.95 13.989 99.9 76.55 8.775 99.9
Fe xxvi .............................. 2s 33.57 3.966 100.0 17.52 2.052 99.9 10.89 1.264 99.9
Fe xxvi .............................. 2p 1/2 10.88 1.042 57.6 4.13 0.378 49.6 2.03 0.179 43.8
Fe xxvi .............................. 2p 3/2 20.44 1.969 56.6 7.67 0.706 47.5 3.74 0.332 40.5
Fe xxv ............................... 1s 2 112.84 13.195 99.9 60.60 7.023 99.9 38.45 4.418 99.9
Fe xxv ............................... 1s2s 3 S 1 23.36 2.761 100.0 12.18 1.427 99.9 7.56 0.878 99.9
Fe xxv ............................... 1s2p 3 P 0 2.44 0.234 57.3 0.92 0.084 49.1 0.45 0.040 43.2
Fe xxv ............................... 1s2p 3 P 1 7.27 0.697 57.2 2.74 0.251 48.9 1.34 0.118 42.9
Fe xxv ............................... 1s2s 1 S 0 8.31 0.982 100.0 4.36 0.512 99.9 2.72 0.316 99.9
Fe xxv ............................... 1s2p 3 P 2 11.48 1.105 56.3 4.28 0.394 46.9 2.08 0.185 39.9
Fe xxv ............................... 1s2p 1 P 1 6.91 0.665 56.4 2.58 0.237 47.1 1.26 0.111 40.2
Fe xxiv .............................. 2s 29.75 3.517 100.0 15.55 1.823 100.0 9.67 1.124 99.9
Fe xxiv .............................. 2p 1/2 8.69 0.831 56.9 3.26 0.297 48.5 1.59 0.140 42.5
Fe xxiv .............................. 2p 3/2 16.37 1.575 55.9 6.07 0.558 46.4 2.94 0.261 39.3
E beam ¼ 14 keV E beam ¼ 17 keV E beam ¼ 20 keV
Fe xxvi .............................. 1s 53.08 6.031 99.8 39.03 4.395 99.8 29.90 3.338 99.8
Fe xxvi .............................. 2s 7.44 0.857 99.9 5.41 0.617 99.9 4.11 0.465 99.8
Fe xxvi .............................. 2p 1/2 1.16 0.099 39.5 0.72 0.060 36.0 0.48 0.039 33.3
Fe xxvi .............................. 2p 3/2 2.11 0.182 35.0 1.31 0.110 30.5 0.87 0.072 26.7
Fe xxv ............................... 1s 2 26.71 3.041 99.9 19.65 2.218 99.8 15.06 1.685 99.8
Fe xxv ............................... 1s2s 3 S 1 5.16 0.594 99.9 3.75 0.428 99.9 2.85 0.322 99.8
Fe xxv ............................... 1s2p 3 P 0 0.26 0.022 38.8 0.16 0.013 35.4 0.11 0.009 32.7
Fe xxv ............................... 1s2p 3 P 1 0.76 0.065 38.4 0.48 0.040 34.9 0.32 0.026 32.1
Fe xxv ............................... 1s2s 1 S 0 1.87 0.215 99.9 1.36 0.155 99.9 1.03 0.117 99.8
Fe xxv ............................... 1s2p 3 P 2 1.17 0.101 34.4 0.73 0.061 29.8 0.48 0.040 26.0
Fe xxv ............................... 1s2p 1 P 1 0.71 0.061 34.8 0.44 0.037 30.4 0.29 0.024 26.8
Fe xxiv .............................. 2s 6.61 0.762 99.9 4.80 0.549 99.9 3.65 0.413 99.8
Fe xxiv .............................. 2p 1/2 0.90 0.077 38.1 0.56 0.047 34.7 0.38 0.031 32.0
Fe xxiv .............................. 2p 3/2 1.66 0.143 33.7 1.03 0.086 29.2 0.68 0.056 25.4
Notes.---Total cross sections # are in units of 10 #24 cm 2 . Differential cross sections #
(# , #) are for observations perpendicular to the electron beam direction in
units of 10 #24 cm 2 sr #1 . Polarizations P are given in percentages.
WARGELIN ET AL.
