JOURNAL OF GUIDANCE, CONTROL, AND DYNAMICS

Vol. 38, No. 10, October 2015

Dynamic Mechanisms for Spacecraft Disposal

from SunEarth Libration Points
Zubin P. Olikara and Gerard Gmez
University of Barcelona, 08007 Barcelona, Spain
and
Josep J. Masdemont
Universitat Politcnica de Catalunya, 08028 Barcelona, Spain

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

DOI: 10.2514/1.G000581
This work investigates end-of-life trajectories for spacecraft in orbit about the sunEarth L1 and L2 libration
points. A decommission plan is often required during the mission design process. The spacecrafts natural departure
dynamics are studied in both the circular restricted three-body problem and an ephemeris model. In particular, the
roles of the unstable manifold and forbidden regions in determining disposal outcomes are considered. A close
correspondence is observed between the models when computing the probabilities of return to the Earth
neighborhood. A simple maneuver scheme, which uses regions partitioned by the zero-velocity surfaces to prevent
returns to the Earth vicinity, is also analyzed. Disposal maneuver costs less than 50 ms are possible for many initial
libration point orbits during a decommissioning phase lasting less than one year.

I.

heliocentric orbit, and will not return to the Earth vicinity for at least
300 years. A similar scheme was performed for Planck in
October 2013 to raise the spacecraft into a long-term parking orbit.
These disposals set a precedent for current missions, such as the
Gaia space observatory launched in December 2013 en route to a
sunEarth L2 Lissajous orbit.**
Constraints present during the decommissioning process are
distinguishable from other mission phase requirements. These
constraints play a fundamental role in the end-of-life analysis and the
determination of feasible trajectories. In the current study, the
primary requirements considered are the following:
1) For protection of the Earth and its satellite orbit zones, a spacecraft
passing through protected regimes such as low Earth or geosynchronous
orbits has strict requirements to mitigate the possibility of collisions [1].
Furthermore, if a spacecraft reenters the Earths atmosphere, a detailed
analysis is required to ensure it does so in a safe manner. For many
spacecraft, such as those using radioisotope power generation, strict
avoidance of the Earth is necessary.
2) It is required that only limited fuel remain aboard the spacecraft.
After a spacecraft completes its primary mission, a majority of its fuel
will be spent. It is expected, however, that future missions will have
some fuel budgeted specifically for the end-of-life phase. In this
analysis, 100 ms of v is assumed to be about the maximum
maneuver capability, though potentially less may be available.
3) Because of mission operation costs, it is desirable for
decommissioning to be completed in the least amount of time possible,
and so a restricted time window for final maneuvers is required. Ideally,
all maneuvers would be completed within 36 months, though this
may not always be feasible. A decommission period of about one year
is considered to be the maximum practical time available.
4) A robust end-of-life scheme is required. If very careful planning
and maneuvering are necessary for a disposal scheme, the associated
costs and risks may make it less desirable than a simpler, though
perhaps nominally more expensive, option. A less complicated
scheme is preferred to maximize the chance of successful disposal.
Individual mission needs will introduce additional constraints to
the design process. Logistically, the availability of resources, such as
ground stations and operations teams, will play a role. Furthermore,
orbit and attitude determination will be needed before and after any
disposal maneuvers, and a small clean-up maneuver may be required

Introduction

N END-OF-LIFE plan is a fundamental aspect of current and

future space missions. This is a well-recognized requirement for
Earth-orbiting satellites given the need to minimize the accumulation
of objects in densely populated orbital zones [1]. Libration point
orbits are also important regions of interest for upcoming missions
[2], but they have significantly different characteristics. The most
commonly used libration points, the sunEarth L1 and L2 points,
have an unstable dynamic component, causing spacecraft to naturally
depart this regime. These departure trajectories, however, can have
many possible outcomes, including returns to the Earth vicinity.
Thus, for spacecraft originating near a libration point, consideration
should be given to the selection of the desired outcome subject to
mission constraints.
The design of libration point missions has evolved over the last
several decades. The first spacecraft to orbit a libration point was
NASAs International SunEarth Explorer launched in 1978. After
its original mission at an L1 halo orbit and extended mission as a
cometary explorer were complete, it was decommissioned into a
heliocentric orbit and made a passage by the Earth in August 2014
[3]. Dynamic systems techniques, which are leveraged in the
current study, were proposed for the trajectory design of the Solar
and Heliospheric Observatory [4,5], a joint ESA and NASA
mission launched in 1995. Although these techniques were not
ultimately used in the final design, they have been applied to
subsequent libration point missions. Recent examples include
ESAs Herschel and Planck spacecraft launched together
in 2009. They were placed into halo and Lissajous orbits,
respectively, about the sunEarth L2 libration point. Herschel was
decommissioned in May 2013 with a maneuver of 114 ms into a
Received 21 February 2014; revision received 8 May 2015; accepted for
publication 14 May 2015; published online 7 September 2015. Copyright
2015 by Institut d'Estudis Espacials de Catalunya. Published by the American
Institute of Aeronautics and Astronautics, Inc., with permission. Copies of this
paper may be made for personal or internal use, on condition that the copier
pay the $10.00 per-copy fee to the Copyright Clearance Center, Inc., 222
Rosewood Drive, Danvers, MA 01923; include the code 1533-3884/15 and
$10.00 in correspondence with the CCC.
*Marie Curie Fellow, IEEC, Departament de Matemtica Aplicada i
Anlisi, Gran Via 585; zubin@maia.ub.es.

Professor, IEEC, Departament de Matemtica Aplicada i Anlisi, Gran Via

585; gerard@maia.ub.es.

