The Astrophysical Journal, 725:20382058, 2010 December 20


C 2010.

doi:10.1088/0004-637X/725/2/2038

The American Astronomical Society. All rights reserved. Printed in the U.S.A.

THE THREE-DIMENSIONAL STRUCTURE OF CASSIOPEIA A

Tracey DeLaney1 , Lawrence Rudnick2 , M. D. Stage3 , J. D. Smith4 , Karl Isensee2 , Jeonghee Rho5 , Glenn E. Allen6 ,
Haley Gomez7 , Takashi Kozasa8 , William T. Reach5 , J. E. Davis6 , and J. C. Houck6
1

Department of Physics and Engineering, West Virginia Wesleyan College, 59 College Avenue, Buckhannon, WV 26201, USA; delaney_t@wvwc.edu
2 Astronomy Department, University of Minnesota, Minneapolis, MN 55455, USA; larry@astro.umn.edu, isensee@astro.umn.edu
3 Mount Holyoke College, Department of Astronomy, 206 Kendade, 50 College Street, South Hadley, MA 01075, USA; mstage@mtholyoke.edu
4 Ritter Astrophysical Observatory, University of Toledo, Toledo, OH 43606, USA; jd.smith@utoledo.edu
5 Spitzer Science Center, California Institute of Technology, MS 220-6, Pasadena, CA 91125, USA; rho@ipac.caltech.edu, reach@ipac.caltech.edu
6 Massachusetts Institute of Technology, Kavli Institute for Astrophysics and Space Research, 77 Massachusetts Avenue, NE80, Cambridge, MA 02139, USA;
gea@space.mit.edu, houck@space.mit.edu, davis@space.mit.edu
7 School of Physics and Astronomy, Cardiff University, Queens Building, The Parade, Cardiff CF24 3AA, UK; haley.morgan@astro.cf.ac.uk
8 Department of Cosmosciences, Graduate School of Science, Hokkaido University, Sapporo 060-0810, Japan; kozasa@mail.sci.hokudai.ac.jp
Received 2009 August 11; accepted 2010 October 20; published 2010 December 3

ABSTRACT
We used the Spitzer Space Telescopes Infrared Spectrograph to map nearly the entire extent of Cassiopeia A
between 5 and 40 m. Using infrared and Chandra X-ray Doppler velocity measurements, along with the locations
of optical ejecta beyond the forward shock, we constructed a three-dimensional model of the remnant. The structure
of Cas A can be characterized into a spherical component, a tilted thick disk, and multiple ejecta jets/pistons and
optical fast-moving knots all populating the thick disk plane. The Bright Ring in Cas A identifies the intersection
between the thick plane/pistons and a roughly spherical reverse shock. The ejecta pistons indicate a radial velocity
gradient in the explosion. Some ejecta pistons are bipolar with oppositely directed flows about the expansion center
while some ejecta pistons show no such symmetry. Some ejecta pistons appear to maintain the integrity of the
nuclear burning layers while others appear to have punched through the outer layers. The ejecta pistons indicate
a radial velocity gradient in the explosion. In three dimensions, the Fe jet in the southeast occupies a hole in
the Si-group emission and does not represent overturning, as previously thought. Although interaction with the
circumstellar medium affects the detailed appearance of the remnant and may affect the visibility of the southeast
Fe jet, the bulk of the symmetries and asymmetries in Cas A are intrinsic to the explosion.
Key words: infrared: ISM ISM: individual objects (Cassiopeia A) ISM: supernova remnants X-rays: ISM
Online-only material: animation, interactive PDF

through and being compressed and heated by the reverse shock

(Morse et al. 2004; Patnaude & Fesen 2007). In a few locations,
the position of the reverse shock itself has been identified just
inside of the Bright Ring from the rapid turn-on of optical ejecta
(Morse et al. 2004).
While most of the observed ejecta at all wavebands are concentrated on the Bright Ring, optical ejecta are also identified at and beyond the forward shock. The forward shock is
visible as a thin ring of X-ray filaments approximately 300
in diameter (Gotthelf et al. 2001). The canonical explanation
for the far outer knots is that they are dense optical ejecta
(ne 400010,000 cm3 ; Fesen & Gunderson 1996) which
cool quickly (tc 2030 yr; Kamper & van den Bergh 1976)
after passing through the reverse shock. They are reheated after
they overtake the forward shock and encounter slow-moving
ambient material which drives a slow shock into the supersonic
dense ejecta and produces optical emission. In addition to the
largely symmetric-appearing Bright Ring and forward shock,
jets of Si- and S-rich ejecta extend out to large distances in the
northeast and southwest with optical ejecta identified at a distance of 200 beyond the forward shock radius (Hwang et al.
2004). The jets may denote an axis of symmetry of the supernova explosion (e.g., Fesen et al. 2006). The optical knots are
rapidly expanding with proper motions of thousands of km s1
on the Bright Ring and up to 14,000 km s1 for the optical ejecta
at the tip of the jets (Kamper & van den Bergh 1976; Fesen et al.
2006). To the southeast, X-ray images show Fe-rich ejecta at a

1. INTRODUCTION
The three-dimensional structure of core-collapse supernova
explosions is of considerable interest, with implications for the
explosion mechanisms (e.g., Akiyama et al. 2003; Wang et al.
2007), the role of rotation (e.g., Moiseenko & Bisnovatyi-Kogan
2007), pulsar natal kicks (e.g., Scheck et al. 2004), and for the
subsequent interaction of ejecta with the circumstellar medium/
pre-supernova wind (Schure et al. 2008). Asymmetries in the
explosion and subsequent fallback have implications for mixing
and the subsequent progression of explosive nucleosynthesis
(Joggerst et al. 2009). Jet-induced scenarios may play a role
in forming supernova remnants such as Cassiopeia A (Cas A)
(Wheeler et al. 2008), and at their extreme, jet-dominated
explosions may be responsible for gamma-ray bursts (e.g.,
Zhang et al. 2004; Mazzali et al. 2005). There is plentiful
evidence for asymmetric supernova explosions from structural
and spectral data (Wang et al. 2002) and from spectropolarimetry
(e.g., Wang et al. 2001; Tanaka et al. 2008).
Cas A provides a unique opportunity to study the threedimensional supernova explosion structure because of the high
velocities of its clumpy ejecta, seen both through Doppler and
proper motions. It is the result of a core-collapse supernova
approximately 330 years ago (Fesen et al. 2006) and is bright
across the electromagnetic spectrum. The brightest emission
in Cas A is concentrated onto the 200 diameter Bright Ring
where ejecta from the explosion are illuminated after crossing
2038

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

greater radius than Si-rich ejecta which has been interpreted as

an overturning of ejecta layers during the supernova explosion
(Hughes et al. 2000).
Large-scale Doppler mappings have been carried out at
optical wavelengths primarily using S and O emission lines
(Lawrence et al. 1995; Reed et al. 1995). These studies showed
that the ejecta defining the Bright Ring outline a shell. Assuming
that the shell is spherical, the Doppler velocities of ejecta knots
can be converted into equivalent distances from the expansion
center along the line of sight. The bulk of the ejecta are found
to lie nearly in the plane of the sky and they are organized into
distinct velocity structures such as the two complete ringsone
blueshifted and one redshiftedthat form the northern part of
the Bright Ring.
A subsequent census of outlying optical ejecta showed that
the northeast Jet is oriented within 6 of the plane of the sky
with an opening angle of 25 (Fesen & Gunderson 1996). The
remainder of the outer optical knots are mostly located within
20 of the plane of the sky and form a giant ring at the location
of the forward shock (Fesen 2001) although there are gaps in
this distribution to the north and south (Fesen et al. 2006). None
of the optical searches, either imaging or spectral, reveal fastmoving ejecta knots projected near the center of the remnant or
a population of very fast moving outer optical knots projected
onto the Bright Ring. One interpretation offered by Fesen (2001)
for this plane-of-the-sky effect was that a near tangent viewing
angle is required to detect ejecta knots. However, other effects
may play a role in the lack of projected outer ejecta knots.
For instance, imaging searches rely on large proper motions to
detect outer ejecta knots and the filters used for the images
may not have been broad enough to detect the highly redor blueshifted emission. While the spectral searches certainly
had the wavelength range to detect fast-moving projected ejecta
knots, the outer optical knots tend to be very faint unless they
are encountering dense ambient media (Fesen 2001; Fesen et al.
2006). Given the small amount of H emission projected against
Cas A (Fesen 2001), any projected outer ejecta knots may be too
faint to have been detected in the spectral searches. Therefore,
despite the absence of a population of projected ejecta near the
center of Cas A, either with velocities typical of outer ejecta
or velocities typical of Bright Ring ejecta, the Bright Ring and
the outer optical knot ring were believed to be limb-brightened
shells (Lawrence et al. 1995; Fesen 2001).
Spectra of the X-ray-emitting ejecta have also been mapped
using a number of telescopes. The first mapping with the Einstein X-ray Observatory (Markert et al. 1983) and a subsequent mapping with ASCA (Holt et al. 1994) were at very low
spatial resolution (1 2 ) and showed large-scale asymmetries
in Doppler structure. These asymmetries were confirmed at
higher spatial resolution with XMM-Newton (20 ; Willingale
et al. 2002, 2003) and the first 50 ks ACIS observation with
the Chandra X-ray Observatory (4 ; Hwang et al. 2001). At
their improved spatial resolution, Willingale et al. (2002) found
that the Si-rich ejecta forms the same general set of ring structures as the optical emission but the two distinct rings to the
north are difficult to identify. The Fe-rich emission is concentrated in three main areas to the north, southeast, and west, but
Willingale et al. (2003) determined that most of the Fe mass
is concentrated within what they describe as a bipolar, double
cone that is oriented at 55 from north and 50 out of the plane
of the sky. Hughes et al. (2000) suggest an overturning in the
explosion in the southeast, with the Fe emission extending out
to the forward shock. Willingale et al. (2002) also found that in

2039

the north the centroid of the Fe-rich ejecta appears behind the
Si-rich ejecta perhaps identifying another overturning region. A
higher spatial resolution (1 ) analysis using the Chandra ACIS
1 Ms data set plus two archival 50 ks data sets also shows the
same optical ring structures in Si-rich emission and shows the
strong Fe-K regions as dynamically distinct (Stage et al. 2004;
Davis et al. 2005). A more recent analysis using Chandra High
Energy Transmission Grating (HETG) spectra of the Si-He
triplet showed that, on 1 spatial scales, the Si-rich ejecta has
a great deal of substructure that is reminiscent of the variations
seen in the optical data (Lazendic et al. 2006).
In this paper, we present a Doppler analysis of Cas As
ejecta using infrared data from the Spitzer Space Telescope
(Ennis et al. 2006) and X-ray data from the archival Chandra
ACIS 1 Ms observation plus two other 50 ks observations
(Stage et al. 2006; Davis et al. 2005). For the first time, we
combine multiwavelength data into a full three-dimensional
reconstruction of Cas A. To this model, we add previously
published X-ray results from Lazendic et al. (2006) and the
jets and outer optical knots from Fesen (2001) and Fesen &
Gunderson (1996). Based on this model, we draw conclusions
about the multiple kinematic components and several major
asymmetries in Cas As explosion.
2. SPITZER OBSERVATIONS AND DATA ANALYSIS
The Spitzer Infrared Spectrograph (IRS) was used on 2005
January 13 to spectrally map nearly the full extent of Cas A
with portions of the outer structures missing from some slits
as shown in Figure 1. Low-resolution spectra (resolving power
of 60128) were taken between 515 m (shortlow module,
SL) and 1538 m (longlow module, LL) with each module
including two orders of wavelength. The long-wavelength
(1538 m) spectra were taken in a single large map with 4
91 pointings, using a single 6 s ramp at each position. To achieve
the spatial coverage with the short-wavelength (515 m) slit, a
set of four quadrant maps were made, two with 4 87 pointings
and two with 3 87 pointings, using a 6 s ramp at each position.
The mapped area ranged from 6. 3 5. 9 (SL) to 11. 0 7. 8 (LL),
with offsets between the maps produced in each of the two orders
in each module of 3. 2 (LL) and 1. 3 (SL), along the slit direction.
The effective overlap coverage of all modules and orders is
4. 9 5. 8. The data used here were processed with the S12
version of the IRS pipeline, using the CUBISM package (Smith
et al. 2007a) to reconstruct the spectra at each slit position,
subtract the sky background, and create three-dimensional data
cubes as described in Smith et al. (2009). The statistical errors at
each position in the data cube are calculated using standard error
propagation of the BCD-level uncertainty estimates produced by
the IRS pipeline.
2.1. The Bright Ring and Diffuse Interior Emission
There are a number of bright infrared ionic emission lines in
Cas A from elements such as Ar, Ne, Si, S, Fe, and the 26 m
blend of Fe and O. For the most part, images made in these
emission lines bear a close resemblance to optical images that
primarily contain S and O emission (Ennis et al. 2006). In order
to characterize the Bright Ring, we have chosen the 6.99 m
[Ar ii] line and the 12.81 m [Ne ii] line. The [Ar ii] line is
relatively bright everywhere, even in Ne and O bright regions,
and it is relatively isolated from other emission lines except
for very weak 6.63 m [Ni ii] and possibly 6.72 m [Fe ii],

2040

DELANEY ET AL.

