Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Celest Mech Dyn Astr (2010) 106:245–259

DOI 10.1007/s10569-010-9257-7

Formation of the extreme Kuiper-belt binary 2001 QW322


through adiabatic switching of orbital elements

Antonio Gamboa Suárez · Daniel Hestroffer ·


David Farrelly

Received: 1 September 2009 / Revised: 1 December 2009 / Accepted: 1 February 2010 /


Published online: 27 February 2010
© Springer Science+Business Media B.V. 2010

Abstract Binaries in the Kuiper-belt are unlike all other known binaries in the Solar
System. Both their physical and orbital properties are highly unusual and, because these
objects are thought to be relics dating back to the earliest days of the Solar System, under-
standing how they formed may provide valuable insight into the conditions which then pre-
vailed. A number of different mechanisms for the formation of Kuiper-belt binaries (KBBs)
have been proposed including; two-body collisions inside the Hill sphere of a larger body;
strong dynamical friction; exchange reactions; and chaos assisted capture. So far, no clear
consensus has emerged as to which of these mechanisms (if any) can best explain the observed
population of KBBs. Indeed, the recent characterization of the mutual orbit of the symmetric
(i.e., roughly equal mass) KBB 2001 QW322 has only served to complicate the picture because
its orbit does not seem readily explicable by any of the available models. The binary 2001
QW322 stands out even among the already unusual population of KBBs for the following
reasons: its mutual orbit is extremely large (≈105 km or about 30% of the Hill sphere radius),
retrograde, it is inclined ≈120◦ from the ecliptic and has very low eccentricity, i.e., e ≤ 0.4
(and possibly e ≤ 0.05). Here we propose a hybrid formation mechanism for this object
which combines aspects of several of the mechanisms already proposed. Initially two objects
are temporarily trapped in a long-living chaotic orbit that lies close to a retrograde periodic
orbit in the three-dimensional Hill problem. This is followed by capture through gravitational
scattering with a small intruder object. Finally, weak dynamical friction gradually switches
the original orbit “adiabatically” into a large, almost circular, retrograde orbit similar to that
actually observed.

A. Gamboa Suárez · D. Farrelly (B)


Utah State University, 0300 Old Main Hill, Logan, UT 84322, USA
e-mail: david.farrelly@usu.edu; david.farrelly@aggiemail.usu.edu
A. Gamboa Suárez
e-mail: angamboa@yahoo.com

D. Hestroffer
IMCCE/Paris Observatory, Paris, France
e-mail: daniel.hestroffer@obspm.fr

123
246 A. Gamboa Suárez et al.

Keywords Kuiper-belt binary · Binaries · Formation mechanisms ·


Chaos-assisted capture · 2001 QW322 · Irregular satellites · FLI maps

1 Introduction

Interest in the trans-Neptunian part of the Solar System, the Kuiper-belt, stems, in part, from
the recognition that this region is a unique laboratory for testing planetary formation models
(Noll 2003). The Kuiper-belt is thought to consist of leftovers from the accretional phase of
the Solar System (Kenyon and Luu 1999; Gomes 2009) and so, caught up in the amber of time,
trans-Neptunian objects (TNOs) provide a window into the early days of the Solar System.
Because of their remoteness, however, knowledge of the dynamical, physical, and chemical
properties of TNOs remains quite limited. Nevertheless, progress towards characterizing the
properties of TNOs is accelerating thanks to the discovery of a significant number of binary
TNOs (Noll 2003).
Kuiper-belt binaries (KBBs) are of special interest because of their unusual physical and
orbital properties, at least when compared to the properties of Near-Earth and main-belt
asteroid binaries. For example, KBBs tend to have partner mass ratios of order unity whereas
Near-Earth and Main Belt binaries typically have mass ratios in the range m r ∼ 10−3 –10−4
where m r = m 2 /m 1 and m 1 and m 2 are the masses of the binary partners with m 2 ≤ m 1 .
The orbits of KBBs are also peculiar in (i) having very large mutual orbit semi-major axes
(a) when compared to the physical radii (r1 , r2 ) of the binary components themselves and
(ii) moderately large eccentricities (Noll 2003). Taken together, these properties place rather
tight constraints on binary formation mechanisms which, in turn, may be used to constrain
formation models of the Solar System itself.
Several mechanisms have been proposed to explain the physical and orbital properties
of KBBs. However, none of them alone seems able to explain the recently characterized
orbit of the KBB 2001 QW322 (Petit et al. 2008): the orbit of 2001 QW322 is extremely
large (≈105 km or about 30% of the Hill sphere radius), retrograde, inclined ≈120◦ from the
ecliptic and with eccentricity e ≤ 0.4 (and possibly e ≤ 0.05). The main objective of this
paper is to demonstrate that a hybrid mechanism, one which contains aspects of several of
the previously proposed capture scenarios, can account for the highly unusual orbit of 2001
QW322 . To that end it is useful to summarize briefly the KBB mechanisms that have already
been proposed.
The first mechanism proposed was that of Weidenschilling (2002). In this model two
bodies accrete after a mutual collision within the Hill sphere of a third, larger, body. They
then remain bound as a single object orbiting the larger body, thereby producing a binary.
The main objections (Goldreich et al. 2002) to this model have to do with the presumed scar-
city of large objects in the primordial Kuiper-belt ( KB). However, Petit and Mousis (2004)
discuss this issue at length and conclude that the mechanism of Weidenschilling might well
have operated, possibly in tandem with other mechanisms. More recently, Petit et al. (2008)
suggest that this mechanism is the most credible way of producing 2001 QW322 . However,
there seems to be no compelling reason why the collisional mechanism of Weidenschilling
(2002) should specifically produce a very wide, and almost circular, mutual orbit. Of course,
non-circular orbits can be effectively circularized by tidal dissipation but this leads to a reduc-
tion in the semi-major axis. That is, the Weidenschilling (2002) mechanism would seem to
require that either (i) the large, almost circular orbit of 2001 QW322 was produced essen-
tially directly or that (ii) a very much larger, initially non-circular orbit was produced which
was then circularized by tidal friction (Ferraz-Mello et al. 2008). Because the mechanism of