Department Chair and Professor, IEEC, Departament de Matemtica

Aplicada I, ETSEIB, Diagonal 647; josep@barquins.upc.edu.

Data available online at http://sci.esa.int/herschel/52797-herschel-statusreport-05-2013 [retrieved 4 September 2013].

Data available online at http://www.esa.int/Our_Activities/Operations/

Planck_on_course_for_safe_retirement [retrieved 20 January 2014].
**Data available online at http://sci.esa.int/gaia/53536-esa-pr-44-2013liftoff-for-esas-billion-star-surveyor [retrieved 20 January 2014].

1976

1977

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

OLIKARA, GMEZ, AND MASDEMONT

to correct any errors. Before the final system shutdown, the tanks
should be empty of any propellant.
The aim of this paper, however, is to give an initial analysis of the
dynamics, as well as a simple disposal technique, including
associated costs. Given the limited fuel remaining and time to
perform maneuvers, natural dynamics will drive the motion and
understanding that provide a framework for the end-of-life design.
This work could be viewed as a starting point for a particular
missions end-of-life study. Independently, recent studies have
explored the feasibility of disposal options from libration point orbits,
including the use of solar sails and alternatives such as Earth reentry
and moon collisions [6,7].
The paper is organized as follows. The primary systems of interest,
circular restricted three-body problem (CR3BP), and ephemeris model
are introduced, and computational aspects relevant for later analysis are
discussed. Natural dynamics of a spacecraft originating in a libration
point orbit are then considered. Because of the dynamic similarities
between the equilibria, the focus is on departure dynamics near L2.
Numerical verification demonstrates that the relevant dynamics are
captured by the CR3BP and that the flow away from the initial orbit is
dominated by the unstable manifold. In addition, the role of the stable
manifold and the zero-velocity surface (ZVS) for bounding the
spacecrafts motion are discussed, and investigation of the associated
probabilities of a return to Earths vicinity is performed. To reduce this
outcomes probability, performing a simple maneuver to modify the
ZVS geometry is considered. The cost associated with various
maneuver locations are presented. Because the costs are based on the
CR3BP, discussion is provided on how these maneuvers can be applied
to a more realistic ephemeris model, and validation of the long-term
orbit lifetimes is conducted.
Note that, during the course of this analysis, simple impulsive
maneuvers are used. This helps to simplify the analysis, while also
reflecting the desire to avoid complex maneuver schemes during the
mission decommissioning phase. Although other means of thrust,
particularly low-thrust schemes, are not considered, one of the
primary drivers of this analysis is the limited fuel remaining during
the end-of-life phase. Thus, natural dynamics play a fundamental role
and the particular thrust method is of lesser importance during the
preliminary design.

II.

Modeling and Numerical Techniques

In this section, the principal models, namely, the CR3BP and a

planetary ephemeris model, used in the end-of-life analysis are
presented. For each model, the nominal libration point orbits from
which the study begins are provided. In addition, relevant numerical
techniques, particularly the computation of the natural departure
directions and the transformation between models, are given. These
tools provide a framework for the investigation of the underlying
dynamics and in the design of a disposal scheme.

primaries are located at x  and x  1 , respectively), and z^ is a

unit vector along the z axis, the direction of the systems angular
momentum. To make the sunEarth model more coherent, the second
primary is in fact the Earth system barycenter, and the mass parameter
 3.0404234 106 includes the contribution from the moons mass.
An important consequence of this model (1) is that it admits an
integral of motion, which is connected to the existence of forbidden
regions that are fundamental to the later analysis. One version of this
integral is known as the Jacobi constant
Cr; v 2Ur kvk2

(3)

which is a function of the spacecrafts state [note that some authors

include a constant term 1  in this definition]. When the
equations of motion are derived from a Hamiltonian perspective, an
equivalent integral H C2 is obtained. This integral serves as an
energy-like constant, and so an increase in the Jacobi constant
corresponds to a decrease in the spacecrafts energy.
The CR3BPs five equilibrium points are the L1 L5 libration
points. The L1 libration point lies between the sun and the Earth on
the x axis. The L2 libration point lies beyond the Earth on the x axis.
The L1 and L2 libration points each lie approximately 1.5 106 km
from the Earth and have stability of type center center saddle.
Additional discussion of the CR3BP can be found in many references
including, notably [8].
A set of nominal libration point orbits in the sunEarth system
serve as the starting point for the analysis. These orbits are
representative of the range of spacecraft states at the end of the regular
mission and before the end-of-life phase. In the plane of the primary
bodies, consider the Lyapunov family of periodic orbits about L2
shown in Fig. 1 with y amplitude increments of 100,000 km. Note that
the plot axes use astronomical units (1 AU  1.496 108 km),
which are equivalent to the nondimensional length units used in
Eq. (1). When the family crosses a y amplitude of approximately
600,000 km, it bifurcates into the out-of-plane L2 halo periodic orbit
family. Some members of the northern branch of the family are
shown in Fig. 1 with z amplitude increments of 100,000 km. The
southern branch has an identical structure mirrored across the
primaries plane. All the orbits have a period of about six months.
Note that these families of periodic solutions represent just a small
part of the libration points center manifold. Nevertheless, they capture
much of the relevant motion of nearby quasi-periodic solutions. Thus,
for this initial analysis, a primary focus will be on spacecraft originating
in a periodic orbit. Given the interest in Lissajous-type trajectories,
however, the quasi-periodic solutions shown in Fig. 2 will also be
considered. These square Lissajous orbits have approximately equal
y and z amplitudes, one smaller orbit with amplitude 100,000 km, and
another larger orbit with amplitude 600,000 km.
For simplicity and to limit the scope of the current end-of-life
analysis, the L2 families of orbits are primarily considered. However,
given the small mass parameter of the sunEarth system, the L1

