Content uploaded by Toshiya Ueta
Author content
All content in this area was uploaded by Toshiya Ueta
Content may be subject to copyright.
arXiv:astro-ph/0411556v1 19 Nov 2004
TO BE P UBLI SHED IN THE ASTRO NOMI CAL JOU RNAL
Preprint typeset using L
A
TEX style emulateapj v. 6/22/04
HST/NICMOS IMAGING POLARIMETRY OF PROTO-PLANETARY NEBULAE:
PROBING OF THE DUST SHELL STRUCTURE VIA POLARIZED LIGHT
TOSHIYA UETA1, 2,3 , KOJI MURAKAWA4,5 , MARGARET MEIXNER6
(Received Aug 24, 2004; Revised Oct 13, Nov 12, 2004; Accepted Nov 18, 2004)
To be published in the Astronomical Journal
ABSTRACT
Using NICMOS on
HST
, we have performed imaging polarimetry of proto-planetary nebulae. Our objective
is to study the structure of
optically thin
circumstellar shells of post-asymptotic giant branch stars by sepa-
rating dust-scattered, linearly polarized star light from unpolarized direct star light. This unique technique
allows us to probe faint reflection nebulae around the bright central star, which can be buried under the point-
spread-function of the central star in conventional imaging. Our observations and archival search have yielded
polarimetric images for five sources:
IRAS
07134+1005 (HD 56126),
IRAS
06530−0213,
IRAS
04296+3429,
IRAS
(Z)02229+6208, and
IRAS
16594−4656. These images have revealed the circumstellar dust distribution
in an unprecedented detail via polarized intensity maps, providing a basis to understand the 3-D structure of
these dust shells. We have observationally confirmed the presence of the inner cavity caused by the cessation of
AGB mass loss and the internal shell structures which is strongly tied to the progenitor star’s mass loss history
on the AGB. We have also found that equatorial enhancement in these circumstellar shells comes with various
degrees of contrast, suggesting a range of optical depths in these optically thin shells. Our data support the
interpretation that the dichotomy of PPN morphologies is due primarily to differences in optical depth and sec-
ondary to the inclination effect. The polarization maps reveal a range of inclination angles for these optically
thin reflection nebulae, dispelling the notion that elliptical nebulae are pole-on bipolar nebulae.
Subject headings: circumstellar matter — stars: AGB and post-AGB— stars: mass loss — planetary nebulae:
general — reflection nebulae
1. INTRODUCTION
The proto-planetary nebula (PPN) phase is a relatively short
(∼103years) stage of stellar evolution for low to intermediate
initial mass (∼0.8−8 M⊙) stars between the asymptotic giant
branch (AGB) and planetary nebula (PN) phases (e.g., Kwok
1993; Van Winckel 2003). During the PPN phase the post-
AGB central star increases its surface temperature from a few
to a few tens of 103K, while the circumstellar dust shell -
created by the AGB mass loss and physically detached from
the central star at the end of the AGB phase - simply coasts
away from the central star. Therefore, PPNs are important
stellar objects in which to investigate the nature of dusty mass
loss during the AGB phase, because the most pristine history
of AGB mass loss is imprinted and preserved in their density
distribution.
The AGB mass loss history can provide crucial clues for
the shell structure formation. While the circumstellar shells of
AGB stars initially assume spherically symmetric shape when
they are formed, they seem to develop largely axisymmetric
structure by the time the AGB mass loss is terminated (e.g.,
Balick & Frank 2002 for a review). Meanwhile, the AGB
mass loss history can also yield information about the internal
evolution of the central star. AGB mass loss may show tempo-
ral variations due to the alternative burning of hydrogen and
helium in two distinct layers via the mechanism called ther-
1Royal Observatory of Belgium, Ringlaan, 3, B-1180, Brussels, Belgium
2Current Address: NASA Ames Research Center/SOFIA, Mail Stop 211-
3, Moffett Field, CA 94035, USA; tueta@mail.sofia.usra.edu
3NRC Research Associate
4Subaru Telescope, National Astronomical Observatory of Japan, 650
North A’ohoku Place Hilo, HI 96720, USA; murakawa@subaru.naoj.org
5Current Address: ASTRON, P.O. Box 2, 7990 AA, Dwingeloo The
Netherlands
6Space Telescope Science Institute, 3700 San Martin Drive, Baltimore,
MD 21218, USA; meixner@stsci.edu
mal pulsation (Iben 1981), while a sudden enhancement of
mass loss near the end of the AGB phase - the so-called su-
perwind (Renzini 1981; Iben & Renzini 1983) - may remove
almost the entire surface layer from the star terminating the
AGB evolution. Thus, the AGB mass loss history is strongly
linked to both the internal and external evolution of the cen-
tral star, and the effects of AGB mass loss can manifest them-
selves in the density distribution in the PPN dust shells.
The density distribution in PPNs can be observationally
studied in two ways: either directly via thermal dust emission
arising from the shells or indirectly via scattered light through
the dusty shells. PPN imaging surveys were conducted us-
ing both methods, and the combined results suggested that
PPNs were intrinsically axisymmetric due to equatorially-
enhanced mass loss near the end of the AGB phase, which
probably coincides with the superwind phase (Meixner et al.
1999; Ueta, Meixner, & Bobrowsky 2000). These surveys
also found a morphological dichotomy among PPNs, show-
ing one-to-one correspondence between the mid-IR and op-
tical morphologies: an optical bipolar reflection nebula is al-
ways found with a mid-IR emission nebula of a single core
surrounded by an elliptical low-emission halo (DUPLEX-
core/elliptical PPNs), while an optical elliptical nebula is usu-
ally associated with a mid-IR emission nebula harboring two
emission peaks as evidence for limb-brightened edge-on dust
torus (SOLE-toroidal PPNs). Subsequent radiative transfer
calculations showed that the optical depth of the shell would
play a more important role in determining the morphology
of the shell than the inclination effect (Meixner et al. 2002;
Ueta & Meixner 2003).
In both morphological cases, the source of axisymmet-
ric structure is an equatorial density enhancement present
in the innermost regions of the shell. Hence, the AGB
mass loss history represented by this part of the shell holds
2 Ueta, Murakawa, & Meixner
the critical piece of information to enhance our understand-
ing of mass loss processes. Taking advantage of their op-
tically thin nature, SOLE-toroidal PPNs have been studied
by high-resolution imaging of thermal dust emission at mid-
IR in order to reveal the density distribution in these shells
(Dayal et al. 1998; Ueta et al. 2001; Kwok, Volk, & Hrivnak
2002; Gledhill & Yates 2003). However, such studies have
been difficult to perform at sufficient spatial resolution be-
cause of the diffraction-limited nature of the mid-IR imaging
and the intrinsically compact nature of PPNs.
In this paper, we use near-IR imaging polarimetry
to investigate the structure of five SOLE-toroidal PPN:
IRAS
07134+1005 (HD 56126),
IRAS
06530−0213,
IRAS
04296+3429,
IRAS
(Z)02229+6208, and
IRAS
16594−4656.
Below, we describe the technique of near-IR imaging po-
larimetry and motivate our use of it in this study (§2) and
outline our observational procedure and data processing steps
(§3). In §4, we present the results, and discuss the individual
sources in §5. We then discuss some of the implications of
the results (§6) and summarize the conclusions (§7).
2. IMAGING POLARIMETRY OF PPNS
Gledhill et al. (2001) used near-IR imaging polarimetry as
an alternative technique to directly capture the structure of the
PPN dust shells. The imaging polarimetry can separate dust-
scattered light (as the linearly polarized component) from di-
rect star light (as the unpolarized component). Hence, one
can easily detect faint, dust-scattered light from PPNs that is
otherwise buried beneath the dominant point-spread-function
(PSF) of the central star: the PPN structure can thus be re-
vealed as in the mid-IR imaging, but at one order better res-
olution in the near-IR. For example, the shell structure of
HD 235858 (
IRAS
22272+5435), which was not fully re-
solved at mid-IR until observed with the 10 m Keck telescope
(Ueta et al. 2001), was resolved by near-IR imaging polarime-
try with the 4 m UKIRT (Gledhill et al. 2001). Therefore,
we performed imaging polarimetry with the Hubble Space
Telescope (
HST
) to investigate the structure of very compact
PPNs by exploiting the unique ability of this novel observing
method and the high resolution capabilities of the telescope.
Polarimetric characteristics of astronomical objects can be
obtained by means of the Stokes parameters, (I,Q,U),
which are computed from the measurements of the beam
intensity passing through linear polarizing elements. The
near-IR camera and multiple object spectrometer (NICMOS;
Malhotra et al. 2002) on
HST
is equipped with three polariz-
ers, and the Stokes parameters can be derived by a matrix con-
version method that is elucidated by Sparks & Axon (1999)
and Hines, Schmidt, & Schneider (2000). Using the Stokes I
(the total intensity) along with the Stokes Qand U, we can
then express the intensity of linearly polarized light, Ip, the
polarization strength, P, and the polarization PA, θ.
However, one must exercise caution when computing Pbe-
cause Ican be affected by the (unpolarized) PSF of the central
source (e.g., Gledhill & Takami 2001). The PSF of the central
star can (1) accompany enormous spider structures that spa-
tially affect the detectability of the circumstellar shell and (2)
induce very large Iwith respect to potentially small Ip, espe-
cially when the optical depth of the surrounding matter is low.
In such cases, the PSF contribution, Ipsf, has to be removed
from Ito obtain the
PSF-corrected
polarization strength, Pcorr;
Pcorr =Ip
I−Ipsf .(1)
In order for the PSF correction to be effective, PSF refer-
ence observations must be carefully designed. The PSF ob-
servations should achieve a similar S/N to the source obser-
vations, yielding spatially and quantitatively equivalent PSF
structures. The removal of the PSF effect is difficult even with
good PSF reference data, since there is no way to know
a pri-
ori
how much matter exists along the line of sight in front
of the central source. The scaling of the PSF reference data
with respect to the target data would always be a problem, and
hence, Pas a ratio of Ipto I(irrespective of the PSF correc-
tion) should be considered to roughly define the
lower
limit.
In the following, we refer to Pcorr as P. On the contrary, Ipand
θcan be extracted even from the PSF-affected data because
these quantities depend only on Qand Uwhich are free from
the unpolarized component by definition. Therefore, we will
make use of the Ipmaps in addition to the Pmaps to investi-
gate the shell structure in PPNs.