692 Vol. 634

For run L, the n ¼ 1 RR peak ( H­like ! He­like Fe) and
n ¼ 2 and 3 RR peaks (dominated by He­like ! Li­like Fe)
were used to determine the energy scale of 34:27 # 0:03 eV
channel #1 , and the position of the Fe xxv K# line (7878.5 eV )
was determined with an accuracy corresponding to #1.5 eV. With
the absolute energy calibration established, the electron beam
energy centroid was measured to be 9217 # 3 eV.
Run IJ had a more balanced mix of He­like, H­like, and bare
Fe ions and thus a complicated blend of lines in the n ¼ 2 and
3 RR peaks (with the relative intensities of each ion's RR emis­
sion fixed at 56 : 29 : 15, respectively, based on the n ¼ 1 RR fits,
Ly# / K# ratio, and theoretical cross sections for EIE and RR), so
the IJ energy scale (34:62 # 0:06 eV channel #1 ) was less well
calibrated than that for the L run. Note that the energy scales for
the two runs are not expected to be identical because of thermal
drifts in the signal­processing electronics. The Fe xxv K# line
position was slightly less well measured (#2 eV ) than in run L
because of the presence of Lyman series emission, and the elec­
tron beam energy was determined to be 17; 207 # 16 eV.
4.2. CX Spectral Fitting
For the CX spectra ( Fig. 2) we first fitted the L data and then
used those results as a template for the He­like spectrum when
fitting the combined He­like and H­like spectra in the IJ data.
As was done for the beam­on fits, we modeled the He­like and
H­like Cr spectra by fixing their n # 3 line intensities at a set
fraction of their Fe counterparts, based on the fitted Cr K# / Fe
K# ratio (0.067 for run L and 0.045 for run IJ ) and Cr Ly# / Fe
Ly# ratio (0.12 for run IJ ).
For the L spectrum, all line energies were fixed except for
Cr and Fe K# and Cr and Fe K limit . The Fe K# and K limit lines
also had free widths (with linked Cr line widths); all other line
widths were linked to Fe K#. The background was fixed at a
constant level based on its value at energies above 9.2 keV. Two
weak Ne­like Ba lines at #5290 and #6200 eV were also in­
cluded. Some slightly stronger Ne­like Ba n ¼ 3 ! 2 lines were
seen in the second Ge detector around 4390 and 4550 eV, giv­
ing us confidence in these line identifications. As noted earlier,
Ba is always a contaminant in the EBIT­II and comes from the
electron gun filament, but its emission has very little effect on the
Fe fits.
For the IJ fits, the He­like Cr and Fe spectra derived from
the run L fits were simply normalized to the fitted intensities of
Cr K# and Fe K# with no other free parameters. All Cr and
Fe Lyman line energies and widths were fixed except for the
Ly limit lines, with the Cr lines scaled to their Fe counterparts as
described above.
Strictly speaking, the Fe xxv spectra will be different in runs
L and IJ because of the contribution of double­electron capture
( DEC ) in run IJ but not in run L, which did not have any bare
Fe ions. However, as explained in x 5.4, DEC has a significantly
lower cross section than single­electron capture (SEC ), and the
Fe xxv spectra that result from SEC and DEC are expected to be
very similar. The effect on the Fe xxvi fits will in any case be very
minor, since the Fe xxv emission lines do not overlap with the
main features of the Fe xxvi spectrum, which are Ly# and the
high­n Lyman peak.
As in the L fit, two Ne­like Ba lines were included below
Fe K# , and a flat background was assumed. Given the higher
beam energy in run IJ (17.21 keV, compared with 8.33 keV for
the ionization potential of Ne­like Ba), there is also likely to be
some emission from other more highly charged species of Ba.
Fig. 2.---Spectra from beam­off (CX ) phase. Fit results from run L ( He­like spectrum) were scaled relative to K# when fitting the run IJ spectrum (combined He­
like and H­like spectra). Uncertainties for Fe xxvi intensities listed in Table 3 are primarily due to uncertainties in the Fe xxv intensities and the assumed background
level.
Fe CHARGE­EXCHANGE SPECTRA 693
No. 1, 2005

Indeed, there are small but probably real excesses of emission
just below and above the Fe K# + Ly# peak that we attribute
to the n max peaks of O­like ! F­like and F­like ! Ne­like Ba
charge exchange. As seen in Figure 2, however, particularly in
the top linear­scale plots, contaminant emission exists at a very
low level and has an essentially negligible effect on the fitted
intensities of the Fe lines.