A. Circular Restricted Three-Body Problem

The CR3BP considers the motion of two massive primary bodies

(for this study, the sun and the Earth) that are assumed to move in a
circular orbit about their center of mass. Let a third body (the
spacecraft) be positioned at r  x; y; z in a barycentric rotating
frame defined such that the primary bodies are fixed along the x axis.
The third bodys six-dimensional state r; v is governed by the
nondimensional equation of motion
  

r_
v

(1)
v_
r U  2v z^
which includes the gradient of a force potential-like function
Ur

1 1 2
  x  y2 
r1
r2 2

(2)

The parameter 0; 0.5 relates the primaries masses, r1 and r2 are
the distances of the third body to each primary (the larger and smaller

Fig. 1 L2 planar Lyapunov (left) and out-of-plane halo (right) periodic

orbits (lengths in AU).

1978

OLIKARA, GMEZ, AND MASDEMONT

B. Ephemeris Model

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

Fig. 2 L2 square Lissajous quasi-periodic orbits with 100,000 (left) and

600,000 (right) km amplitudes (lengths in AU).

families exhibit similar trends. The differences are noted where

applicable. For example, a departure trajectory outside the Earths
orbit that originates near L2 and a departure trajectory inside the
Earths orbit that originates near L1 may revolve around the rotating
frame barycenter in opposite directions, but qualitatively they display
similar behavior.
Because the stable and unstable manifolds play a fundamental role
in the dynamic structure, a brief discussion on their computation for
the case of a periodic orbit follows. These manifolds consist of
trajectories that asymptotically approach the periodic orbit in forward
or backward time. A simple, numerical approximation of each
manifolds tangent direction at the periodic orbit is available from the
state-transition matrix . It can be computed by integrating a matrix
differential equation
_  Axt

(4)

where A fx with initial condition 0  I, along with the

equations of motion x_  fx. After propagating over one period of
the periodic orbit, this gives the monodromy matrix T. From the
center manifold theorem [9], the monodromy matrixs eigendecomposition provides a local basis for the motion near the periodic
orbit. Assume that a pair of real eigenvalues s;u exists off the unit
circle: js j < 1 and ju j > 1. Then, the corresponding stable
eigenvector, denoted ws 0, points along states that asymptotically
approach the periodic orbit. The corresponding unstable eigenvector,
denoted wu 0, points along states that asymptotically depart the
periodic orbit.
These eigenvectors can be mapped around the periodic orbit using
the state-transition matrix: ws;u t  tws;u 0. For convenience,
this vector is normalized and reparameterized in terms of an angle:
^ s;u 2tT ws;u tkws;u tk. Because this is a local reprew
sentation, a basic approach to approximate a state on the true stable or
unstable manifold is to make a perturbation in one of these directions
^ s;u 
xs;u
0   x0   w

(5)

where is a small parameter. A simple choice for the sunEarth

system is to select such that the dimensionalized magnitude of the
position perturbation is 200 km. To globalize the manifold, the
perturbed initial condition is propagated forward or backward in
time. For the included simulations, numerical integration is
performed using a BulirschStoer algorithm [10] with error
tolerances of 1012 (or tighter). Note that the perturbation can be
added or subtracted corresponding to the two halves of each
manifold. For the L1 and L2 libration points, the orbits of interest will
have one-half heading toward the smaller primary and one-half
heading away from it. The procedure for generating the manifold of a
quasi-periodic orbit with a saddle dynamic component is similar
except that the eigenvectors depend on a pair of angles. A numerical
procedure is available in [11].

Although the CR3BP captures much of the relevant dynamics, for

higher-fidelity end-of-life analysis, the precise positions of solar
system bodies provided in NASA Jet Propulsion Laboratorys
DE422 ephemeris are used. In particular, the positions of the sun,
Earth, moon, Jupiter, and Saturn are included. Note that only their
point mass gravitational influences are modeled, neglecting any
nonuniformities in their gravitational field. The resulting model for
the spacecrafts motion is a restricted n  1  6-body problem. An
arbitrarily selected epoch of 18:00 Universal Coordinated Time on
1 January 2015 is used for this preliminary study.
The equations of motion describe the evolution of the spacecrafts
position vector, denoted qt, in an inertial frame centered at the
larger primary, the sun. The ephemeris provides the position vectors
Ri t of the other solar system bodies relative to the sun, as well as all
their masses mi . Defining qi t qt Ri t as the position of the
spacecraft relative to each perturbing body, the spacecrafts dynamics
are governed by
q  Gm1



n
X
q
qi t
Ri t

Gm

i
kqk3 i2
kqi tk3 kRi tk3

(6)

where G is the gravitational constant. This system differs in some

notable ways from Eq. (1). First, the spacecrafts state is no longer
defined in a barycentric rotating frame of reference but rather in a suncentered inertial frame. Second, the system no longer admits an
integral of motion. Third, the trajectories are dependent on the choice
of epoch, and periodic solutions no longer exist. Nevertheless, the
following tools enable one to establish relationships between
dynamic structures in the CR3BP and trajectories computed in the
ephemeris model.
In appropriate regimes, the ephemeris model can be considered a
perturbation of the CR3BP system. However, the equations of motion
(1) and (6) are formulated with respect to different reference frames,
and their time and state variables are scaled differently. One can view
the independent variable in the CR3BP as the true anomaly f
(alternatively, the mean or eccentric anomaly could be used) of the
Earthmoon barycenter relative to the sun. The CR3BP trajectory
r; v is first reparameterized in terms of physical times t starting
at a specified epoch. The dimensional position vector qt in a suncentered inertial frame is then related to the nondimensional position
vector rf in a barycentric sunEarth rotating frame, using a
transformation of the form
^
qt  CtRtrft  x