3. OBSERVATIONS AND DATA REDUCTION
3.1. Observations
We obtained high-resolution near-IR polarimetric data from
four PPNs using NICMOS on-board
HST
through a gen-
eral observer (GO) program 9377 (PI: T. Ueta) during Cy-
cle 11, after the NICMOS cooling system (NCS) was in-
stalled during the servicing mission 3B (SM3B). We selected
the target sources (
IRAS
07134+1005,
IRAS
06530−0213,
IRAS
04296+3429, and
IRAS
[Z]02229+62087) based on
their morphological classification in a previous WFPC2
survey (Ueta, Meixner, & Bobrowsky 2000): they are all
SOLE-toroidal PPNs. We observed our targets with Cam-
era 1 (NIC1), which provides 11′′ ×11′′ field of view at
0.
′′043 pixel−1scale, and short wavelength polarizers (POL0S,
POL120S, and POL240S, covering 0.8 to 1.3 µm). We used
the MULTIACCUM (non-destructive readout) mode to re-
cover information in saturated pixels from unsaturated read-
outs.
To remove the effects of bad pixels and improve the spatial
sampling of the PSF, we employed the dithering technique
and observed each source at five or four dithering positions
(i.e., four-point spiral pattern with or without the pattern cen-
ter) with offsets by non-integer pixels. At each dither position
we took images using all three polarizers, allocating equal
amount of exposure time to each polarizer. The observations
were designed this way to make the visits as efficient as pos-
sible. However, taking multiple exposures at each dither po-
sition can cause a problem particularly when a wide dynamic
range has to be covered, since image persistence due to very
bright parts of the source (the central star, in our case) may
leave photon persistence that can affect the subsequent expo-
sures (e.g., Dickinson et al. 2002). To reduce the effect of
persistence, we took two 5.158 sec BLANK exposures after
each exposure with a polarizer.
In addition to the four target sources, we also obtained data
from a PSF reference star, HD 12088, since we concluded
from our previous experience with WFPC2 that the best way
to reduce the PSF effects from
HST
data was to use ob-
served PSF data (Ueta, Meixner, & Bobrowsky 2000). This
high proper-motion star was selected as a PSF reference be-
cause (1) most unpolarized stars are known to be high proper-
7Hereafter, we refer to each object by the RA part of the
IRAS
designation.
The prefix “Z” for
IRAS
02229 is given by Hrivnak & Kwok (1999) due to
the fact that the source was found in the Faint Source Reject File in the
IRAS
Faint Source Survey.
HST/NICMOS Imaging Polarimetry of PPNs 3
TABL E 1
LOG OF
HST
/NICMOS OBSERVATIONS
DITHER EXPTIMEbORIENTATc
Source Date FILTERaSAMP_SEQ NSAMP PATTERN NPTS (sec) (deg) Ref.
New Data
IRAS 07134+1005 2003 Mar 29 POL-S STEP8 19 SPIRAL 5 519.5079 −129.432
IRAS 06530−0213 2003 Mar 29 POL-S STEP8 19 SPIRAL 5 519.5079 −131.793
IRAS 04296+3429 2003 Mar 28 POL-S STEP8 19 SPIRAL 5 519.5079 −149.313
IRAS (Z)02229+6208d2003 Mar 28 POL-S STEP8 16 SPIRAL 4 319.6832 169.035
HD 12088e2003 Mar 28 POL-S STEP1 19 SPIRAL 5 79.7868 162.51
Archived Data
IRAS 16594−4656 1998 May 2 POL-L STEP32 17 SPIRAL 3 863.8767 78.9531 2
BD +32◦3739f1997 Sep 1 POL-L STEP1 9 NONE . .. 41.86784 −93.5733 1
2002 Sep 9 POL-S SCAMRR 9 SPIRAL 4 12.992 −108.88
2003 Jun 8 POL-S SCAMRR 9 SPIRAL 4 12.992 1.32
REFERENCES. — 1. Hines, Schmidt, & Schneider (2000), 2. Su et al. (2003)
aPOL-S: short wavelength polarizers (POL0S, POL120S, and POL240S); POL-L: long wavelength polarizers (POL0L, POL120L, and
POL240L)
bTotal exposure time per polarizer.
cThe ORIENTAT header parameter refers to PA of the image +y axis (degrees E of N).
dThe “Z” prefix in the IRAS designation is given to indicate the fact that this object was found in the Faint Source Reject File in the
IRAS
Faint Source Survey (Hrivnak & Kwok 1999).
ePSF standard for our target sources.
fPSF (and photometric) standard for the archived data.
motion stars and (2) this star lies in the continuous viewing
zone (CVZ) and can be observed in the same orbit as one of
the target sources that is also in the CVZ without spending an
additional orbit just for PSF observations. The same instru-
mental set-up was used for the PSF reference observations.
We summarize the observing parameters in Table 1.
We are interested in the morphologies of the circumstel-
lar shells revealed by the dust-scattered, polarized light. Al-
though polarimetric data have been acquired from evolved
stars with NICMOS in the past, no study has been done to
extract the spatial information from the data by alleviating the
PSF effects of the bright central star. Therefore, we have an-
alyzed all the archived NICMOS polarimetric data obtained
from evolved stars (mostly done with NIC2) to study the
shell structure. Here, we included data from a PPN,
IRAS
16594−4656, whose Ipmap provides more spatial information
of the shell structure than previously reported (Su et al. 2003).
IRAS
16594−4656 was originally observed in a GO program
7840 (PI: S. Kwok), in which Camera 2 (NIC2; 19.
′′2×19.
′′2
field of view at 0.
′′075 pixel−1scale) and long wavelength po-
larizers (POL0L, POL120L, and POL240L, covering 1.9 to
2.1 µm) were used with the three-point spiral dither pattern.
Table 1 also shows observing parameters for the additional
sources.
3.2. Data Reduction
We used the standard set of NICMOS calibration pro-
grams provided in the latest version (Version 3.1) of
IRAF/STSDAS8at the time of data reduction. The CAL-
NICA calibration routines in STSDAS perform zero-read sig-
nal correction, bias subtraction, dark subtraction, detector
non-linearity correction, flat-field correction, and flux calibra-
tion. After pipeline calibration, we realized that our data were
8STSDAS is a product of the Space Telescope Science Institute, which is
operated by AURA for NASA
affected by the “pedestal effect”, a variable quadrant bias, and
the “Mr. Staypuft”, the amplifier ringing and streaking due
to bright targets (Dickinson et al. 2002). The pedestal effect
was removed by first manually inserting the STSDAS task
bi-
aseq
in the middle of the CALNICA processes (before flat-
fielding) and then employing the STSDAS task
pedsub
after
the CALNICA processes. The former removes the non-linear
components of the bias drift, while the latter takes care of the
linear components. To remove the “Mr. Staypuft” anomaly,
we applied “undopuft” IDL routines provided by the STScI
NICMOS group to the raw data before using any of the CAL-
NICA routines. Any remaining stripes along the fast read-
out direction associated with “Mr. Staypuft” were subtracted
by extracting the anomalous stripes from the block-averaged
blank sky.
The calibrated, anomaly-cleaned frames of dithered im-
ages were then combined into a single image by applying
the variable-pixel linear reconstruction algorithm (the STS-
DAS package
dither
, Version 2.0; Koekemoer et al. 2002).
This method would interlace (“drizzle”) each pixel in multi-
point dithered frames according to the statistical significance
of each pixel. The drizzled images were sub-pixelized by a
factor of 2 during the process: the reduced images would have
pixel scales of 0.
′′0215 pixel−1(NIC1) and 0.
′′0375 pixel−1
(NIC2). Cosmic-rays were also removed by the drizzling al-
gorithm.
The NICMOS arrays are slightly tilted with respect to the
focal plane, and thus the NICMOS pixel scales along the two
spatial axes of each camera are not identical. Although its
effect is small, pixels have to be properly rectified to accu-
rately determine the orientation of the polarization vectors.
Since the geometric distortion parameters have been found to
be identical before and after SM3B, we simply applied the
geometrical transformation within the drizzle process, using
the geometric distortion coefficients provided by the STScI
NICMOS group.
4 Ueta, Murakawa, & Meixner
TABL E 2
PHOTOMETRIC AND PO LARI METR IC CALI BRATION PARAMETERS
PHOTFNU (Jy sec DN−1)tk
Polarizer Pre-SM3B Post-SM3B Pre-SM3B Post-SM3B
POL0S 6.996 E−06 4.31 E−06 0.7766 0.7760
POL120S 6.912 E−06 4.19 E−06 0.5946 0.5935
POL240S 6.914 E−06 4.15 E−06 0.7169 0.7181
POL0L 7.626 E−06 6.17 E−06 0.7313 0.8774
POL120L 7.530 E−06 6.10 E−06 0.6288 0.8381
POL240L 7.517 E−06 6.04 E−06 0.8738 0.9667
3.3. Derivation of the Stokes Parameters
At the end of the processes described above, images would
have units of the default count rate for
HST
data (DN sec−1;
“DN” = data number). These count rates can be translated into
appropriate physical units by means of photometric conver-
sion factors, PHOTFLAM and PHOTFNU, before the Stokes
parameters are derived from the data. However, these factors
depend on the actual detector sensitivities, which were altered
after the NCS installation due to the 15 K raise of the NIC-
MOS detector operating temperature. Since post-SM3B pho-
tometric conversion factors for the polarizers were not avail-
able at the time of our analysis, we estimated them following
the method described below.
First we retrieved archived NICMOS calibration data of
an unpolarized standard star, BD+32◦3739, from CAL pro-
grams executed before and after SM3B (PIDs: 7692, 7958,
and 9644). Then, we reduced the data following the same
procedure as our scientific data and performed aperture pho-
tometry on the calibration data using the
apphot
task in the
NOAO/DIGIPHOT package in IRAF. From the measured
count rates of the pre-SM3B data and the pre-SM3B pho-
tometric conversion factors (PHOTFNUs), we computed the
flux of the standard star through each polarizer. Finally,
we obtained the post-SM3B PHOTFNUs from the derived
flux of the standard star and the measured count rates of the
post-SM3B data. The measured count rates of the scientific
data were then converted into Janskys. Table 2 summarizes
PHOTFNU values we used.