Vol. 725

Figure 1. Slit positions for the Spitzer/IRS mapping overlaid on the Chandra broadband X-ray image. The wavelength ranges in microns are indicated for each of the
shortlow (SL) and longlow (LL) orders.

although the latter transition is a 4 F9/2 a 6 D7/2 which we expect

to be negligible even given the [Fe ii] detected in Cas A at other
infrared wavelengths (Smith et al. 2009; Gerardy & Fesen 2001;
Rho et al. 2003). The [Ne ii] line was used in order to calculate
the [Ne ii]/[Ar ii] ratio, which varies considerably from placeto-place on the Bright Ring and may indicate inhomogeneous
mixing between the oxygen-burning and carbon-burning layers
of the progenitor star (Ennis et al. 2006; Smith et al. 2009).
The disadvantage of the [Ar ii] and [Ne ii] lines is that the SL
mapping did not extend out to the full extent of the Jet and
Counterjet. The [Ar ii] and [Ne ii] images are shown in Figure 2.
In addition to the Bright Ring, there is Diffuse Interior
Emission seen in the 10.51 m [S iv] line, the 18.71 m [S iii]
line, the 33.48 m line of [S iii], and the 34.82 m line of [Si ii]
(Rho et al. 2008; Smith et al. 2009). Bright interior emission
is seen at 26 m as well. Since we detect no emission from
any of the other infrared Fe lines (5.35 m, 17.9 m, etc.) in
the Diffuse Interior Emission, we believe that the 26 m line
is [O iv] with little or no contribution from [Fe ii]. Support for
this interpretation is provided by the high-resolution Spitzer
spectra in Isensee et al. (2010).9 The [Si ii] image, shown in
Figure 2, shows both emission associated with the Bright Ring
that is presumably shocked supernova ejecta and emission in
the interior. The Diffuse Interior Emission bears a striking
resemblance to the freefree absorption observed at 74 MHz
(Kassim et al. 1995). That absorption is thought to occur due to
cold, photoionized ejecta that have not yet crossed the reverse
shock (e.g., Hamilton & Fesen 1998). We discuss the physical
conditions in the unshocked ejecta more fully in Section 4.7.
This Diffuse Interior Emission is brightest and covers most
of the remnant in the [O iv] and in the long wavelength [S iii]
9

Note that the 26 m line is bright in the interior and on the Bright Ring. We
now know from the high-resolution Spitzer spectra that in the interior, the
26 m line is [O iv], however, on the Bright Ring, there is emission from both
[Fe ii] and [O iv] (K. Isensee 2009, private communication).

and [Si ii] lines. We have chosen to use both the 33.48 m [S iii]
line and the 34.82 m [Si ii] line to map the interior emission.
These lines are not entirely free of contamination from nearby
linesthere is a small probability of [Fe iii] at 33.04 m and
a reasonable probability of [Fe ii] at 35.35 m. However, these
potentially contaminating lines lie on opposite sides of the [S iii]
and [Si ii] lines, making it easier to detect their presence. The
[O iv] line is less suitable for mapping the remnant because,
although [O iv] dominates in the interior, on the Bright Ring
there is the possibility of contamination from [Fe ii] emission.

2.2. Preview of Velocity Structure: The First Moment Map

Figure 3 shows the first moment map of the [Ar ii] line,
calculated over the range 7100 to +10,600 km s1 . This image
presents a preview of the Doppler structure of Cas A to be
discussed in more detail below. The color scale ranges from
4000 to +6000 km s1 from blue to red. In approximately
1/3 of the locations, there is more than one Doppler component
along the line of sight, so the first moment represents only a
weighted average of those components. The [Ar ii] first moment
map is very similar to the optical Doppler map of Lawrence
et al. (1995) which was reconstructed from the fitting of spectra.
The double ring structures to the northone redshifted and
one blueshiftedare clearly defined as are the blueshifted
parentheses southeast of center and the distinct Doppler
structures at the base of the northeast jet. The velocity range
is the same as found for the optical ejecta (Lawrence et al.
1995; Reed et al. 1995) meaning that the infrared ejecta have
experienced very little or no deceleration and are effectively in
free expansion as Thorstensen et al. (2001) note for the optical
ejecta. We do not construct a similar first moment map for [Si ii]
because 3/4 of the lines of sight contain more than one Doppler
component.

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

2041

Figure 2. Left: 6.99 m [Ar ii] image and center: 12.81 m [Ne ii] image showing the Bright Ring. Right: 34.8 m [Si ii] image showing the Diffuse Interior Emission.
For reference, the location of the central compact object (CCO) is denoted by the crosses. The [Ar ii] spectrum in Figure 4 was extracted from the small box region on
the southwest Bright Ring. The [Si ii] spectrum in Figure 4 was extracted from the central box near the CCO.

Figure 3. First moment map for [Ar ii] showing the major Doppler structures
that define the Bright Ring. Note that in locations with more than one Doppler
component, the color represents a brightness-weighted average velocity.

2.3. Gaussian Fitting

In order to determine velocities for the structures in the
remnant, we fit Gaussian profiles to each selected emission line
as described below. The data cubes were first spatially binned
by 2 pixels in the x- and y-directions to improve signal to noise,
resulting in effective resolutions of 3. 7 for the [Ar ii] emission
and 10. 2 for the [Si ii] emission. Rather than concentrate on
individual knots or clumps of ejecta, we systematically fit
every spatial pixel in our binned cubes that had an emission
line detection at greater than 3 . The spectral resolution is
approximately the same for all of our selected emission lines
(R 100). The corresponding velocity resolution varies from
2500 to 3000 km s1 .
The Gaussian fitting was carried out using the Interactive
Data Language and the routines MPFIT10 for the Gaussian
component(s) and POLY-FIT to model the local continuum
as a straight line. We began with an automated routine in
which only 1 or 2 Gaussian components were allowed per
emission line with no constraint on their positions with respect to
zero velocityi.e., they could both be negative, both positive,
or one negative and one positive. The [S iii] and [Si ii] lines
10

http://cow.physics.wisc.edu/craigm/idl/fitting.html

were jointly fit because they were close enough in wavelength

that highly redshifted S could blend with highly blueshifted
Si. The automated process only allowed fits that met the
following criteria: (1) Gaussian full width at half-maximum
within 500 km s1 of the spectral resolution of the line being fit,
(2) a signal-to-noise ratio greater than 3, (3) statistical velocity
errors less than 1000 km s1 , (4) velocity values less than
20,000 km s1 , and (5) a reduced 2 value less than 70 for
[S iii] and [Si ii] and less than 25 for [Ar ii] and [Ne ii]. The
automated fits were then examined by hand for accuracy and
some fitting was redone manually to reject false positives and
recover false negatives.
The top panel of Figure 4 shows a typical 2-Gaussian
fit to the [Ar ii] emission at a location on the Bright Ring
with two major velocity structures superposed along the line
of sight. Typical formal velocity errors from the Gaussian
fitting averaged about 100 km s1 , however the actual velocity
uncertainties, determined by experiments with different binning
and Gaussian width constraints, are closer to 200 km s1 . For
the weaker [Ne ii] line, the actual velocity uncertainties were
about 400 km s1 . The [Ar ii] and [Ne ii] fits were compared for
each location and where their velocities agreed within 1 , the
ratio between the fitted Gaussian heights was computed to find
the regions where [Ne ii] is relatively strong.
The bottom panel of Figure 4 shows a typical 2-Gaussian
fit to the [S iii] and [Si ii] emission in a region near the center.
Actual velocity uncertainties for the [Si ii] emission average
about 300 km s1 and for the much weaker [S iii] line, the
actual velocity uncertainties are about 700 km s1 . The fitted
velocities for the corresponding [S iii] and [Si ii] components
generally agree within the velocity errors and there is no
systematic pattern in the velocity differences that would indicate
strong contamination by either the 33.04 m [Fe iii] line or the
35.35 m [Fe ii] line.
All of the final accepted fits are well described by either 1 or
2 line-of-sight Gaussian components. However, we know from
the optical images of Cas A that there is a great deal of structure
on small scales with typical knot sizes between 0. 2 and 0. 4
(Fesen et al. 2001). Our high-resolution infrared spectra taken
with Spitzers shorthigh and longhigh IRS modules show that
the spatial substructure exhibits velocity substructure as well.
Isensee et al. (2010) show that small regions of the Bright Ring
and Diffuse Interior Emission have many velocity components
with total Doppler ranges of as much as 2000 km s1 . At angular
resolutions of 4 10 and spectral resolutions from 2500 km s1
to 3000 km s1 we are insensitive to the small-scale substructure.
However, based on optical Doppler maps (Lawrence et al. 1995)
and our own Doppler map in Figure 3, major structures are

2042

DELANEY ET AL.

Vol. 725

Figure 5. High-resolution [Si ii] spectrum (green) from same region as the
right panel of Figure 4 showing two primary Doppler components with smaller
velocity structures extending to the center of the remnant. The low-resolution
spectrum is plotted in black for comparison and the low-resolution line centroids
are indicated by the vertical dashed lines. The vertical dotted line denotes the
rest wavelength of [Si ii].

Figure 4. Top: Gaussians plus continuum fit (solid line) to the [Ar ii] emission
in a region on the Bright Ring (see Figure 2) with two line-of-sight Doppler
components (dashed red and blue lines). Bottom: joint Gaussians plus continuum
fit to the [S iii] and [Si ii] emission in a region near the center of the remnant
(see Figure 2), each with two line-of-sight Doppler components. The velocities
of the blue (vb ) and red (vr ) components are indicated with 1 statistical errors.
The rest wavelengths of the [Ar ii], [S iii], and [Si ii] lines are indicated by the
vertical dotted lines. The wavelength range plotted in Figure 5 is indicated at
the bottom of the plot.

separated by many thousands of km s1 , so even at low spatial

and spectral resolution we are mapping the gross dynamic
behavior in Cas A. To illustrate this point, we show in Figure 5
the high-resolution (R = 600, velocity resolution = 500 km s1 )
spectrum of the [Si ii] emission displayed in the right panel of
Figure 4. The high-resolution spectrum is, indeed, dominated
by two major velocity components, but there is also weaker
emission extending into the center of Cas A. The vertical dashed
lines indicate the Gaussian centers of the low-resolution fits
and demonstrate the systematic uncertainties of averaging over
multiple velocity components. A more detailed comparison
between the low- and high-resolution spectra near the center
of Cas A is presented in Isensee et al. (2010).
3. CHANDRA DATA ANALYSIS
A spatially resolved spectral analysis was carried out using
the nine individual ACIS pointings from the million-second

observation of Cas A in 2004 (Hwang et al. 2004) and two

additional 50 ks ACIS observations from 2000 (Hwang et al.
2000) and 2002 (DeLaney & Rudnick 2003). The initial analysis
of the Chandra ACIS data is described fully in Davis et al.
(2005) and Stage et al. (2006). Briefly, standard processing and
filtering were applied to the 11 individual pointings and the
data were re-projected onto a common sky-coordinate plane,
maintaining separate files for each pointing. A grid was created
with 1 spacing, approximately 450 by 450 , centered on Cas
A. Regions in this grid were adaptively sized to contain at least
10,000 counts and ranged from 1 1 to 7 7 . The regions
were allowed to overlap so that fits to regions larger than 1 1
are not independent.
One spectral model for each sky region was jointly fitted to
the set of 11 spectra using the Interactive Spectral Interpretation
System (Houck & Denicola 2000). Each spectrum was associated with its own effective exposure, response functions, and the
background spectrum from the same pointing. The background
was not subtracted from the source spectrum, but was instead
added to the model. The background is dominated by photons
from Cas A in the wings of the point-spread function of the
telescope and in the CCD readout streak. The background for
every source spectrum was drawn from the same representative
sky region.
The model consisted of 15 Gaussian components for the Helike K lines of O, Ne, Mg, Si, S, Ar, Ca, Ti, and Fe, the He-like
K lines of Si and S, the H-like K line of Si, and L lines
for multiple Fe ions. The model included interstellar absorption
column density and used a bremsstrahlung continuum. We chose
to use Gaussian+continuum fits rather than non-equilibrium
ionization (NEI) models for several reasons. First, our primary
purpose was to determine the strength and position (shift) of
the line emission, not to model the emission history. Second,
the available NEI models, while allowing variable element
abundances, use a single redshift for the plasma and they do
not account for the dynamical evolution of the shocked gas.
While we could have followed the examples of Laming &
Hwang (2003) and Hwang & Laming (2003, 2009) by tracing