(7)

The term x^ shifts the origin to the sun. The scaling factor
Rt kRtk dimensionalizes the length using the ephemeris
vector Rt pointing from the sun to the Earthmoon barycenter. The
rotation matrix Ct converts the position vector from the rotating
frame to the inertial frame. Note that the rotating frame is oriented
such that the x axis is aligned with Rt, the z axis is aligned with
_
angular momentum Rt Rt,
and the y axis is selected orthogonal
to both, such that the reference frame is right-handed. Computing the
_ is simply a matter of computing the derivative of
inertial velocity qt
Eq. (7) using the product and chain rules. More detailed discussion of
transformation approaches can be found in [5]. Applying the process
in reverse, one can transform back into nondimensional rotating
frame states. This is useful for representing the ephemeris solutions in
a familiar rotating reference frame and will also be used for
determining the maneuver direction in the later results.
Transforming solely the initial state from the CR3BP to the
ephemeris model and then propagating forward in time, a trajectory
will quickly leave the vicinity of the libration point. However, by
looking at a collection of transformed states on the trajectory, they can
all be adjusted slightly such that the ephemeris dynamics are
satisfied, while still maintaining the qualitative structure of the

Data available online at http://ssd.jpl.nasa.gov/?planet_eph_export

[retrieved 14 August 2015].

1979

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

OLIKARA, GMEZ, AND MASDEMONT

original orbit in the CR3BP. This procedure can be accomplished

using a predictorcorrector method. The prediction step involves
transforming a sequence of times and states along the CR3BP
trajectory into the dimensional inertial frame, using the procedure
described earlier. The second step of the method requires correcting
the states such that a continuous natural trajectory connects them. A
simple procedure to do this is via multiple shooting. More details can
be found in, for example, [12]. Integrating each segment for
approximately two months leads to position discontinuities less than
1 cm and velocity discontinuities less than 1 mm/s. In practice, these
discontinuities are negligible compared with limitations in a
spacecrafts state determination and modeling uncertainties.
By applying this approach, one can begin to study the dynamics in
an ephemeris model. The converted versions of the ephemeris
solutions can be viewed as the initial orbits from which the highfidelity end-of-life trajectories originate. For the moment, consider
the L2 families of Lyapunov and halo orbits. These are periodic orbits
in the CR3BP as shown in Fig. 1. The ephemeris model, however, is
time dependent, and periodic solutions no longer exist. Nevertheless,
very similar orbits can be generated, which are shown in Fig. 3,
transformed back into the nondimensional rotating frame for four
periods covering approximately two years, though the orbits do not
exactly repeat.
Although the exact periodic or quasi-periodic structure does not
remain in the transition to a high-fidelity model, the unstable dynamic
component persists. One can numerically approximate the direction of
fastest departure from the ephemeris trajectory. This is the direction in
which a perturbation will take the spacecraft as far away (following the
linearized dynamics) as possible from the nominal trajectory. Instead of
using the monodromy matrix, which depends on having periodicity, a
power-like approach is used for the computation [5].
Consider an ephemeris trajectory having several revolutions, but let
the state-transition matrix t; ti  be partitioned over limited intervals
t ti ; ti1 , for example, a single revolution. This matrix maps a
perturbation wti  at the beginning of the time interval to the perturbation
wt at time t. The collection of these matrices gives the complete statetransition matrix t; t0   t; ti ti ; ti1  t1 ; t0 , but for
numerical overflow reasons, it is not computed directly.
The principal idea of the approach is that a random initial
perturbation wt0  will have some component in the unstable
direction. An alternative choice would be to take the transformed
initial unstable manifold direction from the CR3BP. Assuming a
dominant eigenvalue (Lyapunov exponent), the linear evolution of
this vector should align itself with the unstable direction (i.e., the
direction of greatest expansion) in a short period of time. Its evolution
can be sequentially (i  0; 1; : : : ) computed as follows:
wt  t; ti 

wti 
i

for t ti ; ti1 

(8)

Fig. 3 Ephemeris trajectories associated with Lyapunov and halo orbit

families.

where i kwti k measures the growth of the vector over each

interval. Division by i at each step prevents the vector wt from
growing more than the largest eigenvalue of matrix ti1 ; ti . One
^ of this vector such that it
can then define a normalized version wt
has unit magnitude at each instant of time.
Because some time is required for the unstable component of the
randomly selected initial vector to dominate the other components, an
initial portion of the ephemeris trajectory is removed. For the
examples considered in this paper, two revolutions appears sufficient.
Note that the preceding computations are performed in the inertial
frame of the ephemeris trajectory. Analogous to the CR3BP setting, a
small perturbation with either sign can be added and then propagated
using the full equations of motion to generate the two halves of the
unstable manifold substitute. The trajectories can then be converted
into the rotating frame.

III.

