L161
The Astrophysical Journal, 545:L161–L164, 2000 December 20
䉷 2000. The American Astronomical Society. All rights reserved. Printed in U.S.A.
EMISSION-LINE INTENSITY RATIOS IN Fe xvii OBSERVED WITH A MICROCALORIMETER
ON AN ELECTRON BEAM ION TRAP
J. M. Laming
1
I. Kink,
2
E. Takacs,
3
J. V Porto,
2
J. D. Gillaspy,
2
E. H. Silver,
4
H. W. Schnopper,
4
S. R. Bandler,
4
N. S. Brickhouse,
4
S. S. Murray,
4
M. Barbera,
5
A. K. Bhatia,
6
G. A. Doschek,
1
N. Madden,
7
D. Landis,
7
J. Beeman,
7
and E. E. Haller
7
Received 2000 June 16; accepted 2000 October 11; published 2000 December 6
ABSTRACT
We report new observations of emission line intensity ratios of Fe xvii under controlled experimental conditions,
using the National Institute of Standards and Technology electron beam ion trap (EBIT) with a microcalorimeter
detector. We compare our observations with collisional-radiative models using atomic data computed in distorted
wave and R-matrix approximations, which follow the transfer of the polarization of level populations through
radiative cascades. Our results for the intensity ratio of the 15.014 A
˚
line to the
61 5 1 61
2pS–2p 3dP 2pS–
01 0
15.265 A
˚
line are and at beam energies of 900 and 1250 eV, respectively.
53
2p 3dD 2.94 Ⳳ 0.18 2.50 Ⳳ 0.13
1
These results are not consistent with collisional-radiative models and support conclusions from earlier EBIT work
at the Lawrence Livermore National Laboratory that the degree of resonance scattering in the solar 15.014 A
˚
line
has been overestimated in previous analyses. Further observations assess the intensity ratio of the three lines between
the configurations to the three lines between the configurations. Both R-matrix and distorted
65 65
2p–2p 3s 2p–2p 3d
wave approximations agree with each other and our experimental results much better than most solar and stellar
observations, suggesting that other processes not present in our experiment must play a role in forming the Fe xvii
spectrum in solar and astrophysical plasmas.
Subject headings: atomic data — methods: laboratory — stars: individual (Capella) — Sun: corona —
techniques: spectroscopic — X-rays: general
The high elemental abundance of Fe and the closed shell
structure of Ne-like ions cause Fe
⫹16
to be one of the dominant
ions in forming emission-line spectra from plasmas with tem-
peratures ∼ K. Fe xvii lines will be the dominant lines
6
5 # 10
in most spectra obtained by gratings on Chandra and XMM-
Newton from a wide variety of objects. For example, in the
Capella spectra acquired by Chandra (Brinkman et al. 2000;
Canizares et al. 2000), Fe xvii contributes four or five of the six
strongest lines observed. In conditions of ionization equilibrium,
Fe xvii is unique in providing an electron temperature diagnostic
between two sets of these strong lines (Raymond & Smith 1986).
Solar flare observations have long pointed to discrepancies be-
tween the observed intensity ratios among the strong emission
lines of Fe xvii in the 15–17 A
˚
region. The Capella spectra
referred to above hint that similar problems may exist for astro-
physical sources as well. Among the six strong lines arising from
transitions between the ground state and the excitedstates
61
2pS
0
, , , and , , and , the strongest line
5133 5 13 3
2p 3sPPP 2p 3dP D P
112 1 1 1
from at 15.014 A
˚
often appears diminished in intensity
51
2p 3dP
1
relative to other features. This has led to suggestions that it may
be affected by resonance scattering that removes photons pre-
1
E. O. Hulburt Center for Space Research, US Naval Research Laboratory,
Washington, DC 20375.
2
National Institute of Standards and Technology, 100 Bureau Drive, Gaith-
ersburg, MD 20899.
3
Massachusetts Institute of Technology, 77 Massachusetts Avenue, 26-239,
Cambridge, MA 02139-4307; and University of Debrecen, Debrecen, Bem ter
18/A, H-4026, Hungary.
4
Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cam-
bridge, MA 02138.
5
Osservatorio Astronomico G. S. Vaiana, Piazza del Parlamento 1, 90134
Palermo, Italy.
6
Laboratory for Astronomy and Solar Physics, NASA Goddard Space Flight
Center, Greenbelt, MD 20771.