123
Formation of the extreme Kuiper-belt binary 2001 QW322 247

Weidenschilling (2002) does not seem to make an ab initio prediction of simultaneously very
wide and almost circular binaries it seems reasonable to check whether other mechanisms
exist which can explain these two noteworthy features of the orbit of 2001 QW322 .
The second mechanism proposed is due to Goldreich et al. (2002): in this model two
potential binary partners interpenetrate their mutual Hill sphere and this transient binary then
becomes permanently trapped through its interaction with a sea of planetesimals through
dynamical friction. In a variant of this model, gravitational scattering with a large third
body may replace dynamical friction in the permanent capture of the transient binary. In
the simulations of Goldreich et al. (2002) these two pathways tend to produce eccentricities
significantly larger than that observed for 2001 QW322 . The recently proposed chaos-assisted
capture (CAC) (Astakhov et al. 2003, 2005; Lee et al. 2007; Makó and Szenkovits 2004)
mechanism also relies on the temporary capture of binary partners but specifically in cha-
otic layers of phase space inside the binary Hill sphere. Because temporary trapping can
be very long lived, as opposed to initial trapping in ballistic capture (Circi and Teofilatto
2005), the opportunity for permanent capture through dynamical friction or gravitational
scattering is greatly enhanced. The CAC mechanism was originally developed to explain
the capture of irregular moons at Jupiter and Saturn and was able to provide an explana-
tion for the different relative populations of retrograde and prograde moons between these
two planets (Astakhov et al. 2003). In the case of KBB formation, stabilization in the CAC
mechanism is the result of sequential gravitational scattering with multiple, small “intruder”
bodies. This gravitational scattering model, which involves multiple scattering events, is in
contrast to that proposed by Goldreich et al. (2002) which involves a single encounter with
a relatively large intruder. The creation of binaries having similarly sized partners is an ab
initio prediction of the model together with the prediction of moderate to high eccentrici-
ties. In the case of KBBs previous calculations predict orbits with 0.01 < a/R H < 0.05.
Certainly, while these binaries are wide in comparison to asteroid binaries, they are not
wide compared to 2001 QW322 . However, it should be noted that after initial capture in
the CAC mechanism, the binary orbit can be significantly wider than that of 2001 QW322
(Astakhov et al. 2005; Lee et al. 2007). Subsequent orbit reduction (e.g., through dynamical
friction) then produces a distribution of binary mutual orbit semi-major axes. The post-cap-
ture orbit reduction mechanism in the context of the CAC model is the main focus of this
paper.
An alternative mechanism has been suggested by Funato et al. (2004) and is based on
exchange reactions. The first step is the formation of a binary through a two-body dissipative
encounter. This can happen in at least two ways; (i) tidal disruption of an object during a
close encounter followed by accretion of the resulting fragments to produce a binary; or, (ii)
a “giant” impact in which coagulating debris after the collision produces a satellite orbit-
ing a larger object. Generally, the result is a binary with an extreme mass ratio and a tight,
circular orbit. In the exchange reaction mechanism, the mass ratio of such proto-binaries in
the KB is progressively increased through later inelastic gravitational scattering encounters
with a third object originating (in the simulations) at infinity. Funato et al. (2004) modelled
this by performing extensive scattering simulations between strongly asymmetric-mass bina-
ries already following bound, compact, circular orbits and large incoming intruder masses.
However, this mechanism leads almost exclusively to high eccentricity mutual orbits and,
therefore, is unlikely to have produced the majority of known KBBs which tend to have
moderate or, in the case of 2001 QW322 , very low eccentricities.
The paper is organized as follows: Sect. 2 briefly describes the CAC mechanism in the Hill
approximation. For illustration, the planar (two-dimensional, 2D) limit is considered. In order
to visualize the structure of phase space in the 3D problem the Fast Lyapunov Indicator (FLI)

123
248 A. Gamboa Suárez et al.

is used. However, visualization of the FLI is problematic as explained in Sect. 3. There an


approximate visualization method is introduced which relies on the construction of the nor-
mal form approximation to the Hill Hamiltonian. Our results and conclusions are described
in Sect. 4.

2 Chaos-assisted capture in the Hill problem

In our model the initial stages of binary formation involve temporary capture within the
Sun-binary Hill sphere through particles getting entangled in chaotic regions of phase space,
i.e., chaos-assisted capture (Astakhov et al. 2003). Because of the mass ratios involved, we
study the problem in the Hill approximation. The coordinate system (x, y, z) is rotating with
constant angular velocity ! = (0, 0, 1) which corresponds (in scaled units) to the mean
motion of the binary centre of mass. In this coordinate system the centre-of-mass of
the binary is at the origin. The Hill sphere occupies the region between the 3D physical
space projections of the Lagrange saddle points L 1 and L 2 ; these points, in the rotating
frame, are located on the surface of zero-velocity (Danby 1992). We work in scaled Hill
units such that the Hill sphere radius is given by R H = (1/3)1/3 . Energies are defined in the
rotating frame. Including terms to describe dynamical friction (Goldreich et al. 2002), the
equations of motion are,

ẍ − 2 ẏ − 3x = − rx3 − D& ẋ;


ÿ + 2 ẋ = − ry3 − D& ( ẏ + 3x/2) (1)
z̈ + z = − rz3 − D⊥ ż.