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

the shock evolution along with the NEI evolution, we would still
have had to model multiple plasmas along the line of sight to
decouple for instance Si-dominated ejecta from overlapping and
dynamically distinct Fe-dominated ejecta. Multiple-component
plasma models often have essentially unconstrained line fits
which can lead to artifacts that plague automated fitting routines
such as the linear structures found in the analysis of Yang
et al. (2008). Using Gaussians essentially unties the elements
to allow for multiple Doppler shifts and allows for modelindependent Doppler shift determinations. The joint best-fit
Gaussian component centers were recorded for each line in
each sky region and used to create FITS images of the fitted line
centers for each line.
Given that the Si He line is the brightest emission line in
the X-ray data, it is the natural choice for Doppler analysis.
However, we found significant variations between the velocities derived from the ACIS data and the velocities derived from
the 2001 HETG observation for the 17 regions measured by
Lazendic et al. (2006) at high spectral resolution. As shown in
Figure 6, there were large velocity variations (1500 km s1 )
between the data sets as well as both a systematic velocity scaling and offset. The ACIS data are plotted using a rest wavelength
of 6.648 corresponding to the Si xiii He resonance line that
should dominate the spectrum (note that the forbidden line is at
6.740 so that the expected weighted average rest wavelength
should be between the resonance and forbidden lines). In order
for the ACIS data to match the HETG data, a rest wavelength of
6.6169 is required and the ACIS velocities must be scaled by
1.67. The velocity discrepancies between the HETG and ACIS
data might possibly be the result of ACIS energy calibration
issues near the Si xiii line (G. Allen 2009, private communication), although a follow-up analysis is required to confirm this
hypothesis. We will revisit the ACIS Si data in Section 4.1 to
further demonstrate the unsuitability of these data for our study.
We did not pursue the analysis of the S, Ar, or Ca emission lines
from the Si group of elements because they would likely have
the same velocity issues. Since the X-ray Si-group emission
traces out the same Bright Ring structures as the infrared [Ar ii]
emission (Ennis et al. 2006) and the gross Doppler structure of
the X-ray Si emission is the same as for the optical emission
(Hwang et al. 2001; Willingale et al. 2002; Stage et al. 2004;
Davis et al. 2005), we feel that the Bright Ring is well sampled
without the ACIS Si (or Si-group) data.
We therefore chose to focus on the Fe-K line as a complement
to the infrared lines. It is distributed differently than the Sigroup and O/Ne emission (Hwang et al. 2004) and is a separate
dynamical component from the other X-ray ejecta (Willingale
et al. 2002, 2003). The Fe-K is preferable to the three Fe-L
lines from our Gaussian fitting because it is not contaminated
by nearby Ne or O lines nor any other lines in the Fe-L forest.
The energy calibration near the 6.6 keV Fe-K line should also be
quite accurate due to the 5.9 keV and 6.4 keV Mn-K emission
from the onboard energy calibration sources mounted in the
Science Instrument Module on Chandra.
The left panel of Figure 7 shows an image of the Fe-K
emission in Cas A made using spectral tomography as described
in DeLaney et al. (2004). The spectral tomography technique
is designed to separate overlapping spectral structures and
to spatially visualize the spectral components. The technique
involves taking differences between images from two different
energies with a scale factor chosen to accentuate features of
interest: M(f ) M(67) f M(46) , where f is the scale factor
and M(f ) is the residual image at that scale factor. The two input

2043

Figure 6. HETG-derived Si Doppler velocity compared to ACIS-derived Si

Doppler velocity for the sample of 17 ejecta knots measured by Lazendic et al.
(2006). The best-fit straight line to the data is shown and indicates both a
velocity scaling and offset between the two data sets. The rest wavelength used
to compute the ACIS velocities was 6.648 .

images to the spectral tomography analysis were a 46 keV

image, M(46) , to model the continuum emission (both thermal
and non-thermal), and a 67 keV image, M(67) , that contains
emission from the Fe-K line as well as continuum emission
either from thermal or non-thermal processes. We chose to
scale the 46 keV image by 1/6 prior to subtraction based
on the histogram of log(M(46) /M(67) ), which is Gaussianshaped with a tail at high ratios. The specific choice of scale
factor, f = 1/6, was made to isolate regions where Fe-K
emission dominates (low values of log(M(46) /M(67) )) from
other regions. This difference is shown in the left panel of
Figure 7 (M(1/6)) which has been smoothed to 3 resolution.
Regions that are bright (have positive residuals) in Figure 7
represent ejecta with bright Fe-K emission. Regions that are
dark (have negative residuals) in Figure 7 represent features
where the continuum is bright but there is no or weak Fe-K
emission. Note that the negative residuals occur in regions with
strong non-thermal emission as well as Si-rich (thermal) regions
that are relatively Fe-poor (e.g., DeLaney et al. 2004). Thus, the
tomography technique has separated Fe-K-rich regions from
Fe-K-poor regions. The Fe-K-rich emission is localized to three
regionsthe southeast, the north, and the west as noted by e.g.,
Hwang et al. (2004). The right panel of Figure 7 shows the
fitted line centers for the bright Fe-K regions (Stage et al. 2004).
The velocity pattern matches the Doppler image of Willingale
et al. (2002) almost exactly with a redshifted lobe of emission
to the north and a blueshifted lobe of emission to the southeast.
The approximate velocity range in the image is 4000 km s1
without accounting for ionization effects.
Hwang & Laming (2003) and Laming & Hwang (2003) show
that the ionization age of the X-ray plasma varies between
0.2 and 12 1011 s cm3 from place to place in Cas A. The
centroid of the Fe-K line shifts blueward at higher ionization
age because the line is actually a line complex dominated
by the He forbidden, intercombination, and resonance line
triplet as well as the Ly line. At a higher ionization age,

2044

DELANEY ET AL.

Vol. 725

Figure 7. Left: spectral tomography image showing the Fe-K emission as positive brightness. The spectra from regions A12 and A13 of Hwang & Laming (2003) are
shown in Figure 8. Right: effective Doppler map showing the fitted Gaussian line centers measured in Angstroms. The approximate velocity range is 4000 km s1 ,
however ionization effects are not accounted for. The contours represent brightness levels on the spectral tomography image.

the Ly line becomes brighter relative to the He triplet and

so the blended line shifts to higher energies simulating a
blueshift. Since we only fit a single Gaussian to the Fe-K line
complex, our centroids are affected by ionization as well as
Doppler shift. We demonstrate the importance of this ionization
effect in Figure 8 where we show spectra between 6 and 9 keV
(containing Fe-K and Ni-K lines) extracted from the millionsecond data set for regions A12 and A13 of Hwang & Laming
(2003). These two regions are identified in the left panel of
Figure 7 and have ionization ages of 81011 and 11011 s cm3 ,
respectively (Hwang & Laming 2003). A wavelength shift of
only 0.017 (0.0635 keV, the separation between the Fe-K
He forbidden and resonance lines) results in a velocity shift of
2746 km s1 . In Section 4.4, we correct for ionization effects in
the southeast Fe lobe by artificially collimating that structure.
We make no corrections for ionization anywhere else.
4. RECONSTRUCTING THE THREE-DIMENSIONAL
DISTRIBUTION
The Doppler velocities and sky positions of the infrared and
X-ray data can be used to reconstruct their three-dimensional
positions. In order to convert Doppler velocity to equivalent
line-of-sight spatial position we must assume that, prior to
encountering the reverse shock, the ejecta were freely expanding
from a single point in time, so that distance velocity. This
behavior is expected for a point explosion in space, however
if distance velocity, then this project is not feasible because
there would be no way to map velocity into distance. We must
also choose a geometry, and for this we adopt the simplest
physical picture, a spherical shell, consistent with the optical
(Reed et al. 1995) and infrared data on the Bright Ring. We
do not pursue more complex models involving ellipticity or
disjoint hemispheres because Reed et al. (1995), using better
spatial and spectral resolution, did not find compelling evidence
that those sorts of models better described their optical data. In
all cases, we measure projected radius on the sky with respect to
the optical ejecta expansion center derived by Thorstensen et al.
(2001). This location is 7 north of the central compact object
(CCO) and we note that it is not at the geometrical center of

Figure 8. X-ray spectra for regions A12 and A13 of Hwang & Laming (2003)
(marked in Figure 7) showing the Fe-K and Ni-K emission lines. The higher
ionization age of region A12 results in an apparent blueshift in the spectrum. The
rest wavelengths of the dominant Fe-K and Ni-K He forbidden (dashed line)
and resonance (dotted line) lines and Ly lines (dot-dashed line) are indicated.
At the bottom of the plot, we show that a shift of 0.017 corresponds to an
apparent Doppler shift of 2746 km s1 .

the Bright Ring. Reed et al. (1995) considered several different

centers in their analysis and found that it had minimal impact
on their results.
4.1. Converting Doppler Velocity to Line-of-sight Distance
In order to convert Doppler velocity to line-of-sight distance
independently from the optical determinations, we plot Doppler
velocity versus projected radius on the sky as shown for the
[Ar ii] (red) and [Si ii] (gray) data in Figure 9. The error boxes
at the bottom of the figure indicate the spatial resolution and the
mean Doppler velocity errors for the two populations. The [Ar ii]

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

2045
1

Figure 9. Doppler velocity vs. projected radius for [Ar ii] in red and [Si ii] in
gray. The mean velocity errors and spatial resolution are indicated by error boxes
at the bottom of the figure. The red semicircle represents the best-fit spherical
expansion model. The black semicircle represents the reverse shock and the
projected radius of the forward shock is also indicated. Note that the center of
the spherical expansion model is not at zero velocity.

data are concentrated onto a relatively thin, semi-circular shell

with about a 15% scatter in thickness. The Doppler velocities
range from 4000 to 6000 km s1 . The velocity distribution is
not centered at 0 km s1 , but is offset toward positive velocities.
The velocity range and offset and the detailed structure of the
[Ar ii] distribution are virtually the same as the optical data in
Figure 3 of Reed et al. (1995).
The [Ar ii] emitting regions can all be fit on the same spherical
shell, consistent with the scenario where ejecta become visible
after being heated by the reverse shock and later fade from
view with observed lifetimes 30 yr (van den Bergh & Kamper
1983). Since the shell is narrow, only 10002000 km s1 thick,
we can relate Doppler velocity to spatial distance by fitting a
spherical expansion model. In velocity versus radius space, the
model is a semicircle which we have chosen to parameterize in
terms of the center of the velocity distribution, vc , the minimum
velocity at which the semicircle crosses the velocity axis, vm ,
and a scale factor, S, that relates the velocity axis to the spatial
axis:
(rp /S)2 + (vD vc )2 = (vc vm )2 ,
where rp is the observed projected radius and vD is the
observed Doppler velocity. The scale factor, S, is used to
directly convert Doppler velocity into distance in or out of
the plane of the sky. Using a least-squares fit to the data, we
find vc = 859 100 km s1 , vm = 4077 200 km s1 ,
and S = 0. 022 0. 002 per km s1 . Our values agree with
the Reed et al. (1995) values of vcReed = 770 40 km s1 ,



vmReed = 4520 100 km s , and SReed = 0. 019 0. 001

per km s1 , although we do expect these values to evolve over
time as the remnant evolves and the reverse shock encounters
slower and slower ejecta. Our best-fit model is shown as a red
semicircle in Figure 9. We note that the adoption of a different
value of vc would not change the geometry of the remnant, but
would result in a translation of the entire remnant along the line
of sight.
Also shown in Figure 9 is a fiducial semi-circle representing
the location of the reverse shock in the velocity coordinates of
the [Ar ii] ejecta. The projected radius of the reverse shock is
taken from Gotthelf et al. (2001) and is plotted to be concentric to
the spherical model that describes the ejecta. The reverse shock
location was identified based on X-ray Si emission and indicates
the location where ejecta turn-on after being shock heated. A
spatial analysis of X-ray Si emission and infrared Si and S
emission determined that the infrared and X-ray ejecta turn
on at the same radius indicating that the timescales for shock
heating up to infrared or X-ray emitting temperatures are short
(Smith et al. 2009). One might expect to see a progression from
infrared/optical emission to X-ray emission as ejecta ionize
up to He/H-like states, and this may happen for individual
knots, but some ejecta knots ionize so quickly to X-ray emitting
temperatures that the inside edge of the Bright Ring as defined by
optical/infrared emission is the same as for the X-ray emission.
Although ejecta knots continue moving downstream from the
reverse shock, new ejecta continuously encounter the reverse
shock and turn on thus maintaining a demarcation line at the
reverse shock location (Morse et al. 2004; Patnaude & Fesen
2007).
We identify the forward shock by a small line segment at a
radius of 153 . Since the forward shock and the Bright Ring do
not share the same projected geometric center (Gotthelf et al.
2001), we make no assertion as to any offset between these
structures along the velocity axis. Note that the velocities at
which the reverse shock model crosses the velocity axis do
not represent the actual space velocity of that shock, which is
not constrained by the current analysis. Rather, it represents
the velocity that an undecelerated ejecta knot would have at
the spatial position of that shock at the epoch of observation,
assuming our distance-to-velocity scale factor.
Some of the [Si ii] data are found on the same shell as the
[Ar ii], but most of the [Si ii] is found interior to the fiducial
reverse shock. Since the [Si ii] do not form a well-defined shell in
Figure 9 we cannot independently fit for its scale factor between
Doppler velocity and spatial distance. Henceforth, we will treat
the [Si ii] as though it were in free expansion with the [Ar ii].
We justify our free expansion assumption and discuss the [Si ii]
distribution more fully in Section 4.7.
In order to plot Doppler velocity versus projected radius for
the X-ray Fe-K data, we must first convert the fitted line centers
to Doppler velocities. This is not just a simple mathematical calculation because the Fe-K line is composed of the unresolved
He forbidden, intercombination, and resonance triplet of lines,
the barely resolved Ly line, and, depending on the ionization
age, lines from lower ionization Fe-K species. Because we modeled the Fe-K line complex with a single Gaussian component,
we must determine what average rest wavelength to use for the
velocity calculations and we must account for ionization variations that will change the brightness ratios of these lines and
mimic changes in Doppler velocity.
One way to approach this problem is to assume that the
spherical expansion model derived from the infrared ejecta also