Using the photometric calibrated data for each polarizer, we
derived the Stokes parameters following the matrix inversion
method Hines, Schmidt, & Schneider (2000); Dickinson et al.
(2002). For the post-SM3B data, we used revised matrix co-
efficients provided by the STScI NICMOS group, of which
only the tkcoefficients (which are related to the throughput of
the polarizers) have been updated from the pre-SM3B values.
The tkvalues are listed in Table 2.
3.4. PSF Subtraction
To measure correct polarization strengths, the PSF effects
have to be removed. Thus, we need to confirm the polarization
free nature of our PSF standard star, HD 12088. Fig. 1 shows
the Iand Ipmaps of HD 12088 and of an archived polari-
metric standard star, BD+32◦3739. HD 12088 does not ap-
pear to be extended (besides the more pronounced PSF spikes
and polarizer ghosts) with respect to BD+32◦3739 in I(Figs.
1a and 1b). The Ipmap (Fig. 1c) does not seem to indicate
the presence of intrinsic polarization from HD 12088 other
than the known polarizer ghosts appearing at (0.
′′23,1.
′′19)
and (−0.
′′16,0.
′′93) in comparison with BD+32◦3739 (Fig.
1d). The excess Ipstructure is likely due to photon shot noise
FIG. 1.— The I(top) and Ip(bottom) maps of our PSF standard, HD12088
(left), and an unpolarized standard, BD+32◦3739 (right). The maps are cen-
tered at the position of the star and the tickmarks indicate the RA and DEC
offsets in arcsec. The data +y direction points up and the data +x direction to
the left: the PSF structure and polarizer ghost show at the same location.
caused by the bright star and is restricted mainly to <
∼0.
′′6
from the star. The measured polarization is 1.5% in a 0.
′′6 di-
ameter aperture (about ×3 larger than the FWHM of the PSF)
and 2.8% in pixels registering >10σsky in the Imap for HD
12088, while that for BD+32◦3739 is 1.8% and 1.4%, respec-
tively. Hines, Schmidt, & Schneider (2000) reported <
∼1%
instrumental polarization. Thus, our >
∼1% polarization of
BD+32◦3739 is likely due to systematics in the data reduc-
tion procedure. Slightly higher polarization of HD 12088 in
the >10σsky pixels is likely due to the polarizer ghosts and
pixels affected by the PSF spikes. Therefore, we consider that
polarization in HD 12088 is negligible for our purposes in de-
tecting polarization much higher than a few % and that such
small polarization would not affect our interpretation of the
data.
Closer inspection of Figs. 1a and 1b indicates that there is a
companion object very close to HD 12088 at (−0.
′′19,0.
′′09).
Although such a companion can be removed by combining
images rotated by a multiple of 90 deg around the central
star, it turned out that this operation would average out the
asymmetric PSF structure to a degree that the PSF subtraction
would not be effective. Thus, we performed PSF-subtraction
simply using the raw HD 12088 maps: the resulting PSF-
subtracted maps would suffer from oversubtraction by this
companion object.
For the PSF removal, we tried deconvolution using the
Lucy-Richardson method (the
lucy
task in IRAF) and the
MCS method (Magain, Courbin, & Sohy 1998). However,
deconvolution did not yield satisfactory results because (1) the
PSF effects were too severe and extensive to be removed and
(2) the companion objects interfered with the deconvolution
algorithm. Thus, we resorted to a much simpler scale-and-
shift subtraction approach. The intensity scaling factor and
shifts between the images were found by iteratively search-
HST/NICMOS Imaging Polarimetry of PPNs 5
ing for the best values using the PSF spikes, since we did not
know a priori how much intrinsic nebular flux there is in ad-
dition to the stellar flux and the pixels over the central star
were often susceptible to large photon shot noise. It should be
noted that the HD 12088 data are only effective for our target
data obtained with NIC1 polarizers after SM3B. For PSF sub-
traction in pre-SM3B NIC2 data, we used the BD+32◦3739
data.
Fig. 2 demonstrates improvements gained from PSF sub-
traction. The raw Imap (Fig. 2a) is severely affected by the
PSF, and it is nearly impossible to gain any spatial informa-
tion about the nebulosity except that it is extended. However,
the PSF-subtracted Imap (Fig. 2b) successfully reveals the
internal
shell structure. The raw Ipmap (Fig. 2c) shows even
more shell structure than the PSF-subtracted Imap: polar-
ized light seems to be concentrated at the periphery of the
nebulosity with local brightness enhancements. However, we
now recognize the intrusive PSF spikes and polarizer ghosts
at (−0.
′′8, −0.
′′9) and (−1.
′′4, −0.
′′8) due to photon shot noise
caused by the extremely bright central star (Fig. 2d) shows
the PSF-subtracted Ipmap in which the shell structure is seen
almost artifact-free. However, the quality of the image is
gravely compromised by the S/N of the PSF standard data.
This is why data from the polarimetric standard star have to
be carefully tailored in imaging polarimetry for optically thin
circumstellar shells surrounding a bright central source. Un-
fortunately, we did not have an independent orbit for PSF ob-
servations. Hence, our PSF data was unable to gain enough
S/N to achieve optimum results from PSF subtraction. In the
following, we use raw Ipdata in order to take advantage of
high S/N unless the polarizer ghosts pose serious problems in
the data analysis. As for the Pmaps, we use raw Pdata for
displaying purposes, but measurements were done with the
PSF-corrected data.
4. RESULTS
In Figs. 3 to 7, we present the polarimetric maps of five
PPNs in a uniform format. The PSF-subtracted Iand Ip(top
left and right, respectively) maps reveals the distribution of
scattering medium in these circumstellar shells. The Pmap
(bottom left) represents the distribution of Ipwith respect to I,
augmenting the Iand Ipmaps from a different point of view.
The θmap shows the orientation of the polarization vectors
by the “rainbow wheel” pattern. We opt to display the θmaps
this way since the high resolution quality of the data can be
compromised by inevitable rebinning in making conventional
vector diagrams. In addition, the pixels affected by image
anomaly (by the ghosts and shot noise near the central star)
have been masked out in the θmaps.
In Table 3 we summarize the results of the observations in-
cluding the measured coordinates of the object, Iand Ipfluxes,
mean Pand its standard deviation, and descriptions of Ipstruc-
ture. The coordinates listed are the observed location of the I
peak, which coincides with the Ippeak and the θcenter if the
shell is sufficiently optically thin. The fluxes are determined
by integrating the surface brightness over the shell where pix-
els register more than one σsky. Although sky emission has
been subtracted in calibration, we removed any residual “sky”
emission if the sky value determined in an annulus around
the object registers more than three σsky. The mean Pand its
standard deviation have been determined by using pixels that
registered more than 10 σsky. The Ipstructure is described by
the overall shape, dimensions (typically major and minor axis
lengths), and PA measured east of north.
FIG. 2.— The I(top) and Ip(bottom) maps of
IRAS
07134 derived from the
raw (left) and PSF-subtracted (right) data. The maps are in the standard ori-
entation (N is up, E to the left) and centered at the central star with tickmarks
showing the RA and DEC offsets in arcsec. Detailed shell structure can be
seen in the PSF-subtracted Imap, and even better in the Ipmaps. While the
PSF-subtracted Ipmap shows almost artifact-free structure of the shell, it is
degraded from the low S/N of the PSF standard data.
The accuracy of the photometric results depends on the
quality of the PHOTFNU values used in our analysis (Table
2). The measured Ifluxes are all consistent with the known
photometric values, given the difference in the filter profiles
between J/Kand short/long wavelength polarizers, except for
IRAS
07134. Our
IRAS
07134 observations have yielded the
Iflux of 9.3 Jy, which is more than a factor of three higher
than recent measurements of 2.9 Jy (mJ= 6.8; e.g. Ueta et al.
2003). While this source suffers from the polarizer ghosts,
they are known to cause only less than 1% in brightness of the
primary source (Hines, Schmidt, & Schneider 2000). Thus,
the ghosts alone could not have introduced this inconsistency.
Given that our PHOTFNU have yielded reasonably consistent
photometric results for the other three targets (within 50% dif-
ference), our estimates of the PHOTFNU values do not seem
to have caused systematic errors. The most likely source of
this inconsistency seems to be the photon persistence. The
data for
IRAS
07134 are affected by severe photon persis-
tence, and the affected pixels (∼0.
′′4 of the star) can remain
affected even two dither positions later (i.e., four exposures
or more later). The persistent signal decay can interfere with
the non-linearity correction algorithm in the pipeline calibra-
tion, leading to inaccurate count rates. We have not, however,
attempted to improve the accuracy of our photometric mea-
surements, since (1) absolute photometric calibration is not
possible without properly calibrated PHOTFNU values, (2)
only relative calibration among the three polarizers is impor-
tant in deriving the Stokes parameters (QandU), and (3) only
the vicinity of the star (∼0.
′′4) is affected by the photon per-
sistence.
With the (P,θ) data set, we see highly centrosymmetric na-
ture of polarization in all PPNs. We can use the polarization
6 Ueta, Murakawa, & Meixner
TABL E 3
SUMMARY OF
HST
/NICMOS POLARIMETRIC RESULTS
Measured Coord. (J2000)aFlux hPicIpStructure
Source RA DEC BandbI(mJy) Ip(mJy) (%) Morphology Size (′′) P.A. (◦)d
IRAS 07134 07 16 10.27 +09 59 48.5 POL-S 9300e2000e55±16 Elliptical/Round Hollow Shell 4.8×4.0 25
IRAS 06530 06 55 31.80 −02 17 28.3 POL-S 380 67 47 ±15 Elliptical Shell +Bipolar Cusp 2.7×1.0 20
IRAS 04296 04 32 56.95 +34 36 13.1 POL-S 340 60 40 ±12 Quadrupolar
Elliptical (extension 1) 2.1×0.7 26
Elliptical (extension 2) 3.5×0.5 99
IRAS 02229 02 26 41.79 +62 21 22.2 POL-S 4800 1200 52 ±19 Elliptical 2.1×1.3 59
IRAS 16594 17 03 10.04 −47 00 27.0 POL-L 620 66 39 ±16 Elliptical +Protorusions 5.0×2.2 104
aOf the Ipeak location.
bPOL-S: 0.8−1.3µm, centered at 1.1µm; POL-L: 1.89−2.1µm, centered at 2.05µm
cMean Pand its standard deviation.
dDegrees E of N.
eThe measured Iflux value is three times higher than the previously observed value; this is in part due to photon persistence. See text for details (§4).