5. RESULTS AND DISCUSSION
5.1. Line Energies
Although the four lines within the Fe xxv K# complex cannot
be resolved, the energy centroid of K# was measured fairly
accurately. In the run L CX spectrum the K# blend energy
was 6666 # 5 eV, in contrast to the beam­on centroid of 6685 #
2:5 eV, a difference of 19 # 4 eV. ( Note that the absolute energy
calibration error is the same for both measurements and thus
is not included in the difference error.) As explained in x 2, a shift
is expected because the forbidden and intercombination lines
are much stronger (relative to the resonance line) in CX spectra
than when excited by electron collisions. A similar but less ac­
curately measured shift was also observed in the IJ fits.
The Fe K limit energy was measured to be 8725 # 25 eV,
corresponding to the energy of the n ¼ 9 # 1 level. The most
likely level of CX electron capture (n max ) is probably a little
higher than n ¼ 9, because the K limit peak is a blend of all lines
with n # 8 and is not resolved from the n ¼ 6 or 7 peaks. The
approximation given by equation (1), n max # 11, is thus quite
good.
The Fe Ly limit peak is much more prominent and narrower than
the K limit peak, and its energy was measured as 9251 # 11 eV,
which corresponds to n ¼ 19 # 3. (The #11 eV includes a sta­
tistical error of #7 eV plus 4 eVof energy calibration error.) This
is significantly higher than the n max # 11:5 (corresponding to
#9210 eV ) expected for CX with N 2 using the approximation
from equation (1). For CX with atomic H, our theoretical cal­
culations (see x 5.3) predict that n max ¼ 12 or 13 (#9220 eV ),
versus #12.3 given by equation (1).
The K­series emission lies well below the Ly limit peak in en­
ergy, and there is no evidence for or reason to expect any sig­
nificant emission from other elements that would shift the Ly limit
centroid in our measurement. Double­electron transfer from N 2
is certainly present at some level but, as explained in x 5.4, is
unlikely to explain the high value of n max .
5.2. Line Intensities
Fit results are listed in Table 3. Intensity uncertainties are based
on counting statistics and sensitivity studies similar to those de­
scribed in x 4.1.2. Some specific variables were the strength of
Cr and Ba contaminant lines and the strength of the Fe xxvi Ly#
line, which blends with the stronger Fe xxv K# line. The most
important variable is the level of the background, which is as­
sumed to be flat and to arise from particle­induced background
in the detector. Even gross changes in the background level,
however, have little effect on the fitted Fe line intensities.
As seen in Table 3, the 2 ! 1 peak dominates the Fe xxv
spectrum, in contrast to the strength of the high­n Fe xxvi lines.
Indeed, the sum of the n # 3 Lyman lines exceeds the intensity
of Ly#. As we discuss below, this relatively large hardness ratio
indicates that a large fraction of the transferred electrons are
captured into low angular momentum states, particularly the
l ¼ 1 p states (see x 2). For comparison, the hardness ratio as­
suming statistically populated l levels (appropriate at high en­
ergies) is #0.02 ( Beiersdorfer et al. 2000).
5.3. Comparison with Theory
Detailed theoretical calculations for CX involving molecular
targets other than H 2 are currently infeasible, and most modeling
has been done for CX with atomic H. Experimental measure­
ments of CX spectra of highly charged ions (all with Z # 10)
and various multielectron target gases (Greenwood et al. 2001;
Beiersdorfer et al. 2003), however, indicate that the major dif­
ference when using different neutral gases is some redistribution
of intensity among high­n lines because of differences in neutral
gas ionization potentials and hence n max . One might therefore
expect that CX spectra using N 2 and atomic H should be fairly
similar.
To compare with our experimental results, we use detailed
classical trajectory Monte Carlo (CTMC ) calculations ( Perez
et al. 2001; Olson 1981) to model Fe +26 colliding with atomic H
at 1, 10, and 100 eVamu #1 . A hydrogenic cascade model is then
used to predict emitted line intensities, which are listed in
Table 3. The predicted hardness ratio is seen to increase as the
collision energy decreases. However, even for collision energies
well below those in our experiment, one can see that this model
predicts a substantially smaller hardness ratio than we measure.