Natural Dynamics

In this section, the natural motions of a spacecraft originating in a

libration point orbit are investigated. Given the limited fuel available
for the end-of-life mission phase, the spacecrafts behavior is
principally governed by the gravitational influence of solar system
bodies. Within the setting of the sunEarth CR3BP, the natural
departure and return mechanisms have a geometric interpretation
with defined boundaries. The role of this structure, particularly the
zero-velocity surfaces and stable and unstable manifolds, is
considered first. Then the observed behaviors and outcomes are
validated in the ephemeris model using a Monte Carlo study.
A. CR3BP Dynamic Structures

For the current study, the primary focus in the classification of the
natural motions is the determination of trajectories that return to the
Earth vicinity. Given the end-of-life requirements, the emphasis is on
minimizing the probability of this outcome. The existence of an
integral of motion and invariant manifolds is leveraged for this
analysis. In particular, this leads to boundaries in position space
separating the flow. The consequence of this structure is that a
spacecraft can only cross the Earths orbit if it passes through defined
gateways.
Fixing a Jacobi constant level C0 and examining Eq. (3), it is
apparent that 2Ur C0 . Regions in position space where this
inequality is violated are referred to as forbidden regions. The
boundary satisfies the relation 2Ur  C0 , which requires that
v  0. This so-called ZVS is a two-dimensional manifold in the
three-dimensional position space. If one considers motion restricted
to the plane of the primaries, the boundary is a one-dimensional zerovelocity curve (ZVC) that forms a barrier in the two-dimensional
position space. The ZVC and forbidden region are illustrated in Fig. 4
for three different Jacobi constant levels. In the first illustration, there
is a bottleneck region around the Earth that allows a spacecraft to
access the interior region about the sun and the exterior region away
from the primaries. As the Jacobi constant is increased (i.e., the
energy is decreased), the opening at the L2 libration point closes as
shown in the second illustration. As it is further increased, the
opening at L1 closes. These boundaries set access routes to the Earth
vicinity and will play a role in the end-of-life design presented.
The other principal tools for analyzing the departure and return
dynamics are the stable and unstable invariant manifolds associated
with the spacecrafts libration point orbit. The unstable manifold
dominates the natural motion leaving the starting orbit. For the
periodic orbits of interest, these manifolds are two-dimensional
objects, and for the quasi-periodic orbits of interest, these manifolds
are three-dimensional objects. The benefit of taking advantage of this
structure is that it allows one to easily parameterize states departing
the initial orbit using two or three variables, without searching in the
full six-dimensional phase space.
Recall that the sign of the perturbation in Eq. (5) used to compute
the starting manifold states is free. Considering departure trajectories,
this allows one to generate two halves of the unstable manifold: one
leaving toward the Earth and one leaving away from the Earth. In
Fig. 5, the half of an L2 orbits unstable manifold departing away

1980

OLIKARA, GMEZ, AND MASDEMONT

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

Fig. 4

Qualitative representations (not to scale) of zero-velocity curves and forbidden regions at sequence of Jacobi constant levels C.

Fig. 5 Unstable manifold departing away from Earth (left) and randomly perturbed trajectories (right) from initial Lyapunov orbit.

from the Earth is shown. Also shown are trajectories starting from
random perturbations on the orbit state. The underlying influence of
the unstable manifold is apparent in these results. As a consequence,
following solely the natural dynamics, there is a 50% chance of
heading toward the Earth (which, unless done intentionally, should be
avoided), and a 50% chance of departing away from the Earth.
However, a small maneuver should be possible to switch between the
outcomes. The fundamental idea is that the projection of the
postmaneuver state onto the unstable subspace should be directed
away from the Earth.
The stable and unstable manifolds also provide a geometric means
for categorizing trajectories by their return outcomes. This structure
can be most easily understood for the case of planar motion in the

Fig. 6
right.

CR3BP. Consider a fixed Jacobi constant level at which there is a

Lyapunov orbit located at the gateway to the bottleneck region. The
stable manifold is a two-dimensional object dividing the threedimensional energy level space. The fundamental result is that any
trajectory from outside the bottleneck must lie in the interior of the
stable manifold to return to the Earth vicinity [13]. The stable
manifold itself is the boundary between return and nonreturn
trajectories. To visualize this behavior, the surface of section y  0
with condition y_ > 0 is used. In Fig. 6, the plot shows the closed
curves representing the first intersections of an L2 Lyapunov orbits
stable and unstable manifolds in the exterior region. The unstable
manifold intersections represent the collection of trajectories
departing the Lyapunov orbit, and the segments lying in the interior of

First intersections of stable (blue) and unstable (red) manifolds with surface of section. Expanded view of return trajectories (black) shown on

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

OLIKARA, GMEZ, AND MASDEMONT

the stable manifold intersection will reenter the bottleneck after one
revolution of the ZVS. The second intersections of the remaining
unstable manifold segments can be used to find returns after two
revolutions, and so forth. The primary challenge with this approach is
that the geometry can become quite complicated, making it difficult
to determine whether segments in fact lie in the interior. Even for the
first intersections shown in Fig. 6, the closed curves are distorted into
narrow spiral-like shapes. The situation becomes even more complex
in the spatial CR3BP, but a similar structure bounds trajectories
returning to the Earth vicinity. However, in the spatial case, the stable
manifold of the center manifold should be considered.
Although the direct application of the stable manifold to the
computation of returns can be challenging, it still provides an
important geometric understanding of the return behavior. Instead of
algorithmically determining whether a segment lies in the interior of
the stable manifold curve, one can equivalently check whether the
trajectory reenters the bottleneck by monitoring its distance from the
Earth. For the results shown in Fig. 6, start with many (say N 104 )
evenly spaced initial conditions perturbed to be on the unstable
manifold from the initial periodic orbit. After propagating, one can
determine which initial conditions return that have a neighbor that
does not return. One can then iteratively locate the boundary of these
intervals between return and nonreturn orbits. In fact, these
boundaries correspond to homoclinic connections.
Some comments on this computational procedure are worthwhile.
It depends on using a stable and accurate numerical integration
scheme, and even then may only be accurate for the first two or three
revolutions of the ZVS (about 30 years). It also may miss return
segments corresponding to a departure interval on the Lyapunov orbit
less than 1N of the period. It provides, however, an initial look at
intervals on the unstable manifold that naturally return to the Earth
region. Refinements of this basic scheme are possible to make it more
robust. For example, instead of considering evenly spaced departure
trajectories from the initial orbit, the step size between the departure
trajectories can be adjusted based on the change in the final state.
Similar ideas could be applied to the study of out-of-plane orbits, but
they are not investigated now.
B. Monte Carlo Return Study