TABL E 4
OFFSETS BETWE EN THE
ILLUMINATION SOURCE
AND THE IPEAK
Offsets
Source (arcsec)
IRAS 07134 0.08±0.27
IRAS 06530 0.04±0.20
IRAS 04296 0.06±0.17
IRAS 02229 0.12±0.31
IRAS 16594 0.06±0.14
vectors to backtrack the position of the illumination source,
i.e., the central star. The center of the vector pattern was de-
rived by minimizing the sum of the square of the distance be-
tween the vector position and the vector pattern center. In this
analysis, we used vectors that are in the pre-defined annulus
centered at the presumed pattern center. We assumed the I
peak to be the pattern center, and iterated the process by using
the updated center position until the shift between the previ-
ous and current centers becomes smaller than the numerical
accuracy of the analysis. The vector pattern center was found
to coincide with the Ipeak position in all cases: the results are
summarized in Table 4. Thus, these nebulae - SOLE-toroidal
PPNs - are indeed optically thin and illuminated by the central
star located at where the Ipeak is.
5. INDIVIDUAL SOURCES
In this section, we describe the individual source structure,
including the 3-D aspects of it as revealed by these polarimet-
ric images.
5.1. IRAS 07134+1005 (HD 56126)
Polarization observations of
IRAS
07134 (HD 56126) are
presented in Fig. 3. The PSF-subtracted Imap (Fig. 3a) shows
a slightly elliptical nebula (4.
′′8×4.
′′0 at PA 25◦), which
clearly possesses some internal structure previously unrecog-
nized in the near-IR. The bulk of surface brightness is concen-
trated in the region close to the minor axis of the nebula (PA
115◦), with the eastern side being brighter and spatially more
extended than the western side. There is an apparent bright-
ness peak on the east side of the nebula at (1′′,−0.
′′5), whereas
the west side does not show any local brightness peak. At
the northern and southern end of the nebula, there is signifi-
cantly less surface brightness. There appears to be a filamen-
tary structure along the periphery of the nebula, delineating
the elliptical tips in the low surface brightness region (also
seen well in the grayscale image, Fig. 2b).
In the Ipmap (Fig. 3b), we can identify at least two re-
gions of enhanced surface brightness on the east and west
side of the nebula with the local peaks located at (1′′,−0.
′′5)
and (−1.
′′5, 0.
′′5). The east peak shows stronger and more
extended brightness distribution than the west peak. The sur-
face brightness of the east peak is 18 mJy arcsec−2which is
about 1.5 times brighter than the western counterpart (see the
profiles in Fig. 9). The region of enhanced surface bright-
ness (>
∼5 mJy arcsec−2) extends from PA 50◦to 205◦on the
eastern side, and from 250◦to 350◦(with an apparent gap at
320◦) on the western side. These brightness-enhanced regions
are connected by the lower brightness (∼3 mJy arcsec−2) re-
gion at the N-S elliptical tips of the nebula, in which filamen-
tary structures outline the edge of the tips (more apparently
in Ipthan in I). Incidentally, the interior region encircled by
this “rim” region shows weak surface brightness (<
∼3 mJy
arcsec−2, in the region <
∼1.
′′2−1.
′′3 from the center). Overall,
the bulk of Ipappears radially confined to the nebula periph-
ery beyond about 1.
′′5 from the central star in all directions.
This is particularly seen well in the northern half of the nebula
where surface brightness of the interior region (<
∼1.
′′5 from
the central star) becomes very small (almost null) without any
contamination by the polarizer ghosts
In the Pmap (Fig. 3c), we immediately see that the high P
(>
∼20%) regions occur near the periphery of the nebula be-
yond about 1.
′′5 from the central star in all directions: the
mean polarization strength is 55%. The high Pregions corre-
spond to the high Ipregions. However, in the Pmap, there is
not so much of a difference in the polarization strength at the
elliptical tips and at the eastern/western edges of the nebula
as in the Ipmap. The northern tip shows somewhat weaker
polarization strengths than the southern tip: this is consistent
with slightly higher Iin the northern tip and almost equal Ip
at the both tips. The θmap (Fig. 3d) illustrates the polariza-
tion PA by the image tone. This particular θmap shows an
almost perfect “rainbow wheel” pattern by the uniform and
symmetric gradation of the image tone in the azimuthal direc-
tion, which depicts the highly centrosymmetric nature of the
polarization.
The past observations of this PPN found a slightly
elliptical nebula via dust-scattered star light in the
HST/NICMOS Imaging Polarimetry of PPNs 7
FIG. 3.— Polarimetric maps of
IRAS
07134+1005: (a) the total intensity (I), (b) polarized intensity (Ip), (c) polarization strengths (P), and (d) polarization PA
(θ), respectively from left to right, top to bottom. The maps are in the standard orientation (N is up, E to the left) and centered at the θcenter with tickmarks
showing the RA andDEC offsets in arcsec. The wedges indicate the scale of the imagetone: mJy arcsec−2in Iand Ip, percentage in P, and degrees east of north
in θ(i.e. PA 0◦means the polarization vector, which is perpendicular to the scattering plane, is oriented in the N-S direction). In displaying the θmap we used
pixels which register S/N of >10 σsky in the Imap.
FIG. 4.— Polarimetric maps of
IRAS
06530−0213. The display convention is the same as Fig. 3.
HST/NICMOS Imaging Polarimetry of PPNs 9
FIG. 7.— Polarimetric maps of
IRAS
16594−4656. The display convention is the same as Fig. 3.
optical (Ueta, Meixner, & Bobrowsky 2000) and its
two-peaked core structure via thermal dust emission
in the mid-IR (Meixner et al. 1997; Dayal et al. 1998;
Jura, Chen, & Werner 2000; Kwok, Volk, & Hrivnak 2002).
The observed morphology has been thought to represent
an almost edge-on ellipsoidal (slightly prolate) shell with
an equatorial density enhancement (i.e., torus) that results
in limb-brightened two-peak core emission in the mid-IR.
This interpretation has been corroborated by a 2-D radiative
transfer model of dust emission (Meixner et al. 1997, 2004).
The present observations have revealed the toroidal structure
of the shell for the first time in dust-scattered star light in
the near-IR at more than a factor of two better resolution
than the past mid-IR imaging. Moreover, the polarization
characteristics of the shell confirm that it is optically thin
at 1 µm having the central star as the illumination source.
This further strengthens the edge-on torus interpretation, in
which such density structure would manifest itself as two
limb-brightened peaks only when the shell is sufficiently
optically thin
.
In Fig. 8, we compare the shell structure seen in dust-
scattered near-IR light (color) and in thermal dust emission
at 10.3 µm (contours; from Kwok, Volk, & Hrivnak 2002).
Both data show similar brightness distribution where the east-
ern side is stronger and covers a larger spatial extent: even
the surface brightness gap on the western side at PA 320◦is
seen in both of the Ipand 10.3 µm maps. Thus, we confirm
that the imbalance of brightness distribution is real, and so
is the isolated emission blob north of the central star seen in
the mid-IR. Geometrically, this blob appears to be part of the
western edge of the toroidal structure that has been broken off
for some reason.
There has been a question of whether this imbalance of mid-
IR peak strengths is due to density or temperature effects of
dust grains. If the dust temperature were the cause for the
peak strength imbalance, we would not have seen the same
imbalance in the Ipmap. Polarization maps are sensitive to
scattered light, whereas thermal dust emission maps are sen-
sitive to warm dust. Thus, the Ipdistribution does not neces-
sarily have to follow the distribution of thermal mid-IR emis-
sion. Together with the mid-IR data, our images suggest that
there is simply more dust grains in the eastern side of the shell
than the western side. Recent CO observations have shown a
similar morphology (Meixner et al. 2004), corroborating that
there is more matter (dust and gas alike) on the eastern side of
the shell in this object.
Meanwhile, there are apparent differences between the Ip
and mid-IR morphologies which can be easily understood by
the differences in the nature of light arising from dust grains.
One difference is that the Ipand mid-IR peaks do not spatially
coincide: the Ippeaks are found more towards the edge of
the shell than the mid-IR peaks. In general, mid-IR emission
arises from the warmest (∼100 to 200 K) dust grains located
near the inner edge of the shell. Thus, we would expect the
mid-IR peak at the edge of the inner cavity where the line of
sight traverses the longest distance in the warmest dust. How-
ever, due to the curvature of the surface of the inner cavity,
the peak mid-IR position tends to be found somewhat closer
to the central star. On the other hand, Ipbecomes the strongest
at the inner boundary and decreases in the radially outward
direction assuming a radially decreasing density profile (e.g.
∝r−2), because scattering geometry dictates the behavior of
10 Ueta, Murakawa, & Meixner
FIG. 8.— Spatial relationship between dust-scattered light (I[top] and Ip
[bottom] maps in color) and thermal dust emission (10.3 µm contours at 10,
30, 50, 70, and 90%; Kwok, Volk, & Hrivnak 2002). The display convention
is the same as Fig. 3.
scattered light. Another difference is that the Ippeaks are
more extended along the nebula edge than the mid-IR peaks.
This is simply because scattering can occur as long as there
is enough incident light and scattering medium. Thus, the Ip
map has unveiled the shell structure in the outer shell where
the region of enhanced density is extended well into the high
latitude part of the ellipsoidal shell.
As a PPN,
IRAS
07134 is expected to have an inner cavity
generated by the cessation of mass loss at the end of the AGB
phase. The emission structure of the mid-IR and CO maps
(e.g. Kwok, Volk, & Hrivnak 2002; Meixner et al. 2004) is
consistent with the presence of such a cavity. If the shell of
IRAS
07134 is hollow and has a radially decreasing density
structure, then we expect that (1) the highest Ipoccurs at the
inner edge of the shell and (2) Pshould radially increase due
to the geometrical effect of scattering angles (confined closer
FIG. 9.— Polarized brightness profiles of
IRAS
07134: N cut (at PA 25◦;
black solid line), NW-NW-SE cut (average of cuts at intermediate PAs 70◦,
160◦, and 340◦; gray solid line), E cut (at PA 115◦; dotted line), and W cut
(at PA 295◦; dashed line).
to 90◦) and becomes the highest at the outer edge of the shell.