Here D& and D⊥ are dimensionless quantities characterizing the strength of the dynamical
friction. Initially we set D& = D⊥ = 0 and consider CAC in the planar limit, i.e., z = 0.
It is useful to make a distinction between gravitational scattering, which involves a close
approach of a fourth body to the binary, and dynamical friction which is caused by a sea of
small bodies that may be remote from the binary. In particular, dynamical friction does not
assume close scattering encounters. This is because of the following: gravitational accelera-
tions due to planetesimals drop off as the square of the distance while the number of these
bodies increases as the cube of the distance. Therefore, the most distant planetesimals (i.e.,
those that are not local to the binary) have the greatest effect on the binary. This is similar to
the situation in plasma physics, except here all of the interactions are attractive and so there
is no Debye shielding. The only thing that determines the maximum size of the “polarization
cloud of planetesimals” is the scale height of the planetesimal disk, which is the root mean
square inclination times the distance from the Sun.

2.1 Transformation to Levi-Civita coordinates

For the numerical integrations it is convenient to transform to regularizing coordinates: in the


2D limit this is know as the Levi-Civita transformation (Levi Civita 1906; Petit and Henon
1986). This procedure transforms Kepler orbits which pass close to the origin from being
almost parabolas into straight lines. Essentially the 1/r singularity at the origin is “blown-
up” and velocities which would otherwise diverge as 1/r remain stable. Doing this has both
theoretical and computational advantages, e.g., numerical integrations in regularized coor-
dinates remain stable as trajectories pass close to, or even through, the origin.

123
Formation of the extreme Kuiper-belt binary 2001 QW322 249

The Hamiltonian for the 2D Hill problem in Cartesian coordinates is,


1! 2 " 1 1
H= px + p 2y − # + ypx − x p y − x 2 + y 2 (2)
2 2
x +y 2 2
where x, y, px = ẋ − y and p y = ẏ + x are a set of canonical variables.
The Levi-Civita (LC) transformation is given by:
$ % $ %$ %
x q1 −q2 q1
= (3)
y q2 q1 q2
for the coordinates and
$ $ % %$ %
2 q1 −q2
px p1
= (4)
py
r q2 q1 p2
#
for the conjugate momenta. Here r = x 2 + y 2 = q1 2 + q2 2 . Besides this transformation,
we make a change of timescale given by:
4dt
dτ = & 2 '. (5)
q1 + q2 2
The new Hamiltonian, HLC (q, p) = (r/4)(H + p0 ), and, written as a function of the
new coordinates becomes (after neglecting an inessential constant term):
& '
p0 q 1 2 + q 2 2 1& 2 '
HLC = + p1 + p 2 2
2 2 2
1& 2 2
'
+ q1 + q2 (q2 p1 − q1 p2 )
2
1& '
− q1 6 − 3q1 4 q2 2 − 3q1 2 q2 4 + q1 6 (6)
4
where p0 = C J /2 and C J = −2H is the Jacobi constant.
The further transformation:
q1 = α Q 1 , q2 = α Q 2 , p1 = β P1 , p2 = β P2 (7)
1 3 −3
yields HLC = γ H( where α = 2c 4 , β = 2c 4 and γ = 41 c 2 with c = p0 /2. This leads to a
Hamiltonian that is free of all parameters and is the one used in the numerical calculations.
1& 2 ' & '
H( (Q 1 , Q 2 , P1 , P2 ) = Q 1 + Q 22 + P12 + P22 + 2 Q 21 + Q 22 (Q 2 P1 − Q 1 P2 )
2
& '
− 4 Q 61 + Q 62 − 3Q 41 Q 22 − 3Q 21 Q 42 (8)

2.2 Surfaces of section and the structure of phase space

The two Lagrange saddle points L 1 and L 2 act as gateways between the interior of the Hill
sphere and heliocentric orbits. Because we are only interested in capture, here we only con-
sider the dynamics inside the Hill sphere. Examination of Poincaré surfaces of section in the
Hill problem (Simó and Stuchi 2000)—or, equivalently, the circular restricted three-body
problem (CRTBP) for small masses (Astakhov et al. 2003)—reveals that at energies above
the Lagrange saddle points L 1 and L 2 all phase space is divided into three parts, one of which
regular Kolmogorov-Arnold-Moser (KAM) orbits inhabit, chaotic orbits another, while the
third consists of direct scattering orbits or, in other words, hyperbolic orbits. All these differ

123
250 A. Gamboa Suárez et al.

0.4

0.4
a b

0.2

0.2
P1

P1
0.0

0.0
−0.2

−0.2
−0.4

−0.4
−0.4 −0.2 0.0 0.2 0.4 −0.4 −0.2 0.0 0.2 0.4
Q1 Q1

0.15
0.4

c d
0.2

0.05
P1

P1
0.0

−0.05
−0.2

−0.15
−0.4

−0.4 −0.2 0.0 0.2 0.4 −1.814 −1.810 −1.806


Q1 Q1
Fig. 1 Poincaré surfaces of section in the 2D Hill problem in the absence of dynamical friction. Each panel
is the result of integrating 50 trajectories at the value of the Jacobi constant, C J , shown. The SOS are shown
using the regularized Levi-Civita coordinates as described in the text. The points on the SOS are shaded
according to the sign of the angular momentum of the trajectory upon its intersection with the SOS. For many
trajectories the sign of the angular momentum is approximately conserved (Simó and Stuchi 2000; Astakhov
et al. 2003). Retrograde orbits are dark grey (red online) while prograde orbits are light grey (purple online).
A large, circular periodic orbit, here labeled R1 , lies at the center of the large retrograde island and this orbit
persists to high energies (small values of the Jacobi constant). Also visible, particularly at higher energies, are
thin layers of chaos separating the KAM islands from the direct scattering (hyperbolic) parts of phase space
(colours online)

from each other in the behaviour of their orbits. The chaotic orbits separate the regular from
the hyperbolic regions. As energy is increased above the saddle points the chaotic regions
consist of relatively thin layers of chaos which cling to the progressively shrinking KAM tori.
Because incoming orbits cannot penetrate the regular KAM tori in 2-dimensions (2D) and
can do so only exponentially slowly in 3D, orbits entering the Hill sphere from heliocentric
orbits must either enter chaotic layers or scatter out of the Hill sphere essentially immediately.
The chaotic, regular and hyperbolic regions of phase space can be visualized using Poin-
caré surfaces of section (SOS) which are defined to have Q 2 = 0 and Q̇ 2 > 0. Composite
SOS are shown in Fig. 1 for four values of the Jacobi constant, C J . These were obtained
by plotting SOS for many trajectories in the same figure. In each frame the trajectories have
the same value of the Jacobi constant but different initial conditions. All of the values of the
Jacobi constant shown in Fig. 1 correspond to energies above the Lagrange points. Clearly
visible is a thin chaotic layer which separates the regular KAM regions from the hyperbolic
regions.

