Radioactive Scandium in the Youngest Galactic Supernova Remnant G1.9+0.3

Abstract

We report the discovery of thermal X-ray emission from the youngest Galactic supernova remnant G1.9+0.3, from a 237 ks Chandra observation. We detect strong Kα lines of Si, S, Ar, Ca, and Fe. In addition, we detect a 4.1 keV line with 99.971% confidence which we attribute to 44Sc, produced by electron capture from 44Ti. Combining the data with our earlier Chandra observation allows us to detect the line in two regions independently. For a remnant age of 100 yr, our measured total line strength indicates synthesis of (1-7) × 10–5 M ☉ of 44Ti, in the range predicted for both Type Ia and core-collapse supernovae (SNe), but somewhat smaller than the 2 × 10–4 M ☉ reported for Cas A. The line spectrum indicates supersolar abundances. The Fe emission has a width of about 28,000 km s–1, consistent with an age of ~100 yr and with the inferred mean shock velocity of 14,000 km s–1 deduced assuming a distance of 8.5 kpc. Most thermal emission comes from regions of lower X-ray but higher radio surface brightness. Deeper observations should allow more detailed spatial mapping of 44Sc, with significant implications for models of nucleosynthesis in Type Ia SNe.

Full text

Page 1

Preprint typeset using LATEX style emulateapj v. 11/10/09
RADIOACTIVE SCANDIUM IN THE YOUNGEST GALACTIC SUPERNOVA REMNANT G1.9+0.3
KAZIMIERZ J. BORKOWSKI,1STEPHEN P. REYNOLDS,1DAVID A. GREEN,2UNA HWANG,3ROBERT PETRE,3KALYANI
KRISHNAMURTHY,4& REBECCA WILLETT4
Submitted to ApJ Letters on June 16, 2010
ABSTRACT
We report the discovery of thermal X-ray emission from the youngest Galactic supernova remnant (SNR)
G1.9+0.3, from a 237-ks Chandra observation. We detect strong Kα lines of Si, S, Ar, Ca, and Fe. In addition,
we detect a 4.1 keV line with 99.971% confidence which we attribute to44Sc, produced by electron capture
from44Ti. Combining the data with our earlier Chandra observation allows us to detect the line in two regions
independently. For a remnant age of 100 yr, our measured total line strength indicates synthesis of (1−7)×
10−5M?of44Ti, in the range predicted for both Type Ia and core-collapse (CC) supernovae, but somewhat
smaller than the 2×10−4M?reported for Cas A. The line spectrum indicates supersolar abundances. The Fe
emission has a width of about 26,000 km s−1, consistent with an age of ∼ 100 yr and with the inferred mean
shock velocity of 14,000 km s−1deduced assuming a distance of 8.5 kpc. Most thermal emission comes from
regions of lower X-ray but higher radio surface brightness. Deeper observations should allow more detailed
spatial mapping of44Sc, with significant implications for models of nucleosynthesis in Type Ia supernovae.
Subject headings: ISM: individual objects (G1.9+0.3) — ISM: supernova remnants — nuclear reactions, nu-
cleosynthesis, abundances — X-rays: ISM
1. INTRODUCTION