Our Ipand Pmaps do show these characteristics exactly as
expected. Although it is still possible that the contamination
by the unpolarized component of the PSF artificially lower P
in the central region, the high P(>
∼20%) region is restricted
near the outer edge of the shell beyond the PSF. This high
Pregions occupy the same portions of the shell as the high
Ipregions, forming a “ring” structure at the rim of the shell.
Therefore, the shell of
IRAS
07134 most likely possesses an
inner cavity, and the part of the shell probed by scattered light
represents a hollow spheroid.
To better illustrate the spatial variation of surface bright-
ness, we have made various cuts in the Ipmap. In Fig. 9, we
show profiles along the northern major axis (N cut; PA 25◦;
solid black line), eastern minor axis (E cut; PA 115◦; dotted
line), western minor axis (W cut; PA 295◦; dashed line), and
intermediate PAs (NE-NW-SE cut; PAs 70◦, 160◦, and 340◦;
solid gray line). These profiles are derived from a linear cut
of 10-pixel width (or the median of multiple cuts, in the case
of the NE-NW-SE cut). We do not include the S and SW cuts
because of the contamination by the polarizer ghosts.
These cuts show a similar profile: there is a steep outer edge
representing dust pile-up, which surrounds the high bright-
ness region of the main shell, and the brightness steeply falls
down to a relatively flat, plateau region. All cuts show the in-
ner plateau inward of <
∼1.
′′3−1.
′′4, except for the E cut with
plateau of <
∼0.
′′9. We interpret that this inner “plateau” pro-
file is due to the inner cavity created by a precipitous drop in
the mass loss rate at the end of the AGB phase. In the con-
text of polarization, the presence of the inner cavity means the
absence of the scattering medium near the plane of the sky, re-
sulting in an abrupt decrease of Ipand P, thereby forming an
inner boundary of the shell. It is, therefore, the direct obser-
vational evidence for the presence of such an inner cavity in
PPNs.
5.2. IRAS 06530−0213
The polarization maps of
IRAS
06530 are displayed in Fig.
4. The PSF-subtracted Imap (Fig. 4a) shows a highly el-
liptical shape of the nebula (2.
′′4×1.
′′1 at PA 20◦), in which
HST/NICMOS Imaging Polarimetry of PPNs 11
FIG. 10.— Polarized brightness cuts of
IRAS
06530: N-S cut (average of
N and S profiles at PA 20◦and 110◦; gray solid line) and E-W cut (aver-
age of cuts along lines perpendicular to the major axis 0.
′′43 N and S of the
equatorial plane; black solid line).
the low brightness (∼5 mJy arcsec−2) elliptical tips extend
beyond the barrel-shaped region of higher surface brightness
(>
∼15 mJy arcsec−2). The Ipmap (Fig. 4b) reveals a cusp-
like (or a sideway x) structure within the barrel region. Such
structure typically indicates the swept-up walls of the bipolar
cavities. Thus,
IRAS
06530 is likely a near edge-on, highly
prolate ellipsoidal shell showing a rather bipolar nature in the
low latitudinal region. Fig. 10 shows the polarized brightness
cuts of the shell made along the major axis (N-S cut; N-S
averaged, thick gray line) and the lines parallel to the minor
axis at 0.
′′43 N and S of the central star (E-W cut; E-W av-
eraged, solid black line). The N-S cut demonstrates the elon-
gated Ipstructure with a gentle slope at the edge, while the
E-W cut shows a peak at around 0.
′′34 that defines a bipolar
cavity wall. The polarization characteristics seen in the Pand
θmaps (Fig. 4c and 4d, respectively) are very similar to those
of
IRAS
07134 (Fig. 3). Pis stronger near the edge of the
shell while it is absent in the central region even beyond the
PSF. The polarization pattern is very much centrosymmetric
as seen from the almost perfect rainbow wheel pattern.
Thus,
IRAS
06530 is a highly prolate spheroidal shell,
which very likely has an inner cavity as in
IRAS
07134. Al-
though the shell is optically thin, there is sufficiently high
concentration of dust grains at the equatorial region so that
the bipolar cusp structure is seen in Ip. The total optical depth
is not large enough to induce the full extinction of star light
expected in typical bipolar PPNs (e.g.,
IRAS
17150−3224;
Kwok, Su, & Hrivnak 1998; Su et al. 2003). This may be why
we do not observe a low Ip“hole” in the central region of the
shell as we saw in
IRAS
07134. However, we do see a possi-
ble inner cavity in the Pmap.
5.3. IRAS 04296+3429
In Fig. 5 we present data from
IRAS
04296. The PSF-
subtracted Imap (Fig. 5a) shows a quadrupolar nebula
(roughly 1′′ ×1.
′′5) with an east-west elongated core having
round protrusions towards PA 26◦and 206◦. The Ipmap (Fig.
5b) successfully unveils the slanted
X
shape of the nebula
more clearly than the Imap. The nebula’s
X
shape is due
to the presence of two axes of elongation. One of the elonga-
tions is oriented at PA 26◦and shows a relatively well-defined
FIG. 11.— Surface brightness radial profiles of
IRAS
04296: of the exten-
sion 1 (at PA 26◦; black solid line), of the extension 2 (at PA 99◦; gray solid
line), and of the least extended part of the shell (the average of the cuts at PA
63◦and 153◦; dotted line).
elliptical shape (2.
′′1×0.
′′7; extension 1). The other elonga-
tion, oriented at PA 99◦, is fainter with its surface-brightness-
limited ends smeared out in the background (3.
′′5×0.
′′5; ex-
tension 2). These elongations are not oriented perpendicu-
lar to each other. Although Sahai (1999) reported that ex-
tension 2 is not straight with a 5◦shift between the eastern
and western tips from the WFPC2 data, we are unable to con-
firm this in our data due to the confusion by the PSF spike
nearly aligned with this extension. While the Ipstructure
is consistent with the one seen in the previous WFPC2 data
(Ueta, Meixner, & Bobrowsky 2000) the central star appears
more prominently in the near-IR data, since the central star
(the nebula) is brighter (fainter) in the near-IR. This is why the
Imap does not clearly show the
X
structure as in the WFPC2
images.
Fig. 11 displays Ipradial profiles of the extensions to better
present the extent of the structures. These profiles are con-
structed by taking linear cuts of 10-pixel width along the ex-
tensions and averaging the values at both ends. For compar-
ison, we also show the “least extended” shell profile, created
from linear cuts at PA 63◦and 153◦(in-between directions
of the extensions). The extension 1 profile (black solid line)
shows Ipexcess in the region close to the central star, but it
suddenly falls to the flux level similar to the least extended
profile (black dotted line) at around 1′′ . On the other hand,
the extension 2 profile exhibits Ipexcess as far out as ∼1.
′′8
before it gradually falls down to the background level.
The Pmap (Fig. 5c) uncovers the structure of the two ex-
tensions even better. In extension 1, strongly polarized light
(>
∼30%) is concentrated near the periphery of the elongation,
especially at the tips, whereas weakly polarized light fills the
cavity surrounded by the high polarization region. Because of
the compactness of the nebula, we can not rule out the pos-
sibility that the central region registers low Pvalues due to
the residual unpolarized component from the central star. The
Pstructure of extension 1 resembles that of the previous two
sources. However, in extension 2, the weak Pcavity is not
very well-defined partly due to confusion by the presence of
falsely high polarization caused by the PSF spikes. Never-
theless, high polarization is observed as far out as 1′′ from
the central star. The θmap (Fig. 5d) exhibits a general cen-
12 Ueta, Murakawa, & Meixner
trosymmetric pattern of an optically thin shell, with only a
marginally shallower gradient over the extensions.
Based on the optical morphology of the nebula, Sahai
(1999) interpreted extension 1 to be a bounded disk that col-
limated outflows manifesting themselves as extension 2. Our
polarization data, however, do not support this disk interpre-
tation. The Ipshape of extension 1 does not support a geomet-
rically thin disk structure. If extension 1 is a thin disk with its
plane inclined 24◦with respect to the line of sight, the spatial
distribution of the disk material in the plane of the sky will be
extremely restricted. This would result in strong scattering in
a geometrically narrow region in the sky, and the Ipmap will
likely be very narrow in the direction perpendicular to the ex-
tension. However, this is not the case. The fact that extension
1 appears elliptical in Ipstrongly suggests that the distribution
of scattering matter is elliptical in the plane of the sky, imply-
ing a prolate spheroidal structure of the shell. The Pmorphol-
ogy of extension 1 shows a possible low Pcavity surrounded
by the region of strong polarization at the nebula edge. Thus,
extension 1 is very likely a prolate spheroid, possibly with
an inner cavity. In terms of polarization characteristics, there
is no significant difference in both extensions except for the
sharpness of the tip structure. In extension 1, we see the high-
est Pat the tips of the elongation. However, in extension 2,
the highest Pdoes not arise from the tips. As the profiles indi-
cate in Fig. 11, extension 2 is extended out to about 1.
′′8. This
may suggest that extension 2 is also a hollow prolate spheroid,
but is inclined with respect to the plane of the sky so that the
region of high Poccurs where the spheroidal shell intersects
with the plane of sky and not at the tips.
5.4. IRAS (Z)02229+6208
We show the polarization maps of
IRAS
02229 in Fig. 6.
The PSF-subtracted Imap (Fig. 6a) shows an elliptically ex-
tended nebula (2.
′′1×1.
′′3 at PA 45◦), which is consistent with
the previous WFPC2 images (Ueta, Meixner, & Bobrowsky
2000). The surface brightness distribution is such that the
southwestern tip is slightly brighter than the northeastern tip.
The Ipmap (Fig. 6b) exhibits the same elliptical shape of the
nebula. Unlike the Imap, the Ipmap does not show any ap-
parent spatial difference in the polarized surface brightness
distribution. Although the low brightness edge (∼50 mJy
arcsec−2) is elongated towards PA 45◦(nearly aligned with
a PSF spike), the high surface brightness core (>200 mJy
arcsec−2) appears to be elongated towards a slightly different
direction (PA 60◦).