The correspondence between the CR3BP and ephemeris dynamics

is a main premise upon which this work is based. As a first
comparison, a Monte Carlo analysis is performed about the family of
reference L2 Lyapunov orbits in each model. Random position
perturbations of magnitude 200 km are applied to states along the
orbit. As long as the perturbation is sufficiently small, the precise
choice of perturbation has minimal influence on the results. A total of
105 trajectories are generated from the set of initial conditions for
each orbit. Because of the unstable manifolds influence, in both
models about half the trajectories are observed to depart toward the

1981

Earth and about half depart away from the primaries. Because of the
end-of-life restrictions in the Earth vicinity, the focus is on this latter
set of trajectories in the exterior region.
To classify the outcome, one can check whether the exterior
spacecraft trajectory eventually returns to the Earth vicinity. This is
defined as passing within 1 106 km of the Earth. When the set of
exterior Monte Carlo trajectories first depart the libration point orbit,
none of the trajectories have entered the Earth vicinity. As longer
intervals of time are considered, an increasing percentage of the
trajectories will have crossed into this region at some point along their
path. This is illustrated in Fig. 7 over a 100 year time period. The first
spacecraft returns appear after approximately 15 years. A second set
of returns appear between 25 and 30 years from their libration point
orbit departure. This trend is seen in both the CR3BP and ephemeris
results. An additional observation is that natural trajectories
departing from larger Lyapunov orbits have a higher probability of
returning to the Earth. It should also be noted that the ephemeris
trajectories considered tend to have a 12% higher chance of return
within 100 years. This is likely due to the influence of the moon,
which is in an orbit of radius 3.85 105 km relative to the Earth
(rather than being located at the Earth system barycenter as modeled
by the CR3BP) and is a direction for future investigation.
Nevertheless, there is a close relation between the return behavior
seen in the CR3BP and ephemeris models.
The results can be understood by considering the forbidden
regions and zero-velocity surface presented earlier. At the Jacobi
constant level of the libration point orbits, the forbidden region makes
a thin circular shape centered about the sun with a small opening at
the Earth. For an exterior trajectory, the semimajor axis is larger than
the Earths and, thus, has a longer two-body period about the sun.
Therefore, in the rotating frame, the trajectory rotates clockwise
while bouncing off the ZVS. Depending on the particular initial
conditions, after one revolution, the trajectory can either enter
through the L2 gateway or make a subsequent revolution around the
ZVS. This is the basis for the jumps at about 15 year intervals seen in
Fig. 7. Note that similar behavior is observed for interior trajectories
departing from an L1 libration point orbit except that their two-body
period is less than the Earths, and thus, they rotate counterclockwise.
The ZVS influences the Earth return results in an additional manner.
Larger libration point orbits correspond to Jacobi constant levels with
larger gateway openings to the Earth regime. Therefore, a trajectory
departing a larger orbit will have a greater probability of returning to
the Earth after a revolution about the ZVS. This explains the trend
observed in Fig. 7 over various orbit sizes.
The focus so far has been on trajectories that depart a libration
point orbit in a direction away from the Earth. For trajectories directed
toward the Earth, as a simple study, the same set of L2 Lyapunov
orbits in the CR3BP are considered. Their probabilities of passing
within the geosynchronous radius over a 10 year interval is

Fig. 7 Monte Carlo study of exterior trajectories returning to Earth vicinity in CR3BP (left) and ephemeris (right) models for L2 Lyapunov orbits with
specified amplitudes.

1982

OLIKARA, GMEZ, AND MASDEMONT

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

Fig. 9 Illustration of an end-of-life scheme: 1) depart along unstable

manifold; 2) perform maneuver to close ZVS; 3) disposed spacecraft
follows natural motion outside ZVS.

Fig. 8 Monte Carlo study in CR3BP of bottleneck trajectories passing

within geosynchronous radius. Initial orbit amplitudes correspond to
Fig. 7.

investigated. In Fig. 8, one can see that, for the larger orbits
considered, well over 20% pass inside this radius within four years.
Although this coarse analysis does not consider the moon, it is
intended to illustrate that passes near the Earth are not an unusual
phenomena in the bottleneck region. The subsequent analysis will
thus focus on disposal to other regions of space.
Overall, it is apparent that there is a close correspondence between
the dynamics in an ephemeris model and the CR3BP outside of the
bottleneck region. In particular, the dynamics are strongly influenced
by the unstable manifold, which provides a set of natural end-of-life
trajectories that are easily parameterized. Some of these trajectories
have more desirable outcomes than others. If the probability of a
successful disposal is not deemed sufficient, maneuvers can be
incorporated to achieve a desired result.

IV.