7
Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA
94720.
dominantly from the line of sight in this transition (Schmelz,
Saba, & Strong 1992; Phillips et al. 1996, 1997; Bhatia & Kastner
1999; Saba et al. 1999). Furthermore, the overall intensity of the
lines from the configuration often appears enhanced rel-
5
2p 3s
ative to the intensity theory would predict when compared with
the lines from the configuration. This has led some authors
5
2p 3d
(U. Feldman 2000, private communication; see Sampson &
Zhang 1987) to suggest that an inner shell ionization of a
2p
electron from the Na-like Fe ion plays a role in forming the
spectrum, producing Ne-like Fe in the excited configuration
. Other explanations suggest that strong dielectric recom-
5
2p 3s
bination occurs (Liedahl et al. 1990) or simply that the theoretical
atomic models do not include a sufficient number of excited
levels to adequately treat all the radiative cascades. In this Letter
we describe initial results of experiments designed to test some
of these ideas.
In a study of tokamak spectra of Fe xvii, Phillips et al. (1997)
found good agreement between observations and synthetic spec-
tra calculated from models including the lowest 37 fine-structure
levels of Fe xvii (i.e., all configurations up to ), in-
6
2s2p 3d
cluding dielectronic and inner shell satellites of Fe xvi. The
collisional data used in Phillips et al. (1997) are similar to some
of those used here. They are computed in the distorted wave
approximation and tabulated in Bhatia & Doschek (1992). To-
kamak plasma sources do not allow the individual atomic pro-
cesses to be experimentally isolated and studied in detail. More
control is possible with the use of an electron beam ion trap
(EBIT). Positive ions are trapped in the space charge of an
essentially monoenergetic electron beam that also ionizes and
excites the ions. The ion charge state is selected by tuning the
electron beam energy. In the case of Ne-like ions, almost com-
plete selectivity is possible. Taking advantage of this, Brown et
al. (1998) made detailed observations of Fexvii at beam energies
of 850, 1150, and 1300 eV. They observed lines originating from
levels with principal quantum number n up to 11. Their mea-
L162 EMISSION-LINE INTENSITY RATIOS IN Fe xvii Vol. 545
Fig. 1.—Section of a microcalorimeter spectrum of Fe L shell lines recorded
at an EBIT beam energy of 2.54 keV. Lines from Fe xvii to Fe xxiii are
prominent.
Fig. 2.—Fit to the six strong Fe xvii lines recorded at an electron beam
energy of 900 eV. The line wavelengths are 15.014, 15.265, 15.456, 16.780,
17.055, and 17.100 A
˚
. The last two transitions are blended together at the
resolution of the microcalorimeter and are fitted as a single component. The
upper panel gives the residuals from the fit.
surement of the intensity ratio of the strongest 15.014 A
˚
line to
the 15.265 A
˚
line originating from the level was
53
2p 3dD
1
, which is in between theoretical values (generally3.04 Ⳳ 0.12
closer to 4) and ratios observed in solar and stellar coronae
(usually in the range 2–2.5). Table 2 of Brown et al. (1998) gives
a useful summary of the various calculations and observations
of this intensity ratio. However, Brown et al. (1998) in their
paper report no results for the intensities of the transitions.3s–3p
Our experimental technique couples the National Institute of
Standards and Technology (NIST) EBIT with a microcalorimeter
detector and is discussed in detail in Silver et al. (2000). This
offers a combination of high throughput, wide bandpass, and
sufficient spectral resolution to allow most lines to be resolved
(see Kelley et al. 2000; Silver et al. 2000). One modification
from previous experimental runs was to move the microcalorim-
eter to a new observation port on the EBIT, which allowed much
closer access to the trapped ions in the electron beam. Thisfurther
improved the high-energy bandpass and allowed us to dispense
with the X-ray optic. Another important feature is that the mi-
crocalorimeter when used in this configuration is polarization
blind. Thus, the only correction required for this effect is for the
degree of spatial anisotropy in the emission of polarized light;
the measured intensity will depend on the degree of polarization
through this anisotropy, which is different for lines of different
polarization. As we discuss below, such effects are rather small
and can be adequately modeled. A part of a spectrum taken at
a beam energy of 2500 eV is shown in Figure 1 to illustrate the
capabilities of the detector. These data were collected in only
15 minutes. The complete spectrum extends to much higher
energies and would include the Fe K line complex if the beam
energy was sufficiently high. Of course, in this case the Fe L
spectrum would not be so strong, and it is these lines, and spe-
cifically Fe xvii, that are the main focus of this Letter.