The Pmap (Fig. 6c) appears quite differently with respect
to other Pmaps: strong polarization is not concentrated near
the periphery in this nebula. Instead, we see a band of low
polarization (<
∼30%) in the middle of the shell aligned with
PA 150◦, which separates the region of medium polarization
(∼30−40%) on the southwest side and the region of high po-
larization (>
∼40%) on the northeast side. The θmap (Fig. 6d)
shows a generally centrosymmetric pattern. However,the gra-
dient of the vector angle seems to be steep in the low polariza-
tion band, and shallow in the elliptical tips of the shell. This
indicates that more vectors in the elliptical tips are aligned
parallel to the low Pband which most likely represents the
equatorial plane. Note also that the orientation of the low
polarization band is not perpendicular to the direction of the
elongation of the shell, but to that of the core elongation (PA
60◦). The polarization characteristics of this object is very
distinct with respect to those of the previous objects.
The peculiar Pmap can be understood in terms of the in-
ner structure and inclination of the shell. The presence of the
low Pband suggests that dust density is more equatorially
enhanced within a rather geometrically narrow region in this
object. If the near side of this inner torus is tilted towards the
northeastern direction, the southwestern side of the shell (i.e.,
the near side of the ellipsoidal shell) appears more illuminated
by the direct star light in our viewing angle. This is consistent
with the Imap showing the brighter southwestern tip than the
northeastern tip (Fig. 6a). However, assuming a spheroidal
density distribution of the shell there will be no difference in
terms of the amount of scattering medium in the plane of the
sky, and hence, the Ipbrightness will be the same on both sides
of the shell (Fig. 6b). Since Pis a ratio of Ipto I, we would
see lower (higher) Pon the southwestern (northeastern) side
of the shell. Therefore,
IRAS
02229 seems to be similar to
IRAS
06530 with a spheroidal shell with a relatively stronger
equatorial density enhancement. However, the shell orienta-
tion of
IRAS
02229 is most likely more pole-on compared to
IRAS
06530.
IRAS
02229 was marginally resolved in the mid-IR
in our previous imaging survey (Meixner et al. 1999),
and has recently been observed at the Gemini North
(Kwok, Volk, & Hrivnak 2002). Fig. 12 shows the Ipand P
maps overlaid with 10.3 and 18.0 µm contours. 10.3 µm emis-
sion is extended with PA of 20◦, which is off by 25◦with
respect to the Ielongation (Fig. 12a). We have also seen a
slight shift of PA between the outer and inner Istructure (45◦
and 60◦, respectively). Assuming these parts of the shell with
different PAs represent distinct portions of the shell structure,
this shift of PA may indicate that the inner torus is precess-
ing/rotating in the counterclockwise direction. Such preces-
sion/rotation of the inner torus has already been suspected in
a SOLE-toroidal PPN, HD 161796 (Gledhill & Yates 2003).
The low Pband also does not seem to show strong spatial
correlation to the 10.3 µm emission region (Fig. 12b). How-
ever, the 18.0 µm map seems to be elongated in the direc-
tion of the low Pband (Fig. 12c). Kwok, Volk, & Hrivnak
(2002) suspected that the 18.0 µm map may have suffered
from variable sky conditions, If
IRAS
02229 has geometri-
cally narrow density distribution in the innermost region of
the shell, it is possible that the mid-IR continuum emission
would be elongated along the equatorial plane as we see in
Fig. 12c. If this geometry causes the 18.0 µm elongation,
then this raises a question as to why the 10.3 µm emission,
another map of continuum, is not elongated as the 18.0 µm
map. The 10.3 µm map, in fact, is morphologically very sim-
ilar to the 11.7 and 12.5 µm maps that represent unidenti-
fied IR feature emission. Although Kwok, Volk, & Hrivnak
(2002) concluded that 10.3 µm emission represented contin-
uum, comparison of the
ISO
spectrum with the mid-IR filter
profiles (see Fig. 4 in Kwok, Volk, & Hrivnak 2002) may in-
dicate that the 10.3 µm filter does capture some emission due
to the same unidentified IR feature emission that caused elon-
gation in the 11.7 and 12.5 µm images. It is also interesting
to note that the 20.8 µm map (representing the unidentified
“21 µm” feature) shows somewhat different elongation with
respect to other maps. Thus, it may be that the 18.0 µm map
is the true continuum emission map that reflects the disk-like
dust distribution of the shell, and that other maps are different
due to distinct dust species responsible for feature emission at
other wavelengths. Further investigation at higher resolution,
preferably spatial spectroscopy, is necessary to determine if
distinct spatial distribution of different dust species can result
in such changes of morphology.
HST/NICMOS Imaging Polarimetry of PPNs 13
FIG. 12.—
IRAS
02229: Spatial relationship between dust-scattered light
and thermal dust emission (color Ipmap and 10.3 µm contours; [a]), the
polarization strength and thermal dust emission (color Pmap and 10.3 and
18.0 µm contours; [b] and [c], respectively). The mid-IR data are from
Kwok, Volk, &Hrivnak (2002), and the contours are at 10, 30, 50, 70, and
90%. The display convention is the same as Fig. 3.
5.5. IRAS 16594−4656
We present the archived 2 µm polarization data of
IRAS
16594−4656 that have been reanalyzed to mitigate the PSF
effects (Fig. 7). The PSF-subtracted Imap (Fig. 7a) reveals
the inner structure of the nebula that is clearly elongated in
the east-west direction (5.
′′0×2.
′′2 at PA 81◦−261◦;>
∼7 mJy
arcsec−2). The Ipmap (Fig. 7b) uncovers the structure of the
shell, in which we see the bipolar cusp structure, similar to
the one in
IRAS
06530 (Fig. 4b), corresponding to the main
Ielongation. We do not see any elliptically elongated tips
surrounding the cusp structure as in the case for
IRAS
06530.
The tips of the cusp are more elongated in this object and
appear to delineate the wall of the bipolar cavities.
The Pstructure resembles those with a hollow shell (Fig.
3c, 4c, and probably 5c), in which there is a region of low
polarization in the middle of the shell encircled by a region of
high polarization. In
IRAS
16594, however, the region of high
polarization does not seem to completely surround the central
low Pregion: it is mainly found southeastern and northwest-
ern ends of the shell. Moreover, it appears that the Pmap also
possesses a low Pband structure in PA 167◦, which is similar
to the one in
IRAS
02229 (Fig. 6c). The θmap is very much
centrosymmetric.
IRAS
16594 has been observed with
HST
many times in
the past. WFPC2 images have shown the object’s multi-
polar reflection nebula that has three extensions on each side
(e.g., Hrivnak, Kwok, & Su 1999). These extensions form
oppositely pointing pairs in the directions of PA 40◦−220◦,
60◦−240◦, and 80◦−260◦. The length of the elongation de-
creases in the counterclockwise direction, while the surface
brightness increases. Thus, the optical structure has been in-
terpreted as oppositely directed material ejected episodically
from a rotating source. Our PSF-subtracted Iand Ipimages
show that the inner shell structure is aligned with the elon-
gation of PA 80◦−260◦, as has been seen only faintly in
heavily PSF affected NICMOS broad to medium band images
(Su et al. 2003). Hence, the episodic ejection interpretation is
consistent with the nebula’s inner shell structure.
The toroidal nature of the innermost shell has recently been
exposed by mid-IR imaging, in which the presence of the
limb-brightened peaks in the emission core has proven that the
shell has an equatorial density enhancement along PA 170◦
(García-Hernández et al. 2004). The elongation of the inner
shell is therefore aligned with the symmetric axis of the torus.
Thus, it seems likely that material ejection is presently chan-
neled into the directions of the inner shell elongation by the
torus or by some collimation mechanism(s) that can generate
such outflows and equatorial density enhancement.
The orientation of the torus is spatially coincident with the
low Pband seen in our data (Fig. 7c). This confirms our in-
terpretation of the presence of low Pband as a manifesta-
tion of dust density enhancement along the equatorial plane
of the system (see Fig. 6c and discussion associated with it).
Su et al. (2003) interpreted that the low Paround the central
star was due to an inclined dust torus at an “intermediate” an-
gle. However, the Pmap of
IRAS
16594 does not suggest an
inclination angle as large as
IRAS
02229 in which the imbal-
ance of Idue to inclination resulted in the imbalance of Pon
the opposing sides of the shell. Thus, the orientation of the
torus in
IRAS
16594 is more likely close to edge-on. A dust
emission model with our 2-D radiative transfer code has indi-
cated the inclination angle of roughly 75◦with respect to the
line of sight (Ueta et al. 2004).
14 Ueta, Murakawa, & Meixner
6. DISCUSSION
6.1. 3-D Shell Structure via (Ip, P)
Our main objective in the present study is to investigate
the structure of
optically thin
PPNs by means of imaging po-
larimetry. We have made use of the (Ip,P) data set to achieve
our goal instead of the more conventional (P,θ) data set with
which
optically thick
regions of the circumstellar shells are
probed through the way polarization vectors are aligned with
respect to the equatorial plane (e.g. Whitney & Hartmann
1993; Su et al. 2003).
In radially decreasing density distribution typical of PPNs,
Ipradially decreases. So, Ipbecomes the strongest at the inner
edge of the shell if the shell has an inner cavity as in the case
of PPNs. In the optically thin regime where single scatter-
ing dominates, Pin general becomes the strongest at the outer
edge of the shell because of the geometrical effect. With the
(Ip,P) data set, we have successfully detected both inner and
outer edges of the shell in
IRAS
07134. The data have shown
that the polarized surface brightness distribution encircles the
inner cavity, forming a complete “ring”. In terms of scatter-
ing geometry, such brightness distribution indicates a hollow
spheroid. The data have also shown equatorial enhancement
in the material distribution.
Our previous studies of the PPN structure independently
confirmed (1) equatorially enhanced (toroidal) dust distribu-
tion in the innermost region of the shell by mid-IR imag-
ing (Meixner et al. 1999; Ueta et al. 2001), (2) the pres-
ence of an elliptical shell surrounding the central torus
(Ueta, Meixner, & Bobrowsky 2000) by optical imaging, and
(3) an equatorially enhanced hollow spheroid (a combination
of the above structures) embedded in a spherically symmetric
outer shell would explain all the morphological characteristics
by numerical modeling (Meixner et al. 2002). With imaging
polarimetry, we have been able to observationally prove that
the PPN structure is a hollow spheroid with a built-in equato-
rial enhancement.