Thesupernovaremnant(SNR)G1.9+0.3hasanageoforder
100 years (Reynolds et al. 2008, Paper I). Its integrated X-ray
spectrum is well described by a model of synchrotron emis-
sion from a power-law distribution with an exponential cutoff
(XSPEC model srcut), with rolloff frequency hνroll= 2.2
keV and a very high absorbing column density NH∼= 5×1022
cm−2(Reynolds et al. 2009, Paper II). The high column sug-
gests a distance of order that to the Galactic Center; we adopt
a nominal distance of 8.5 kpc, at which the mean expansion
rate is vs∼ 14,000 km s−1(Paper I). No thermal emission is
apparent in the integrated spectrum.
A great deal can be learned about the nonthermal emission
from the overall morphology and spatial variations of the non-
thermal spectrum (Paper II). However, without the detection
of thermal emission, basic, crucial information such as the
SN type, distance, elemental abundances, and a better age es-
timate cannot be obtained. Its detection was one of the goals
of a much longer Chandra observation.
2. OBSERVATIONS
We re-observed G1.9+0.3 for 237 ks with Chandra in four
observations between July 13 and 26, 2009, using the ACIS-
S CCD camera (S3 chip). We checked aspect correction
and created new level–1 event files appropriate for VFAINT
mode. No flares occurred during the observation.
correction was applied and calibration was performed using
CALDB version 3.4.0. Finally, the datasets were merged and
weighted response files created. We extracted spectra using
the specextract script. We obtained about 40,000 source
counts, consistent with the rate observed in 2007.
CTI
1Department of Physics, North Carolina State University, Raleigh, NC
27695-8202; kborkow@unity.ncsu.edu
2Cavendish Laboratory; 19 J.J. Thomson Ave., Cambridge CB3 0HE,
UK
3NASA/GSFC, Code 660, Greenbelt, MD 20771
4Department of Electrical and Computer Engineering, Duke University,
Durham, NC 27708
FIG. 1.— Chandra image of G1.9+0.3, smoothed with the spatio-spectral
method of Krishnamurthy et al. (2010). Red, 1 – 3 keV; green, 3 – 4.5 keV;
blue, 4.5 – 7.5 keV. Image size 127??×121??.
Figure 1 shows the 2009 image, smoothed with the spatio-
spectral method of Krishnamurthy et al. (2010). The remnant
has expanded significantly since 2007; a roughly E-W diam-
eter through the brightest parts of the shell gave an estimated
expansion rate corresponding to about 13% over 23 years, but
with large errors, consistent with rates reported in Paper I
(16±3%) and with the radio confirmation (15±2%; Green
et al. 2008). We defer a detailed discussion of proper motions
to a later paper. Here we concentrate on spectroscopy.
Figure 2 compares the radio and X-ray morphologies. The
marked and surprising difference between them suggests that
the radio image may hold a clue to the location of thermal
emission. We extracted the spectrum from the radio-bright
arXiv:1006.3552v1 [astro-ph.HE] 17 Jun 2010

