The Search for Distant Objects in the Solar System Using Spacewatch - Astronomical Journal

THE SEARCH FOR DISTANT OBJECTS IN THE SOLAR SYSTEM USING SPACEWATCH
Jeffrey A. Larsen, Eric S. Roe,1
and C. Elise Albert
Physics Department, US Naval Academy, Annapolis, MD, USA; larsen@usna.edu, m065796@usna.edu, albert@usna.edu
Anne S. Descour2
and Robert S. McMillan
Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ, USA; adescour@lpl.arizona.edu, bob@lpl.arizona.edu
Arianna E. Gleason
Lawrence Berkeley National Laboratory, Berkeley, CA, USA
Robert Jedicke
Institute for Astronomy, University of Hawaii, Honolulu, HI, USA
and
Miwa Block, Terrence H. Bressi, Kim C. Cochran, Tom Gehrels, Joseph L. Montani,
Marcus L. Perry, Michael T. Read, James V. Scotti, and Andrew F. Tubbiolo
Lunar and Planetary Laboratory, Tucson, AZ, USA
Received 2006 May 12; accepted 2006 November 13
ABSTRACT
We have completed a low-inclination ecliptic survey for distant and slow-moving bright objects in the outer solar
system. This survey used data taken over 34 months by the University of Arizona’s Spacewatch Project based at
Steward Observatory, Kitt Peak. Spacewatch revisits the same sky area every three to seven nights in order to track
cohorts of main-belt asteroids. This survey used a multiple-night detection scheme to extend our rate sensitivity to as
low as 0.012 arcsec hrÀ1
. When combined with our plate scale and flux sensitivity (V % 21), this survey was sensitive
to Mars-sized objects out to 300 AU and Jupiter-sized planets out to 1200 AU. The survey covered approximately
8000 deg2
of raw sky, mostly within 10
of the ecliptic but away from the Galactic center. An automated motion-
detection program was modified for this multinight search and processed approximately 2 terabytes of imagery into mo-
tion candidates. This survey discovered 2003 MW12, currently the tenth largest classical Kuiper Belt object. In addition,
several known large Kuiper Belt objects and Centaurs were detected, and the detections were used with a model of our
observational biases to make population estimates as a check on our survey efficiency. We found no large objects at low
inclinations despite having sufficient sensitivity in both flux and rate to see them out as far as 1200 AU. For low in-
clinations, we can rule out more than one to two Pluto-sized objects out to 100 AU and one to two Mars-sized objects to
200 AU.
Key words: Kuiper Belt — minor planets, asteroids — solar system: formation — surveys
Online material: machine-readable table
1. INTRODUCTION
The announcement of several very large solar system objects
in the summer of 2005 ([136199] Eris, 2005 FY9, 2003 EL61, and
2004 XR190) raised considerable interest in the Kuiper Belt. From
its relatively simple theoretical origins (Edgeworth 1949; Kuiper
1951), the Kuiper Belt displays a rich orbital structure that has us
asking new and exciting questions about the formation and early
history of our solar system. One of the most intriguing features of
the Kuiper Belt is the so-called Kuiper Cliff, where the observed
number of classical (low-eccentricity and low-inclination) Kuiper
Belt objects (KBOs) with semimajor axes greater than 50 AU rap-
idly falls to zero (Dones 1997; Jewitt et al. 1998; Trujillo Brown
2001; Allen et al. 2001; Gladman et al. 2001; Petit et al. 2006).
Despite numerous searches, no low-eccentricity objects on the
far side of the gap created by the Kuiper Cliff have been detected
with low inclinations. Some of these searches for smaller objects
have used quite powerful telescopes with rate sensitivity out to
hundreds of AU (Allen et al. 2002; Bernstein et al. 2004). De-
spite their amazing depth (objects as small as 37 km were detect-
able to a distance of 60 AU), classical-type objects on the far side
of the gap have simply not been detected. On the other hand, the
higher inclination and eccentricity scattered disk objects (Brown
2001; Gladman et al. 2002) have been routinely discovered past
the edge of the Kuiper Cliff. Even discoveries like 2004 XR190
(Allen et al. 2006), which was detected within a degree of the
ecliptic plane, has a high inclination. Most objects of appreciable
size are scattered disk objects with higher inclinations.
The origin of the Kuiper Cliff has inspired many theoretical
models that explain with varying degrees of success the rich or-
bital structure we observe. One explanation for the Kuiper Cliff
is that it arises as the consequence of resonance sculpting by a
Mars-sized body just outside the cliff area (Brunini Melita
2002). This proposal was later shown to have several difficulties
inexplaining other orbital characteristics of the Kuiper Belt (Melita
et al. 2004). The meme of larger planets in the outer solar system
continues in other searches for companions to the Sun that may
be more distant than the Brunini Melita candidate. Timing data
from recent accurate astronomical clocks can rule out Jupiter-
sized objects as far as 200 AU. The observed orbital structure might
be due to one or more stellar passages (Ida et al. 2000). Despite
early problems with models describing the very sharp edge, newer
A
1
Current address: Naval Postgraduate School, Monterey, CA, USA.
2
Current address: Arizona Genomics Computational Laboratory, University
of Arizona, Tucson, AZ, USA.
1247
The Astronomical Journal, 133:1247Y1270, 2007 April
# 2007. The American Astronomical Society. All rights reserved. Printed in U.S.A.