Standard imaging data show only the structure of the cir-
cumstellar shells projected to the plane of the sky: the struc-
tural information along the line of sight is degenerate. How-
ever, the (Ip,P) data set can retain the 3-D properties of these
shells because the Ipand Pstrengths depend on the scattering
geometry within the shells. Thus, we can effectively probe the
structure of PPNs, by extracting their 3-D information. In ad-
dition, we have been able to determine the detailed geometry
of our target sources from their polarization properties:
IRAS
04296 has a quadrupolar shell and does not harbor a disk with
collimated outflows and
IRAS
02229 is oriented in a rather
inclined direction with respect to us.
6.2. Morphological Classification Scheme of PPNs
Gledhill et al. (2001) introduced a classification scheme for
their imaging polarimetry of PPNs based on the Ipmor-
phology and other polarization properties. These categories
are
Shells
,
Bipolars
, and
Core-dominated
. Their study con-
firmed the SOLE-toroidal vs. DUPLEX-core/elliptical bifur-
cation found among PPNs and supported the idea that the bi-
furcation originated from the varying degrees of optical thick-
ness of the shell. Our target PPNs are all
Shells
based on their
polarization properties (that is, they are equivalent to SOLE-
toroidals). However, our high resolution (Ip,P) data set has
shown a range of morphologies among these
optically thin
PPNs, from a detached shell structure (
IRAS
07134; Fig. 3b,
Fig. 9) to a bipolar cusp structure (
IRAS
16594; Fig. 7b), and
even the mixture of the two (
IRAS
06530; Fig. 4b, Fig. 10).
This suggests that even for
optically thin
PPNs there are mul-
titudes of optical depths that their shells can assume.
As pointed out by Gledhill et al. (2001), the division of
the SOLE-DUPLEX bifurcation is not clearly defined. These
morphological classes are the both ends of a spectrum in the
domain of optical depth. A given PPN can have any op-
tical depth in this continuous distribution of optical depth,
and therefore, can assume any morphology along this SOLE-
DUPLEX spectrum. The Pmaps of
IRAS
02229 and
IRAS
16594 (Figs. 6c & 7c) have also hinted that inclination an-
gles and highly geometrically thin equatorial density enhance-
ments can leave characteristic signature in the resulting mor-
phology. Thus, we have demonstrated the robustness of our
data, and our results further strengthen the suggestion that the
optical depth of the shell plays a major role in determining the
PPN morphology with an added complexity from the actual
geometry of the equatorial enhancement and the inclination
of the object. In order to quantitatively understand the PPN
morphology in the
optically thin
regime, we need more scat-
tering/polarization models of optically thin shells with con-
sideration of inclination angles.
6.3. Structure of Superwind
PPNs form as a result of stellar mass loss along the AGB
(c.f., Kwok 1993; Balick & Frank 2002). The innermost
shells are created by stellar material that has been ejected
in the most recent mass loss, and thus, represent the lat-
est AGB mass loss history. In other words, the observed
shells likely embody the
superwind
shells (Renzini 1981;
Iben & Renzini 1983). According to the standard scenario
for the AGB mass loss, superwinds are enhanced versions
of AGB winds that occur during the final stage of the AGB
evolution (corresponding to the thermal pulsing AGB phase;
Iben 1995 for a thorough review). Hydrodynamical simula-
tions combined with detailed stellar evolution models have
shown that thermal pulses lead to enhanced mass loss events
(Schröder, Winters, & Sedlmayr 1999) and the formation of
detached dust shells (Steffen, Szczerba, & Schönberner 1998;
Steffen & Schönberner 2000). Therefore, we can use the ob-
served shell structure to glimpse the history of superwind
mass loss at the end of the AGB phase. Here, we will con-
tinue our discussion using
IRAS
07134 data that shows the
density distribution in the greatest detail.
IRAS
07134 has a rather well-defined outer edge at 2.
′′0
to 2.
′′4 from the central star as seen in the Iand Ipimages
(Fig. 3a and 3b) and surface brightness profiles (Fig. 9). At
the edge, the signal registers at least 10 σsky, and hence, the
edge is most likely density bounded. The local surface bright-
ness peak at the northern edge of the shell (Fig. 9; i.e., the
filamentary structure in the Iand Ipmaps, Figs. 3a and 3b)
likely represents a pile-up of matter at the interface between
the faster dust-driven wind and the slower shock-driven wind
co-existing in the AGB shell seen in the model calculations
(Steffen & Schönberner 2000). Thus, the observed shell in
dust-scattered star light very likely embodies the shell cre-
ated by superwind, especially by the wind associated with the
last thermal pulse. Using the distance of 2.4 kpc (Hony et al.
2003) and the wind velocity of 10.5 km s−1(Meixner et al.
2004), the edge represents matter ejected 2000 to 2400 years
ago. We also note the presence of inner cavity of about 1.
′′4
radius, which is most likely due to the precipitous decrease of
mass loss at the end of the AGB phase and defines the
physi-
cally detached
nature of the PPN. The width of the superwind
HST/NICMOS Imaging Polarimetry of PPNs 15
TABL E 5
DYNAMI CAL PROPERTIES OF THE SOURCES
Rsw Rin D vexp tsw tdyn
Source Direction (′′) (′′) (kpc) Ref. (km s−1) Ref. (yrs) (yrs) Comments
IRAS 07134 Pole (N) 2.2 1.3 2.4 1 10.5 2 990 1430
Pole (S) 2.5 1.3a2.4 1 10.5 2 1300 1430
Equator (E) 1.8 0.9 2.4 1 10.5 2 990 990 Prolonged mass loss?
Equator (W) 2.0 1.3 2.4 1 10.5 2 780 1430
IRAS 06530 Pole 1.35 0.9b6.7c3 14 3 1000 2100 Bipolar cavity
Equator 0.5 0.35d6.7c3 14 3 350 800 Bipolar cusp
IRAS 04296 Extension 1 1.05 0.6 4.0 4 12 5 720 960 Co-existing flows?
Extension 2 1.75 0.6e4.0 4 12 5 1800 960 Co-existing flows?
IRAS 02229 Pole 1.05 . .. 2.2f6 13 6 850f0f
Equator 0.65 .. . 2.2f6 13 6 530f0f
IRAS 16594 Pole 1.05 0.6b2.2 7 16 8 220 1400 Precessing/Rotating Torus
Equator 1.1 0.7d2.2 7 16 8 260 460 Bipolar Cusp
REFERENCES. — 1. Hony et al. (2003), 2. Meixner et al. (2004), 3. Hrivnak & Reddy (2003), 4. Meixner et al. (1997), 5.
Omont et al.. (1993), 6. Reddy, Bakker, & Hrivnak (1999), 7. Van de Steene & van Hoof (2003), 8. Loup et al. (1990)
NOTE. — Rsw: superwind shell (Ip) size; Rin: inner radius; D: distance; vexp: expansion velocity; tsw: duration of superwind
mass loss; tdyn dynamical expansion time
aAssumed to be the same as in Pole (N).
bEstimated from the Pprofile.
c2.8 times as far as
IRAS
07134, provided the objects have the same luminosity (Hrivnak & Reddy 2003).
dMeasured at 0.
′′7 above and below the equatorial plane.
eAssumed to be the same as in Extension 1.
fThe lower limit.
gAssuming no cavity.
shell is about 0.
′′8. So, the final mass loss has continued for
about 800 years after the edge matter was ejected, and the
shell has been expanding for about another 1400 years since
the mass loss ceased at the end of the AGB phase. This du-
ration of the final mass loss is consistent with theoretically
derived duration of superwind mass loss (Blöcker 1995).
The surface brightness profiles along the equatorial direc-
tions are evidence for distinct histories of mass loss experi-
enced by the different equatorial regions of the shell. The
observed shell structures suggest that the equatorial density
enhancement does not occur symmetrically. There is larger
amount of dust grains on the eastern side than the western
side. Assuming the constant mass loss velocity of 10.5 km s−1
on both sides of the shell, this imbalance could have resulted
from prolonged superwind mass loss experienced by the east-
ern side: while the western superwind terminated about 1400
years ago, the eastern superwind continued another 400 years.
Alternatively,the eastern side may not have been moved as far
away from the central star as the western side if the wind mo-
mentum is equal on both sides.
The presence of the inner cavity confirms the physically de-
tached nature of the shell in
IRAS
07134. However, other ob-
jects do not show the inner cavity as clearly as
IRAS
07134,
especially in Ip. These shells instead show the bipolar cusp
structure around the central star. Since the cusp structure sug-
gests the presence of matter in the inner cavity, the presence
of the cavity in these shells may not have been seen clearly
in Ipas a “hole”. Their Pstructure is, nevertheless, very sim-
ilar to that of
IRAS
07134 (Fig. 4c for
IRAS
06530, Fig. 5c
for
IRAS
04296, and, Fig. 7c for
IRAS
16594). Thus, we
tentatively suggest their marginally hollow nature. As for the
cause of the cusp structure (and the presence of matter in the
inner cavity), it is not very clear. However, it may be intrin-
sic to PPNs that have optically thicker shells. That is, such
cusp structure within the inner cavity may be
required
to form
as PPNs develop optically thicker shells. The cusp structure
may also be indicative of the presence of a rather flattened
torus, which can participate in forming highly elongated shell
as we see in these objects. However, the degree to which this
shell elongation mechanism works seems to be relatively low
in these objects, and thus, these nebulae do not possess the
prototypical bipolar morphology.
The presence of a single inner shell in
IRAS
16594 indicate
that the multiple shell structure seen in the optical has been
created one elongation at a time while the central region pre-
cesses/rotates. However, the quadrupolar structure of
IRAS
04296 (Fig. 5) suggests two elliptical elongations that can co-
exist. Using the distance of 4 kpc (Meixner et al. 1997) and
the CO outflow velocity of 12 km s−1, we estimate that the
superwind which shaped extension 1 was initiated about 1580
years ago. The absence of the pile-up edges in extension 2
raises an important question about its origin. If the wind ve-
locity along extension 2 is similar to that along extension 1,
the wind that shaped extension 2 must have been initiated an-
other 1580 years earlier than the beginning of extension 1 for-
mation. If the shaping of extension 2 was initiated about the
same time as the shaping of extension 1 by the onset of su-
perwind, the wind velocity along extension 2 must have been
about 24 km s−1. If the amount of dust in the two extensions is
different (e.g. Sahai 1999), two winds with the same velocity
could generate elongations with differing length. Either case,
two elongations seem to have been generated concurrently at
least for some time. Thus, there are likely multiple channels
to create multi-polar shell structures.