Page 2

2
FIG. 2.— Top: 1.4 GHz radio image (Green et al. 2010, in preparation).
Resolution 2.3??×1.4??. Bottom: X-ray image showing regions from which
spectra below were extracted. Blue: North rim. Red: center.
region shown on Figure 2; it is shown in Figure 3. Lines are
apparent. We found lines as well in the central region shown
in Figure 2, and analyzed the regions separately.
3. SPECTRAL ANALYSIS
We combined the 2007 and 2009 datasets for a total ex-
posure of 286 ks. We modeled the background rather than
subtracting it, and used Markov-chain Monte Carlo (MCMC)
methods as implemented in the PyMC software package (Patil
et al. 2010) to determine best parameter values and error
ranges (e.g., van Dyk et al. 2001). MCMC methods require
the specification of priors on parameters to be determined; we
describe them in some detail below.
The spectrum of the northern, radio-bright rim is shown in
the upper panel of Figure 3. Spectral lines typical of strongly
underionized plasma are apparent (such plasma is expected in
G1.9+0.3 because of its youth and the low density of the am-
bient ISM). SN 1006 has a very similar X-ray spectrum; in
the same spectral range of Figure 3, Yamaguchi et al. (2008)
find prominent Kα lines of abundant elements such as Si, S,
Ar, Ca, and Fe, with line centroids at 1.815 keV, 2.36 keV,
3.01 keV, 3.69 keV, and 6.43 keV. (As in SN 1006, O, Ne, and
Mg lines might also be present at lower energies, but cannot
be seen because of the high absorption.) In addition to these
lines produced in hot shocked plasma, the radioactive decay
of44Ti to44Sc and finally to the stable isotope44Ca will re-
sult in the emission of X-ray and γ-ray lines in very young
remnants (44Ti decays with a mean life of 87±2 yr, Görres
et al. 1998). This decay commences via an electron capture
to44Sc, leaving a K-shell vacancy followed rapidly either by
Auger decay or by emission of a fluorescence photon of en-
ergy 4.09 keV (the fluorescent yield is 0.169). Nuclear de-
excitation gamma rays at 1.157 MeV and 78.4 and 67.9 keV
are also emitted. An inspection of the radio-bright rim spec-
trum (Figure 3) reveals the presence of Si, S, Ar, and Fe Kα
lines, and a broad feature near 4 keV that may be a blend of
Doppler-broadened Ca and Sc Kα lines. Lines are generally
weaker in the low-surface brightness interior (lower panel of
Figure 3), with Ar and Sc lines being the most prominent.
We modeled the spectra of the northern, radio-bright rim
and the faint interior with an absorbed power law plus emis-
sion lines of Si, S, Ar, Ca, Sc, and Fe. This simple model
does not account for dust scattering and does not separate the
underlying continuum into thermal and nonthermal compo-
nents, but it suffices for the determination of line strengths,
centroids, and widths. We used a normal (Gaussian) prior
for the absorbing column density NH, with mean (standard
deviation) of 6.89(0.11)×1022cm−2, based on our multire-
gion spectral fit without dust scattering to the 2007 data (Pa-
per II; solar abundances of the absorbing ISM are those of
Grevesse & Sauval 1998, fits with dust scattering resulted in
NHlower by 25%). Noninformative, uniform and logarithmic
priors were assumed for the power-law index Γ and the (un-
absorbed) 5–10 keV continuum flux F5−10 keV, respectively.
We used normal priors for line energies, setting mean ther-
mal line energies equal to the values measured by Yamaguchi
et al. (2008) for SN 1006, and to 4.09 keV for the Sc line. We
allowed for the possible presence of significant Doppler shifts
by choosing a conservative, large (σ = 104km s−1) Doppler
width for these priors. (For numerical stability, these nor-
mal priors were truncated to include only a finite range in line
energies; we verified that our results are not affected by this
procedure.) In view of the 14,000 km s−1blast wave speed,
large line widths are expected. Thermal lines in a young rem-
nant arise in a fast-moving, shocked shell bounded by for-
ward and reverse shocks. An optically and geometrically thin
shell expanding with velocity vshellproduces flat-topped lines
with Doppler widths of 2vshell. (Line widths are likely to be
somewhat less in off-center locations such as the north rim.)
We assumed flat-topped profiles with the same (but unknown)
Doppler width for all thermal lines. A truncated normal prior
was assumed for vshell, with mean of 14,000 km s−1and 1σ
width of 5000 km s−1, extending from 0 to 50,000 km s−1.
The Sc line was modeled by a Gaussian; we assumed a half-
normal prior for its width σSc
for line widths exclude very large (> 50,000 km s−1) widths,
but otherwise provide rather weak constraints.
Some constraints on the Sc line strength are provided by
vwith σ = 0.15 keV. These priors