2046

DELANEY ET AL.

applies to the X-ray ejecta. However, we know from proper

motion measurements that the X-ray ejecta, on average, are
decelerated, with an expansion rate of 0.208% yr1 (DeLaney
& Rudnick 2003). We can use the ratio between the X-ray
expansion rate and the free expansion rate (0.304% yr1 ;
Thorstensen et al. 2001) to scale the spherical expansion model
parameters into the decelerated velocity space appropriate to
the X-ray ejecta. These scaled parameters are Sx = 0. 032
per km s1 , vcx = 589 km s1 , and vmx = 2806 km s1 .
Note that in the reference frame of the decelerated ejecta, the
model vmax and vmin simply scale from the free expansion values.
The center of the velocity distribution, vcx , reflects the midpoint
between vmax and vmin in the decelerated space and is also a
simple scaling of the model. Although the average expansion
rate of the X-ray ejecta is 0.208% yr1 , the X-ray ejecta have
experienced a range of decelerations with a 1 variation of
0.08% yr1 (DeLaney et al. 2004). The resulting effective
velocity error from using the wrong distance-to-velocity scale
factor is less than 1300 km s1 in most cases, which is < 20%
of the total spread in X-ray velocities. In the case of the FeK ejecta, the deceleration error is small (< 1/2) compared to
the ionization error, but in the case of the Si xiii ejecta, the
deceleration error is the dominant error.
The distance-to-velocity scaling for the X-ray ejecta is appropriate under the condition that the X-ray emitting material
was in free expansion until encountering the reverse shock and
the mass wall of swept-up ejecta immediately downstream. The
subsequent deceleration and heating of the ejecta lead to the
X-ray emission. The X-ray ejecta are then assumed to continue
expanding at a slower, but constant, velocity. In reality, the ejecta
should continuously decelerate as they move downstream from
the reverse shock. However, the expansion rate of the X-ray
ejecta is roughly constant with radius according to the proper
motion data of DeLaney et al. (2004). The errors introduced
by the constant velocity assumption are dwarfed by the inherent variation in deceleration experienced by the ejecta at the
reverse shock as discussed above. This is not unexpected based
on hydrodynamic models. For example, in the snapshot in time
shown in Figure 1 of Truelove & McKee (1999), the difference
in ejecta velocity between the reverse shock and contact discontinuity is only a few percent. This expected difference in velocity
is much smaller than the variation observed in the actual data
(DeLaney et al. 2004). Thus, the assumption of constant velocity motion after deceleration at the reverse shock is appropriate,
to first order, for the X-ray ejecta. The X-ray-emitting ejecta
on the Bright Ring are therefore cospatial with the optically
and infrared-emitting ejecta, but are decelerated with respect to
those ejecta and consequently require a different scaling factor
to convert Doppler velocity to distance.
Figure 10 shows the Doppler velocity versus projected radius
for the Fe-K data with the scaled model plotted in red. The FeK data are plotted using a rest wavelength of 1.8615 which
provides the best fit (by eye) to the spherical model. The FeK error box at the bottom of the figure is an indication of the
degree of uncertainty due to ionization differences. The fiducial
reverse shock is indicated by the black semicircle and has been
mapped into the X-ray ejecta velocity space the same way as for
the infrared data in Figure 9, but now the velocities represent
the speed of decelerated ejecta at the reverse shock location.
The distribution of the Fe-K emission in Figure 10 is very
interesting. The northern, redshifted emission appears to form
a partial spherical shell similar to the infrared [Ar ii] emission.
The thickness of this shell is about 20 30 , which is nearly

Vol. 725

Figure 10. Doppler velocity vs. projected radius for Fe-K in green and HETG
Si xiii in black. Different symbols are used to represent the north (open square),
west (filled triangle), and southeast (filled circle) Fe-K distributions. The mean
velocity errors and spatial resolution are indicated by error boxes at the bottom
of the figure. The red semicircle represents the best-fit spherical expansion
model from Figure 9, but scaled appropriately for the decelerated X-ray ejecta.
The black semicircle represents the reverse shock in this decelerated reference
frame. The forward shock projected radius is indicated at 153 . A few Doppler
fingers that are due to ionization effects are identified. These correspond to
fingers 13 in Figure 14.

as thin as the infrared [Ar ii] shell. The blueshifted Fe-K ejecta
in the southeast of the remnant are distributed quite differently.
Rather than a thin, shell-like structure, the blueshifted ejecta
extend as a column, approximately 70 long, and at an angle of
27 out of the plane of the sky.
We also show in Figure 10 the X-ray HETG Si xiii data from
Lazendic et al. (2006). With only 17 small regions, the Si xiii
data sparsely populate the Bright Ring shell. However, since the
velocities of the Si xiii data are determined through dispersed
HETG spectra, they are completely independent of the ACIS FeK data. Therefore, the HETG Si xiii data provide confirmation
that the rest wavelength used to calculate the Fe-K Doppler
velocities is correct.
In Figure 11, we show the Doppler velocity versus projected
radius for the full X-ray ACIS Si data and the X-ray HETG
Si xiii data from Lazendic et al. (2006). The ACIS data are
plotted using a rest wavelength of 6.6169 that was derived
from Figure 6 by forcing the best-fit straight line to cross the
diagonal line at 0 km s1 . In addition, the ACIS velocities have
been multiplied by 1.67 to scale them to the HETG velocities
based on Figure 6. Just like the ACIS Fe-K data, the ACIS
Si data show extended Doppler fingers both toward negative
velocities and toward positive velocities even beyond the top
of the plotted area. This is unexpected because the Si xiii He
line complex was fit with a separate Gaussian component from

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

2047

Figure 12. Three-dimensional projections showing the fiducial reverse shock

(sphere) and CCO (cross) as seen from Earth (left) and from a 90 rotation to
the east (right). The location where ejecta Doppler velocities = 0 is identified
in the right panel. Here and in subsequent three-dimensional projections, the
coordinate system is identified at upper right in each panel with red = east,
green = north, and blue = to Earth. The color coding for this and subsequent
three-dimensional projections are described in Table 1.

(J2000) = 58 48 42. 5 (Fesen et al. 2006), and zero velocity,

although its actual line-of-sight location is unknown, as we
discuss further in Section 5.3. The left panel shows the view
from Earth with the CCO identified by the cross and the right
panel shows the model with a 90 rotation. The reverse shock is
not centered at v = 0, but is shifted into the sky by 859 km s1
(in the reference frame of the freely expanding ejecta) which
corresponds to a displacement of 0.31 pc assuming a distance
of 3.4 kpc to Cas A.
Figure 11. Doppler velocity vs. projected radius for ACIS Si in orange and
HETG Si xiii in black. The mean velocity error and spatial resolution for the
HETG data are indicated by the error box at the bottom of the figure. The
red semicircle represents the best-fit spherical expansion model from Figure 9,
but scaled appropriately for the decelerated X-ray ejecta. The black semicircle
represents the reverse shock in this decelerated reference frame. The forward
shock projected radius is indicated at 153 .

the Si xiv Ly line and so ionization effects should not be an

issue as they are for the ACIS Fe-K data. The extensive amount
of Doppler fingers essentially destroys the fine-scale Doppler
structure of the Bright Ring. Due to the many issues with the
X-ray ACIS Si data, no more useful Doppler information about
the Si-group ejecta can be derived from the ACIS Si data set
than we already have from the infrared [Ar ii].
Using the velocity-to-arcsecond scale factors derived for the
freely expanding ejecta (infrared [Ar ii] and [Si ii]) and for the
decelerated ejecta (X-ray Fe-K and HETG Si xiii), we plot their
three-dimensional distributions as well as place the fiducial
reverse shock and a fiducial CCO into three dimensions with
the data. The data have been smoothed with a three-dimensional
Gaussian to glue neighboring voxels together such that the
symbol size is 15 for [Si ii], 10 for Si xiii, and 6 for
the other ejecta. We did not encode brightness information into
these reconstructions. We discuss the various three-dimensional
ejecta distributions and model components below.
4.2. Illuminating the Spherical Reverse Shock
It is apparent from Figures 9 and 10 that the hot ejecta
are mostly concentrated into a thin shell. The inside edge of
this hot shell defines the location where ejecta are being
heated by passage through the reverse shock. In Figure 12,
we show a three-dimensional model of the sphere representing
the location and shape of the reverse shock in Cas A. The CCO
has been placed in the model at (J2000) = 23h 23m 27.s 94,

4.3. Cas As Rings

To demonstrate how well the shocked ejecta define
the reverse shock surface, we show four projections of
the three-dimensional distributions of [Ar ii] and Si xiii in
Figure 13from Earth, from north, from a 60 rotation to the
east, and from a 120 rotation to the west. These same four
views will be used for all subsequent three-dimensional figures. The blue material in Figure 13 identifies regions where
the [Ne ii]/[Ar ii] ratio is high which will be discussed briefly
in Section 4.6.
The most striking aspect of the [Ar ii], [Ne ii], and Si xiii
emission is that they do not simply form a limb-brightened
spherethey are organized into a cellular structure that appears
as rings on the surface of a sphere. When seen face-on, they
define the Bright Ring of Cas A. In the three-dimensional
views, we can identify two complete rings to the north and
a complete ring to the southwest. Broken or partial rings are
found in other locations and we also identify a parentheses
structure to the southeast of the CCO that forms an incomplete
ring as seen in earlier optical work (Lawrence et al. 1995; Reed
et al. 1995). While the [Ar ii], [Ne ii], and Si xiii distributions
define a roughly spherical surface, there are two locations where
the ejecta stretch to slightly larger radiito the northeast and the
west. In Section 4.5, we show that these two rings are at the bases
of the Jet and the Counterjet. Presumably the reverse shock is
distorted in these two locations accounting for the conical shape
of the heated ejecta found there (Fesen et al. 2006).
4.4. The Fe-K Distribution
In Figure 14, we show four projections of the threedimensional Fe-K data plotted with the reverse shock and CCO.
In the upper right panel of Figure 14 we can best see the effects of ionization. Regions 15 all form Doppler fingers that
extend directly toward or away from Earth. These fingers are

2048

DELANEY ET AL.

Vol. 725

Figure 13. Three-dimensional projections of the infrared [Ar ii] emission (red), X-ray Si xiii emission (black), the fiducial reverse shock (sphere), and the CCO (cross).
Locations where the [Ne ii]/[Ar ii] ratio are high are indicated in blue. The lines in the top left and bottom left panels identify an apparent symmetry axis for the
Ne-rich regions. The inferred CCO proper motion direction is indicated in the top left panel. Major structures discussed in Section 4.3 are indicated. The individual
views are: top left, from Earth; top right, from the north; bottom left, 60 rotation to the east; bottom right, 120 rotation to the west. These same four projections will
be used for all subsequent three-dimensional figures.