For comparison, measured and theoretical Fe xxvi CX spectra
are plotted in Figure 3, along with the measured electron impact
excitation spectrum at 17.2 keV.
5.4. Molecular Targets versus Atomic Hydrogen
Similar disagreements between experimental and theoretical
hardness ratios have been noted for CX of multielectron targets
with other hydrogenic ions with Z k 10, including Ar, Kr, and
Xe ( Beiersdorfer et al. 2000), and our present results confirm the
trends established in that work, as illustrated in Figure 4. The
TABLE 3
Relative Line Intensities
Spectrum Run or Model 2 ! 1 3 ! 1 4 ! 1 4 × ! 1 3 × ! 1
Fe xxv (measured) ................. L (#20 eV amu #1 ) 1 0.074(7) 0.046(5) 0.199(10) 0.273(12)
IJ (#10 eV amu #1 ) 1 0.069(7) . . . . . . . . .
Fe xxvi (measured) ................ IJ (#20 eV amu #1 ) 1 0.12(5) 0.04(4) 1.04(7) 1.17(7)
Fe xxvi (theory) ..................... 1 eV amu #1 1 0.17 0.06 0.38 0.55
10 eV amu #1 1 0.14 0.05 0.28 0.42
100 eV amu #1 1 0.08 0.02 0.10 0.18
Notes.---Measurement errors ( listed for last digits in parentheses) include both statistical and fitting uncertainties described in the
text. Errors for Fe xxvi lines are largely driven by the range of acceptable background levels and by uncertainties in the Fe xxv spectrum.
Theoretical CTMC calculations are for CX with atomic H.
WARGELIN ET AL.
694 Vol. 634

problem, therefore, must be with the theory we are comparing
against, which presents two possibilities: either model predic­
tions of CXwith H are wrong for the Z and collision energies we
are using, or those predictions are not applicable because we are
not using atomic H as the target.
Theoretical modeling at the low energies involved here is less
reliable than at higher energies because of the quasi­molecular
states temporarily formed during slow collisions, but CX of a
bare ion with H is the simplest CX system to model, and errors
should be small. Predictions with the CTMC code used here give
good agreement with experimental measurements of highly
charged ions plus hydrogen, e.g., for Ar ×17 × H / D with a col­
lision energy of 13--40 keV amu #1 ( Beiersdorfer et al. 2005).
We therefore believe that the second possibility is much more
likely. A multielectron target immediately raises the possibility
that multielectron capture (which we henceforth assume means
double­electron capture) is important. However, the funda­
mental issue to keep in mind, whether discussing single­electron
capture or DEC, is the angular momentum distribution of the
captured electron(s). In order to explain the intensity of the
high­n lines, a large fraction of the radiatively decaying levels
must be p states that can decay directly to the 1s ground state.
Unfortunately, there are virtually no experimental publica­
tions that even roughly match the EBIT parameter space with
respect to ion and collision energy, i.e., nearly fully stripped ions
with Z > 10 with a collision energy #10 eV amu #1 . The two
main experimental CX regimes are those of electron cyclotron
resonance ( ECR) sources and heavy­ion storage rings. The
former cannot create bare ions beyond #Ne +10 and generally
operate at collision energies of a few to several tens of keV
amu #1 , and the latter use even higher energies (k1 MeVamu #1 ),
well above the range of applicability to EBIT results. For various
instrumental reasons and because the ions are moving, those
experiments usually focus on total CX cross sections and/or
recoil ion momentum or Auger electron energy spectroscopy.
Naturally, theoretical efforts have focused on available experi­
mental data and are also limited to relatively tractable models of
CX with H, He, and H 2 . Only a small fraction of CX papers of
any kind discuss angular momentum distributions, and even
fewer present photon emission spectra.