models are finding more success (Melita et al. 2005). Kenyon
Bromley (2004) can explain stellar passages stirring up the disk
and creating orbits like (90377) Sedna. The distribution may not
be a product of a single encounter, as numerous stellar passes
might have occurred if the Sun formed in a star cluster (Brown
et al. 2004; Fernańdez Brunini 2000). Stellar encounters may
have stolen planetesimals from the other star system and left them
as our scattered disk population (Morbidelli Levison 2004).
Finally, as discussed by Levison et al. (2004) the proximity of
the Kuiper Cliff to the 1:2 resonance with Neptune may also be
an explanation. They showed that close stellar encounters most
likely did not occur after the formation of the large KBOs, and so
the observed orbit distributions are not explained by its passage,
leaving a collisional cascade as a possible explanation. In this
scenario, the dynamically cold Kuiper Belt could not have formed
large objects from the observed densities of material present.
Levison Morbidelli (2003) argued that the entire belt formed
closer to the Sun and subsequently migrated outward. This mi-
gration stopped (Gomes et al. 2004) by reaching the outer edge
of the protoplanetary disk after ejecting much of its mass. Plan-
etary embryos as large as Earth or Mars outside the initial loca-
tion of Neptune are also ejected by this process. In any event,
Morbidelli et al. (2002) argue that the Kuiper Belt could not
have formed by planetary embryos still in residence, since by
now they should have been detected, given the large number of
searches.
How the Kuiper Belt formed is also a currently debated issue.
Given the relative paucity of material, the current classical Kuiper
Belt cannot have formed from the amount of material present; it
had to be larger in the past for this to happen. One set of formation
models (summarized in Kenyon [2002]) has a denser disk, where
material depletion occurs due to gravitational stirring and a colli-
sional cascade that turned many bodies into dust or removed them
directly. Levison Morbidelli (2003) argued that there are limita-
tions to these models, and accretion cannot have occurred at high
inclinations and eccentricities because of the peculiar velocities
involved; thus, the entire Kuiper Belt formed close to the Sun and
then migrated outward. The scattered disk might be the method
of interaction between Neptune and the Oort cloud (Fernańdez
et al. 2004). Trying to explain Sedna’s distant, eccentric orbit,
Stern (2005) proposed accretion in the very distant solar system
between 75and 100 AU from the Sun, following an encounter that
sent it inward. This theory makes it interesting to find out whether
or not there are large bodies on nearly circular orbits with q
75 AU that have not been excited and hence are primordial.
In any event, studies of other solar systems tempt us with hints
that the outer solar system may have more structure than we can
currently see. Over 40% of stars in the Trapezium cluster have
Kuiper Belt disks larger than 50 AU (Vicente Alves 2005).
While the truncation in our own disk may well be simply the con-
sequence of stellar formation through photoevaporation, colli-
sions, or gravitational sculpting, other evidence exists for systems
with fairly large gaps in their disks (Greaves et al. 2005), which
would be a strong motivator to continue attempts to detect aster-
oids on the other side of the gap.
M. Brown and C. Trujillo have performed some truly stunning
survey work, covering major fractions of the outer solar system
within 10
of theeclipticto MR ¼ 20Y21 (Trujillo Brown2003).
Their survey uses data with an interimage time of several hours
and has successfully detected large objects out to almost 100 AU
and well within the flux limit of the survey (Brown et al. 2004,
2005). With one exception, these have been at relatively high in-
clinations. If we wish to test some of the theories being proposed
for the outer solar system, it is worth expanding the rate sensitiv-
ity of the search to the even slower rates indicative of more dis-
tant objects.
In this paper we present the results of a search of the same region
of sky, but instead we use image time intervals spanning three to
five nights so that we can detect motions as small as 0.01200
Y
0.02800
hrÀ1
. While this area of the ecliptic has already been cov-
ered by previous surveys, local conditions and field crowding can
cause objects to be overlooked. This reduction was performed
without requiring any special observations by Spacewatch and in
terms of resources is basically a ‘‘free’’ project. In addition, Space-
watch goes to similar limiting magnitudes as the works of Brown
and Trujillo. While we propose to search for very distant objects at
relatively shallow magnitudes, it is not unreasonable to do so.
Large objects in the outer solar system seem to have albedos that
increase substantially with size (Noll et al. 2004a, 2004b; Grundy
et al. 2005; Lykawka Mukai 2005) and with an atmosphere may
even reflect a greater portion of incident sunlight.
2. OBSERVATIONS AND SURVEY METHODOLOGY
Our data were collected at the 0.9 meter Spacewatch telescope
(IAU observatory code 691) at the Steward Observatory on Kitt
Peak in Arizona as part of its normal near-Earth asteroid search.
Spacewatch is a group at the University of Arizona’s Lunar and
Planetary Laboratory founded by T. G. and R. S. M. in 1980. The
primary goal of Spacewatch is to explore the various populations
of small objects in the solar system, and study the statistics of as-
teroids and comets in order to investigate the dynamical evolu-
tion of the solar system. CCD scanning studies have been made
of the Centaur (Jedicke Herron 1997), main-belt (Jedicke
Metcalfe 1998), trans-Neptunian (Larsen et al. 2001), and Earth-
approaching asteroid populations (Rabinowitz 1991; Jedicke
1996; Bottke et al. 2002; Jedicke et al. 2003; J. A. Larsen et al.
2007, in preparation). Spacewatch also finds potential targets for
interplanetary spacecraft missions and radar observations (Ostro
et al. 2003), provides follow-up astrometry of such targets, and
finds and follows objects that might present a hazard to Earth.
Fig. 1.—Scale drawing of the focal plane of the 0.9 m mosaic camera, illustrat-
ing the layout of the four E2V CCDs. Gaps between the CCDs are approximately
7000
. For comparison, the previous 2K ; 2K detector used by Spacewatch at the
0.9 m is shown as a region enclosed within a dashed line.
LARSEN ET AL.1248

Fig. 2.—Sample image from the Spacewatch mosaic camera. Inter-CCD gaps are shown to scale.
TABLE 1
Entries for the Pointing History of the Survey
Reduction Name Night N a
Overlapb
(deg2
)
R.A.
(J2000.0)
Decl.
(J2000.0) UT Date UTc
Time Matched Objects
FWHM
(arcsec) Observer
2003.05.10.14.06.............. 1 76 2.8082 14 25 11.2 À14 00 08 2003 May 3 06:37:10.6 30,915 2.5 J. A. Larsen
2 14 25 09.8 À13 59 53 2003 May 10 06:22:59.3 27,252 1.8 R. S. McMillan
2003.05.10.71.02.............. 1 290 2.8635 16 25 12.8 +08 31 48 2003 May 5 07:38:22.4 49,492 2.6 J. A. Larsen
2 16 25 11.4 +08 31 56 2003 May 10 09:02:35.1 41,143 1.7 R. S. McMillan
2003.05.10.71.09.............. 1 220 2.8501 16 17 49.1 +01 35 54 2003 May 5 08:06:52.3 50,807 2.6 J. A. Larsen
2 16 17 46.8 +01 36 05 2003 May 10 09:29:08.6 46,952 1.5 R. S. McMillan
2003.05.11.80.01.............. 1 59 2.8815 17 02 11.6 +06 47 50 2003 May 5 09:32:00.7 71,429 2.3 J. A. Larsen
2 17 02 12.3 +06 47 54 2003 May 11 09:28:46.0 60,335 2.7 R. S. McMillan
2003.05.11.80.02.............. 1 50 2.8757 17 09 35.7 +06 47 50 2003 May 5 09:35:42.8 79,975 2.4 J. A. Larsen
2 17 09 36.3 +06 47 54 2003 May 11 09:32:29.6 71,066 2.3 R. S. McMillan
2003.05.11.80.03.............. 1 35 2.8755 17 02 11.7 +05 03 52 2003 May 5 09:39:26.3 77,277 2.3 J. A. Larsen
2 17 02 12.3 +05 03 57 2003 May 11 09:36:16.7 68,946 2.5 R. S. McMillan
2003.05.11.80.04.............. 1 120 2.8637 17 09 35.8 +05 03 52 2003 May 5 09:43:08.6 89,165 2.2 J. A. Larsen
2 17 09 36.4 +05 03 57 2003 May 11 09:40:01.5 77,327 2.0 R. S. McMillan
2003.05.11.80.05.............. 1 144 2.8767 17 02 11.8 +03 19 52 2003 May 5 09:47:27.6 85,876 2.2 J. A. Larsen
2 17 02 12.0 +03 19 58 2003 May 11 09:44:35.6 82,032 1.8 R. S. McMillan
2003.05.11.80.07.............. 1 192 2.8767 17 02 11.9 +01 35 54 2003 May 5 09:55:16.2 88,383 2.2 J. A. Larsen
2 17 02 12.0 +01 36 01 2003 May 11 09:52:28.5 87,928 1.7 R. S. McMillan
2003.05.11.80.08.............. 1 255 2.8606 17 09 34.7 +01 35 55 2003 May 5 10:55:37.5 97,724 1.7 J. A. Larsen
2 17 09 36.1 +01 36 00 2003 May 11 09:56:12.9 92,498 1.7 R. S. McMillan
Note.—Units of right ascension are hours, minutes, and seconds, and units of declination are degrees, arcminutes, and arcseconds. Table 1 is published in its entirety
(all 3930 regions) in the electronic edition of the Astronomical Journal. A portion is shown here for guidance regarding its form and content.
a
Number of candidate motions generated by SLOSUR.
b
The offset between the two nights as determined by integrating over the image WCS solutions.
c
Time listed is for shutter open. The mid-exposure time is 60 s later.