thus likely not velocity structures, but rather regions where the
ionization state of the X-ray gas is higher (blueshifted) or lower
(redshifted) than the average ionization state of the X-ray Fe
ejecta as a whole. We corrected for these Doppler fingers in
the southeast Fe-K complex (fingers 14) by simply collimating the datai.e., we shifted the data composing the Doppler
fingers, and only the Doppler fingers, in velocity until they
were between the dotted lines in Figure 10. We did not correct
any other Doppler fingers in the Fe-K distribution (i.e., only
the southeast Fe-K was collimated). We show in Figure 15 the
three-dimensional Fe-K distribution after ionization correction.
As we noted in Section 3, the Fe-K emission is localized
to three locationsthe west, the north, and the southeast. In
the west and the north, the Fe-K forms a partial shell on the
spherical surface that defines the fiducial reverse shock, but in
the southeast, the Fe-K emission forms a jet-like structure that
extends outward from the reverse shock sphere. The three strong
Fe-K regions do not fit easily into a bipolar structurethe north
and west regions are 90 from each other and the southeast
extension points back to the center of Cas A and not to either
the north or west regions.
4.5. Ejecta Pistons and Piston Rings
Figure 16 shows the shock heated Si/Ar/Ne rings plotted
with the Fe-K emission. We can immediately see that the three
Fe-K regions are circled by rings of [Ar ii] ejecta. While the
[Ar ii] ring to the north is complete, to the west and southeast
there are only partial [Ar ii] rings. Contrary to reports by
Willingale et al. (2002, 2003) the Fe-K emission does not appear
behind the higher ejecta layers (Si/Ar) to the northrather the
Fe-K emission is ringed by the [Ar ii] emission. Similarly, the
southeast Fe-K extension, which has been interpreted as an

overturning of ejecta layers (Hughes et al. 2000), is not in front

of the [Ar ii] emission, but ringed by the [Ar ii]. Both the FeK and [Ar ii] emission in the southeast turn on at the reverse
shock, but the Fe-K-rich ejecta extend farther downstream than
the [Ar ii]-rich ejecta and thus the Fe-K-rich ejecta are at a larger
average radius than the adjacent [Ar ii]-rich ejecta.
The fact that the Fe-K emission correlates so well with the
rings in the [Ar ii] emission is a confirmation that the rest
wavelength used to calculate the Fe-K Doppler velocities is
correct. Recall that the rest wavelength of the Fe-K line was
determined by a fit of the Fe-K data to the spherical expansion
model in the Doppler velocity versus projected radius plot in
Figure 10. If a different rest wavelength had been used, the
entire Fe-K distribution would have been shifted into or out
of the plane of the sky thus losing alignment with all of the
rings. The alignment of the Fe-K with the [Ar ii] rings is in
no way related to the ionization correction applied to the FeK data. The ionization correction was simply a collimation of
the southeast Fe-K structure thus no ionization correction was
applied to the Fe-K emission to the north or west, so those
alignments with [Ar ii] rings are solely due to the choice of rest
wavelength used to calculate Doppler velocity. To the southeast,
collimation was applied to only the Doppler fingers in the four
locations along the 27 line in the top right panel of Figure 14.
The bulk of the southeast Fe-K data remained untouched. None
of the four locations where the Doppler fingers appear on the
southeast Fe-K structure are on the surface of the fiducial reverse
shocklocations 13 are well outside the fiducial reverse shock
and location 4 is inside. Thus, the collimation of the southeast
Fe-K has no effect on the location where the Fe-K crosses the
reverse shock or the alignment of the Fe-K with the [Ar ii] ring
on the surface of the fiducial reverse shock.

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

2049

Figure 14. Three-dimensional projections of the X-ray Fe-K emission (green) with the reverse shock (sphere) and CCO (cross). Regions 15 in the upper right panel
identify red- and blueshifted Doppler fingers that result from ionization effects. The Doppler fingers in the southeast Fe-K complex (14) were collimated to lie along
the 27 line.

Figure 15. Three-dimensional projections of the ionization corrected X-ray Fe-K emission with the reverse shock (sphere) and CCO (cross).

Cas As well-known Jet and Counterjet are also associated

with shock heated Si/Ar/Ne rings. The prominent Si-rich northeast Jet and the weaker southwest Counterjet are dominated by
a series of linear rays that are visible in optical (e.g., Fesen
& Gunderson 1996), X-ray (Hwang et al. 2004), and infrared
(Hines et al. 2004) images. Proper motions and Doppler measurements of outer optical knots show the high velocity of the

jets and their orientation close to the plane of the sky. (Kamper
& van den Bergh 1976; Fesen & Gunderson 1996; Fesen 2001;
Fesen et al. 2006). The number of detected outer optical ejecta
knots has grown considerably since the first detection of the jet
region by Minkowski (1959). The current count stands at 1825
knots of varying compositions of O, S, and N (Fesen et al. 2006;
Hammell & Fesen 2009). Of these, only a small subset (135)

2050

DELANEY ET AL.

Vol. 725

Figure 16. Three-dimensional projections of the infrared [Ar ii] (red), high infrared [Ne ii]/[Ar ii] ratio (blue), X-ray Si xiii (black), X-ray Fe-K (green), outer optical
ejecta (yellow), fiducial reverse shock (sphere), and CCO (cross). The fastest moving outer optical ejecta defining the northeast Jet and southwest Counterjet may have
moved as much as 12 in the intervening years, which is about twice the size of the plotting symbols.

have measured Doppler velocities (Fesen & Gunderson 1996;

Fesen 2001). In Figure 16, we plot the outer optical ejecta in yellow at their 1988 or 1996 locations, depending on the available
data, using the velocity-to-arcseconds scale factor S = 0. 022
per km s1 which is appropriate for freely expanding ejecta.
Strings of optical knots define the linear features in the Jet while
the Counterjet is not easily identified with this limited sample
of outer optical ejecta. At the bases of the Jet and Counterjet,
where they intersect the spherical reverse shock, we find broken
and distorted rings of [Ar ii] ejecta.
In Section 5.2, we suggest that the Jet, Counterjet, southeast
Fe-K extension, northern Fe-K emission and other less prominent features are all regions where the ejecta have emerged
from the explosion as pistons of faster than average ejecta,
and that the rings on the spherical surface represent the intersection points of these pistons with the reverse shock, similar
to the bow-shock structures described by Braun et al. (1987).
Figure 17 is a schematic of the geometry involved. It represents
a simple extension of our earlier schematic (Figure 10 in Ennis
et al. 2006), where we now focus on one region of significantly
higher velocitythe pistonand the ring-like structure at its
intersection with the reverse shock that would appear in three
dimensions. Emission from the piston itself fades after reverseshock passage, which can also leave empty rings on the spherical
reverse shock surface. The ejecta layering in Figure 17 may remain intact or the piston may break through the outer layers,
as suggested for the Jet and Counterjet (Fesen et al. 2006).
4.6. Bipolarities and Other Symmetries
While the Jet/Counterjet is the most obvious bipolar structure
in Cas A, the strong [Ne ii] emission seen in Figures 13 and 16
also appears to be another example of a bipolar structure, but in a
completely different direction. These appear quite prominently
as the Ne-crescents in the face-on views of Smith et al. (2009)

Figure 17. Cross-section of a simple ejecta piston model in which the layering of
the progenitor star has been preserved. At the intersection between the wedgeshaped piston and reverse shock, shock-heated Fe is enclosed by a ring of
shock-heated Ne/O and Ar/Si emission. Interior to the reverse shock, the Fe
sits within a hole in the unshocked Si distribution. The center of the piston is
moving faster than the edges of the piston accounting for the Fe layer having
crossed the reverse shock in the center while the Ne/O and Ar/Si layers are just
now reaching the reverse shock at the edges. All of the ejecta currently crossing
the reverse shock have velocities of 5000 km s1 .

(their Figure 8) and Ennis et al. (2006) (their Figure 9). These
regions are also illuminated by optical O emission (Fesen 2001)
and distinguished by their distinct lack of X-ray Si emission
(Ennis et al. 2006). Smith et al. (2009) showed that these two
strong Ne regions are approximately symmetric around the
kinematic center of the remnant (Thorstensen et al. 2001) and
the position of the CCO. The symmetry axis is quite close to
the direction of the natal kick inferred by Fesen et al. (2006).
In the bottom panels of Figure 13, we can now also see that
these Ne/O regions are symmetric around the CCO along the
line of sight. While this might appear to be tantalizing evidence
that the Ne-crescents contributed significantly to the natal kick
experienced by the CCO, Smith et al. (2009) determined that
there is not enough mass/energy in the Ne-crescents to affect the
CCO motion. In the context of the Figure 17 schematic, the Necrescents would reflect the passage of a slower-moving piston

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

2051

Figure 18. Three-dimensional projections of the infrared [Si ii] emission (gray), fiducial reverse shock (sphere), and CCO (cross). There are two populations of [Si ii]
ejectaa shocked population that resides on the Bright Ring and an unshocked population that is physically interior to the reverse shock.

through the reverse shock, where only the Ne/O layer has been
illuminated so far. While the Ne-crescents have not caused the
CCO motion, they may perhaps arise due to the same dynamical
asymmetries that led to the CCO motion.
All of the ejecta pistons discussed thus far, whether bipolar or
not, all lie in the same broad plane. Even the outer optical knot
distribution, which forms a giant ring at or beyond the forward
shock with a thickness that is about the same as the distance
between the front and back sides of the [Ar ii] emission, is
roughly oriented along the same plane as the ejecta pistons. In
the next section, we define this broad plane using the infrared
[Si ii] emission.
4.7. The Tilted Thick Disk
Some of the [Si ii] data in Figure 9 map onto the same shell
where the [Ar ii] is found, but most of the [Si ii] is found interior
to the [Ar ii] shell. There are two reasons why the [Si ii] data
might appear interior to the [Ar ii] data: (1) the [Si ii] emission
is spatially interior to the [Ar ii] emission or (2) the [Si ii] is
spatially coincident with the [Ar ii], but decelerated. We reject
the latter explanation because we find no evidence that the
interior [Si ii] is shocked. The freefree absorption seen in
the radio suggests that the interior [Si ii] is cool (1000 K;
Kassim et al. 1995). The 18.7 m/33.48 m [S iii] line ratio
suggests a low density (ne upper limit of 100 cm3 ; Smith
et al. 2009) in the interior and there is no associated X-ray
emission that traces out the same interior structure as the infrared
[Si ii]. These interior conditions are in contrast to the conditions
on the Bright Ring where the temperatures are much higher
(500010,000 K; Arendt et al. 1999), the densities are much
higher (ne 104 cm3 ; Smith et al. 2009), and there are
spatially coincident but decelerated X-ray ejecta (DeLaney et al.
2004). Furthermore, the assumption of free expansion for the
[Si ii] ejecta results in a smooth connection between the interior

ejecta and the Bright Ring features at the spherical reverse shock
shell. In order to achieve this same smooth connection with a
decelerated population of [Si ii], we would have had to adopt
a range of distance-to-velocity scale factors with some [Si ii]
even more decelerated than the X-ray ejecta. Therefore, we
believe that there are two populations of [Si ii] ejecta in Cas
Aa shocked population that resides on the Bright Ring and an
unshocked, photoionized (Hamilton & Fesen 1998) population
that is in free expansion and is physically interior to the reverse
shock. While we believe that the [Si ii] ejecta are indeed in free
expansion, the model reconstruction is highly dependent on this
assumption.
In Figure 18, we show four projections of the threedimensional [Si ii] distribution including the CCO and the reverse shock. The [Si ii] emission is concentrated onto two concave, wavy sheetsone front and one rear. The distribution of
the [Si ii] is not spherically symmetric nor is it centered on zero
velocity. Most of the [Si ii] maps interior to the fiducial reverse
shock with portions extending out onto the Bright Ring. For
comparison, the average three-dimensional radius of the [Ar ii]
ejecta is 110 while the average three-dimensional radius of
the [Si ii] ejecta is 91 . The two sheets are separated by a series of holes around the edges. The plane containing the holes
is not exactly in the plane of the sky, but is oriented with an
25 rotation about the NS axis and an 30 rotation about the
EW axis. Our reconstruction shows that the two [Si ii] sheets are
separated by a much lower density region, where only a little
emission is seen (Figure 18).
We therefore describe the [Si ii] distribution as a tilted thick
disk as shown in Figure 19 where the front and rear faces
of the thick disk are shown as thin gray disks. The area between
the faces contains relatively weaker emission from Si, S, and O
ejecta as demonstrated in Figure 5 and by Isensee et al. (2010).
Presumably there are unshocked Fe ejecta, but emission from
this species has not been detected.

2052

DELANEY ET AL.

Vol. 725

Figure 19. Three-dimensional projections showing the tilted thick disk model (gray and hashed) with the fiducial reverse shock (sphere) and the CCO (cross). The
hash marks between the faces indicate that the region is not empty but contains weak emission from unshocked Si, S, and O and probably undetected unshocked Fe.
Approximate orientations with respect to the northsouth and eastwest axes are indicated.