The paper closest to matching our experimental parameters is
Martin et al. (1994), which presents data on CX of bare N, O, F,
Ne, Na, and Al with various noble gas targets at an energy of about
1 keV amu #1 but does not discuss l or include spectra. Edgu­Fry
et al. (2004) present results on O ×8 × H and H 2 at 14 keVamu #1 ;
although there are no spectra, the ``Q­value'' plot in their Figure 11
indicates that the angular momentum distribution is different for
CXwith H and H 2 . This is illustrated more clearly in their plots for
Ar +8 , although that ion is far from fully stripped. We suspect that
the high­n enhancement we observe is simply due to creation of a
larger p­state population in CX with N 2 than in CX with H, and
that DEC, while present at some level, is not the key to the high­n
emission enhancement, as we now explain.
There are many papers that discuss DEC. The one most rel­
evant for us is that by Chesnel et al. (1999), which discusses
DEC in Ne ×10 × He with collision energies from 50 eVamu #1 to
15 keVamu #1 and includes helpful summaries of various DEC­
related processes such as autoexcitation (AE ), correlated double
capture (CDC ), correlated transfer and excitation (CTE ), and
autotransfer to Rydberg states (ATR). It also presents several
conclusions regarding symmetric autoionizing states (with
n # n 0 ), asymmetric autoionizing states (with n 3n 0 ), and the
contributions to radiative stabilization of those states as a func­
tion of collision energy.
For doubly excited Fe +24 , a typical symmetric state will be 10l
10l 0 , while 9l 12l 0 and 8l 15l 0 are examples of asymmetric states.
Such states can either (1) autoionize to ground (not interesting
here, because no photon is emitted ), (2) autoionize to a singly
excited state resulting in autoionizing double capture (ADC ) and
the emission of one H­like photon, or (3) radiatively stabilize
resulting in true double capture ( TDC ) and the emission of one
n ! 1 He­like satellite of a H­like line and one true He­like line.
In ADC, because continuum levels are more densely populated
at lower energies, the retained electron will tend to fall to a level
such that the ejected electron has just enough energy to escape.
Fig. 3.--- Comparison of measured and theoretical Fe xxvi spectra, all
normalized to the intensity of Ly#. High­n emission is much stronger in the
CX spectra than in the electron impact excitation spectra. The CX high­n peak
is also much stronger than predicted by the CTMC model. Theoretical spectra
are plotted with better resolution for clarity.
Fig. 4.---Hardness ratios for H­like and He­like CX emission as a function of
Z, for collision energies of #10 eV amu #1 . Dashed lines through the H­like
measurements (solid points) and He­like measurements (open points) are drawn
only to guide the eye. Neutral gases used in the experiments are CO 2 (for O), Ne
(for Ne), Ar (for Ar), N 2 (for Fe), and Kr (for Kr). Results from CTMC cal­
culations (crosses) for CX with atomic H are extrapolated to Kr (solid line). The
figure is adapted from that in Beiersdorfer et al. (2000), in which the CTMC
curve was inadvertently shifted slightly downward. The H­like O point is from
Beiersdorfer et al. (2001), and He­like O is from Beiersdorfer et al. (2003).
Fe CHARGE­EXCHANGE SPECTRA 695
No. 1, 2005

The resulting singly excited state will thus still have a medium n
value (#7, for both symmetric and asymmetric autoionizing initial
states), but one that is less than the n for SEC (#12). Unknown,
however, is the typical l value and thus what fraction of the ADC
radiative decays go directly to ground (from p states) and what
fraction end up feeding the yrast cascade yielding Ly# photons.
In the first case, the enhanced medium­n emission will blend with
the high­n peak and broaden it, and also move the centroid to
lower energies. We do not see any significant broadening, how­
ever, and the peak centroid energy is, if anything, higher than
expected from SEC, which would indicate that the fraction of
ADC is fairly small. In the second case, the fraction of Ly#
emission is enhanced over that for SEC, which is the opposite of
the enhanced high­n emission we are attempting to explain.
TDC tends to occur from asymmetric autoionizing states
(either populated directly during DEC or after autotransfer from
more symmetric states via ATR---this appears to be a matter of
some controversy), because configuration interaction between
the two widely separated electrons is minimal and they act nearly
independently. The first radiative decay, usually from the lower n
level because radiative rates scale as roughly n #3 , produces a
He­like satellite line (effectively a H­like line with a high­n spec­
tator electron). Like ADC, this enhances either the medium­n
line emission or Ly#.