The polarimetric data have shown that the presence of equa-
torial density enhancement is prevalent among PPNs with
widely varying degrees of strength. With the presence of
axisymmetry in the late AGB mass loss history being es-
tablished as fundamental, it is then reasonable to assume
that something equally fundamental is responsible as the ori-
gin for axisymmetry - the equatorial enhancement - in the
AGB mass loss. An axisymmetric mass loss process may
16 Ueta, Murakawa, & Meixner
be a natural consequence of very fundamental physics in-
volved in any stellar mass loss and/or any dusty outflow-
ing astronomical phenomena. In Table 5, we summarize
quantities of geometric and dynamic nature of the objects.
We measured the inner and superwind (outer) radii (Rin and
Rsw, respectively) assuming the presence of the inner cav-
ity based on their Pstructure. Using the distance and shell
expansion velocities for these objects taken from the liter-
ature (Loup et al. 1990; Omont et al.. 1993; Meixner et al.
1997; Reddy, Bakker, & Hrivnak 1999; Hony et al. 2003;
Hrivnak & Reddy 2003; Van de Steene & van Hoof 2003;
Meixner et al. 2004), we derived the duration of superwind
(Tsw) and expansion the cessation of mass loss (tdyn). These
quantities would not only characterize the axisymmetric na-
ture of PPNs but also serve as constraining parameters for
any models to generate equatorial enhancements in the AGB
mass loss with magnetic fields/companion objects (e.g., Soker
1998; Mastrodemos & Morris 1999; Matt et al. 2000).
7. CONCLUSIONS
We have performed imaging polarimetry using
HST
/NICMOS to observe optically thin PPNs, with the
aim to reveal their density structure by the polarized intensity,
Ip. In the optically thin regime, the (Ip,P) data set can
effectively probe the presence of scattering bodies distributed
in the vicinity of the bright central illumination source. This
is a great advantage of imaging polarimetry over conventional
imaging, in which the central star (and its PSF structure) will
always dominate in these objects and the detection of faint
nebulosities will not be trivial.
We have found that (1)
IRAS
07134 is an equatorially en-
hanced, prolate hollow spheroidal shell that is nearly edge-
on. The presence of the inner cavity has been observation-
ally confirmed and the peak imbalance seen in the present
study and past study of mid-IR dust emission has been de-
termined to be due to density effects, (2)
IRAS
06530 is a
prolate spheroid with a possible cavity, but its equatorial en-
hancement is more bipolar-like, filling the cavity with some
matter, (3)
IRAS
04296 is a combination of two spheroids in-
tersecting with each other with one of the extensions being
inclined towards us, (4)
IRAS
02229 resembles
IRAS
06530,
but has greater equatorial enhancement and the inclination an-
gle closer to pole-on compared with other objects, and (5)
IRAS
16594 has very bipolar-like morphological characteris-
tics indicating relatively high optical depth of the shell, whose
multi-polar structure is due to a pair of outflows channeled
out from a rotating/precessing torus. These observations have
also indicated that there are multiple channels of evolution
for multi-polar shells: while
IRAS
04296 seems to be a co-
existing quadrupolar nebula, multiple poles of
IRAS
16594
likely result from a rotating/precessing torus.
Our observations have strongly suggested that PPNs do
possess an inner cavity which physically separates the shell
from the central star and represents dynamical time for the
shell expansion since the end of the AGB phase.. The inner
PPN shell seems to be density bounded and its structure is
likely shaped by superwind. In the case of
IRAS
07134, we
have been able to observationally confirm that the inner PPN
structure is a hollow spheroid with some equatorial enhance-
ment. We have also confirmed that SOLE-toroidal PPNs have
optically thin shells and that the varying degree of equatorial
density enhancement (i.e., the optical depth) determines the
detailed shell morphology even among this specific group of
optically thin PPNs.
This research is based on observations with the NASA/ESA
Hubble Space Telescope, obtained at the Space Telescope Sci-
ence Institute (STScI), which is operated by the Association
of Universities for Research in Astronomy, Inc. under NASA
contract No. NAS 5-26555. Authors have been supported by
NASA STI 9377.05-A. Ueta has also been supported by the
project IAP P5/36 financed by FSTC of the Belgian State. The
assistance from E. Roye and A. Schultz during data reduc-
tion was appreciated. We thank K. Volk who generously pro-
vided us with mid-IR data taken at Gemini. F. Courbin is also
thanked for his help using the MCS deconvolution software.
We also appreciate valuable comments from anonymous ref-
eree that helped improving the paper.
REFERENCES
Balick, B., & Frank, A. 2002, ARA&A, 40, 439
Blöcker, T. 1995, A&A, 297, 727
Dayal, A., Hoffmann, W. F., Bieging, J. H., Hora, J. L., Deutsch, L. K., &
Fazio, G. G. 1998, ApJ, 492, 603
Dickinson, M., et al. 2002, in “HST NICMOS Data Handbook Version 5.0”,
ed. B. Mobasher (Baltimore: STScI)
García-Hernández D. A., Manchado, A., García-Lario, P., Benítez Cañete,
A., Acosta-Pulido, J. A., Pérez García, A. M. 2004, in ASP Conf. Ser. Vol.
313, Asymmetric Planetary Nebulae III, eds. M. Meixner, J. Kastner, B.
Balick, & N. Soker (San Francisco: ASP), 367
Gledhill, T. M., Chrysostomou, A., Hough, J. H., & Yates, J. A. 2001,
MNRAS, 322, 321
Gledhill, T. M., & Takami, M. 2001, MNRAS, 328, 266
Gledhill, T. M., & Yates, J. A. 2003, MNRAS, 343, 880
Hines, D. C., Schmidt, G. D., & Schneider, G. 2000, PASP, 112, 983
Hony, S., Tielens, A. G. G. M., Waters, L. B. F. M., & de Koter, A. 2003,
A&A, 402, 211
Hrivnak, B. J., & Kwok, S. 1999, ApJ, 513, 869
Hrivnak, B. J., & Reddy, B. E. 2003, ApJ, 590, 1049
Hrivnak, B. J., Kwok, S., & Su, K. Y. L. 1999, ApJ, 524, 849
Iben, I. Jr. 1981, ApJ, 246, 278
Iben, I. Jr. 1995, Physics Report, 250, 2
Iben, I. Jr., &Renzini, A. 1983, ARA&A, 21, 271
Jura, M., Chen, C., & Werner, M. W. 2000, ApJ, 544, L141
Koekemoer, A. M., et al. 2002, “HST Dither Handbook”, Version 2.0
(Baltimore: STScI)
Kwok, S. 1993, ARA&A, 31, 63.
Kwok, S., Su, K. Y. L., & Hrivnak, B. J. 1998, ApJ, 501, L117
Kwok, S., Volk, K., & Hrivnak, B. J. 2002, ApJ, 573, 720
Loup, C., Forveille, T., Nyman, L. Å, & Omont, A. 1990, A&A, 227, L29
Magain, P., Courbin, F. and Sohy, S. 1998, ApJ, 494, 472
Malhotra, S., et al. 2002, “NICMOS Instrument Handbook”, Version 5.0
(Baltimore: STScI)
Mastrodemos, N. & Morris, M. 1999, ApJ,523, 357
Matt, S., Balick, B., Winglee, R., & Goodson, A. 2000, ApJ, 545, 965
Meixner, M., Ueta, Bobrowsky, M., & Speck, A. K. 2002, ApJ, 571, 936
Meixner, M., Ueta, T., Dayal, A., Hora, J. L., Fazio, G., Hrivnak, B. J.,
Skinner, C. J., Hoffmann, W. F., & Deutsch, L. K. 1999, ApJS, 122, 221
Meixner, M., Skinner, C. J., Graham, J. R., Keto, E., Jerrigen, J. G., & Arens,
J. F. 1997, ApJ, 482, 897
Meixner, M., Zalucha, A., Ueta, T., Fong, D., & Justtanont, K. 2004, ApJ,
accepted
Omont, A., Loup, C., Forveille, T., te Lintel Hekkert, P., Habing, H., &
Sivagnanam, P. 1993, A&A, 267, 515
Reddy, B. E., Bakker, E. J., & Hrivnak, B. J. 1999, ApJ, 524, 831
Renzini, A. 1981, in Physical Processes in Red Giants, eds. I. Iben Jr. & A.
Renzini (Dordrecht: Reidel), 431
Sahai, R. 1999, ApJ, 524. L125
Schröder„ K.-P., Winters, J. M., & Sedlmayr, E. 1999, A&A, 349, 898
Soker, N. 1998, MNRAS, 299, 1242
Sparks, W. B., & Axon, D. J. 1999, PASP, 111, 1298
Steffen, M., & Schönberner, D. 2000, A&A, 357, 180
Steffen, M., Szczerba, R., & Schönberner, D. 1998, A&A, 337, 149
Su, K. Y. L., Hrivnak, B. J., & Kwok, S. 2001, AJ, 122, 1525
Su, K. Y. L., Hrivnak, B. J., Kwok, S., & Sahai, R. 2003, AJ, 126, 848
HST/NICMOS Imaging Polarimetry of PPNs 17
Ueta, T., & Meixner, M. 2003, ApJ, 586, 1338
Ueta, T., Meixner, M., & Bobrowsky, M. 2000, ApJ, 528, 861
Ueta, T., Meixner, M., Hinz, P. M., Hoffmann, W. F., Brandner, W., Dayal,
A., Deutsch, L. K., Fazio, G. G., & Hora, J. L. 2001, ApJ, 557, 831
Ueta, T., Meixner, M., Moser, D. E., Pyzowski, L. A., & Davis, J. S. 2003,
AJ, 125, 2227
Ueta, T., de Vlieger, M., Van de Steene, G. C., Szczerba, R., & Van Winckel,
H. 2004, A&A, submitted
Van de Steene, G. C., & van Hoof, P. A. M. 2003, A&A, 406, 773
Van Winckel, H. 2003, ARA&A, 41, 391
Whitney, B. A., & Hartmenn, L. 1993, ApJ, 402, 605