The Spacewatch mosaic camera (McMillan et al. 2000) re-
placed our earlier drift-scan system (McMillan et al. 1986; Gehrels
et al. 1986; Rabinowitz 1991; Jedicke 1996; Larsen et al. 2001)
on the same telescope and fills a niche in limiting-magnitude
and sky coverage that is unique among all other asteroid surveys,
which typically have a brighter limiting magnitude. The Space-
watch 0.9 m telescope is a corrected prime-focus f/3 system. The
detector uses a Schott OG-515 filter that passes light with wave-
lengths longer than 515 nm up through the wavelength limit of the
CCD, 950 nm. The camera has an effective wavelength of 700 nm
for solar system objects. The mirror has a clear aperture of 0.9 m,
and the image scale at corrected focus is 74.000
mmÀ1
. The
mosaic camera has four thinned and backside-illuminated E2V
Technologies Model CCD42-90-I-941 CCDs, each of 4608 ;
2048 pixels. The pixel size is 13.5 m, leading to an image scale
of 1.0000
pixelÀ1
and an effective field of view of 2.9 deg2. Given
a 120 s tracked exposure under good conditions, the limiting mag-
nitude (50% automated detection) of the 0.9 m mosaic for main-
belt asteroids is a visual magnitude of 21.7 (assuming solar-
colored objects). The layout of the CCDs in the optical plane is
illustrated in Figure 1 and represents a factor of 9 increase in
collecting area over the drift-scanning system previously used.
Each image collected by the system is 81 megabytes in size and
has 37 million pixels. A representative image is shown in
Figure 2.
In 1995, R. J. suggested that Spacewatch embark on a ‘‘revisit’’
strategy in order to extend the arcs of main-belt asteroids. While
this move does not change the near-Earth asteroid detection rates
for appropriately long times between revisits, it does allow a sub-
stantial fraction of main-belt asteroids to be reobserved for further
study. As a result, Spacewatch periodically (every 3Y7 days) revis-
its the same regions of sky as the closer main-belt and near-Earth
Fig. 3.—Relative positions and sizes of the region surveyed by this work on
the J2000.0 sky. The ecliptic is denoted by the red lines with Æ10
boundaries in
orange. The plane of the Milky Way is plotted in cyan, and each of the 3930 re-
gions processed are plotted by to-scale green squares. The star density toward the
Galactic center increases uncomfortably in summer, hence the gap in coverage be-
tween the Milky Way and ecliptic planes. A second region of sparse coverage is at
23h
, corresponding to monsoon season in Tucson.
Fig. 4.—Distribution of the average number between nights of matched ob-
jects in the survey regions. While most of the regions are at relatively high Ga-
lactic latitudes and low star densities, a nontrivial number of objects end up in
more cluttered regions, which can lead to higher false-candidate levels.
Fig. 5.—Comparison of the FWHM of star profiles (seeing) between nights.
Since this survey uses observations between two nights, the limiting sensitivity
of the survey is controlled by the seeing on the worse of the two. Interestingly,
very few fields have large FWHM on both nights of observation.
Fig. 6.—Distribution of the worse FWHM of star profiles (seeing) between
the paired survey regions. These numbers represent a wide variety of observing
conditions, an important factor when considering limiting magnitude.
LARSEN ET AL.1250 Vol. 133

asteroids move out of them. Since distant and slow-moving ob-
jects would more than likely stay in the same survey region as
well, the several-night time-baselinefor detectiongreatly enhances
our slow-rate sensitivity. Until late 2004, revisits were taken on
the same pointing centers as the original night’s observations.
After this time, however, Spacewatch followed J. A. L.’s sug-
gestion that the pointing centers drift along with the mean motion
of the main-belt objects in order to maximize their yield. This
reduced the amount of usable sky for slow movers and must be
considered in the analysis.
In the normal operation mode of the telescope, three images
are taken of the same region of sky over a 40Y60 minute interval.
From these images, main-belt and near-Earth asteroids are de-
tected. For this distant survey, we have loosened the criteria re-
quired to match an object between images (so that slow-moving
objects will be detected as an unmoving object) and then searched
for motions of apparently stationary objects seen in a position on
one night but seen in a different position during a later revisit. As
such, each region we have processed involves six images taken
over two nights. We have processed 3930 regions in the course of
the survey for a grand total of 23,580 separate images (or 1.9 tera-
bytes of mosaic image data).
All Spacewatch images were stored on DVD-Rs and hard
drives and were periodically copied and transported to a lab at the
US Naval Academy for this project. The US Naval Academy lab
consists of seven computers running Fedora Core 4 Linux (kernel
ver. 2.6.11-1.1369_FC4), each equipped with an Intel Model 630
Pentium 4 3.0 GHz EM64T processor, 2 gigabytes of DDR2 RAM,
a 150 gigabyte SATA system hard drive, and three 400 gigabyte
removable hard drive bays, yielding a net storage capacity of
1.2 terabytes per computer. All six computers were cross-connected
on an dedicated 1000 Base-T network.
Table 1 gives the pointing history for the survey. For each night,
the table presents the image information for the middle image ac-
quired by Spacewatch. The primary quantities presented in the
table include a region designation, the number of motion can-
didates produced by the motion-detection software, the overlap
between the six passes in square degrees (determined by numer-
ically calculating the overlap between the World Coordinate Sys-
tem [WCS] solutions), the sky coordinates and observation times
for each middle image of the three taken that night, the number of
matched objects detected between the three images on the night,
the average full width at half-maximum (FWHM) of the stellar
proﬁle between the images, and the observer identity.
The relative sky coverage of this survey is shown in Figure 3.
The main points to notice are that the survey avoids the Galactic
center in summer, has sparse coverage during monsoon season,
and otherwise is relatively tightly constrained within just a few
degrees of the ecliptic. In the summer of 2004 Spacewatch de-
cided to survey closer to the fundamental plane of the ecliptic as
well; hence, our distant planet survey becomes biased against
high-inclination objects. The properties of the data are further
explored in Figure 4, which shows the net distribution of regions
by average number of matched objects between nights. Figure 5
compares the stellar proﬁle (FWHM) between paired nights
of observations. Because the worse FWHM of the two nights
Fig. 7.—Layout of capsule review candidates. A candidate sits at a different position on each night. Each night has three separate pictures that can be extracted from
it. Position 1 is shown in the top row and position 2 in the bottom row. Avalid candidate would be seen in position 1 on the three passes of night 1 only and position 2 on
the three passes of night 2 only.
Fig. 8.—Example of a good candidate motion: (20000) Varuna, observed 2005 November 30 at V ¼ 20:1 mag. Compare with the layout in Fig. 7.
SPACEWATCH OUTER SOLAR SYSTEM SURVEY 1251No. 4, 2007