123
Formation of the extreme Kuiper-belt binary 2001 QW322 251

Fig. 2 Fast Lyapunov indicator


cJ = 4.329

0.4
(FLI) map obtained as described
in the text at the same value of 50
the Jacobi constant, C J , as

0.2
Fig. 1a. In the FLI map the points 40
are shaded according to the value
of the FLI as shown in the grey 30
scale (colours online)

0.0
P1
20

−0.2
10

−0.4
−0.4 −0.2 0.0 0.2 0.4
Q1

In the CAC model the first step is the formation of long-lived quasi-bound binaries corre-
sponding to orbits which have become trapped in these chaotic layers. Because the orbits are
chaotic any quasi-bound or transient binary which is formed will eventually break apart. How-
ever, the lifetimes of these quasi-bound objects can be sufficiently long that, in the interim,
energy loss may occur. Since the chaotic layers in the Hill problem lie adjacent to regular
KAM regions, then any energy loss tends to capture the binary permanently. Astakhov et al.
(2005) and Lee et al. (2007) proposed that capture and subsequent hardening of three-body
Hill orbits ( i.e., energy loss and Keplerization) then proceeds through chaotic inelastic grav-
itational scattering with low-mass intruders which happen to transit the Hill sphere. Once a
binary has been captured initially then it was estimated that further encounters with intrud-
ers are probable—about once every 3,000–10,000 years—and this can eventually produce
binaries whose mutual orbits are Kepler ellipses. Empirically the process is found to be most
efficient for equal mass binaries and relatively low-mass intruders having ∼2% or less of
the total binary mass. Previous simulations in the Hill problem by Astakhov et al. (2005)
and Lee et al. (2007) suggest that this picture also holds in three-dimensions. However,
while this mechanism can produce very wide binaries, it tends to produce mutual orbits with
eccentricities e > 0.4 which likely rules it out as the formation mechanism of 2001 QW322 .
A possible clue to the origin of the large, retrograde, apparently almost circular orbit of
2001 QW322 may be seen in Fig. 1d. For small values of the Jacobi constant (high energies),
C J , only retrograde orbits are stable. As already noted by Simó and Stuchi (2000), a circular,
retrograde periodic orbit, which we label R1 , persists to extremely high energies. Depend-
ing on the energy, this orbit (in 2D) may lie at the center of a small KAM island. In turn,
surrounding this island, is a narrow and “sticky” chaotic region (Perry and Wiggins 1994;
Astakhov et al. 2003; Kandrup and Sideris 2002; Frœschlé et al. 2006) in which binaries
may become trapped for many periods—see, e.g., frames (c) and (d) (plots showing typical
distributions of trapping times in these chaotic layers are contained in Fig. 2 of Lee et al.
(2007)). These observations raise the questions; did the progenitor of 2001 QW322 originally
lie close to R1 ? And, if so, how was it stabilized?
A complication is that the orbit of 2001 QW322 is inclined and, therefore, the simulations
should be done in 3D. This, of course, precludes using SOS in order to visualize the dynamics.
Actually, it is not absolutely necessary to have a picture of the 3D phase space at all because
it is easy to demonstrate by explicit simulation that long living 3D trajectories exist close
to the periodic orbit R1 . These trajectories appear regular for many periods of time but are

123
252 A. Gamboa Suárez et al.

ultimately chaotic and eventually escape. Nevertheless it is useful to try to devise a way of
visualizing the structure of phase space close to R1 . We do this by using the Fast Lyapunov
Indicator coupled with construction of the normal form approximation to the Hill problem
using methods developed by Deprit and co-workers (1991).

3 Fast Lyapunov indicators and the normal form

The objective now is to seek a means of visualizing the phase space in the 3D Hill problem
in the vicinity of the large retrograde periodic orbit, R1 , shown in Fig. 1d. In general, the
Fast Lyapunov Indicator (FLI) can often be used to map chaotic and regular regions of phase
space when Poincaré SOS cannot readily be computed (Frœschlé et al. 2000, 2006) as is the
case in the 3D Hill problem. Given an n-dimensional continuous-time dynamical system,

dx/dt = F(x, t), x = (x1 , x2 , . . . , xn ), (9)

the Fast Lyapunov Indicator is defined (Frœschlé et al. 2000)