We observed Fe xvii transitions at a variety of electron beam
energies between 900 and 4000 eV. Comparing spectra at beam
energies of 900 and 1250 eV allows us to study what effect (if
any) enhanced radiative cascades at the higher energy might
have. At 900 eV, we are also well below the thresholds for
excitation by inner shell ionization of Fe xvi (Sampson & Zhang
1987) should any be present in our trap (see below). For these
two beam energies, experimental intensity ratios were obtained
by fitting a sum of Gaussian functions to the experimental data,
and the area underneath a particular Gaussian was treated as a
line intensity. The Gaussian functions slightly underestimate the
“wing” of the spectral lines, and better resemblance is achieved
with Voigt functions. However, the results from these different
fits remained well within quoted error bars, determined from the
statistical quality of the fit (dominant uncertainty) and systematic
uncertainties (much smaller contributions) due to the detector/
window efficiency, and therefore the results from Gaussian fits
are presented. The data, fit, and residuals for the 900 eV beam
energy spectrum are shown in Figure 2. At these beam energies,
there is no evidence of inner shell transitions in Fe xvi. The
strongest of these lines, the , cal-
262 2522
2s 2p 3sS –2s 2p 3sP
1/2 3/2
culated to be at a wavelength between 17.29 and 17.31 A
˚
(Phillips et al. 1997; Bautista 2000), would be easily visible in
our spectra, resolved from the Fe xvii 17.055 and 17.100 A
˚
lines. Using higher resolution crystal spectrometers observing
one of the Lawrence Livermore National Laboratory EBITs un-
der similar conditions, Brown et al. (1998) also saw no evidence
for these transitions.
The spectra were corrected for the transmission efficiencies
of the microcalorimeter windows. (The quantum efficiency of
the detector is 100% for energies below 5 keV.) The window
transmission has been well defined for the entire microcalor-
imeter bandpass of 0.1–10 keV. The three windows are made
of polymide (800 A
˚
) and aluminum (1100 A
˚
), and their thick-
nesses have been measured to an accuracy of 0.5% (Powell et
al. 1997). The efficiency curve is smooth and featureless in the
energy band that includes the Fe L emission. We have taken
great care to minimize potential contamination of these win-
dows, but over long periods of time (days) it is possible that
minute amounts of hydrocarbons could collect on the outer
window (the most susceptible since it views the vacuum con-
nection to the EBIT). Our calculations show that only several
monolayers of nitrogen or oxygen would freeze out on the
window given the cleanliness of the EBIT vacuum connection
and the extremely low pressures at the initiation of the cooling
cycle. Since the Fe xvii lines were measured over a period of
No. 2, 2000 LAMING ET AL. L163
TABLE 1
Observational and Theoretical Fe xvii Intensities Relative to the 15.014 A
˚
Line
Line Wavelength
(A
˚
)
Line
Energy
(eV)
Upper
Level Data (0.9 keV)
Theoretical (0.9 keV)
a
Data (1.25 keV)
Theoretical (1.25 keV)
a
Distorted Wave R-Matrix Distorted Wave R-Matrix
15.265 ................ 812.5
53
2p 3dD
1
0.34 Ⳳ 0.02
b
0.25 0.25 0.40 Ⳳ 0.02 0.26 0.26
15.456 ................ 802.5
53
2p 3dP
1
0.10 Ⳳ 0.01 0.047–0.036 0.059–0.046 0.09 Ⳳ 0.01 0.035–0.027 0.034–0.027
16.780 ................ 739.1
c53
2p 3sP
1
0.45 Ⳳ 0.04 0.46–0.41 0.47–0.41 0.40 Ⳳ 0.02 0.51–0.44 0.47–0.42
17.055 ................ 727.1
c51
2p 3sP
1
0.54–0.47 0.60–0.53 0.58–0.50 0.55–0.47
17.100 ................ 725.0
53
2p 3sP
2
0.34–0.30 0.43–0.38 0.41–0.35 0.39–0.33
17.055⫹17.100
d
...... 0.88 Ⳳ 0.09 0.88–0.77 1.03–0.91 0.92 Ⳳ 0.12 0.99–0.85 0.94–0.80
a
Theoretical range goes from unpolarized to maximum polarization case. R-matrix cross sections are from Mohan et al. 1997 substituted for first 26 excited
levels.
b
Uncertainties correspond to 1 j.
c
Some authors swap the LS coupling notation for these levels, with 16.780 A
˚
originating from and 17.100 A
˚
from .