Closing Zero-Velocity Surface

In this section, performing a simple maneuver to forbid a future

return to the Earth vicinity will be considered. If such a maneuver is
possible, the protection requirements for the Earth and its satellite
zones would be satisfied. The associated maneuver costs and times
will be analyzed to determine whether this is a feasible end-of-life
scenario. The primary idea will be to take advantage of the dynamic
structure, specifically the unstable manifold and the zero-velocity
surface, that exists in the CR3BP. It will then be verified that a shadow
of this structure persists when considering a more accurate
ephemeris model.
For a spacecraft orbit about L1 or L2, the libration point gateway
through the corresponding energy levels ZVS must be open. As
discussed in the previous section, this means that trajectories
departing the orbit have the possibility of eventually returning to
Earth. A maneuver, however, allows one to change the Jacobi
constant (or, equivalently, the energy level), effectively modifying the
geometry of the zero-velocity surface. If the level is changed
sufficiently such that the Earth region and the spacecrafts region are
disconnected by a forbidden region, a return to the Earth vicinity is
impossible under the CR3BP dynamics.
Before performing this maneuver, the unstable manifold will drive
the dynamics. To separate the spacecraft from the Earth region,
consider the half of the manifold departing away from the Earth
toward the interior region for an L1 spacecraft and toward the exterior
region for an L2 spacecraft. The regions are denoted in Fig. 4. A small
departure burn could be used to accomplish this, but this is not
currently studied in detail; however, some discussion is warranted.
The departing trajectory will initially shadow the nominal libration
point orbit as shown in Fig. 9 before leaving the vicinity. Depending
on the mission scenario, the science objectives could continue to be
met for a few months after the maneuver (recall that the orbit period is
about six months). In fact, as a contingency toward the end of the

primary or extended mission phase, the spacecrafts state could be

slightly biased from the nominal orbit onto the unstable manifold
directed away from the Earth. Then, in case of failure, the spacecraft
would naturally depart the Earths vicinity. This trajectory could be
selected so as not to return for a long period of time, though an
eventual return to the region is possible. Performing the departure
maneuver while science objectives are still being met has the added
benefit of reducing the effective decommissioning phase length. For
this reason, the disposal time periods presented can be viewed as
conservative. Furthermore, increasing the magnitude of the departure
burn (the current analysis corresponds to departure maneuvers less
than 1 ms) could reduce the decommissioning time.
For this preliminary analysis, as well as in the interest of disposal
simplicity, following the departure from the initial libration point
orbit, the most basic option is used: a single impulsive maneuver to
change the Jacobi constant value. Before the maneuver is performed,
the Jacobi constant value at any point along the unstable manifold
matches the initial orbits value C0 . Recalling the geometry of the
ZVS shown in Fig. 4, the postmaneuver Jacobi constant value must
be increased (energy is decreased) to at least the libration points
value to close the surface: C1 CLi (i  1; 2). Returning to Eq. (3),
for a spacecraft with position and premaneuver velocity r; v0 , the
maneuver v must satisfy
C1  2Ur kv0  vk2

(9)

One can first consider when a solution to this equation exists. If

C1 > 2Ur, there is no possible maneuver to close the gateway to
the Earth at this position. Physically, this means that the current
position r lies in the forbidden region at the Jacobi constant value C1 ,
and thus, this is not an appropriate place for a maneuver. Even if a
solution exists, certain positions r correspond to a state in the Earths
region after the maneuver and not in the interior or exterior region.
This is an undesirable outcome for an end-of-life disposal trajectory,
and these solutions are discarded.
A simple illustration of the geometry is shown in Fig. 9. A
trajectory departing away from the primaries on the unstable
manifold of an L2 libration point orbit is shown in blue. The zerovelocity surface at this Jacobi constant level C0 is shown in dark gray,
and in light gray is the ZVS at the Jacobi constant level C1  CL2 of
the libration point. Initially, the position r is in the forbidden region
relative to this new ZVS, then passes through the exterior region, the
forbidden region again, and finally out into the exterior region. The
end-of-life scheme described can be applied at any point on this
trajectory in the exterior region. It is not applicable for any of the other
segments.
Once appropriate positions on the unstable manifold have been
identified, the maneuver cost is calculated. From an inspection of
Eq. (9), it can be observed that the most efficient maneuver to increase
the Jacobi constant will be in the direction v0 . This allows the
maneuver magnitude to be computed as
v  v0

p
2Ur C1

(10)

OLIKARA, GMEZ, AND MASDEMONT

which is well defined for positions outside of the forbidden region.

Recall that, if C1  CLi , this maneuver closes the gateway such that
the Earth and spacecraft regions share a single point Li . Larger
maneuver magnitudes increase the width of the forbidden domain
separating these regions.
Further insight into the cost can be obtained by incorporating the
initial libration point orbits Jacobi constant value. This value C0 is
conserved along the unstable manifold, and so C0  2Ur v20 .
Substituting into Eq. (10) allows one to represent the maneuver
magnitude strictly in terms of position,
v 

p p
2Ur C0 2Ur C1

(11)

or equivalently, in terms of velocity,

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

v  v0

q
v20 C1 C0 

(12)