The second TDC radiative decay results in a He­like photon
that originates from a singly excited high­n state, often with n
higher than is the case for SEC. For the same reasons as in the SEC
spectrum (because of the #S ¼ 0 selection rule; see x 2), the re­
sulting He­like emission will be dominated by 2 ! 1 emission
(i.e., K# ). The signature of DEC in the He­like spectrum will
therefore be very hard to discern, and indeed we see no difference
in the run L (pure SEC ) and run IJ Fe xxv (SEC + DEC) spectra.
To summarize, DEC should either increase the relative in­
tensity of Ly# (which is not inconsistent with the observed
spectrum but would not explain the large high­n peak) or in­
crease the medium­n Lyman emission (which would manifest
itself as broadening of the high­n peak and move its centroid to
a lower energy, which is not seen in our measurements). We
therefore conclude that DEC is not the source of the high­n
emission Lyman enhancement. Instead, we believe that CX of
high­Z fully stripped ions yields very different excited­state
angular momentum distributions depending on whether the neu­
tral target has one or many electrons, with a much larger fraction
of p states in the latter case. Measurements using higher reso­
lution detectors, which we plan to conduct over the next few
years, will address this and other questions.
6. CONCLUSIONS
We have presented experimental charge­exchange spectra
of Fe +26 and Fe +25 interacting with N 2 at collision energies of
#10 eV amu #1 . The resulting H­like and He­like spectra show
significant enhancement of high­n emission with respect to elec­
tron impact excitation spectra. This high­n excess is especially
pronounced in the Fe xxvi spectrum and, as has been observed in
other measurements of moderate­ and high­Z H­like emission at
low collision energies, is much stronger than predicted by clas­
sical trajectory Monte Carlo models of CX with atomic H. Our
measurements indicate that this is likely to be because the l dis­
tribution of the captured electron(s) depends on whether the neu­
tral target has one or many electrons.
CX emission may be detectable in the Galactic ridge and
Galactic center, arising from low­energy cosmic rays or highly
ionized thermal gas interacting with neutral clouds in the Ga­
lactic plane. Two key diagnostics of this emission are strong
enhancement of forbidden and intercombination line emission
in the He­like K# complex (exemplified by a 19 eV shift in the
Fe xxv K# blend centroid in our experiment), and enhancement
of high­n emission in the H­like Lyman spectrum, particularly if
the emission is from thermal ions. Line widths can be used to
discriminate between cosmic­ray CX (with widths greater than
100 eV FWHM for Fe) and thermal CX ( less than 10 eV widths).
The XRS microcalorimeter on the just­launched Astro­E2 there­
fore should be able, given sufficiently deep observations, to clearly
identify the spectral signatures of CX in diffuse X­ray emission
from the Galactic center and Galactic ridge, and perhaps in some
supernova remnants.
This work was supported by NASA's Space Astrophysics and
Analysis program under Grant NAG5­10443. B. W. was also
supported by NASA contract NAS8­39073 to the Chandra
X­Ray Center. Work at the University of California Lawrence
Livermore National Laboratory was performed under the aus­
pices of the US Department of Energy under contract W­7405­
ENG­48.
REFERENCES
Beiersdorfer, P., Bitter, M., Marion, M., & Olson, R. E. 2005, Phys. Rev. A, 72,
032725
Beiersdorfer, P., Lisse, C. M., Olson, R. E., Brown, G. V., & Chen, H. 2001,
ApJ, 549, L147
Beiersdorfer, P., Osterheld, A. L., Decaux, V., & Widmann, K. 1996a, Phys.
Rev. Lett., 77, 5353
Beiersdorfer, P., Schweikhard, L., Crespo Lo ’pez­Urrutia, J., & Widmann, K.
1996b, Rev. Sci. Instrum., 67, 3818
Beiersdorfer, P., et al. 2000, Phys. Rev. Lett., 85, 5090
---------. 2003, Science, 300, 1558
Bussard, R. W., Ramaty, R., & Omidvar, K. 1978, ApJ, 220, 353
Chesnel, J.­Y., Fre ’mont, F., Sulik, B., Merabet, H., Bedouet, C., Husson, X.,
Grether, M., & Stolterfoht, N. 1999, Nucl. Instrum. Meth. Phys. Res. B, 154,
142
Cravens, T. E. 2002, Science, 296, 1042
Ebisawa, K., Maeda, Y., Kaneda, H., & Yamauchi, S. 2001, Science, 293,
1633
Edgu­Fry, E., Wech, A., Stuhlman, J., Lee, T. G., Lin, C. D., & Cocke, C. L.