controls how effective the region is processed, the distribution
of the survey regions in terms of the worse FWHM is tabulated
in Figure 6.
3. SOFTWARE
Following Spacewatch’s pioneering development of a practical
real-time moving-object detection program (Rabinowitz 1991),
J. A. L. has developed several generations of detection software
for the Spacewatch project. These programs have been used for
drift scanning at the 0.9 and 1.8 m telescopes, as well as on the
more conventional tracked images now generated by the 0.9 m mo-
saic camera. Slow-moving object-detection programs were also
written for Spacewatch during the drift-scanning epoch (Twitch,
described in Larsen et al. [2001]). Avery nice discussion of the is-
sues relevant to slow-moving object software can be found in Petit
et al. (2004).
The latest version of the software used by the Spacewatch
mosaic is the combination of two programs that run on a small
cluster of computers at the telescope site: MOSAF (MOSaic
Astrometry Finder) and MOSSUR (MOSaic SURvey). These
programs still use a catalog-based search, which is fairly efficient
in terms of execution time versus candidates found.MOSAF takes
raw mosaic pixel data, performs all necessary flat-field, dark, bias,
and fringe corrections, creates an object catalog of all detections in
a manner very similar to SExtractor (Bertin Arnouts 1996), and
finally creates raw and processed astrometrically calibrated MEF
FITS images using the cfitsio libraries of Pence (1999), the
WCSLIB libraries of Mink (2002), and the USNO-A2.0 astrometry
catalog.3
The created catalogs contain many image parameters,
such as the shape, position, flux, moments, and the parameters of
a simple ellipse fit. MOSAF is customized to deal with the Space-
watch imaging system and is integrated into the image-creation
pipeline. MOSSUR uses the object catalogs created by MOSAF
to search for moving objects and create a nearYreal-time valida-
tion review for the observer at the telescope. This system has been
operational for 3 years and will be described in an upcoming paper
on the mosaic’s near-Earth asteroid results.
SLOSUR (SLOw SURvey) is a MOSSUR variant written in C
in the Linux environment that runs as a postprocessing step on
archival data. It takes two nights of MOSAF catalogs and images
and finds the stationary (matched) objects on each night (which
were foundin all three passes). Sincewe use a verygenerous match
window (radius 300
), slow-moving objects are counted as stationary
objects, while more rapid main-belt and near-Earth asteroids re-
main unmatched. The matched objects are then matched between
the two nights and a list of ‘‘unmatched matched’’ objects are
created, which appeared stationary on one night but were absent
on the other night. These matched objects are compared between
nights to find slow moving candidates with sky plane rates be-
tween 3 pixels and 0.03
dayÀ1
, less than a 2 mag brightness
Fig. 9.—Example of a false positive motion. The combination of a spike off a bright star and pointing error eliminates one candidate as a match, while the other
is borderline in signal-to-noise ratio.
Fig. 10.—Example of another false candidate motion. Borderline weather conditions combined with image fragmentation in a bright star.
3
VizieR Online Data Catalog, 1252, 0 (D. B. A. Monet et al., 1998).

difference, valid pixel locations seen on all six images, moving
along the ecliptic in retrograde motion at an angle to the ecliptic
of less than 30
, and whose net signal-to-noise ratio was 3Y5
(variable with conditions).
Valid moving object candidates had 12 postage-stamp-sized
snapshots generated from the relevant images (six images of two
positions, as depicted in the layout of Fig. 7) and were placed in
an encapsulated review that could be downloaded to a reviewer’s
workstation or laptop for easy access. Each candidate had its image
and motion parameters embedded in its FITS header. An ex-
ample of a good (real) candidate motion is shown in Figure 8,
while two bad (false-positive) motions are presented in Figures 9
and 10. Because we use a list-based detection search and stel-
lar spikes can fragment easily, they form the bulk of the false de-
tections. Objects similar to these were subsequently manually
rejected by human reviewers.
4. PROCESSING AND PRELIMINARY RESULTS
All 3930 regions (revisited images between 2003 March and
2006 March) shown in Table 1 were processed by SLOSUR and
reviewed by human reviewers. These 1.92 terabytes of raw data
resulted in 1.37 billion objects detected, which were reduced to
only objects that matched on a single night but which appeared to
be moving between nights. A total of 434,996 candidates needed
visual validation (a process taking 1Y2 s per image on a laptop/
graphical workstation). Of these candidates, 668 were deemed by
the reviewer to be worthy of further study. The breakdown of can-
didates is presented in Table 2. From the 668 objects, 17 were real
and in some cases were multiple reimages of the same object
caught in multiple repetitions of the same images.
The small bug mentioned in Table 1 refers to a problem with
2003 MarchYApril Spacewatch images in which the astrometry
solution of the vertical CCD in Figure 1 would have a small (300
)
systematic error for the right ascension in the corners of the im-
ages. While this bug was rapidly repaired, the archived images still
had the error and it resulted in some stars being falsely reported
as unmatched between nights. The nature of the bug added false
candidates and an extra review burden but would not have re-
moved any actual moving objects. After several reviews, we
reprocessed all astrometry by default, and these kinds of false
candidates disappeared from all subsequent reviews.
As can be seen from Table 1, a raw sky coverage of 7790.6 deg2
was processed. This number is smaller than the 10,600
of solid
angle covered by the telescopes because of the moving field cen-
ters discussed in x 2. Also, due to another change in the Space-
watch survey strategy in 2004, fields are much closer to the
ecliptic plane than was originally planned and repeat more often.
Given a region of 14,400 deg2
between Æ10
ecliptic latitude our
raw sky coverage is approximately 55% of the total available,
neglecting other effects.
Our survey commenced with images taken in 2003 March.
Over the course of the survey, we discoveredonly a single new ob-
ject, 2003 MW12 (H ¼ 3:8, a ¼ 45:9 AU, e ¼ 0:137, i ¼ 21:5
),
which was in our data almost from the beginning of the survey
(2003 May). The discovery images for the object are shown in Fig-
ure 11. For real objects and to aid in searching images for promising
TABLE 2
Breakdown of SLOSUR Candidates That Required Further Analysis
Number Status of Detection
79..................... Weather/focus effects caused some objects to not be matched between nights.
359................... Strange artifacts from the CCD/flat-fielding effects, spikes.
49..................... Objects planted in the reviews to calculate observer efficiency.
17..................... Detections and redetections of outer solar system objects.
137................... One or both candidates were asteroids at their stationary point.a
11..................... Objects are candidates because of a small bug in the astrometry-matching code in SLOSUR (see text).
16..................... Single-position objects that had no valid paired object at the second position. Manual searches could not find any matching valid candidate
position within the survey motion limit. As such, we concluded these represented transient events.
a
While we surveyed at stationary points for main-belt asteroids, it should be noted that phase effects in the outer solar system are still minimal at these elongations
from opposition.
Fig. 11.—Discovery raw images of 2003 MW12, V ¼ 20:7, from 2003 May 23.

TABLE 3
Objects Detected and Expected Objects in the Survey
Date Asteroid Name R.A. Decl. Predicted Magnitude
k Rate
(deg dayÀ1
)