Page 3

RADIOACTIVE SCANDIUM IN YOUNGEST GALACTIC SNR G1.9+0.33
IBIS/ISGRI onboard INTEGRAL. The 68 and 78 keV lines
have been detected by IBIS/ISGRI in Cas A (Renaud et al.
2006) but not in G1.9+0.3 (Renaud et al. 2009). We use the
Cas A detection to bound priors. We chose a weakly infor-
mative gamma prior for the Sc line strength, defined as the
expected line counts in a 286 ks exposure with Chandra, with
shape parameter α = 1.2 and scale parameter β = 0.01 (see
van Dyk et al. (2001) for discussion of the use of gamma pri-
ors in modeling emission lines in X-ray spectra). Its mean of
α/β =120 counts is comparable to the square root of the vari-
ance, equal to (α/β2)1/2= 110 counts. This prior disfavors
a strong Sc line that was searched for but not found in Cas
A (Theiling & Leising 2006), but otherwise provides rather
weak constraints. We assumed the same weakly informative
prior on line strength for thermal lines of Si, S, Ar, Ca, and
Fe.
The sky background was determined by a fit to the back-
ground spectrum extracted from a large area on the S3 chip.
The sky background was modeled by two absorbed power-
laws, while the particle background model involved a combi-
nation of power-laws with exponential cutoffs and narrow flu-
orescent lines. We allowed for spatial variations in the particle
background, with a logarithmic prior imposed on the particle
normalization.
Table 1 contains results of spectral fits; models are plotted
in Figure 3. Except for fluxes, values quoted in Table 1 are
standard means of the MCMC draws. We used a geometric
mean for F5−10 keV, and mean line fluxes F were evaluated as
(?F1/2?)2(according to recommendations by van Dyk et al.
2001). The 90% confidence intervals were computed using
the 0.05 and 0.95 quantiles of the draws. The Sc line width is
poorly constrained, so we provide only upper limits based on
the 90% highest probability density (HPD) interval for σSc
The Sc line and thermal lines of all abundant elements have
appreciable strengths in the spectrum of the northern rim.
Thermal lines are very broad, with a full width of 28,000 km
s−1, consistent with the estimated shock velocity of 14,000 km
s−1. The Fe Kα line provides the strongest constraints on line
widths, but Doppler broadening is important for other lines
as well. Photon statistics are insufficient to study variations
in Doppler widths between different elements. This includes
Sc, with a width that does not seem to be different from ther-
mal line widths. But Sc forms a line blend with Ca, so errors
are particularly large for these elements. In the interior, the
two strongest lines are Ar and Sc. The 90% confidence inter-
vals for Si, S, Ca, and Fe extend to low flux levels ? 5×10−8
ph cm−2s−1, so only one thermal line (Ar) may have been
unambiguously detected in the interior. Notwithstanding this
decline in the overall strength of thermal lines, the Sc line re-
mains strong (it is only a factor of 2 weaker than in the north-
ern rim). The prominence of Sc in the interior (relative to
thermal lines) is consistent with its origin in the unshocked
ejecta.
In both spectral regions independently, we examined the
significance of the Sc line detection using a likelihood ratio
test as described in Protassov et al. (2002). The null model
consisted of an absorbed power law plus Si, S, Ar, Ca and Fe
lines described by flat line profiles with equal (but unknown)
Dopplerwidths, togetherwiththesame priorsthatweused for
our spectral fits. The alternative model is obtained by adding
an additional line at 4.09 keV to the null model. Large sam-
ples from the posterior distribution of the null model were
obtained using MCMC simulations. From these samples we
v.
FIG. 3.— Top: Spectrum of N rim (radio-bright region) shown in Figure 2.
Lines of Ca and Sc blend together because of the large Doppler widths. Bot-
tom: Spectrum of interior region. In both cases, background has been mod-
eled rather than subtracted; source and background models are described in
the text. The two black lines are the background and source models; the red
line is the total.
simulated synthetic datasets, fit each dataset with the null
model and the alternative model, and computed the likelihood
ratio statistic T. We find that the posterior predictive p-values
for the null model are 0.0098 and 0.030 for the northern rim
and interior, respectively. The p-value is equal to 0.00029 for
the null model to be valid in both of these regions. We can
therefore reject the null model, i.e., claim the detection of the
4.1 keV line, at 99.971% significance.
The total Sc line flux was derived by combining the spectra
shown in Figure 3, and fitting the combined dataset separately
with the same spectral model as before. We obtain the total
Sc line flux of 1.2×10−6ph cm−2s−1, with a 95% confidence
interval of (0.35,2.4)×10−6ph cm−2s−1. The column density
of 6.9×1022cm−2implies that X-rays of 4.1 keV are attenu-
ated by about 33% due both to absorption and scattering. The
Sc line fluxes have not been corrected for this, but the cor-
rection has been made for the44Ti masses that are discussed
next.
We expect 0.169 fluorescence photons per44Sc atom, be-
fore decay to44Ca (mean life 5.4 hr) (Hiraga et al. 2009).
Thus the line flux, along with the source age and distance,
gives directly the amount of44Ti synthesized in the explo-
sion. Including the branching ratio for the emission of the
1.157 MeV photon, we expect the flux of X-ray photons to