F L I (x(0), v(0), t) = ln |v(t)|, (10)

where v(t) is a solution of the system of variational equations


$ %
dv ∂F
= v. (11)
dt ∂x

Use of the FLI (Frœschlé et al. 2006) is enhanced if the value of the indicator can be pro-
jected onto a plane, e.g., of initial conditions, as was done by Astakhov and Farrelly (2004)
in an application of this technique to the elliptic three-body problem. For example, Fig. 2
makes a comparison between the FLI map and the Poincaré surface of section for the 2D
Hill problem. Of course, in this limit the FLI is unnecessary because a SOS can be computed
directly. Here the comparison is done simply to illustrate that the FLI indicator is capable
of distinguishing between regular and chaotic regions of phase space and also of detecting
resonances. For more than 2 degrees of freedom Poincaré SOS are problematic. For exam-
ple, the elliptic planar restricted three-body (ERTBP) problem is a 2.5 degree of freedom
problem for which the FLI has proved to be a very useful tool for examining the structure
of phase space (Astakhov and Farrelly 2004). In the ERTBP Astakhov and Farrelly (2004)
took advantage of the fact that, by choosing the initial phase judiciously, the ERTBP can be
made to coincide exactly with the circular restricted three-body problem (CRTBP) at time
t = 0. Thus, at t = 0 initial conditions can be launched from a plane which corresponds to
an exact SOS of the CRTBP. The FLI is then computed along each trajectory and its value
projected onto the original plane of initial conditions at t = 0. As shown in Astakhov and
Farrelly (2004) and Astakhov et al. (2005) using the FLI in this way provides a very good
overview of the structure of phase space in the ERTBP. The situation in the 3D Hill problem
is, however, even more complicated because the system has 3—rather than 2.5—degrees of
freedom. A consequence is that the FLI cannot be projected exactly onto a plane.
We have developed an approximate approach in which the FLI is projected onto a surface
which corresponds to a SOS not of the 3D Hill problem but of an approximation to that
problem obtained by construction of the normal form (Deprit and Williams 1991).

123
Formation of the extreme Kuiper-belt binary 2001 QW322 253

3.1 The normal form

The Kepler problem and, by extension, the Hill problem can be expressed in terms of the
Delaunay variables {L , G, H, &, g, h} (Deprit and Williams 1991). The normal form is a
lower dimensional approximation to, in this case, the Hill problem obtained by averaging
over the mean anomaly &. The resulting system is then only a function of the set of (trans-
formed) variables {L , G, H, g, h} with L being a constant of the motion. Thus, the phase
space is 4D and, in principle, a SOS can be constructed. However, computation of a SOS of
the normal form itself is not our objective. Rather, working on the assumption that the nor-
mal form provides a reasonable approximation to the dynamics, we project the FLI obtained
from integration of the exact 3D equations of motion onto an appropriately defined - and
approximate—plane of the averaged problem. This is obviously an inexact procedure and we
expect that the FLI maps so obtained will have some “thickness” but, nevertheless, it does
turn out to provide a means of visualizing the long living chaotic orbits that lie close to R1 .
The procedure is as follows:
In Cartesian coordinates, the Hill Hamiltonian in 3D is given by
p2 1 r2 3
F= − + − (x p y − y px ) − x 2 (12)
2 r 2 2
#
where r = x 2 + y 2 + z 2 . First we perform the Kustaanheimo-Stiefel (KS) transformation
which can be thought of as the generalization of the LC transformation to the 3D Kepler
problem (Stiefel and Scheifle 1971). Explicitly it is given by,
R = Tu (13)
where R = (x, y, z, 0) and u = (u 1 , u 2 , u 3 , u 4 ) and
 
u 1 −u 2 −u 3 u4
 u2 u1 −u 4 −u 3 
T=  u3
. (14)
u4 u1 u2 
u 4 −u 3 u2 −u 1
The Hamiltonian that results is a perturbed 4D isotropic harmonic oscillator.
1& 2 '
F =4= Pu + ω2 u2 − 4H |u|2 + 2|u|6 − 6|u|2 (u 1 u 3 + u 2 u 4 )2 . (15)
2
. √
where |u| = u 21 + u 22 + u 23 + u 24 and ω = 2 C J where C J = −2F is the Jacobi constant
and H is the z−component of the angular momentum.
Next we perform a Birkhoff-Gustavson normalization which corresponds, in the case of
Delaunay variables, to averaging over the mean anomaly &. After various canonical trans-
formations the normalized Hamiltonian can be expressed in terms of the components of the
two pseudo-angular momenta (Deprit and Williams 1991):
1 1
J= (G + A); K = (G − A) (16)
2 2
where G is the physical orbital angular momentum and A is the Runge-Lenz vector. The
effective Hamiltonian, after expansion to 6th order in coordinates and momenta, is, in terms
of these angular momenta variables:
H( = −4C J (Jz + K z ) + 4(Jz 2 + K z 2 )
−12Jy 3 − 12Jx 2 Jy + 12K y Jy − 12K y 3 − 24Jx K x − 12K x 2 K y − 16Jz K z . (17)

123
254 A. Gamboa Suárez et al.

Using Eq. 16 the averaged 3D Hill Hamiltonian can now be written in terms of the
canonical variables {G, h} and {A z , φz }. Furthermore, in terms of the Delaunay variables
{L , G, H, &, g, h}, we have
G x = G sin i sin h A x = Le(cos g cos h − sin g cos i sin h)
G y = G sin i cos h A y = Le(cos g sin h + sin g cos i cos h (18)
Gz = H A z = Le sin g sin i
where e is the eccentricity and i the inclination. Setting e = 0 and i = π corresponds to
a circular, retrograde orbit in the plane, i.e., the orbit R1 . In the {A z , φz } phase plane, for
a fixed and appropriately chosen value of the angle h the periodic orbit R1 , appears at the
origin. The idea is to choose initial conditions of the full 3D problem in the {A z , φz } plane
with fixed angle h, compute the FLI for these trajectories, and then project these values back
onto the (approximate) {A z , φz } plane. This is an approximation because, in reality, the mean
anomaly is varying and so the projection has some width, i.e., it will appear to be somewhat
fuzzy because each point in the {A z , φz } plane will, in general, belong to more than one
trajectory.
The FLI maps were computed as follows:

1. Initial conditions were picked randomly inside the Hill sphere and integrated until they
escaped the Hill sphere. Trajectories were considered to be permanently bound, and so
eliminated, if they did not escape in a time T = 10000 where T is the physical time in
scaled Hill units, i.e., 10000/2π periods or ∼ 0.5 Myr at 45 au. Of course, some of these
orbits will eventually escape at later times (i.e., they are not truly bound). However, this
cutoff time is large enough that increasing it does not materially affect our conclusions
or the FLI maps.
2. Trajectories which escaped were then back-integrated in time from their initial coordi-
nates until they escaped from the Hill sphere. The coordinates and momenta of these
trajectories on the boundary of the Hill sphere were stored.
3. The stored coordinates were then used as initial conditions and the trajectory was inte-
grated forward in time until it again left the Hill sphere. At that point the value of the
FLI was recorded.
4. In constructing the FLI map only trajectories with lifetimes greater than 1000/(2π)
periods were considered. Each intersection with the {A z , φz } plane was recorded. After
the integration was completed each intersection with this plane was assigned the same
and final value of the FLI assigned to the trajectory. The values of the FLI were then
color coded so as to construct the FLI map.
5. This procedure was repeated for a large number of trajectories and the results combined
into a single map.