4.8. Three-dimensional Relationships among Infrared, X-ray,

and Optical Emission
In Figure 20, we show the full compilation of data (this
figure is also available as an mpeg animation and a threedimensional PDF in the online version of the journal). It is
immediately clear that the rings of [Ar ii] match the holes of
[Si ii] supporting the interpretation of the Bright Ring as the
intersection point between the spherical reverse shock and the
flattened ejecta distribution. [Si ii] emission is found interior to
the strong [Ne ii] regions and three of the [Si ii] holes are filled
with Fe-K emission. To the southeast, specifically, the hole in
the [Si ii] distribution indicates that there is no unshocked [Si ii]
immediately interior to the Fe-K extension. Had there been a
spatial inversion of ejecta layers to the southeast, one might
expect to see Si-rich ejecta physically interior to the Fe-rich
ejecta. Thus, the spatial relationship between the [Ar ii] rings,
the [Si ii] holes, and the Fe-K emission supports the piston
schematic outlined in Figure 17 in which the whole column
of ejecta is displaced outward along these directions. While
there is a [Si ii] hole at the base of the northeast Jet, there is no
Fe-K filling that hole. To the west, there is an Fe-K-filled [Si ii]
hole and an empty [Si ii] hole in the southwest. Without a better
mapping of the southwest Counterjet, we cannot tell at this time
what the relationship is between the Counterjet and these west/
southwest [Si ii] holes. Similarly, we cannot tell whether the
west Fe-K is related to the Counterjet, although the distortion of
the reverse shock to the west is a strong indicator of Counterjet
interaction.
The complete outer optical knot distribution shown in Fesen
et al. (2006) forms a giant ring at the location of the forward
shock and the limited Doppler velocity information indicates
that this ring is oriented in the same broad plane defined by the
[Si ii] emission. The outer optical ejecta show layering with

S-rich ejecta closer to the reverse shock and N-rich ejecta

at larger radii. Previous interpretations of the outer ejecta
distribution are that it is a limb-brightened shell (Lawrence
et al. 1995; Fesen 2001). However, given that the inner ejecta
distribution is a flattened disk, the outer S- and O-rich optical
ejecta distribution may simply be an extension of the inner Sigroup and O/Ne layers out to large radii. We would not expect
to see outer S- and O-rich optical ejecta projected across the
front and back of Cas A because those layers are still interior to
the reverse shock in those directions. However, we might expect
to see fast-moving N-rich ejecta projected across the disk of Cas
A. In Section 5.3, we will discuss the nature of the missing
ejecta toward the front and back of Cas A. There are gaps in the
outer optical knot distribution to the north and south as shown in
Fesen et al. (2006). The northern gap is in the same direction as
the northern Fe-K cap and [Si ii] hole while the southern gap is
in the same direction as a closed region in the [Si ii] emission
where no large holes exist. The nature of these relationships
cannot be determined with the limited sample of outer optical
knots displayed here.
5. DISCUSSION
The picture that emerges from the three-dimensional models
presented in Figures 1220 is of a flattened explosion where
the highest velocity ejecta were expelled in a broad plane in the
form of jets or pistons. As discussed below, the overall flattened
ejecta structure is intrinsic to the explosion, with only small
effects from later circumstellar medium (CSM) interactions. In
contrast to the flattened ejecta distribution, the Bright Ring,
which is formed by the intersection of the ejecta with the
reverse shock, maps onto a roughly spherical surface (Reed et al.
1995). Furthermore, the diffuse radio emission is well modeled
by a limb-brightened spherical shell (Gotthelf et al. 2001)

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

2053

Figure 20. Three-dimensional projections of the infrared [Ar ii], high [Ne ii]/[Ar ii] ratio, and [Si ii] emission, the X-ray Fe-K and Si xiii emission, and the outer
optical knots. The color coding is described in Table 1.
(An animation and a three-dimensional PDF of this figure are available in the online journal. Interactive PDF requires a current version of Adobe Acrobat Reader.)

and the forward shock appears nearly circular (Gotthelf et al.

2001).
In this discussion, we suggest a coherent dynamical picture of
the explosion, including the various observed symmetries and
asymmetries, and the contrasting structures seen in the shocked
versus un-shocked components. This picture should constrain
and guide the next generation of simulations, and perhaps even
provide insights into how a star, or at least Cas A actually
explodes.
5.1. The Flattened Explosion and Thick Disk
Since the unshocked ejecta have not interacted with the
reverse shock, they are in a pristine state and give direct clues
as to the actual explosion geometry. Previous indications of the
intrinsic nature of the asymmetries in Cas A come from the
first observations showing the northeast Jet (Minkowski 1959),
from later X-ray analyses showing the Fe emission distributed
into two lobes oriented nearly 90 to the jet axis (Willingale
et al. 2002), and finally from the presumed natal kick of the
CCO (Fesen et al. 2006). The observations presented here of
the pre-reverse shock thick disk show that Cas As explosion led
to a flattened ejecta distribution at least for the inner layers of
the star (the possible geometry of the outer (He/C) layers will
be discussed in Section 5.3). The origins of this flattening, and
resultant disk, are unclear. It could simply be the remaining
structure interior to the reverse shock from a highly oblate
explosion in which the faster moving jets and pistons in the
symmetry plane have already driven most material beyond
the reverse shock. Since the breaking of spherical symmetry
seems essential in order to produce viable supernovae (Burrows
et al. 2007), the oblate explosion geometry could result from
instabilities such as the Spherical Accretion Shock Instability
(SASI; Blondin et al. 2003), although this instability occurs
very early in the explosion and its geometric signature might

be lost at later times. Other instabilities also show oblate

geometry, such as the neutrino-driven core-collapse explosion
model presented in Burrows (2009) which was calculated using
a three-dimensional adaptive mesh refinement code.
Alternatively, if the explosion were jet-induced, the thick disk
could be a perpendicular toroidal structure, such as proposed by
Wheeler et al. (2008). Similar to earlier work by Burrows et al.
(2005) and Janka et al. (2005), they propose that the prominent
northeast and weaker southwest jets (Fesen & Gunderson 1996;
Hwang et al. 2004; Hines et al. 2004) are actually secondary to
the main energy outflow. Wheeler et al. (2008) suggest that
the primary energy-carrying jet is associated with the irondominated ejecta in the southeast, approaching us at an angle
of 40 50 . Although the observations here confirm the Fe
southeast jet structure, their model fails because it predicts
an expanding torus roughly perpendicular to the jet direction,
which would contain the secondary northeast and southwest
jets and other ejecta features. By contrast, our observations show
that all of these structures, are in a single toroidal-like structure
in the same broad plane, not perpendicular, to the primary
southeast Fe jet. At the same time, the Wheeler et al. (2008) work
suggests important possible instabilities in the explosion that
are worth further exploration. A similar geometry, consisting of
bipolar and equatorial mass ejection, is suggested by Smith &
Townsend (2007) for the Homunculus Nebula around Carinae,
and Smith et al. (2007b) for SN2006gy.11
In a more speculative vein, the thick disk may be responsible
for the appearance of the light from Cas As explosion. It was
underluminous as seen from Earth (van den Bergh & Dodd
1970; Chevalier 1976; Shklovsky 1979), which would occur if
the optical depth were high toward the surfaces of the disk (in
the direction of Earth), whereas light could escape more easily
in the jet/piston plane. This curious idea may be supported by
11

See illustration at http://chandra.harvard.edu/photo/2007/sn2006gy/.

2054

DELANEY ET AL.

the light echoes which appear to mostly lie in the plane of the
sky (Rest et al. 2008, 2010). In two locations, echoes are seen
out of the plane of the sky, but this apparent discrepancy may
again be related to the explosion structure. The position angles
of the out-of-plane echoes are along (1) the northeast Jet and
(2) the north Fe-illuminated piston. So the disk surfaces could
be swept-up, (optically) thick material, with lower densities in
its jet/piston containing plane. A calculation of the physical
conditions for this scenario would be of great interest, but is
beyond the scope of the current work.
5.2. Jets, Pistons, and Rings
From the first optical observations describing the Jet as a
flare (Minkowski 1959) to recent X-ray and infrared images
showing a fainter Counterjet to the southwest (Hwang et al.
2004; Hines et al. 2004), Cas As striking jets have been a topic
of much discussion. The jets are not the only ejecta pistons
in Cas A, however, they are simply the Si-group-illuminated
ones (Ar, Si, S), which made them easily visible in X-ray, optical, and infrared emission. There are Fe-illuminated pistons and
Ne/O-illuminated pistons as well. The Ne/O-illuminated pistons show a bipolarity similar to the jets, but the Fe-illuminated
pistons show no such symmetry. The differences in composition
of the various pistons cannot be due to temperature, ionization
state, density, etc., since these differences appear consistently in
the infrared ionic lines (Ennis et al. 2006), optical (e.g., Chevalier & Kirshner 1979) and X-ray emission lines (e.g., Hughes
et al. 2000), and in the dust composition (Rho et al. 2008). In
this paper, the description of the observed structures in terms of
pistons is a simple extension of the description in Ennis et al.
(2006), where the ejecta in different directions were inferred to
be moving at different velocities. All of the pistons currently
visible are in the same broad plane defined by the interior thick
disk.
In order to see pistons dominated by different elements, the
hydrostatic nucleosynthetic layers must remain somewhat intact, as illustrated in Figure 17. In addition, there must be a
significant radial velocity gradient across the layers during the
explosion, since otherwise they would arrive virtually simultaneously, and would never appear segregated, at the Bright
Ring. Remembering that all undecelerated ejecta currently encountering the Bright Ring must be moving at 5000 km s1 ,
this means that each piston has a different range of velocities.
For example, if in one direction, the velocities varied between
1000 km s1 (Fe) to 5000 km s1 (Ne), then the Ne layer would
be just encountering, and be visible at the Bright Ring, but Si
would not. If, in another direction, the velocities varied from
1000 km s1 (Fe) to 10,000 km s1 (Ne), then the Ne would be
beyond the Bright Ring, but Si-group emission would be visible.
Thus, we conclude that the Ne-illuminated pistons have slower
average velocities than the Fe-illuminated pistons, although the
currently visible portions are all around 5000 km s1 .
The pistons represent directions where relatively fastermoving ejecta were expelled. The energy in the pistons can be
estimated by the mass in the piston and the velocity of the ejecta.
The Fe-illuminated piston to the north, for example, contains
about 3.2 M
of visible material (ejecta+CSM) (Willingale
et al. 2003). Moving at a velocity of 3000 km s1 , the total
kinetic energy would be about 3 1050 erg. An estimate for the
northeast Jet based on hydrodynamic models places the total
energy there at about 1050 erg (Laming et al. 2006). Thus, the
pistons represent a significant fraction of the remnants energy
budget.

Vol. 725

Any piston pushing out through less dense and chemically

distinct layers above would be subject to a series of instabilities such as RayleighTaylor, RichtmeyerMeshkov, and
KelvinHelmholtz, which would naturally result in some overturning of the layers (e.g., Kifonidis et al. 2006). Certainly
small-scale mixing between layers has occurred, and perhaps
even some incomplete mixing on large scales, however the
instabilities did not lead to a homogenization of the ejecta
(Hughes et al. 2000). Furthermore, Fesen et al. (2006) show,
based on their analysis of outer optical ejecta, that the original
layering of the star is still intact in some regions. In addition,
the onion-skin nucleosynthetic layers must be remarkably preserved in order to reproduce the observations, as indicated in
Figure 17. The observed size of the rings, which subtend solid
angles up to a steradian, require large-scale, low-order mixing
modes. Therefore, a large-scale instability or plume mechanism
(e.g., the Ni-bubbles of Woosley 1988 as advocated by Hughes
et al. 2000 to explain the southeast Fe structures), that does
not result in whole-scale mixing or ejecta layer inversion, is required by the observational data. Although simulations are beginning to uncover both the larger and the finer scale instabilities
(Hammer et al. 2010), it is not yet clear under what conditions
the low-order mixing modes should dominate without producing
an overturning of the layers.
The pistons become visible upon their encounter with the reverse shock, where they are compressed, heated, and ionized.
The physical conditions in the interacting regions are very inhomogeneous, resulting in X-ray (107 K), optical (104 K), infrared
ionic line (103 K), and dust (102 K) emission. The appearance of
the pistonshock interaction region is illustrated in Figure 17,
which shows the formation of rings (e.g., of Ar) and the filling of the rings by material from deeper (e.g., Fe) layers. This
model is reminiscent of the geometry described by Braun et al.
(1987) to explain their radio bow shock structures and most
of their paraboloid extensions (see their Figure 2) correspond
to our ejecta pistons. The ejecta cool after the reverse shock
passage, so that the emitting material is mostly confined to a
shell above the reverse shock. The inside edge of the shell thus
indicates the location and shape of the reverse shock and the
thickness of the shell is related to the ejecta lifetime. Radiative
cooling times for the dense optical knots are quite short (days to
months), so their observed 30 yr lifetimes (van den Bergh &
Kamper 1983; van den bergh & Kamper 1985) are likely due to
continued heating during the passage of the shock through the
knot (Morse et al. 2004). At typical speeds of 5000 km s1 ,
this results in a thickness of 10 for the illuminated
shell.
X-ray emission will also fade after reverse shock passage,
but the dominant physical processes are not yet clear. The
Fe emission in the north, e.g., is <20 thick, translating to a
lifetime of <90 years at a typical space velocity of 3500 km s1
(DeLaney et al. 2004). However, it should be much thicker if
radiative cooling were dominant, since at the low densities of
the X-ray emitting material (100.5 102.5 cm3 ) these timescales
are of order 107 108 yr, even for high metallicities (Gnat &
Sternberg 2006). Since the fading of the X-ray ejecta is clearly
not from radiative cooling, this would suggest that the thin
X-ray shell must be thin because the knots are quickly disrupted,
but the models of Hwang & Laming (2003) suggest dynamical
survival times of 200300 yr. This discrepancy between the
relatively short apparent lifetimes of the X-ray knots and the
long lifetimes predicted by dynamical models remains to be
resolved.