Page 4

4
be related to that in 1.157 MeV photons by Fx= 0.15Fγ. The
gamma-ray flux is related to the mass of44Ti by
?
with the mean life τ = 87 yr (Görres et al. 1998). (If the44Ti
is ionized beyond the He-like state, the effective lifetime is
longer, but the low ionization age of Fe implied by the line
centroid reported in Table 1 means that this effect should be
negligible.) We used these relations with an age of 100 yr
to obtain the following44Ti masses: 2.8(1.1,5.1)×10−5M?
in the north, 1.3(0.36,2.6) × 10−5M? in the interior, and
3.3(0.95,6.5)×10−5M?for the total. (Errors are 90% con-
fidence limits for the two subregions, but 95% limits for the
total.) These masses have been corrected for absorption and
scattering by being multiplied by a factor of 3/2. For an age
of 140 yr (corresponding to no deceleration at all, and hence
an upper limit), the masses would be larger by an additional
factor of 1.54.
We also modeled the spectrum of the northern, radio-bright
rim with a plane shock non-equilibrium ionization (NEI)
model (vpshock in XSPEC), again without dust. We find
NH= 6.88(6.72,7.03)×1022cm−2, consistent with earlier de-
terminations; kTe= 3.6±0.4 keV; and ionization timescale
τ ≡ net = 1.4(0.78,3.1)×109cm−3s. (Cited uncertainties
are 90% confidence intervals.) This extremely low ioniza-
tion age is very unusual for SNRs, but appropriate for an un-
precedentedly young object. Abundances have substantial er-
rors, but solar abundances are ruled out: compared to solar,
Si (set equal to S) = 3.4 (2.0, 4.6); Ar = 17 (2.8, 37), and
Fe = 4.1(2.4, 5.7). A blueshift of line centers is required,
vr= −4,300(−7,600,−1,300) km s−1. We find a FWHM of
26,500(17,600,34,600) km s−1, consistent with an estimated
shock velocity of 14,000 km s−1.
Fγ=3.6×10−6M(Ti)
10−5M?
D
8.5 kpc
?−2?τ
yr
?−1
e−t/τph cm−2s−1
4. DISCUSSION
We summarize our results as follows:
1. Clearexpansioncanbeseenbetweenthe2007and2009
X-ray images, consistent with rates inferred between
1985 radio and 2007 X-ray, and between 1985 radio
and 2008 radio.
2. The radio-bright N rim region shows strong Kα lines of
Si, S, Ar, Ca, and Fe (much weaker thermal emission is
also present in the shell interior).
3. The spectral lines are both broadened and shifted.
Widths of thermal lines are 26,000±4,000 km s−1with
blueshifts of about 4,000 km s−1.
4. Global spectral fits with a plane-shock model imply
strong overabundances of the elements we detect, con-
sistent with their interpretation as ejecta.
5. We detect for the first time with high significance a line
at 4.1 keV which we attribute to scandium-44. The line
strength of about 1.2×10−6ph cm−2s−1implies a mass
of44Ti of 3.3(0.95,6.5)×10−5M?, assuming an age of
100 yr and a distance of 8.5 kpc.
6. We detect44Sc separately in the northern shell and the
interior, with greater strength in the shell.
We believe the discovery of44Sc emission to be the most
important new result. A feature at 4.1 keV identified with44Sc
has been reported previously in G266.2–1.2 (RX J0852.0-
4622) by Tsunemi et al. (2000), Iyudin et al. (2005), and
Bamba et al. (2005). However, Slane et al. (2001) and, more
comprehensively, Hiraga et al. (2009) have failed to confirm