Figure 3 is such a FLI map for a set of chaotic orbits in 3D which lie close to R1 . This
map is typical of the structure of phase space proximal to the periodic orbit R1 which lies at
the origin in Fig. 3. The blank (white) region surrounding R1 corresponds to a regular KAM
island. Chaotic orbits which are trapped in sticky KAM regions may appear to be quasi-peri-
odic for many periods before eventually escaping (Perry and Wiggins 1994; Astakhov et al.
2003; Kandrup and Sideris 2002; Frœschlé et al. 2006) and it is this phenomenon which
explains the structure apparent in the FLI map. For clarity, regular (i.e., permanently bound
by the criteria outlined above) orbits have been excluded from this map. The map appears
fuzzy because the surface onto which the FLI is projected is not an exact SOS of the full 3D
Hill problem whereas the trajectories themselves are integrated in the fully 3D problem. This

123
Formation of the extreme Kuiper-belt binary 2001 QW322 255

0.5
cJ = 0.500
200

0.25
150

0.0
Az

100
−0.25

50
−0.5

−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5


φz
Fig. 3 Fast Lyapunov Indicator maps for chaotic (escaping) trajectories projected onto the A z , φz plane for
the value of C J shown in the figure. In the FLI map the points are shaded according to the value of the FLI
as shown in the grey scale (color bar online). The periodic orbit R1 lies at the origin and the blank (white)
area surrounding it corresponds to a regular KAM island. Even though chaotic orbits ultimately escape they
may appear to be quasi-periodic in the short term: it is this which is responsible for the structure apparent
in the FLI map (Astakhov et al. 2003). For clarity regular orbits have been excluded from this map (colours
online)

approximate, hybrid, procedure is therefore, not, a priori, guaranteed to work: however, in


this case and for the parameters chosen, the KAM region surrounding R1 is clearly visible as
is the sea of chaotic orbits which, for long periods, appear to be glued to this island (Astakhov
et al. 2003).
We wish to emphasize that none of the capture simulations to be described rely on Fig. 3 or
on the normal form analysis. For example, the results in the next section and Fig. 4 are based
on numerical simulations of the full problem without any approximations. Fig. 3 is simply
a way of illustrating the fact that long-living chaotic orbits exist close to the periodic orbit
R1 . As noted previously, the numerical simulations themselves—without the normal form
analysis—demonstrate that these orbits exist. However, the normal form analysis provides
a physical explanation for why binaries get captured close to R1 in terms of two physical
quantities: the angular momentum and the Runge-Lenz vector which behave as adiabatically
conserved action variables under the action of dynamical friction. It is worth noting that a
similar type of “fuzzy” projection (to Fig. 3) has been used to visualize the phase space of
multidimensional systems by Lichtenberg and Lieberman (1992).

4 Results and conclusions

So far the calculations have focussed on trying to visualize phase space and, in particular,
to demonstrate that the retrograde periodic orbit of Fig. 1d is surrounded by a chaotic layer
in 3D as it is in 2D. In this section we describe the results of our capture simulations which
(i) do not rely on the previous, approximate analysis and (ii) involve integrations of the fully
3D Hill problem.
An important question, which is addressed in Lee et al. (2007), as well as in a more recent
paper by Schlichting and Sari (2008), concerns the initial probability of capture into a chaotic
zone. The interesting paper by Schlichting and Sari (2008) criticizes the CAC model on the
basis that the probability of entering a chaotic zone is quite small—see, e.g., Fig. 3 of their

123
256 A. Gamboa Suárez et al.

Fig. 4 Properties of the three

0.8
most likely observed orbits of
2001 QW322 (heavy black lines,
green/blue online) compared to
simulations (grey lines, black/red
online). The radial distance from
the origin represents the

0.6
semi-major axis, a, as a fraction
of R H . The length of each line
shows the pericenter-to-apocenter
distance and reflects the

a RH sin(i)
eccentricity of the orbit. For
clarity each of the three possible

0.4
“best fit” observed orbits (which
differ mainly in eccentricity and
semi-major axis rather than
inclination) has been randomly
assigned a different inclination
within the range of inclinations
0.2

estimated by Petit et al. (2008).


Time averaged instantaneous
orbital elements are used.
D& = D⊥ = 0.00015 and initial
conditions were chosen randomly
(colours online)
0.0

−0.8 −0.6 −0.4 −0.2 0.0

a RH cos(i)

paper. While we agree that the capture probability is not large and, in fact, declines as the
longevity of the CAC orbits increases we also note that Schlichting and Sari (2008) conclude
(in the last paragraph before Sect. 8 of their paper) that:

“. . . longevity of the transient binary [as discussed by Astakhov et al. (2005)] with
t3R H ≥ 15*−1 becomes only important for very weak dynamical friction (i.e., D ≤
0.002) and is most likely not crucial for KBB formation.”