Rate
(deg dayÀ1
)
Delta
(AU)
2003 May 17................ 2003 MW12 16.3130 À2.064 20.7 À0.017 0.004 47.56
2003 May 23................ 2003 MW12
a
16.3119 À2.060 20.7 À0.017 0.004 47.56
2003 Jun 21 ................. 2003 MW12
a
16.2796 À1.995 20.7 À0.015 0.000 47.65
2003 Jul 7 .................... 2003 MW12 16.2645 À2.004 20.7 À0.013 À0.002 47.79
2003 Oct 24 ................. (79360) 1997 CS29 8.4141 19.356 21.6 0.005 À0.001 43.60
2003 Dec 22 ................ (79360) 1997 CS29 8.3875 19.440 21.5 À0.016 0.004 42.73
2004 Apr 13................. (26181) 1996 GQ21 14.4632 À8.892 21.2 À0.018 0.008 38.90
2004 Apr 19................. 2003 MW12 16.4172 À2.236 20.7 À0.013 0.007 47.63
2004 May 9.................. 2002 GH166 12.9049 À6.689 21.6 À0.018 0.007 31.82
2004 Jun 11 ................. 2002 GP32 15.1184 À15.931 21.8 À0.020 0.005 31.18
2004 Jun 12 ................. 2003 MW12
a
16.3593 À1.996 20.7 À0.016 0.001 47.48
2004 Jun 17 ................. 2003 MW12
a,b
16.3538 À1.991 20.7 À0.016 0.001 47.51
2004 Jun 17 ................. 2000 KK4
a,b
16.4557 À2.223 22.5 À0.017 0.003 43.48
2004 Oct 4 ................... (55637) 2002 UX25 1.6632 9.743 19.9 À0.018 À0.008 41.44
2004 Oct 7 ................... (84719) 2002 VR128 2.2698 17.352 21.3 À0.022 À0.005 35.06
2004 Oct 7a
.................. (48639) 1995 TL8 2.2184 13.572 21.8 À0.017 À0.006 42.09
2004 Oct 8 ................... (84522) 2002 TC302 1.7968 16.082 20.7 À0.018 À0.004 46.85
2004 Oct 8 ................... (55637) 2002 UX25 1.6584 9.710 19.9 À0.018 À0.008 41.43
2004 Oct 10 ................. 2004 VT75 1.3184 6.847 21.5 À0.022 À0.008 35.44
2004 Oct 11 ................. (84719) 2002 VR128 2.2639 17.330 21.3 À0.022 À0.006 35.04
2004 Oct 11 ................. (48639) 1995 TL8 2.2136 13.548 21.7 À0.018 À0.006 42.07
2005 Mar 13 ................ (82155) 2001 FZ173 12.4587 À4.252 21.4 À0.022 0.008 31.83
2005 Apr 12................. (26181) 1996 GQ21 14.5867 À9.104 21.2 À0.018 0.008 39.13
2005 Apr 12................. 2003 GH55 14.2018 À12.111 21.8 À0.019 0.007 39.74
2005 Apr 12................. 2004 PR107 14.2005 À11.933 21.8 À0.016 0.005 51.53
2005 May 4.................. 2003 MW12
a
16.4744 À2.135 20.7 À0.015 0.006 47.41
2005 May 6.................. 2003 MW12
a
16.4724 À2.124 20.7 À0.016 0.006 47.40
2005 May 10................ 2003 MW12
a
16.4681 À2.103 20.7 À0.016 0.005 47.38
2005 Jun 6 ................... 2002 GP32 15.2898 À16.629 21.8 À0.022 0.005 31.10
2005 Jun 11 ................. 2002 GP32 15.2827 À16.603 21.8 À0.021 0.005 31.13
2005 Jul 5 .................... 2001 QY297 21.0993 À17.301 21.8 À0.016 À0.005 41.94
2005 Aug 28................ 2003 QW90 0.2074 À1.930 20.9 À0.015 À0.007 43.41
2005 Aug 28................ 2001 QG298 0.1127 À1.472 21.8 À0.020 À0.010 30.94
2005 Sep 23................. (26308) 1998 SM165 1.5670 4.722 21.4 À0.018 À0.009 35.39
2005 Sep 27................. (26308) 1998 SM165 1.5619 4.684 21.4 À0.019 À0.010 35.37
2005 Oct 22 ................. (19521) Chaos 4.2843 24.782 21.2 À0.017 À0.001 41.18
2005 Oct 25 ................. (42301) 2001 UR163 1.4991 10.235 21.1 À0.016 À0.006 48.78
2005 Oct 25 ................. (26308) 1998 SM165 1.5233 4.419 21.4 À0.021 À0.009 35.36
2005 Oct 27 ................. (48639) 1995 TL8 2.2854 13.910 21.7 À0.019 À0.006 42.21
2005 Jan 27.................. (19521) Chaosa
4.2784 24.776 21.2 À0.018 À0.001 41.13
2005 Oct 31 ................. (42301) 2001 UR163 1.4926 10.197 21.2 À0.016 À0.006 48.80
2005 Oct 31 ................. (48639) 1995 TL8
a
2.2803 13.884 21.7 À0.019 À0.006 42.21
2005 Oct 31 ................. (15874) 1996 TL66 3.3560 12.400 20.9 À0.021 À0.009 34.24
2005 Nov 4.................. (33340) 1998 VG44
a
4.7958 19.790 21.3 À0.024 À0.003 28.87
2005 Nov 29................ (20000) Varuna 7.2333 25.038 20.1 À0.016 0.004 42.52
2005 Nov 30................ (20000) Varunaa
7.2322 25.042 20.1 À0.016 0.004 42.51
2005 Dec 21 ................ (19521) Chaos 4.2008 24.634 21.2 À0.020 À0.003 41.06
2005 Dec 22 ................ (79360) 1997 CS29 8.5629 18.722 21.5 À0.016 0.004 42.73
2005 Dec 23 ................ (19521) Chaos 4.1982 24.627 21.2 À0.019 À0.003 41.07
2005 Dec 24 ................ (82075) 2000 YW134 8.4410 17.993 21.4 À0.016 0.005 42.62
2005 Dec 26 ................ (79360) 1997 CS29 8.5585 18.737 21.5 À0.017 0.004 42.69
2006 Jan 4.................... (79360) 1997 CS29 8.5478 18.774 21.5 À0.019 0.004 42.62
2006 Jan 4.................... (82075) 2000 YW134 8.4282 18.050 21.4 À0.018 0.006 42.54
2006 Jan 20.................. 2001 CZ31 9.5592 15.266 21.8 À0.018 0.007 40.08
2006 Jan 21.................. (79360) 1997 CS29 8.5258 18.851 21.5 À0.020 0.005 42.56
2006 Jan 21.................. (26181) 1996 GQ21 14.7369 À9.695 21.4 0.010 0.001 40.43
2006 Jan 31.................. 2000 CN105 10.6181 11.724 21.7 À0.016 0.007 45.33
2006 Feb 20................. (82075) 2000 YW134 8.3683 18.319 21.4 À0.017 0.005 42.60
Note.—Units of right ascension are hours, and units of declination are degrees.
a
This object detected in our survey.
b
This detection occurred because of an increased exposure time to reobserve a Spacewatch near-Earth asteroid detection, SW40E6. It is coincidental that 2003 MW12
showed up in the same exposure.

candidates we made extensive use of the Orbfit program by
Bernstein Khushalani (2000). In the case of our sole discovery,
Orbfit provided predictions sufficient to gain recovery on week-,
month-, and 2 yrYlong timescales from our images for the dis-
covery. Later the object was independently redetected six additional
separate times. As part of E. C. S.’s Trident project, he and J. A. L
made a successful follow-upvisit to the Spacewatch1.8 m telescope
in 2006 March in order to confirm the object. It has also been
recovered by another asteroid survey group (Near-Earth Asteroid
Tracking [NEAT]), which found it in images from 2002 May 30
and 2002 June 7 by stacking survey images.
The rest of our bright object detections are given in Table 3
(marked by footnote a). There are 10 detections besides 2003
MW12, although one of them, 2000 KK4, is below the magnitude
limit of the survey and was detected in a series of long expo-
sure time images required for the recovery of another asteroid.
Although the rest of our objects were already detected, they are
intrinsically bright objects and tell us about the limiting mag-
nitude of the survey. The detection of an object such as (26308)
1998 SM165 is significant because, with an absolute magnitude
of 5.8, it has an apparent visual magnitude of 21.4, illustrating
some sensitivity in both parameter spaces. Given these detections,
we can conclude that this survey should be capable of detecting an
object with the magnitude of 2003 MW12 out to 100 AU.
5. EFFICIENCY
It has long been realized (e.g., Jedicke Herron 1997; Trujillo
Jewitt 1998) that motion-detection software must be calibrated
to understand the probability that it can make a particular detec-
tion. This calibration, the efficiency, must be measured under the
full range of operating conditions for the system. This is especially
true for Spacewatch, given frequent observations under light cloud
cover and poor seeing. The overall efficiency of this survey is de-
termined by two factors: the ability of the SLOSUR program to
detect faint objects under a wide range of conditions, and the
reviewer’s efficiency in recognizing them during validation.
To determine the net efficiency of the system, test motions
were generated and objects were planted in the survey imagery
to be detected. Because part of the testing is the human reviewer,
these objects must be indistinguishable from the other objects in
the frame. With that in mind, efficiency determination is relatively
interactive. A series of bright, unsaturated template stars are found
in each of the six images that constitute a given region. Postage-
stamp subimages are made of the star, and these stamps are di-
vided down to represent fainter objects with the same optical
characteristics as the parent image. These images are reintroduced
into the survey images in a pattern consistent with a reasonable
motion. An example of a planted object is shown in Figure 12.
The software calibration and validation process was performed,
and the net efficiency was calculated as the fraction of successful
detections.
Fig. 12.—Example of planting test objects into SLOSUR processed mosaic images. The left image is original raw data; the right image is the same image as the one
on the left, but with an artificial object inserted in the left-center portion of the image. Because the artificial objects were extracted from the same images in which they
were planted it is exceedingly difficult to recognize them as false.
Fig. 13.—Net efficiency surface for this survey in terms of number vs. mag-
nitude. The magnitude distribution for the planted objects is shown as an open
histogram. Objects detected by the software are shown as a hatched histo-
gram, and objects validated by the reviewer as being real are shown as a filled
histogram.
TABLE 4
Final Probability of Detection as a Function of Magnitude
Magnitude Range
Probability
of Detection
Minimum Probability
of Detection
V 19 .................................. 0.92 0.72
19 V 20......................... 0.81 0.20
20 V 21......................... 0.53 0.40
21 V 22......................... 0.11 0.06
SPACEWATCH OUTER SOLAR SYSTEM SURVEY 1255