13
PP
11
d
Experimental result is for the sum of 17.055 and 17.100 intensities; theoretical results are for each line separately and their sum.
1 hr (as opposed to several days), the chances of a contami-
nation buildup occurring is unlikely. Furthermore, we point out
that the Fe xvii line ratios measured by our instrument are
relatively insensitive to even a large amount of contamination.
For example, 3000 A
˚
of nitrogen or oxygen ice formed on the
outer cryostat window could reduce transmission by 25% but
only alter the ratios for the lines by 9%; the line3s/3d 2p–3d
ratios would remain unchanged. This scenario is highly unlikely
since 3000 A
˚
of N or O ice is 2 times thicker than the window
itself and would undoubtedly break it.
We compare our observations with theoretical results com-
puted using a model Fe xvii ion comprising the configurations
up to , i.e., 73 levels, with extra configurations
6
2s2p 4d
, , and for and 6 included
25 25 6
2s 2pns2s 2pnd 2s2pnp np 5
for their radiative cascades. This brings the total up to 113 levels.
Excitation rates are calculated for monoenergetic electron beams
between 0.9 and 4.0 keV energy, including only those levels that
are energetically accessible in each case. The calculations use
impact excitation cross sections computed in a distorted wave
approximation as described in Bhatia & Doschek (1992) for all
transitions among the lowest 73 levels. As an alternative, R-
matrix cross sections from the ground state to the configurations
up to (i.e., the first 26 excited levels) can be substituted
5
2p 3d
from Mohan, Sharma, & Eissner (1997). Energy levels and ra-
diative decay rates are taken from the distorted wave target cal-
culation, supplemented with those forhigher lying configurations
taken from a computation with the HULLAC (Hebrew Univer-
sity Lawrence Livermore) Code for Atomic Physics (M. Kla-
pisch, A. Bar Shalom, W. H. Goldstein, E. Meroz, A. Chon, &
M. Cohen 1988, unpublished; Goldstein et al. 1988). Collisional
excitation rates to these higher levels are scaled from those for
the lower lying ones.
For the purposes of modeling emission from the EBIT, we
follow the analysis in Takacs et al. (1996) and expand our 113-
level model to the 457 magnetic sublevels present. This is done
in order to account for the polarization arising from the excitation
by the electron beam. For the upper levels of the strongest tran-
sitions (i.e., those with ), excitation cross sections to in-j p 1
dividual magnetic sublevels are given in the relativistic distorted
wave approximation by Zhang, Sampson, & Clark (1990). For
transitions where a nonrelativistic approach is valid, these data
are supplemented with simple results from a Coulomb-Bethe
approximation.
All levels with are assumed to be unpolarized. Morej ≥ 3
accurate dependent cross sections will be substituted as theym
j
become available, but the current procedure should model the
transmission of polarization through radiative cascades suffi-
ciently accurately to allow us to evaluate the anisotropy of the
emission from the EBIT. More precise work would be necessary
with crystal spectrometers.
In Table 1 we compare our experimental Fe xvii line intensity
ratios with those calculated by the methods outlined above for
beam energies of 0.9 and 1.25 keV. The theoretical results come
from the distorted wave cross sections calculated herein and from
the same distorted wave results but with the cross sections for
excitation from the ground state to the first 26 excited levels
replaced by the R-matrix results of Mohan et al. (1997). The
range of theoretical ratios extends from results calculated in the
limit of zero polarization to those in the limit of maximum po-
larization. The finite distribution of pitch angles in the electron
beam causes the true theoretical result to lie somewhere between
these two limits. This depolarization can be estimated from re-
sults given in Gu, Savin, & Beiersdorfer (1999). Even in the
limit of no depolarization, the polarization correction required
for the microcalorimeter is rather small, less than would be the
case for a crystal spectrometer, but remains the dominant sys-
tematic uncertainty in our experiment at these two beam energies.