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

While the ejecta layering appears to be retained in some

pistons, in others there is evidence for a breakthrough of the
outer layers. For instance, the rings at the base of the Jet and
Counterjet are noticeably distorted and lift off of the fiducial
reverse shock, outlining the surface of a cone-like structure. The
northeast cone is well defined while in the west only the top edge
of the cone is evident. These cones are completely consistent
with the scenario put forth by Fesen et al. (2006) where the
jets represent streams of high-speed ejecta that were expelled
up through the outer layers of the progenitor star, distorting and
perhaps obliterating the reverse shock in the process.
The projected forward shock does not show large asymmetries. If indeed the forward shock is spherical, the deformities
in the reverse shock, which developed as a reflection of the
forward shock, might have occurred after the reverse shock
formed. However, since the forward and reverse shocks are now
well separated and dynamically distinct, any deformations in
the early forward shock shape may have been smoothed out by
interaction with the CSM. In this case, the reverse shock might
reveal the original shape of the forward shock. Alternatively,
the forward shock may indeed still have large deformities. For
instance, Laming et al. (2006) speculate that the forward shock
may indeed extend around the northeast Jet but would be effectively invisible because the plasma flow is oblique with respect
to the shock resulting in less efficient diffusive shock acceleration. Given the size of Cas A, Laming et al. (2006) further
speculate that the forward shock at the Jet tip, where the plasma
flow would be perpendicular to the shock, may actually be outside the field of view of Chandras ACIS-S3 chip.
5.2.1. Southeast Fe Piston and CSM

The X-ray Fe emission in the southeast is of considerable

interest for models of supernova explosions and the subsequent
instabilities in the shocked ejecta. The southeast was thought to
be a region where the Si and Fe ejecta layers had overturned
(Hughes et al. 2000) because the Fe-rich ejecta were at a larger
radius than the Si-rich ejecta. The subsequent spectral analyses
of Laming & Hwang (2003) and Hwang & Laming (2003, 2009)
showed that the Fe-rich ejecta to the southeast are at a higher
ionization state on average than the Si-rich ejecta and thus the
Fe-rich ejecta crossed the reverse shock at an earlier time and are
indeed at a larger radius, on average, than the Si-rich ejecta. Our
three-dimensional reconstruction is completely consistent with
the geometry inferred by the earlier spatial and spectral analyses,
however our hypothesis is that, rather than a spatial inversion
of ejecta layers to the southeast, the whole ejecta column along
that direction has been displaced outward.
If there had been an inversion of the ejecta layers, we
would expect to see Si-rich ejecta, either shocked or unshocked,
immediately interior to the Fe-rich ejecta along that explosion
direction; this is not observed. The Fe-K structures all turn on at
the reverse shock, implying that there is a source of unshocked
Fe-rich ejecta supplying those regions. Unfortunately, these
unshocked Fe-rich ejecta are not observed here, or anywhere
in the interior of the remnant. Furthermore, the total Fe detected
in Cas A is only a few percent of what is predicted by
nucleosynthesis models (Hwang & Laming 2003), so it is likely
that there is a supply of unshocked Fe-rich ejecta interior to the
reverse shock. We note that, if the unshocked Fe were rather
low density, as observed for the unshocked ejecta in the Type Ia
remnant SN1006 (Hamilton et al. 1997), then it would be well
below our detection limit. A low-density Fe layer might also
explain why the Fe-rich knots are typically appear more diffuse

2055

Figure 21. Gray scale: 34.48 m [Si ii] emission smoothed to 20 resolution; red
contours: X-ray Fe-K emission; blue contours: optical H and [N ii] emission
from diffuse and clumpy CSM. The green lines denote the linear northeast jet
structures and the green circle identifies the location of the outermost optical
ejecta observed by Fesen et al. (2006). The edges of the LL1 slit mapping shown
in Figure 1 are indicated.

than the Si-rich knots (Hughes et al. 2000). Other scenarios

for why we have not detected Fe in the unshocked ejecta are
explored by Isensee et al. (2010).
There do appear to be N, O, and S optical knots projected in
front of the Fe-rich ejecta (Fesen et al. 2006), but in our threedimensional picture, the few outer optical knots in the southeast
with Doppler velocity measurements are off to the side rather
than directly in front of the Fe ejecta. A Doppler mapping of
the complete set of detected outer optical ejecta knots to the
southeast is needed to determine their relationship with the Ferich ejecta.
The southeast Fe (and the Jet/Counterjet) appear very longlived compared to the rest of the Bright Ring, and, in particular,
the other bright Fe regions. Assuming a static reverse shock,
the ejecta at the tip of the southeast column, with total space
velocities of 6000 km s1 (DeLaney et al. 2004), were shocked
at least 190 years ago, completely consistent with the long
lifetimes derived by Laming & Hwang (2003) for knots in this
region of Cas A. The proper motions indicate that the X-ray
Fe has been decelerated; had it remained in free expansion, the
southeast Fe jet would extend out as far as the northeast Jet.
The long lifetimes of the southeast Fe-rich ejecta are certainly
consistent with the radiative cooling timescale for an X-ray
plasma, however, it is not clear why the ejecta should live longer
in the southeast than elsewhere on the Bright Ring. We note that
there is an unusual structure to the CSM in this direction. The
ionization age of the Fe jet drops rapidly at large radius, where
it may be interacting with the forward shock and the CSM.
Figure 21 shows [Si ii] observed beyond the forward shock to
the east of Cas A. The gray-scale [Si ii] image has been smoothed
to 20 resolution. The outer [Si ii] emission extends as an 30
(0.5 pc) wide filament from the eastern edge of the remnant
out to a distance of 6 from the center of Cas A where it
meets a broken northsouth ridge or shell. The observed ridge
is 5 long, but its true extent is unknown because it extends
beyond our field of view. The three linear structures defining
the northeast jet and the outermost optical ejecta observed by
Fesen et al. (2006) are plotted in green in Figure 21. We also
observe 33.48 m [S iii] and 15.56 m [Ne iii] in these outer
structures. The outer emission correlates very well with an H

2056

DELANEY ET AL.

cloud, identified by the blue contours in Figure 21, as well as

very faint H filaments farther to the east of Cas A (Fesen
& Gunderson 1996; Fesen 2001). Doppler velocities for this
outer material are small, only a few 100 km s1 with errors
typically of the same magnitude, indicating that it is almost
certainly CSM.
5.3. The Offset Spherical Component
Despite the large-scale asymmetries of the flattened explosion, and the prominent jets, and pistons, the forward and reverse
shocks appear to be spherical. This issue was addressed many
years ago by Bisnovatyi-Kogan & Blinnikov (1982), who asked,
[Must] all young remnants whose shape is nearly spherical have
been engendered by a spherically symmetric explosion? By no
means! They found that propagating shocks tend to isotropize
with time, although the time to approach sphericity depends on
factors such as density gradients and possibly shells through
which the shocks propagate (e.g., Chevalier & Soker 1989). It
would be very useful to use the observed transition from the
interior flattened explosion to the spherical shock structures in
Cas A to calculate the physical structure of the pre-SN stellar
atmosphere and winds. From the standpoint of enabling stars to
explode, asymmetries such as we observe appear to be essential
in, e.g., restarting a stalled forward shock (e.g., Rampp & Janka
2000).
The relationship between the structures observed here and
the position of the CCO can also be helpful in constraining the
nature of the explosion. In the plane of the sky, the CCO is
inferred to be in motion based on its offset of 7 nearly due
south of the optical expansion center (Thorstensen et al. 2001).
The derived transverse velocity is 350 km s1 in the direction
of the southern gap in the outer optical ejecta (Fesen et al. 2006)
and is roughly aligned with the Ne-illuminated pistons, however
there is insufficient mass in these pistons to generate such a kick
(Smith et al. 2009). The projected motion of the CCO could be a
recoil from the Fe-illuminated ejecta piston to the north, which
has no counter piston, and this scenario should be explored
further. Note that as long as the optical ejecta are undecelerated,
this observed offset cannot be due to a density gradient in the
CSM; such a gradient can shift the apparent center of the shock
structures (Reed et al. 1995; Dohm-Palmer & Jones 1996), but
not the expansion center.
In our analysis, we have positioned the CCO at zero velocity
along the line of sight, although there is no information on
its actual velocity in that direction. If the CCO were moving
very rapidly away from us (1000 km s1 ), then it could be
centrally placed between the surfaces of the thick disk (see
Figure 9), making the flattened explosion frontback symmetric.
However, there is evidence that the observed velocity asymmetry
between the surfaces of the thick disk could well be intrinsic
to the explosion. In Figures 4 and 5, the blueshifted [Si ii]
emission is significantly brighter than the redshifted emission.
In all of the Gaussian fits to the [Si ii] emission toward the
center, the near side was found to be brighter than the far
side. In addition, Isensee et al. (2010) show that there are
major structural differences between the front and back interior
surfaces. These front and back differences in brightness and
structure could be due to different masses/densities/energies in
the two different directions.
The existence of flattened inner ejecta surrounded by a
spherical reverse shock raises another important issue related to
the inward motion (with respect to the outward moving ejecta)
of the reverse shock. The flattened ejecta model requires that

Vol. 725

the ejecta being encountered by the reverse shock on the front

and back sides of Cas A are from an outer layer of the star while
the inner layers of the star are encountering the reverse shock
around the edges. We now know from Cas As light echoes that
these layers were indeed intact at the time of explosion and
that the progenitor even had a thin H envelope (Krause et al.
2008). The diffuse radio emission is well modeled by a limbbrightened spherical shell indicating that the reverse shock is
indeed encountering ejecta and particles are being accelerated
at least to low energies on the near and far sides of Cas A
(Gotthelf et al. 2001). Furthermore, if the reverse shock were
encountering nothing on the near and far sides of Cas A,
it would propagate much faster those directions and lose its
spherical shapewhich is not observed. The question is where
is the thermal emission from these ejecta that the reverse shock
must be encountering?
The X-ray data would be ideal for testing the hypothesis of
whether the reverse shock was currently encountering C-rich
ejecta on the near and far sides of Cas A, as our model requires,
rather than ejecta from an inner layer of the star. Unfortunately,
there are several technical issues. First, the absorption toward
Cas A is quite high, NH 1022 cm2 such that the C lines are
absorbed away. Second, the grain scattering halo in the direction
of Cas A (Predehl & Schmitt 1995), in conjunction with the
CCD readout streak (Stage et al. 2006), essentially ensures
that every location in Cas A is contaminated by a scaled-down
version of the global spectrum. This means that the diffuse X-ray
emission near the center of Cas A appears relatively abundant
in Si because there is no appropriate background spectrum that
can be used. This background contamination is not a problem
for discrete features because nearby diffuse regions can be used
to model the background and effectively subtract it away. When
this is done, the only discrete features projected near the center
of Cas A are quasi-stationary flocculi (CSM clumps) and nonthermal filaments (e.g., DeLaney et al. 2004).
Why, then, do we not see discrete ejecta knots, in any
waveband, at the reverse shock location on the near and far
sides, especially given that there are fast-moving N-rich ejecta
seen at large radii in the plane of the sky? In standard supernova
remnant ejecta profile models, the He and C layers should have
much lower densities than the Ne/O and Si/S layers which are
visible in X-rays, infrared, and optical emission (see discussion
of density profiles in, e.g., Truelove & McKee 1999). Perhaps
the densities in the He and C layers are simply too low to
identify discrete knots. A low density would allow the reverse
shock to propagate quickly in the He and C layers, leading
to a flattened reverse shock, which is not observed. However,
initially steep ejecta profiles might be made much shallower or
flat by the passage of the first reverse shock generated while
the forward shock is still within the atmosphere of the star (see,
e.g., Figure 4 of Kifonidis et al. 2003). In such a case, the ejecta
density profile of the evolved supernova remnant would also
be shallower and the currently observed second reverse shock
would remain spherical. However, the absence of discrete ejecta
knots from the He and C layers then remains a puzzle.
6. SUMMARY
In summary, we find a spherical component to the explosion,
a tilted thick disk, and ejecta pistons in the plane of the thick
disk. The gross morphology of Cas A must be shaped by the
explosion rather than a CSM interaction. The Si- and Ne/Oilluminated pistons are bipolar and the Fe-illuminated pistons
are not. The southeast Fe jet does not represent an overturned

No. 2, 2010

THREE-DIMENSIONAL STRUCTURE OF CAS A

2057

Table 1
Color Code for the Three-dimensional Graphics in Figures 1216 and 1820
Component

Telescope/
Instrument

Color/
Symbol

Comments

[Ar ii]
High [Ne ii]/[Ar ii] ratio
[Si ii]
Fe-K
Si xiii
Outer optical knots

Spitzer/IRS
Spitzer/IRS
Spitzer/IRS
Chandra/ACIS
Chandra/HETG
Ground based

Red
Blue
Gray
Green
Black
Yellow

Fiducial CCO

...

Cross

Fiducial reverse shock

...