We selected 30 regions from those available to represent the
full range of seeing, focus, weather, and crowding from back-
ground objects. Images were selected from each region and then
planted back in, irrespective of weather conditions, field crowd-
ing, focus, seeing, and other conditions. The reviewer was un-
aware of exactly which fields contained false objects.
We find that some of the software nondetections occur due to
centroiding effects when the moving object is only traveling 2Y
3 pixels. Unlike surveys for faster moving objects, we find no real
dependence on sky-plane rate after this. Since a motion of 2Y
3 pixels represents the lowest rate that is measurable for our sur-
vey, we use it (1 pixel dayÀ1
for a typical field) to set a lower limit
on our sensitivity.Our net rate sensitivityis between4.500
hrÀ1
down
to 0.012Y0.02800
hrÀ1
. In terms of rate sensitivity, this is a factor of
40Y80 better than existing surveys. It also sets our sensitivity for
object detection as up to 1200 AU for very large objects, deter-
mined solely by the size of the object and the flux received from
it.
Figure 13 shows the results of the efficiency determination. It
is obvious that this curve does not look like the typical efficiency
surface, but that is easily explained (Fig. 5) by varying condi-
tions between two separate nights. Since a good night can be
compared with either a good or bad second night, the efficiency
flattens out at the faint end and does not immediately drop to zero
based on the ratio of signal to noise. While test objects were
planted over all expected rates of motion, efficiency as a function
of rate of motion is not explicitly presented. Objects that were
missed in the efficiency determination showed no rate-based trends.
In addition, there was little reason to expect a strong rate depen-
dence given that the slowest rates of motion searched were on
the order of twice the size of the seeing profiles, minimizing the
chance of a centroiding error causing a loss in detection. For higher
rates of motion, field crowding was the dominant effect, and since
the objects moved several seeing disks (onto effectively indepen-
dent regions of the sky), there were no rate-based trends in detection
efficiency.
Two layers of testing are represented in this figure. The open
portion of the histogram represents the magnitude distribution of
the 154 planted objects. Originally the planted objects were sup-
posed to be 1 mag brighter, but a coding error caused too many
obviously faint images to be planted. The hatched portion of the
histogram represents the number detected in software and the
solid portion was what was caught by the reviewer. A total of
53 objects of the 154 planted were detected by the software and
49 of those were recognized as such by the reviewer. With a
formal error being represented as the minimum probability of de-
tection, the results of Figure 13 are presented in Table 4.
While the efficiency is only roughly determined, we can per-
form a quick check to ascertain whether it is representative of
the survey. We tabulate the detections of Table 3 by magnitude in
Table 5. From this, we apply the detection probability of Table 4
to generate an expected number of objects by magnitude, which
is then compared with the actual detections. In this determina-
tion, we neglect the two detections denoted in Table 3 as having
been exposed longer than the standard exposure time. We find
reasonable agreement between the two and conclude that we can
use this efficiency curve in x 6.
The efficiency has one final use. Figure 14 considers our de-
tection probability as a function of H and distance from the Sun.
We find that a given image has a 90% chance to detect a Mars-
sized object out to 120 AU and a 75% chance to find a Pluto-
sized object at the same distance. At 100 AU, it seems that the
smallest H detectable is between 1 and 2.
6. OBSERVATIONAL BIAS: SURVEY SIMULATIONS
According to Table 1, our sole large discovery, 2003 MW12,
was available multiple times through the survey. Likewise, it is
obvious that other asteroids were missed due to a myriad of rea-
sons, such as not being in a survey field or having moved be-
tween regions and being lost in the ‘‘picket-fence’’ effect. The
problem with this is trying to determine what fraction of the space
was successfully searched for each class of objects, or more appro-
priately, to determine what orbital parameters were biased against
in the measurements. The resulting calculation can be used to es-
timate either the population size or the fraction of the population’s
phase space that has been searched.
A superior introduction to the method and technique of bias
determination can be found in Jedicke et al. (2002), and so this
paper restricts itself to the details of implementation. This con-
cept is not new to outer solar system work, however. Brown
(2001) developed a bias-free technique to study the radial distri-
bution of the Kuiper Belt, and this technique was later extended
by Levison et al. (2004) to probe stellar encounters on its outer
edge.
One of the easiest, most straightforward ways to determine the
bias of an asteroid survey is to simulate the survey using a set of
synthetic elements generated to represent the asteroids and ex-
amine the properties of the ones that were detected. The engine
for the simulator is the program SEARCHMOSAIC written by
J. Scotti and based on the FORTRAN libraries of D. Tholen.
Around the engine J. A. L. wrote a controller that recreates the
time progression and conditions of the survey and converts the
asteroids expected in each region into sorted lists of detection
opportunities for each. It should be noted that we are performing
our simulations using absolute magnitude instead of diameter so
TABLE 5
Application of Detection Probability to Expected and Detected Objects
Magnitude Range
Number
in Field
Number
Expected
Number
Detected
V 19 .................................. 0 0 0
19 V 20......................... 2 1:6 Æ 0:4 0
20 V 21......................... 14 7:4 Æ 1:8 7
21 V 22......................... 40 4:4 Æ 2:0 7
Fig. 14.—Probability of detection in the outer solar system by this survey as a
function of absolute magnitude.

that we can ignore the myriad of pitfalls inherent in the albedo-
diameter conversion.
After collecting all the possible times that a given asteroid could
be detected by the synthetic survey we sort them as a function of
rate, magnitude, location, and observing conditions. The efficiency
function of x 5 is then used to compute the expectation value of
the detection. Using the uncertainties and classic propagation of
error we can very simply determine the relative confidence limits
on the detection. We find that due to multiple repetitions of our
survey regions and generous overlaps, the uncertainties in the ef-
ficiency determination are not a major factor in discovering the
fraction of phase space surveyed.
In the following sections we perform studies of our bias for four
populations. The first two are for the known Plutino and classical
KBO populations, in order to establish the general effectiveness of
our technique.We then move to two unknown populations (that is,
Fig. 15.—Orbit element distribution for classical KBOs and plutinos used to generate our simulated population. Data are from the IAU’s Minor Planet Center Web site.