Schlichting and Sari (2008) use values of D ≈ 0.1 − 0.2 (see Eq. 1 above) with * ≈ 1/
40 yr−1 . They define t3R H to be the time over which the separation between the two proto-
binary partners is less than 3R H for partners that approach to within at least R H . Several
comments can be made: although these authors are obviously criticizing our model their
simulations, in fact, agree with ours in finding similar capture probabilities. In particular,
they agree with us that the CAC model will be important in the case of weak dynamical
friction—i.e., the case considered here. We have performed simulations using values of D
comparable to Schlichting and Sari (2008) and find, as do they, that permanent capture occurs
very quickly—on the order of 50–400 years. In their simulations Schlichting and Sari (2008)
allow dynamical friction to operate for varying, presumably randomly chosen, times so as
compute a time, tt yp which they define to be the typical time for a binary to be captured.
They estimate that tt yp ≈ 80 years for D = 0.2. Our simulations agree. However, we also
find that unless D is then turned off very quickly, the binary mutual orbit will continue to
spiral-in, quickly resulting in a collision between the binary partners in essentially all cases.
Therefore, in order for the Schlichting and Sari (2008) mechanism to be viable, it seems to
us that dynamical friction with D ≈ 0.1 must have operated over a very short window of

123
Formation of the extreme Kuiper-belt binary 2001 QW322 257

time. That is, while the relatively large values of D used by Schlichting and Sari (2008) lead
to efficient capture, unless D is then rapidly reduced to zero over a period of a few years, the
captured binaries go on to self-destruct. It is for this reason that we employ a much smaller
value of D for which, according to Schlichting and Sari (2008) and our previous simulations
(Lee et al. 2007), capture through the CAC mechanism becomes efficient.
In this regard, we again emphasize that Fig. 3 is not intended to reflect the frequency of
capture into chaotic zones but rather, to illustrate that long living chaotic orbits exist close to
the periodic orbit R1 in the phase space of the 3D Hill problem. Studies of collisional veloc-
ities and rates of capture can be found elsewhere, e.g., Kenyon and Luu (1999), Goldreich
et al. (2004), Queck et al. (2007).
The Monte Carlo simulations are identical to those described previously by Astakhov et al.
(2005) and Lee et al. (2007) with the exception that the binary is first captured through grav-
itational scattering with a single small intruder after which dynamical evolution is assumed
to proceed by weak dynamical friction instead of gravitational scattering.
Our simulations were done as follows:

1. Initially a cohort of quasi-bound chaotic binary orbit initial conditions was harvested by
integrating Hill’s equations in the absence of dynamical friction. Initial conditions were
chosen randomly inside the three-body (Sun-binary) Hill sphere for randomly chosen
values of the Jacobi constant, C J , in the same range as that used in Fig. 1. If an orbit
remained inside the Hill sphere for a time T1 < TH < T2 then its initial conditions were
stored. We refer to TH as the orbit’s “natural” Hill lifetime. In scaled units we chose
T1 = 15 (∼ 600 years) and T2 = 1500. While much longer (and shorter) living orbits
are possible selected integrations for such orbits produce comparable results.
2. Intruder particles were then fired at each binary orbit with the phase of the binary chosen
randomly. This was done in the four-body Hill approximation (Astakhov et al. 2005;
Lee et al. 2007). If a binary was captured, and this was determined by integrating the
equations of motion for very long times, then the trajectory was saved.
The reason for assuming that small-intruder scattering is responsible for capture is that
previous computations (Astakhov et al. 2005; Lee et al. 2007) show that this mecha-
nism favors symmetric binaries by actively discriminating against asymmetric binaries.
Schlichting and Sari (2008) argue that the L 2 (dynamical friction) mechanism also favors
symmetric binaries. However, inspection of the equations of motion including dynam-
ical friction (Eq. 1) readily reveals that the mass enters only as the total mass through
the dynamical friction term. Therefore, the mass ratio of the binary partners is irrelevant
at a given total mass. On the other hand, in the four-body Hill equations of motion the
binary mass ratio appears.
The argument of Schlichting and Sari (2008) seems to be that there is more dynamical
friction acting when the total mass of the binary is the largest and, therefore, from a
population with a distribution of masses one might expect to see higher total masses
favored which would, in turn, favor more symmetric binaries. However, using the rela-
tively large dynamical friction parameters they use, and assuming a population with a
distribution of diameters up to several hundred kilometers, we found this effect to be
quite weak, i.e., using D ≈ 0.1 it is easy to capture binaries if at least one of them is
massive enough. Furthermore, Schlichting and Sari (2008) do not explicitly consider the
probabilities of encounters between two large bodies being less probable than between
unequal sized bodies, because there will tend to be more small bodies present. These
issues could best be resolved by future simulations.

123
258 A. Gamboa Suárez et al.

3. The trajectory was then integrated subject to dynamical friction. The dynamical friction
terms in Eq. 1 were switched on and off adiabatically using a smooth switching function.
The integrations were stopped when the semi-major axis of the mutual orbit was roughly
comparable to that of 2001 QW322 as a fraction of R H .

As noted, in these simulations we found that if the dynamical friction terms are large,
comparable, say, to those used by Schlichting and Sari (2008), then the semi-major axis of
the orbit shrunk very quickly and, unless the dynamical friction is applied only for a window
of a few tens to a few hundreds of years a collision almost always occurs. This seems to
argue against strong dynamical friction as being a source of wide binaries, that is, unless the
dynamical friction is applied only for a relatively short time.
In contrast, using a very small value for the parameters D& and D⊥ , e.g., D& = D⊥ =
0.00015 leads to a gradual reduction in the semi-major axis. Because the dynamical friction
term is a small perturbation then the orbital elements tend to retain their original values—this
is an example of adiabatic switching (Skodje and Cary 1998). Figure 4 shows the results of
these simulations and demonstrates that this model is capable of producing mutual orbits
rather similar to that observed for 2001 QW322 .
In conclusion, the CAC capture mechanism, appropriately modified, is able to explain
each of the possible orbits proposed by Petit et al. (2008) for 2001 QW322 . This suggests
that, perhaps, KBBs are not produced by a unique formation mechanism and that, as proposed
by Astakhov et al. (2005) and Lee et al. (2007), likely each of the proposed formation mech-
anisms may have operated separately or in combination to produce the observed population
of KBBs.