Brown et al. (1998) measured intensity ratios for the 15.014/
15.265 A
˚
lines of , , , and2.93 Ⳳ 0.16 3.15 Ⳳ 0.17 2.77 Ⳳ 0.19
from observations at beam energies of 1150, 1150,3.00 Ⳳ 0.20
850, and 1300 eV, respectively. They then average these values
to give a mean result of , compared with theoretical3.04 Ⳳ 0.12
predictions of around 4. Explicit in this argument is an assump-
tion that the polarizations of these two lines are the same, since
the observations were made with crystal spectrometers at Bragg
angles where they are polarization sensitive. Such an assumption
is supported by theory (Zhang et al. 1990). However, the result
obtained using this assumption demonstrates that this same the-
ory gives an intensity ratio between these two lines inconsistent
with experiment. While it is quite possible that the calculations
get the polarizations right and the intensity ratio wrong, it is
certainly not guaranteed. The microcalorimeter employed in the
present work is not polarization sensitive, and our experimental
values for the intensity ratio— at 0.9 keV and2.94 Ⳳ 0.18
at 1.25 keV—are more robust, being sensitive only2.50 Ⳳ 0.13
to the possible difference in the anisotropy of the emitted ra-
diation. The first value is consistent with the average result of
Brown et al. (1998), while the second is slightly lower. These
results support the suggestion of Brown et al. (1998) that the
effect of resonance scattering in the 15.014 A
˚
resonance line in
solar spectra has hitherto been overestimated.
Intensity ratios from our data for the three lines originating
from to those from are shown in Figure 3. For
55
2p 3s 2p 3d
energies greater than 1.25 keV, they are obtained by summing
L164 EMISSION-LINE INTENSITY RATIOS IN Fe xvii Vol. 545
Fig. 3.—Variation of the intensity ratio (I ⫹ I ⫹ I )/(I ⫹
16.780 17.055 17.100 15.014
with electron beam energy. The first three lines originate in theI ⫹ I )
15.265 15.456
configuration and the second three in the . For comparison, results
55
2p 3s 2p 3d
from model calculations using the distorted wave results (see Bhatia & Doschek
1992) are shown, and where appropriate the R-matrix results of Mohan et al.
(1997) are overlaid. The range given for each calculation corresponds to the
limits of unpolarized (upper curve) and completely polarized (lower curve)
emission. The observational ratios from various solar and stellar observations
are shown for reference on the same plot. For these, the energy scale on the
x-axis should be disregarded. Fe xvii is generally formed in plasmas at a
temperature of K, and electrons with energies close to threshold, i.e.,
6
5 # 10
less than 1 keV, dominate the excitation rate. The references are (1) Blake et
al. (1965); (2) Parkinson (1975); (3) Hutcheon, Pye, & Evans (1976);
(4) McKenzie et al. (1980); (5) Phillips et al. (1982); (6) Acton et al. (1985);
(7) Brinkman et al. (2000); and (8) Canizares et al. (2000).
the number of detected photons over the spectral regions where
the and lines appear, respectively,
65 65
2p–2p 3s 2p–2p 3d
correcting for the presence of the Fe xix lines at 15.114 and
15.210 A
˚
(820 and 815 eV, respectively), by measuring the strong
Fe xix lines at 13.465, 13.507, and 13.521 A
˚
(920
43
2p–2p 3d
eV) and subtracting off the appropriate ratio determined from
calculations with HULLAC. Since Fe xix has a ground-state,
3
P
2
polarization corrections are negligible. The calculated ratios, in
the range 0.25–0.33 for beam energies 2–4 keV, also agree well
with higher resolution observations of solar active regions (re-
cently reassessed by Phillips et al. 1999). The error bars are
estimated from the Poisson counting statistics and uncertainties
arising from background subtraction (i.e., the Fe xix lines) and
in the detector/window quantum efficiency but are dominated
by a Ⳳ20% error in the measured intensity of the Fe xix blend
at 920 eV. Also shown in Figure 3 are observational ratios of
the three lines to the three lines taken from various solar3s 3d
and stellar observations.
The experimental ratios for the electron beam energy of 0.9
keV suggest that the models including the R-matrix crosssections
are less preferable than those using only distorted wave cross
sections. The former cross sections appear to overestimate
slightly the emission in the 17.055 and 17.100 A
˚
lines. Elsewhere
we have a small preference for the R-matrix results, although
we caution here that a small contribution to the Fe xvii emission
may arise by recombination from Fe
⫹17
into excited states of
Fe xvii, and until this effect is adequately modeled firm con-
clusions might be premature. Our experience with similar spectra
of Kr xxvii (Kink et al. 2000) suggests that the inclusion of
recombination would increase the theoretical ratios. More3s/3d
importantly, the difference between the two theoretical ap-
proaches and our experiment is significantly smaller at all en-
ergies than the discrepancies between the solar and astrophysical
observations summarized in Figure 3 (with the exception of
Phillips et al. 1982) and either theory. The basic electron impact
excitation theory for these lines in Fe xvii appears to be correct,
in contrast with that for the 15.265/15.014 ratio (see Table 1),
and one must look to other processes to model the solar and
astrophysical observations satisfactorily. These may invalidate
the electron temperature diagnostic of Raymond & Smith (1986).