populations that have no corresponding currently known mem-
bers). These two populations represent the ‘‘Planet X’’ candidate
of Brunini Melita (2002) (too nicely deﬁned to avoid testing)
and a second population whose orbital parameters are the same
as the ﬁrst, but whose semimajor axes increase dramatically to
represent very distant, very large objects that may have formed
with the solar system, as per Stern (2005).
6.1. Observational Bias: Plutino and Classical
KBO Population
To simulate the Plutino and classical KBO populations, we
generated 799,850 sets of orbital elements for H 8 popula-
tions. The generated population was drawn from the actual orbital
element distribution for the population as shown in Figure 15 but
randomized slightly.4
In general, the semimajor axis was allowed
to vary between 32 and 49 AU, the eccentricity was allowed to
vary from 0 to 0.4, the inclination from 0
to 47
, the absolute
magnitude from 0 to 8 mag, and the three orbital angles were
randomized on the range from 0
to 360
. Element sets were
excluded from the simulation if they were very different in the
Fig. 16.—Estimate of bias for H 8 classical KBOs and plutinos. Note that the scale of the error estimate has been expanded by a factor of 5.
4
Data on orbital element distributions were taken from the IAU Minor Planet
Center Web site, available at http://cfa-www.harvard.edu/iau/mpc.html.

entire set of orbital elements from examples in the general
population.
The simulation over the entire survey of Table 1 generated
1,162,175 detection possibilities. The detection lists were sorted
so that all detection opportunities for each individual asteroid
were considered as a unit, and the net conditional probability of
detection with error was computed using the values in Table 4.
The ratios of detected asteroids to the original population (in other
words, the completion estimate) is presented in Figures 16Y18.
The completeness estimate is shown as a binned function of each
orbital parameter versus the semimajor axis. Each plot is accom-
panied by the estimated error on each bin. No completeness es-
timate was uncertain by more than 0.1, owing mostly to the large
number of objects simulated and the tendency of objects to have
multiple chances at detection.
From Figure 16 we can see that our survey was most complete
and independent of semimajor axis for H 4. This complete-
ness was approximately 50%. After this, the completeness drops
quickly as objects get smaller and fainter, diminishing to es-
sentially zero for objects with absolute magnitudes less than
6.5. For the inclination there is an unsurprising bias that il-
lustrates that most of our ﬁelds are more tightly constrained to
the ecliptic than 10
(with the bulk being less than 5
away).
The other four orbital parameters show no or weak trends with
Fig. 17.—Same as Fig. 16.

semimajor axis. In general, our most uncertain bins are those
that are closer to the Sun. Besides their relative rarity (Fig. 15)
the higher motion of these objects tends to give them fewer de-
tection opportunities.
This survey detected only ﬁve distinct classical KBOs and
plutinos, two each with 3 H 4 and one each in the ranges 4
H 5; 5 H 6, and 6 H 7. Using our net complete-
ness estimates, including error, and linearly interpolating over
the bias surfaces, we present a population estimate for the larger
objects in Table 6.
With fairly large uncertainties, the prediction agrees with what
is already known to be in the belt and indicates that surveys are
TABLE 6
Estimated Size of the Classical/Resonant KBO
and Scattered Disk Population
Magnitude Range
Number
Detected
Population
Estimate
Number
Known
3 H 4........................... 2 6:3 Æ 2:4 7
4 H 5........................... 1 9:4 Æ 7:8 15
5 H 6........................... 1 150 Æ 60 60
6 H 7........................... 1 500 Æ 250 226
LARSEN ET AL.1260

Fig. 19.—The orbit element distribution for Centaurs and scattered disk objects used to generate our simulated population. Data are from the IAU’s Minor Planet
Center Web site.

Fig. 20.—Estimate of bias for H 8 scattered disk objects. Note that for the error estimates the scale of the vertical axis has been exaggerated by a factor of 5.
1262

1263

approaching completeness for the larger objects. Many smaller
objects (H 5) remain to be discovered.
6.2. Observational Bias: Scattered Disk
Brown (2001) showed that the scattered disk objects have
very high inclinations on average. Because this survey is tightly
confined to the plane of the ecliptic, one would expect heavy bi-
ases against the scattered disk. To simulate the Centaur and scat-
tered disk KBOs, we generated 799,994 sets of orbital elements
for H 8 populations. The generated population was drawn from
the actual orbital element distribution as shown in Figure 19 and
was then randomized slightly. In general, the semimajor axis was
allowed to vary from 30 to 500 AU, the eccentricity was allowed
to vary from 0.1 to 0.85, the inclination from 0
to 78
, absolute
magnitude from 0 to 8, and once again the three orbital angles
were randomized on the range from 0
to 360
. Element sets were
excluded from the simulation if they were very different in the
entire set of orbital elements compared to examples in the general
population. While this method of selecting elements may reflect a
bias of current surveys, we are already heavily biased against the
population and seek only a quick estimate of our completeness for
scattered disk objects.
As expected, the scattered disk simulation produced less than
half the number of detection possibilities (450,979) as the clas-
sical belt. The ratios of unique discoveries (or the sum of the de-
tection probabilities for each object that could have been detected)
to the entire simulated population are presented in Figures 20Y22.
Because our element distributions were faithfully sampled from
the true distribution, gaps in the inclination and semimajor axis
space appear in the figures. As before, the histograms represent a
binned function of each orbital parameter versus the semimajor
axis. Each plot is also accompanied by the estimated error on the
completeness for each bin.
From Figure 20 we can see that there are no really uniform
coverage regions in absolute magnitude space, and the peak com-
pleteness only reaches approximately 40%. The completeness is
strongly a function of semimajor axis, with the most complete pa-
rameter space being the objects that spend a lot of time in the space
of the classical Kuiper Belt. The inclination, unsurprisingly, shows
a very strong bias against distant and more highly inclined classical
KBOs. As before this is due to the bias toward the ecliptic.
Over the course of this survey we detected only a single scat-
tered disk KBO. Since the population of large objects goes so
much farther away than our survey can cover, our estimate of the
size distribution is bound to be in error, but we estimate that there
are approximately 20 scattered disk objects with 5 H 6
within 55 AU of the Sun, with 10 already known.
6.3. Observational Bias: Brunini Melita Candidate
While the candidate of Brunini Melita (2002) has been ruled
out (Melita et al. (2004) and almost certainly been surveyed
through the observations of Trujillo Brown (2003), its orbital
characteristics form a good starting place for exploring a pop-
ulation for which no discoveries have yet been made.
The region of orbit parameter space originally thought to con-
tain the candidate is shown in Figure 23. Since there was only a
single candidate postulated, we did not randomly simulate the
population from a distribution but instead created a uniform sam-
ple grid across the main orbital parameters. This grid contained
600,000 evenly spaced element sets. For the candidates, we kept
the parameters studied by Brunini Melita (2002). The semi-
major axis was allowed to vary from 55 to 80 AU, the eccentric-
ity was allowed to vary from 0.0 to 0.3, and the inclination was
kept within 10
of the ecliptic. As before, the three orbital angles
were randomized in the range 0
to 360
.
The results of this simulation are found in Figures 24Y26. In
terms of Pluto- or Mars-sized objects, we eliminate more than
half the available phase space (closer to 90% if we restrict the
search to within 5
of the ecliptic, ignoring the summer Milky
Way and summer skies from Tucson). This preference for smaller
distances from the ecliptic is obvious in Figure 24 as well. The
other figures show no dependencies on the orbital parameters;
the completeness basically comes from the flux limitations and
distance of the object. From these figures, we conclude that the
number of objects with H 8 in the ecliptic but more distant
than 50 AU must be less than five in the best case, and that no
more than two objects with H 2 can be expected within 10
of the ecliptic.
6.4. Observational Bias: Very Distant Gas Giants and Planets
Many of the early Planet X searches were for gas giantYsized
planets, and references to possible binary companions to the Sun
are still made. In fact, considering the orbits of 2000 CR105 and
(90377) Sedna, Morbidelli Levison (2004) considered sce-
narios in which these objects were captured from low-mass stars
or brown dwarf stars that encountered the Sun. If these kinds of
captures are possible, it is conceivable that a larger planet-sized
object could also have been weakly captured and bound to the
Sun. Current astrometric sensitivity tells us that they are not close
(Zakamska Tremaine 2005), but it will require further missions
like Gaia to completely eliminate the possibility. Distant objects
may also have accreted in the outer solar system (Stern 2005) that
are definitely below the rate limits of the completed outer solar
system surveys even if they were visible at the time.
An object as large as Jupiter (H % À9) would be easily visible
to 200 AU but would rapidly decrease in brightness as the distance
increased.By the time a Jupiter-sizedobject reaches 2500AU, it is
25 mag in the visual, below the flux limit of this survey. Some-
where around 1400 AU, Jupiter-sized objects will reach the limit-
ing magnitude of this survey.
A quick calculation can estimate how much of the sky we have
successfully searched for these objects. As in x 6.3 we generate
600,000 more sets of orbital elements with orbit characteristics
like Brunini Melita (2002) (within 10
of the ecliptic) but with
semimajor axes ranging from 100 to 2500 AU and H ¼ À1 to
À9. As before, our elements are evenly spaced in a, e, i, and H
Fig. 23.—Eccentricity vs. semimajor axis showing the region of space in
which the 2002 candidate could lie. Data are from the IAU’s Minor Planet Center
Web site.