2004, Phys. Rev. A, 69, 052714
Erickson, G. W. 1977, J. Phys. Chem. Ref. Data, 6, 831
Freeman, P. E., Doe, S., & Siemiginowska, A. 2001, Proc. SPIE, 4477, 76
Fulks, G. J. 1975, J. Geophys. Res., 80, 1701
Greenwood, J. B., Williams, I. D., Smith, S. J., & Chutjian, A. 2001, Phys. Rev.
A, 63, 062707
Janev, R. K., & Winter, H. 1985, Phys. Rep., 117, 265
Kaneda, H., Makishima, K., Yamauchi, S., Koyama, K., Matsuzaki, K., &
Yamasaki, N. Y. 1997, ApJ, 491, 638
Kharchenko, V., Rigazio, M., Dalgarno, A., & Krasnopolsky, V. A. 2003, ApJ,
585, L73
Koyama, K., Maeda, Y., Sonobe, T., Takeshima, T., Tanaka, Y., & Yamauchi, S.
1996, PASJ, 48, 249
Levine, M. A., Marrs, R. E., Henderson, J. R., Knapp, D. A., & Schneider,
M. B. 1988, Phys. Scr., T22, 157
Martin, S., Bernard, J., Denis, A., De ’sesquelles, J., Chen, Li, & Ouerdane, Y.
1994, Phys. Rev. A, 50, 2322
Masai, K., Dogiel, V. A., Inoue, H., & Scho˜nfelder, V. 2002, ApJ, 581, 1071
Mitsuda, K., Kunieda, H., Inoue, H., & Kelley, R. 2004, Proc. SPIE, 5488, 177
Muno, M. P., et al. 2004, ApJ, 613, 326
Olson, R. E. 1981, Phys. Rev. A, 24, 1726
Parpia, F. A., Fischer, C. F., & Grant, I. P. 1996, Comput. Phys. Commun., 94,
249
Perez, J. A., Olson, R. E., & Beiersdorfer, P. 2001, J. Phys. B, 34, 3063
Plante, D. R., Johnson, W. R., & Sapirstein, J. 1994, Phys. Rev. A, 49, 3519
Rule, D. W., & Omidvar, K. 1979, ApJ, 229, 1198
WARGELIN ET AL.
696 Vol. 634

Ryufuku, H., & Watanabe, T. 1979, Phys. Rev. A, 20, 1828
Scofield, J. H. 1989, Phys. Rev. A, 40, 3054
Silk, J., & Steigman, G. 1969, Phys. Rev. Lett., 23, 597
Smith, A. J., Beiersdorfer, P., Reed, K. J., Osterheld, A. L., Decaux, V.,
Widmann, K., & Chen, M. H. 2000, Phys. Rev. A, 62, 012704
Smith, R. K., Edgar, R. J., Plucinsky, P. P., Wargelin, B. J., Freeman, P. E., &
Biller, B. A. 2005, ApJ, 623, 225
Snowden, S. L., Collier, M. R., & Kuntz, K. D. 2004, ApJ, 610, 1182
Tanaka, Y. 2002, A&A, 382, 1052
Tanaka, Y., Koyama, K., Maeda, Y., & Sonobe, T. 2000, PASJ, 52, L25
Tanaka, Y., Miyaji, T., & Hasinger, G. 1999, Astron. Nachr., 320, 181
Vainshtein, L. A., & Safronova, U. I. 1985, Phys. Scr., 31, 519
Wargelin, B. J., Markevitch, M., Juda, M., Kharchenko, V., Edgar, R., &
Dalgarno, A. 2004, ApJ, 607, 596
Watson, W. D. 1976, ApJ, 206, 842
Wise, M. W., & Sarazin, C. L. 1989, ApJ, 345, 384
Wong, K. L., Beiersdorfer, P., Reed, K. J., & Vogel, D. A. 1995, Phys. Rev. A,
51, 1214
Fe CHARGE­EXCHANGE SPECTRA 697
No. 1, 2005