these reports, at a level below the previous detection claims. It
appears that G1.9+0.3 shows the first definite detection of this
transition. Our line strength implies a flux in the 1.157 MeV
line of about 8×10−6ph cm−2s−1, well below the COMP-
TEL limit of about 2×10−5ph cm−2s−1(Dupraz et al. 1997).
While our inferred44Ti mass is nominally in conflict with the
upper limit of 2×10−5M?reported by Renaud et al. (2009),
the errors are large enough to accommodate it.
44Ti, while a trace product of explosive nucleosynthesis,
carries important information about the details of the explo-
sion (Diehl & Timmes 1998; The et al. 2006). It is primar-
ily produced in “α-rich freezeout” conditions, in which rapid
cooling causes departures from nuclear statistical equilibrium
and a high concentration of free α particles. In core-collapse
(CC) SNe, its production is sensitive to the location of the
mass cut (within which material becomes the neutron star),
and in most models its production is correlated with that of
56Ni, and therefore with the supernova luminosity. Predicted
yields range from 3 to 9×10−5M?(see summary in The et
al. 2006). Traditional SN Ia models undergo α-rich freezeout
under somewhat different conditions and do not produce large
quantities of44Ti. However, recent simulations of off-center
delayed-detonation explosions have shown substantially in-
creased yields (Maeda et al. 2010). Maeda et al. (2010) quote
a mass of44Ti of 1.6×10−5M?, compared to 2×10−6M?
for a centrally-ignited pure deflagration and 3×10−6M?for a
central delayed detonation. (For comparison, the benchmark
W7 Type Ia model [Nomoto, Thielemann, & Yokoi 1984]
gives 8×10−6M?, while a series of more recent simulations
by Iwamoto et al. (1999) give a range from (1−5)×10−5M?.)
Galactic constraints on the supernova rate can be obtained
from the observed abundance of44Ca, and the absence of ob-
vious discrete sources of 1.157 MeV γ-ray emission, with the
important exception of Cas A, the only firm detection to date
(Iyudin et al. 1994, 1999). The accompanying hard X-ray
lines have also been seen from Cas A (Renaud et al. 2006,
and references therein), with an inferred mass of44Ti of about
2×10−4M?, rather more than expected from most CC models
((6−14)×10−5M?; Timmes et al. 1996), though asymmetric
models can do better (e.g., Nagataki et al. 1998).
Since the 4.1 keV inner-shell transition of44Sc formed from
electron capture in44Ti does not require that the scandium be
ionized, we are sensitive to both shocked and unshocked ma-
terial. This is evident in the interior spectrum in Figure 3, in
which the44Sc line is as prominent as in the north rim, but the
other lines are weaker. The greater abundance of44Sc in the
north, then, represents a true spatial distribution – surprising,
considering that the44Sc is expected to be produced in the
neighborhood of the Fe-peak elements, that is, the innermost
ejecta. These properties suggest that the explosion was sub-
stantiallyasymmetric, perhapsconsistentwiththegreaterpro-
duction of44Ti found in asymmetric explosion models such
as Maeda et al. (2010). A substantially longer observation
should allow better spatial localization of the44Sc emission,
and an unprecedented view into supernova nucleosynthesis.