Acknowledgments We thank J.-M. Petit and an anonymous referee for their insightful and helpful reviews
of the manuscript. This work was supported, in part, by a grant from the US National Science Foundation.

References

Astakhov, S.A., Burbanks, A.D., Wiggins, S., Farrelly, D.: Chaos-assisted capture of irregular
moons. Nature 423, 264–267 (2003)
Astakhov, S.A., Farrelly, D.: Capture and escape in the elliptic restricted three-body problem. Mon. Not. Roy.
Astron. Soc. 354, 971–979 (2004)
Astakhov, S.A., Lee, E.A., Farrelly, D.: Formation of Kuiper-belt binaries through multiple chaotic scattering
encounters with low-mass intruders. Mon. Not. Roy. Astron. Soc. 360, 401–415 (2005)
Circi, C., Teofilatto, P.: Effect of planetary eccentricity on ballistic capture in the solar system. Celes. Mech.
Dyn. Astron. 93, 69–86 (2005)
Danby, J.M.: Fundamentals of Celestial Mechanics. pp. 255–270. 2nd edn. Willmann-Bell, Richmond,
VA (1992)
Deprit, A., Williams, C.A.: The Lissajous transformation IV. Delaunay and Lissajous variables. Celes. Mech.
Dyn. Astron. 51, 271–280 (1991)
Ferraz-Mello, S., Rodríguez, A., Hussmann, H.: Tidal friction in close-in satellites and exoplanets: The Darwin
theory re-visited. Celes. Mech. Dyn. Astron. 101, 171–201 (2008)
Frœschlé, C., Guzzo, M., Lega, E.: Graphical evolution of the Arnold web: from order to chaos.
Science 289, 2108–2110 (2000)
Frœschlé, C., Lega, E., Guzzo, M.: Analysis of the chaotic behavior of orbits diffusing along the Arnold
web. Celes. Mech. Dyn. Astron. 95, 141–153 (2006)
Funato, Y., Makino, J., Hut, P., Kokubo, E., Kinoshita, D.: The formation of Kuiper-belt binaries through
exchange reactions. Nature 427, 518–520 (2004)
Goldreich, P., Lithwick, Y., Sari, R.: Formation of Kuiper-belt binaries by dynamical friction and three-body
encounters. Nature 420, 643–646 (2002)
Goldreich, P., Lithwick, Y., Sari, R.: Planet formation by coagulation: A focus on Uranus and Neptune. Annu.
Rev. Astron. Astrophys. 42, 549–601 (2004)

123
Formation of the extreme Kuiper-belt binary 2001 QW322 259

Gomes, R.S.: On the origin of the Kuiper belt. Celes. Mech. Dyn. Astron. 104, 39–51 (2009)
Kandrup, H.E., Sideris, I.V.: Chaos in cuspy triaxial galaxies with a supermassive black hole: a simple toy
model. Celes. Mech. Dyn. Astron. 82, 61–81 (2002)
Kenyon, S.J., Luu, J.X.: Accretion in the early outer solar system. Ap. J. 526, 465–470 (1999)
Lee, E.A., Astakhov, S.A., Farrelly, D.: Production of trans-Neptunian binaries through chaos-assisted
capture. Mon. Not. Roy. Astron. Soc. 379, 229–246 (2007)
Levi Civita, T.: Sur la résolution qualitative du probléme restreint des trois corps. Acta. Math. 30,
305–327 (1906)
Lichtenberg, A.J., Lieberman, M.A.: Regular and Chaotic Dynamics. pp. 380–386. 2nd edn. Springer, New
York (1992)
Makó, Z., Szenkovits, F.: Capture in the circular and elliptic restricted three-body problem. Celes. Mech. Dyn.
Astron. 90, 51–58 (2004)
Noll, K.S.: Transneptunian binaries. Earth Moon Planets 92, 395–407 (2003)
Perry, A.D., Wiggins, S.: KAM tori are very sticky: rigorous lower bounds on the time to move away from an
invariant Lagrangian torus with linear flow. Physica D 71, 101–121 (1994)
Petit, J.-M., Henon, M.: Series expansions for encounter-type solutions of Hill’s problem. Cel. Mech. 38, 67–
100 (1986)
Petit, J.M., Mousis, O.: KBO binaries: How numerous were they?. Icarus 168, 409–419 (2004)
Petit, J.M., Kavelaars, J.J., Gladman, B.J., Margot, J.L., Nicholson, P.D., Jones, R.L., Parker, J.W., Ashby,
M.L.N., Bagatin, A.C., Benavidez, P., Coffey, J., Rousselot, P., Mousis, O., Taylor, P.A.: The extreme
Kuiper-belt binary 2001 QW322 . Science 322, 432–434 (2008)
Queck, M., Krivov, A.V., Sremc̆ević, M., Thébault, P.: Collisional velocities and rates in resonant planetesimal
belts. Celes. Mech. Dyn. Astron. 99, 169–196 (2007)
Schlichting, H.E., Sari, R.: Formation of Kuiper belt binaries. Ap. J. 673, 1218–1224 (2008)
Simó, C., Stuchi, T.J.: Central stable/unstable manifolds and the destruction of KAM tori in the planar Hill
problem. Physica D 140, 1–32 (2000)
Skodje, R.T., Cary, J.R.: An analysis of the adiabatic switching method - Foundations and applications. Comp.
Phys. Rep. 8, 223–292 (1998)
Stiefel, E.L., Scheifle, G.: Linear and Regular Celestial Mechanics. Springer, Berlin (1971)
Weidenschilling, S.J.: On the origin of binary transneptunian objects. Icarus 160, 212–215 (2002)

123

You might also like