Fig. 24.—Estimate of bias for H 8 objects similar to Brunini Melita’s (2002) planet candidate. Note that the scale of the error estimates has been exaggerated
by a factor of 5.
1266

1267

with random angles. While the inclination distribution limit seems
very tight, it is chosen to maximize the strength of the conclusions
we can make about the existence of large objects in the solar sys-
tem given that our survey fields are tightly constrained to the
ecliptic. Physically it also means we are specifically checking for
Stern’s proposedobjects, whichshould be neartheinvariable plane
while restricting our ‘‘captured Jupiters’’ to the small subset that
would end up occupying orbits in the ecliptic. The estimated com-
pleteness of this survey for these objects is presented in Figure 27.
Within 10
of the ecliptic we can rule out large numbers of Pluto-
sized objects out to 150 AU and large numbers of Mars-sized
objects to within 200 AU, and we have covered 50% of the low-
inclination sky available for Jupiter-sized objects out to 700 AU.
7. CONCLUSIONS AND DISCUSSION
We have completed the first multinight survey of the ecliptic
searching for distant and slow objects. We discovered one new
large KBO and detected six others, indicating that new sweeps of
the sky are productive even if they have been previously covered
simply due to the complexities of running large surveys over
many nights and variable conditions.
While we have not found any larger planets, we have com-
puted limits on their existence within the constraints of our sur-
vey. The next generation of surveys will obviously approach this
problem with much greater sensitivity. In the plane of the eclip-
tic, the space beyond the Kuiper Cliff is devoid of large objects
out to fairly large distances depending on size. We have made
some population estimates for larger members of the Kuiper Belt
and have shown that large numbers of new planets are unlikely.
In particular, we have ruled out more than one to two Pluto-sized
objects out to 100 AU, and one to two Mars-sized objects to
200 AU. We have searched a substantial portion of the space within
700 AU and on the ecliptic that could be occupied by Jupiter-sized
objects. This survey’s detection in either case has the same con-
sequences as the observational limits set by Allen et al. (2002)
but extends the distance considerably, from 60 to 150 AU for
larger objects.
E. S. R. gratefully acknowledges the Office of Naval Research
for partial support of this work, via the Naval Academy Trident
Scholar Program, on funding document N0001406WR20137.
J. A. L. also gratefully acknowledges the Office of Naval Research
for partial support of this work, via the Naval Academy Research
Council, on funding document N0001405WR20153.
Spacewatch is funded by grants from NASA’s Near-Earth
Object Observation Program, NASA’s Planetary Astronomy
Program, the Brinson Foundation of Chicago, Illinois, and by
donations from private individuals. Support was also provided
recently by the Packard Foundation, the Kirsch Foundation, the
Paul G. Allen Foundation, and the US Air Force Office of Sci-
entific Research.
While not regular observers, Spacewatch appreciates the ef-
forts of K. L. S. Larsen and T. N. Woodell, who assisted in some
of the data collection. We also would like to thank Debora Katz
for several very useful discussions.
REFERENCES
Allen, R. L., Bernstein, G. M., Malhotra, R. 2001, ApJ, 549, L241
———. 2002, AJ, 124, 2949
Allen, R. L., Gladman, B., Kavelaars, J. J., Petit, J.-M., Parker, J. W.,
Nicholson, P. 2006, ApJ, 640, L83
Bernstein, G., Khushalani, B. 2000, AJ, 120, 3323
Bernstein, G. M., Trilling, D. E., Allen, R. L., Brown, M. E., Holman, M.,
Malhotra, R. 2004, AJ, 128, 1364
Bertin, E., Arnouts, S. 1996, AAS, 117, 393
Bottke, W. F., Jr., Morbidelli, A., Jedicke, R., Petit, J.-M., Levison, H. F., Michel,
P., Metcalfe, T. S. 2002, Icarus, 156, 399
Brown, M. E. 2001, AJ, 121, 2804
Brown, M. E., Trujillo, C., Rabinowitz, D. 2004, ApJ, 617, 645
———. 2005, ApJ, 635, L97
Brunini, A., Melita, M. D. 2002, Icarus, 160, 32
Dones, L. 1997, ASP Conf. Ser. 122, From Stardust to Planetesimals, ed. Y. J.
Pendleton A. G. G. M. Tielens (San Francsico: ASP), 347
Edgeworth, K. E. 1949, MNRAS, 109, 600
Fernańdez, J. A., Brunini, A. 2000, Icarus, 145, 580
Fernańdez, J. A., Gallardo, T., Brunini, A. 2004, Icarus, 172, 372
Gehrels, T., Marsden, B. G., McMillan, R. S., Scotti, J. V. 1986, AJ, 91, 1242
Gladman, B., Holman, M., Grav, T., Kavelaars, J., Nicholson, P., Aksnes, K.,
Petit, J.-M. 2002, Icarus, 157, 269
Gladman, B., Kavelaars, J. J., Petit, J.-M., Morbidelli, A., Holman, M. J.,
Loredo, T. 2001, AJ, 122, 1051
Gomes, R. S., Morbidelli, A., Levison, H. F. 2004, Icarus, 170, 492
Greaves, J. S., et al. 2005, ApJ, 619, L187
Grundy, W. M., Buie, M. W., Spencer, J. R. 2005, AJ, 130, 1299
Ida, S., Larwood, J., Burkert, A. 2000, ApJ, 528, 351
Jedicke, R. 1996, AJ, 111, 970
Jedicke, R., Herron, J. D. 1997, Icarus, 127, 494
Jedicke, R., Larsen, J., Spahr, T. 2002, in Asteroids III, ed. W. F. Bottke, Jr.,
et al. (Tucson: Univ. Arizona Press), 71
Jedicke, R., Metcalfe, T. S. 1998, Icarus, 131, 245
Jedicke, R., Morbidelli, A., Spahr, T., Petit, J.-M., Bottke, W. F., Jr. 2003,
Icarus, 161, 17
Jewitt, D., Luu, J., Trujillo, C. 1998, AJ, 115, 2125
Kenyon, S. J. 2002, PASP, 114, 265
Kenyon, S. J., Bromley, B. C. 2004, Nature, 432, 598
Fig. 27.—Estimate of bias for very large and distant objects.

The Search for Distant Objects in the Solar System Using Spacewatch - Astronomical Journal

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to The Search for Distant Objects in the Solar System Using Spacewatch - Astronomical Journal (20)

The Search for Distant Objects in the Solar System Using Spacewatch - Astronomical Journal