Peach

...
...
...
...
Lazendic et al. (2006)
Fesen & Gunderson (1996)
Fesen (2001)
Placed at v = 0 km s1
pink in the three-dimensional PDF
...

region; it sits in a hole in the Si emission with no unshocked

Si emission immediately interior to it. The southeast Fe ejecta
knots are unusually long-lived compared to the Fe-rich ejecta to
the north and west, perhaps the result of an interaction with the
CSM structures in this direction.
In Isensee et al. (2010), we explore the detailed structure
of the thick disk, and the relationship between the Si and O
layers before they have been influenced by the reverse shock.
Further optical mapping of the outer knots (Milisavljevic &
Fesen 2010) will be useful in seeing if they match up with
the Fe pistons. The lack of visible Fe in the interior of Cas
A remains a puzzle, which is important both for the dynamics
and our understanding of nucleosynthesis. Long wavelength
infrared observations (40200 m) with Herschel and Sofia
are needed to search for unshocked Fe emission ([Fe ii], [Fe iii])
as well as emission from the presumed outer C and N layers
([C ii], [N ii], [N iii]) that should be interacting with the forward
shock on the near and far sides of the remnant. Finally, whether
our piston hypothesis is true or false, our unprecedented data
set and the three-dimensional geometry presented here will
invite theoretical work for a deeper understanding of supernova
explosion mechanisms and nucleosynthesis.
This work is based on observations made with the Spitzer
Space Telescope, which is operated by the Jet Propulsion
Laboratory, California Institute of Technology under NASA
contract 1407. Partial support for this work was provided
by NASA/JPL through award 1264030 to the University of
Minnesota. This work is also based on observations made
with the Chandra X-ray Observatory. Partial support for this
work was provided by NASA through Chandra Award AR56008X to the University of Minnesota. During the course of
this work, T.D. received partial support from NASA through
SAO grant GO3-4063A while a post-doc at Harvard. At MIT,
T.D. received support from NASA through SAO contract SV373016 to MIT for support of the Chandra X-ray Center and
Science Instruments, which is operated by SAO for and on
behalf of NASA under contract NAS8-03060. While at WV
Wesleyan College, T.D. received partial support through a
Faculty Research Enhancement Grant from the NASA-WV
Space Grant Consortium.
The three-dimensional graphics were made using the 3D
Slicer program which is a medical imaging tool being adapted
for astronomical use by the Astronomical Medicine Project
which is a part of Harvards Initiative in Innovative Computing.12 T.D. thanks Megan Watzke for connecting her with the
12

http://am.iic.harvard.edu

Astronomical Medicine Project and Michelle Borkin for providing valuable assistance with 3D Slicer. T.D. also thanks Mike
Noble for help with the three-dimensional PDF graphics and
Dan Dewey for valuable conversations. T.D., L.R., and K.I.
thank Alex Heger for guidance and valuable conversations concerning supernova models. Finally, we thank the referee for a
thorough review of the manuscript which has resulted in significant improvements in the presentation of our results.
APPENDIX
THREE-DIMENSIONAL PDF GRAPHICS
Figure 20 is available in the online version of the journal
as a three-dimensional PDF file. The PDF file contains the
full three-dimensional graphics information and allows the user
to directly manipulate the graphics by rotating the figure and
R
R
zooming in and out. The free Adobe
Reader
program version
8 or newer is required to view the three-dimensional PDF file.
By clicking on the graphic, the three-dimensional aspect is
enabled and a model tree opens allowing the user to turn on
and off individual elements of the figure. Both the original and
ionization corrected Fe-K emission are provided in the threedimensional PDF version of Figure 20 so that the readers can
judge for themselves what effect collimation has had on the
southeast Fe-K structure. The color coding for the individual
components of the model is described in Table 1. Coordinate
axes are provided to help keep track of orientation. The red axis
points east, the green axis points north, and the blue axis points
to Earth.
REFERENCES
Akiyama, S., Wheeler, J. C., Meier, D., & Lichtenstadt, I. 2003, ApJ, 584, 954
Arendt, R. G., Dwek, E., & Moseley, S. H. 1999, ApJ, 521, 234
Bisnovatyi-Kogan, G. S., & Blinnikov, S. I. 1982, SvA, 26, 530
Blondin, J. M., Mexxacappa, A., & DeMarino, C. 2003, ApJ, 584, 971
Braun, R., Gull, S. F., & Perley, R. A. 1987, Nature, 327, 395
Burrows, A. 2009, KITP Conference: Stellar Death and Supernovae (2009
August 1721), http://online.kitp.ucsb.edu/online/sdeath_c09/burrows/
Burrows, A., Dessart, L., Ott, C. D., & Livne, E. 2007, Phys. Rep., 442, 23
Burrows, A., Walder, R., Ott, C., & Livne, E. 2005, in ASP Conf. Ser. 332, The
Fate of the Most Massive Stars, ed. R. M. Humphreys & K. Z. Stanek (San
Francisco, CA: ASP), 350
Chevalier, R. 1976, ApJ, 208, 826
Chevalier, R., & Kirshner, R. 1979, ApJ, 233, 154
Chevalier, R., & Soker, N. 1989, ApJ, 341, 867
Davis, J. E., Houck, J. C., Allen, G. E., & Stage, M. D. 2005, in ASP Conf.
Ser. 347, Astronomical Data Analysis Software and Systems XIV, ed. P.
Shopbell, M. Britton, & R. Ebert (San Francisco, CA: ASP), 444
DeLaney, T., & Rudnick, L. 2003, ApJ, 589, 818

2058

DELANEY ET AL.

DeLaney, T., Rudnick, L., Fesen, R. A., Jones, T. W., Petre, R., & Morse, J. A.
2004, ApJ, 613, 343
Dohm-Palmer, R., & Jones, T. W. 1996, ApJ, 471, 279
Ennis, J. A., Rudnick, L., Reach, W. T., Smith, J. D., Rho, J., DeLaney, T.,
Gomez, H., & Kozasa, T. 2006, ApJ, 652, 376
Fesen, R. A. 2001, ApJS, 133, 161
Fesen, R. A., & Gunderson, K. S. 1996, ApJ, 470, 967
Fesen, R. A., Morse, J. A., Chevalier, R. A., Borkowski, K. J., Gerardy, C. L.,
Lawrence, S. S., & van den Bergh, S. 2001, AJ, 122, 2644
Fesen, R. A., Pavlov, G. G., & Sanwal, D. 2006, ApJ, 636, 848
Fesen, R. A., et al. 2006, ApJ, 645, 283
Gerardy, C. L., & Fesen, R. A. 2001, AJ, 121, 2781
Gnat, O., & Sternberg, A. 2007, ApJ, 168, 213
Gotthelf, E. V., Koralesky, B., Rudnick, L., Jones, T. W., Hwang, U., & Petre,
R. 2001, ApJ, 552, L39
Hamilton, A. J. S., & Fesen, R. A. 1998, ApJ, 327, 178
Hamilton, A. J. S., Fesen, R. A., Wu, C.-C., Crenshaw, D. M., & Sarazin, C. L.
1997, ApJ, 481, 838
Hammell, M. C., & Fesen, R. A. 2008, ApJS, 179, 195
Hammer, N. J., Janka, H.-Th., & Mueller, E. 2010, ApJ, 714, 1371
Hines, D. C., et al. 2004, ApJS, 154, 290
Holt, S. S., Gotthelf, E. V., Tsunemi, H., & Negoro, H. 1994, PASJ, 46,
L151
Houck, J. C., & Denicola, L. A. 2000, in ASP Conf. Ser. 216, Astronomical
Data Analysis Software and Systems IX, ed. N. Manset, C. Veillet, & D.
Crabtree (San Francisco, CA: ASP), 591
Hughes, J. P., Rakowski, C. E., Burrows, D. N., & Slane, P. O. 2000, ApJ, 528,
L109
Hwang, U., Holt, S., & Petre, R. 2000, ApJ, 537, L119
Hwang, U., & Laming, J. M. 2003, ApJ, 597, 362
Hwang, U., & Laming, J. M. 2009, ApJ, 703, 883
Hwang, U., Szymkowiak, A. E., Petre, R., & Holt, S. S. 2001, ApJ, 560,
L175
Hwang, U., et al. 2004, ApJ, 615, L117
Isensee, K., et al. 2010, ApJ, 725, 2059
Janka, H.-Th., Scheck, L., Kifonidis, K., Muller, E., & Plewa, T. 2005, in ASP
Conf. Ser. 332, The Fate of the Most Massive Stars, ed. R. M. Humphreys &
K. S. Stanek (San Francisco, CA: ASP), 363
Joggerst, C. C., Woosley, S. E., & Heger, A. 2009, ApJ, 693, 1780
Kamper, K., & van den Bergh, S. 1976, ApJS, 32, 351
Kassim, N. E., Perley, R. A., Dwarakanath, K. S., & Erickson, W. C. 1995, ApJ,
455, L59
Kifonidis, K., Plewa, T., Janka, H.-Th., & Muller, E. 2003, A&A, 408,
621
Kifonidis, K., Plewa, T., Scheck, L., Janka, H.-Th., & Muller, E. 2006, A&A,
453, 661
Krause, O., Birkmann, S. M., Usuda, T., Hattori, T., Goto, M., Rieke, G. H., &
Misselt, K. A. 2008, Science, 320, 1195
Laming, J. M., & Hwang, U. 2003, ApJ, 597, 347
Laming, J. M., Hwang, U., Radics, B., Lekli, G., & Takacs, E. 2006, ApJ, 644,
260

Vol. 725

Lawrence, S. S., MacAlpine, G. M., Uomoto, A., Woodgate, B. E., Brown, L.

W., Oliverson, R. J., Lowenthal, J. D., & Liu, C. 1995, AJ, 109, 2635
Lazendic, J. S., Dewey, D., Schultz, N. S., & Canizares, C. R. 2006, ApJ, 651,
250
Markert, T. H., Canizares, C. R., Clark, G. W., & Winkler, P. F. 1983, ApJ, 268,
134
Mazzali, P., et al. 2005, Science, 308, 1284
Milisavljevic, D., & Fesen, R. 2010, BAAS, 42, 469
Minkowski, R. 1959, in IAU Symp. 9, Paris Symp. on Radio Astronomy, ed. R.
N. Bracewell (Stanford, CA: Stanford Univ. Press), 315
Moiseenko, S. G., & Bisnovatyi-Kogan, G. S. 2007, Ap&SS, 311, 191
Morse, J. A., Fesen, R. A., Chevalier, R. A., Borkowski, K. J., Gerardy, C. L.,
Lawrence, S. S., & van den Bergh, S. 2004, ApJ, 614, 727
Patnaude, D., & Fesen, R. 2007, AJ, 133, 147
Predehl, P., & Schmitt, J. H. M. M. 1995, A&A, 293, 889
Rampp, M., & Janka, H.-T. 2000, ApJ, 539, L33
Reed, J. E., Hester, J. J., Fabian, A. C., & Winkler, P. F. 1995, ApJ, 440, 706
Rest, A., et al. 2008, ApJ, 681, L81
Rest, A., et al. 2010, ApJ, submitted (arXiv:1003.5660)
Rho, J., Reynolds, S. P., Reach, W. T., Jarrett, T. H., Allen, G. E., & Wilson, J.
C. 2003, ApJ, 592, 299
Rho, J., et al. 2008, ApJ, 673, 271
Scheck, L., Plewa, T., Janka, H.-Th., Kifonidis, K., & Muller, E.
2004, Phys. Rev. Lett., 92, 011103
Schure, K. M., Vink, J., Garcia-Segura, G., & Achterberg, A. 2008, ApJ, 686,
399
Shklovsky, I. S. 1979, Nature, 279, 703
Smith, J. D., et al. 2007, PASP, 119, 1133
Smith, J. D., et al. 2009, ApJ, 693, 713
Smith, N., & Townsend, R. 2007, ApJ, 666, 967
Smith, N., et al. 2007, ApJ, 666, 1116
Stage, M. D., Allen, G. E., Houck, J. C., & Davis, J. E. 2004, BAAS, 36, 948
Stage, M. D., Allen, G. E., Houck, J. C., & Davis, J. E. 2006, Nat. Phys., 2, 614
Tanaka, M., Kawabata, K., Maeda, K., Hattori, T., & Nomoto, K. 2008, ApJ,
689, 1191
Thorstensen, J. R., Fesen, R. A., & van den Bergh, S. 2001, AJ, 122, 297
Truelove, J. K., & McKee, C. F. 1999, ApJS, 120, 299
van den Bergh, S., & Dodd, W. W. 1970, ApJ, 162, 485
van den Bergh, S., & Kamper, K. 1983, ApJ, 268, 129
van den Bergh, S., & Kamper, K. 1985, ApJ, 293, 537
Wang, L., Baade, D., & Patat, F. 2007, Science, 315, 212
Wang, L., Howell, D., Hoflich, P., & Wheeler, J. C. 2001, ApJ, 550, 1030
Wang, L., et al. 2002, ApJ, 579, 671
Wheeler, J. C., Maund, J. R., & Couch, S. M. 2008, ApJ, 677, 1091
Willingale, R., Bleeker, J. A. M., van der Heyden, K. J., & Kaastra, J. S.
2003, A&A, 398, 1021
Willingale, R., Bleeker, J. A. M., van der Heyden, K. J., Kaastra, J. S., & Vink,
J. 2002, A&A, 381, 1039
Woosley, S. E. 1988, ApJ, 330, 218
Yang, X.-J., Lu, F.-J., & Chen, L. 2008, Chin. J. Astron. Astrophys., 8, 439
Zhang, W., Woosley, S., & Heger, A. 2004, ApJ, 608, 365