Page 5

RADIOACTIVE SCANDIUM IN YOUNGEST GALACTIC SNR G1.9+0.35
TABLE 1
SPECTRAL FITS
RegionNH
Γ
F5−10 keV
vshell
(104km s−1)
σSc
v
Line energies (keV) and line strengths (10−7ph cm−2s−1)
SulfurArgon(1022cm−2)(10−13ergs cm−2s−1)SiliconCalciumScandiumIron
North rim6.88 2.403.001.4 1.11.832.352.993.724.086.49
(6.72, 7.05) (2.29, 2.51) (2.80, 3.22)(0.9, 1.9) < 1.9 (1.78, 1.88) (2.31, 2.40) (2.92, 3.04) (3.59, 3.87) (4.00, 4.15) (6.38, 6.61)
4.410.4
(2.1, 7.0)(5.9, 15.6)(3.1, 13)
2.10.91.832.32
(1.2, 3.0) < 1.6 (1.72, 1.95) (2.17, 2.46) (2.87, 3.05) (3.52, 3.94) (4.02, 4.17)
1.2 3.1
(0.1, 4.0)(0.4, 8.3)(1.5, 13)
7.75.21018
(1.2, 11)
3.74
(4.0, 19)
4.09
(11, 27)
6.4
(6.0, 6.8)
3.0
(0.4, 8.4)
Interior6.892.71 1.352.96
(6.72, 7.06) (2.54, 2.86) (1.22, 1.51)
5.91.94.6
(0.2, 6.0)(1.3, 9.5)
NOTE. — Line energies (strengths) are in rows 1 and 5 (3 and 7), with 90% confidence limits listed in adjacent rows.
Our discovery of these spectral lines, and the44Sc line
in particular, remains consistent with the possibility that
G1.9+0.3 resulted from a Type Ia event, a possibility we find
increasingly likely. The bilaterally symmetric synchrotron–
X-ray morphology, the extremely high shock velocities we
both infer and measure from line broadening, and the promi-
nent Fe K emission, all support a Type Ia origin. We have not
yetruledoutaCCorigin, however; whilestrongironemission
is unusual in CC remnants, Cas A is a counterexample, and it
is possible that a CC origin can explain the other features as
well. What is certainly clear is that G1.9+0.3 still has a great
deal to teach us about the evolution of very young SNRs and
about the supernovae that produced them.
This work was supported by NASA through Chandra Gen-
eral Observer Program grant SAO GO6-7059X.
REFERENCES
Bamba, A., Yamazaki, R., & Hiraga, J.S. 2005, ApJ, 632, 294
Diehl, R., & Timmes, F.X. 1998, PASP, 110, 637
Dupraz, C., et al. 1997, A&A, 324, 683
Görres, J., et al. 1998, PRL, 80, 2554
Green, D.A., et al. 2008, MNRAS, 387, L54
Grevesse, N., & Sauval, A. J. 1998, Space Science Reviews, 85, 161
Hiraga, J.S., et al. 2009, PASJ, 61, 275
Iwamoto, K., et al. 1999, ApJS, 125, 439
Iyudin, A.F., et al. 1994, A&A, 284, L1
Iyudin, A.F., et al. 1999, Astrophys.Lett.Comm., 38, 383
Iyudin, A. F., et al. 2005, A&A, 429, 225
Krishnamurthy, K., Raginsky, M., & Willett, R. 2010, SIAM J.Imaging.Sci.,
in press
Maeda, K., et al. 2010, ApJ, 712, 624
Nagataki, S., et al. 1998, ApJ, 492, L45
Nomoto, K., Thielemann, F.-K., & Yokoi, K. 1984, ApJ, 286, 644
Patil, A., et al. 2010, submitted to Journal of Statistical Software
Protassov, R., et al. 2002, ApJ, 571, 545
Renaud, M., et al. 2006, ApJ, 647, L41
Renaud, M., et al. 2009, in Proceedings of the 2nd International Simbol-X
Symposium, Simbol-X: Focusing on the Hard X-ray Universe. AIP Conf.
Proc., 1126, 359
Reynolds, S.P., et al. 2008, ApJ, 680, L41 (Paper I)
Reynolds, S.P., et al. 2009, ApJ, 695, L149 (Paper II)
Slane, P,O., et al. 2001, ApJ, 548, 814
The, L.-S., et al. 2006, A&A, 450, 1037
Theiling, M. F., & Leising, M., D. 2006, New Astronomy Reviews, 50, 544
Timmes, F. X., et al. 1996, ApJ, 464, 332
Tsunemi, H., et al. 2000, PASJ, 52, 887
van Dyk, D.A., et al. 2001, ApJ, 548, 224
Yamaguchi, H., et al. 2008, PASJ, 60, 141