These measurements are a subset of a larger survey of spec-
troscopy experiments performed with the microcalorimeter on
the NIST EBIT. Future work will concentrate on acquiring
spectra with significantly higher statistical quality, simply by
increasing the integration time at each beam energy. This might
also allow the simultaneous observation of radiative recom-
bination features from the EBIT.
A. K. B., G. A. D., and J. M. L. were supported by NASA
contract W19539 (Applied Information Systems Research Pro-
gram); G. A. D. and J. M. L. were also supported by the NRL/
ONR Solar Magnetism and the Earth’s Environment 6.1 Re-
search Option. E. H. S., H. W. S., and S. R. B. acknowledge
support in part by NASA grant NAG5-5104, and I. K. ac-
knowledges support from the Swedish Foundation for Coop-
eration in Research and Higher Education (STINT). We also
acknowledge the comments of an anonymous referee, partic-
ularly in regard to the Fe xix lines.
REFERENCES
Acton, L. W., et al. 1985, ApJ, 291, 865
Bautista, M. A. 2000, J. Phys. B, 33, 71 (erratum 33, 2833)
Bhatia, A. K., & Doschek, G. A. 1992, At. Data Nucl. Data Tables, 52, 1
Bhatia, A. K., & Kastner, S. O. 1999, ApJ, 516, 482
Blake, R. L., Chubb, T. A., Friedman, H., & Unizicker, A. E. 1965, ApJ, 142,
1
Brinkman, A. C., et al. 2000, ApJ, 530, L111
Brown, G. V., Beiersdorfer, P., Liedahl, D. A., Widman, K., & Kahn, S. M.
1998, ApJ, 502, 1015 (erratum 532, 1245 [2000])
Canizares, C. R., et al. 2000, ApJ, 539, L41
Goldstein, W. H., Oreg, J., Zigler, A., Bar Shalom, A., & Klapisch, M. 1988,
Phys. Rev. A, 38, 1797
Gu, M. F., Savin, D. W., & Beiersdorfer, P. 1999, J. Phys. B, 32, 5371
Hutcheon, R. H., Pye, J. P., & Evans, K. D. 1976, MNRAS, 175, 489
Kelley, R. L., et al. 2000, Nucl. Instrum. Methods Phys. Res., 444, 170
Kink, I., et al. 2000, Phys. Rev. E., submitted
Liedahl, D. A., Kahn, S. M., Osterheld, A. L., & Goldstein, W. H. 1990, ApJ,
350, L37
McKenzie, D. L., Landecker, P. B., Broussard, R. M., Rugge, H. R., Young,
R. M., Feldman, U., & Doschek, G. A. 1980, ApJ, 241, 409
Mohan, M., Sharma, R., & Eissner, W. 1997, ApJS, 108, 389
Parkinson, J. H. 1975, Sol. Phys., 42, 183
Phillips, K. J. H., et al. 1982, ApJ, 256, 774
Phillips, K. J. H., Greer, C. J., Bhatia, A. K., Coffey, I. H., & Keenan, F. P.
1997, A&A, 324, 381
Phillips, K. J. H., Greer, C. J., Bhatia, A. K., & Keenan, F. P. 1996, ApJ, 469,
L57
Phillips, K. J. H., Mewe, R., Harra-Murnion, L. K., Kaastra, J. S., Beiersdorfer,
P., Brown, G. V., & Liedahl, D. A. 1999, A&AS, 138, 381
Powell, F. R., et al. 1997, Proc. SPIE, 3113, 432
Raymond, J. C., & Smith, B. W. 1986, ApJ, 306, 762
Saba, J. L. R., Schmelz, J. T., Bhatia, A. K., & Strong, K. T. 1999, ApJ, 510,
1064
Sampson, D. H., & Zhang, H.-L. 1987, Phys. Rev. A, 36, 3590
Schmelz, J. T., Saba, J. L. R., & Strong, K. T. 1992, ApJ, 398, L115
Silver, E., et al. 2000, ApJ, 541, 495
Takacs, E., et al. 1996, Phys. Rev. A, 54, 1342
Zhang, H. L., Sampson, D. H., & Clark, R. E. H. 1990, Phys. Rev. A, 41,
198