From the first relation (11), the cost decreases monotonically as Ur
increases inside the admissible region, 2Ur C1 > C0 . Considering positions away from the Earth region, the force potential (2) can
be approximated as
1 1
Ur  r2
r 2
taking into account that jzj r for practical sunEarth libration point
orbits and their unstable manifolds. The maneuver magnitude is thus
approximately minimized at the perihelion in the interior region (r < 1)
and the aphelion in the exterior region (r > 1). Note that this
corresponds to maximizing the rotating-frame velocity magnitude v0 ,
which minimizes the equivalent second relation (11) for v. Roughly
speaking, the maneuver cost decreases as distance from the ZVS
increases. This discussion does not, however, consider the premaneuver
time spent along the unstable manifold, which is an important practical
consideration in the following analysis.
Maneuver costs for various families of initial libration point orbits
can now be considered. In Fig. 10, costs for various size planar L2
Lyapunov orbits with amplitudes Ay ranging from 100,000 to
600,000 km are included. Note that the unstable manifolds associated
with the periodic orbits studied are two dimensional. For presenting
the results, the manifold is parameterized in the following manner. On
the vertical axis, an angle uniformly parameterizing the initial
periodic orbit in time is used. On the horizontal axis, the time
traveling along the unstable manifold is shown. Each trajectory on the
unstable manifold from a departure point on the periodic orbit traces
out a horizontal line. The coloring corresponds to the cost computed
using Eq. (10) to close the ZVS at that point along the manifold.
Uncolored regions correspond to states on the manifold that are either
in the Earth bottleneck or the forbidden region at the new Jacobi
constant level. It should be noted that these costs are best case
results because they correspond to the idealized sunEarth CR3BP
model. In the following section, adding a safety factor to this
maneuver size will be considered to account for perturbations.
Examining Fig. 10, one can observe that for each orbit initially
there is only a narrow segment that allows the ZVS to be closed and
the spacecraft to be trapped in the exterior region. Furthermore, the
v maneuvers along this band are on the order of hundreds of meters
per second for the larger Lyapunov orbits considered. Once a
spacecraft passes approximately eight months on the unstable
manifold, however, a maneuver closing the ZVS is possible for all the
trajectories. The cost reduces to the order of meters per second within
a year.
The cost profile for L2 halo periodic orbits close to their bifurcation
with the Lyapunov orbits follows a similar trend, as shown in the
upper left of Fig. 11. For halo orbits with larger out-of-plane
components, however, the initial orbit is completely contained within
the forbidden region at the Jacobi constant level CL2 . A spacecraft
needs to get a sufficient distance along the unstable manifold away
from the initial halo orbit to close the ZVS. Nevertheless, after about

1983

nine months the scheme can be applied, and a cost of less than 50 ms
is possible at 12 months.
Next, consider the Lissajous orbits about the L2 libration point
shown in Fig. 2. As before, a spacecraft can depart these orbits along
the unstable manifold and then perform a maneuver to close the zerovelocity surface at the L2 libration point. However, the manifold is
three dimensional, and so three variables are needed to parameterize it.
The sequence of images shown in Fig. 12 all correspond to a single
initial Lissajous orbit of amplitude 100,000 km. Figure 13 corresponds
to a single initial orbit of amplitude 600,000 km. Each plot uses two
angles to parameterize the initial state on the torus from which the
manifold trajectory departs (subject to a small perturbation). The
sequence is for different amounts of time along the manifold. This
allows one to explore the full space of maneuvers for this particular
initial Lissajous orbit. Initially, there are just small regions that
correspond to potential maneuvers (analogous to the situation with the
Lyapunov orbits). These maneuvers can have a prohibitively high cost.
However, the space of potential maneuvers increases as the spacecraft
departs along the manifold. Furthermore, the associated cost decreases.
For the smaller Lissajous orbit in Fig. 12, the cost is on the order of a
few meters per second at six months. For the larger Lissajous orbit in
Fig. 13, maneuvers on the manifold are possible from any initial
departure location after about eight months. The cost ultimately
decreases to less than 50 ms at 12 months.
For illustrative purposes, as a modification of the current scheme,
consider the half of the unstable manifold heading toward the Earth
rather than away from it. In this case, one can consider trajectories
from an L2 orbit that pass through the L1 gateway and into the interior
region. A maneuver can then be performed to trap the spacecraft in
this region away from the Earth. The results for a 500,000 km L2
Lyapunov orbit are shown in Fig. 14. The situation is much more
complex in this region. Only certain segments on the manifold have
maneuvers possible and the time for these maneuvers is at least one
year after departing the initial orbit. In red highlight, trajectories
passing within the geosynchronous radius are shown, which could
lead to additional end-of-life constraints. Furthermore, the moons
influence is not included, which could significantly affect these
transit trajectories. This example serves to emphasize the simplicity
gained by considering trajectories that do not pass through the Earth
bottleneck where the dynamics are much less predictable.

V.

Ephemeris Model Simulations and Results

The preceding maneuver analysis is based on the geometry of the

forbidden regions present in the CR3BP. Closing the ZVS at a
libration point strictly forbids a return only within this simplified
model. However, the earlier investigation of natural trajectories
suggests that a shadow of this barrier remains in an ephemeris model.
The current section studies two aspects that arise. First, adaptation to
the disposal scheme for application in an ephemeris model will be
presented. Second, the validity of the lifetimes in the higher-fidelity
model will be investigated. This will include the use of a safety factor
scaling the maneuver magnitude to further increase the Jacobi
constant and decrease the energy.
To transition the results to an ephemeris model, each component of
the end-of-life trajectory must be updated: the initial libration point
orbit, the departure trajectory, the disposal maneuver, and finally, the
long-term motion in the solar system. The first steps proceed as
described in Sec. II.B. A prediction of the initial orbit in the
ephemeris model is generated, and then it is corrected such that a
nearby continuous trajectory satisfying equations of motion (6) is
obtained. Although the unstable manifold is no longer a well-defined
surface, an unstable direction can nevertheless be computed using the
power-like approach presented. Analogous to the CR3BP setting, a
perturbation aligned with this direction can be propagated forward in
time to produce a trajectory lying on the unstable manifold substitute.
Finally, at a selected time, an impulsive maneuver is performed to put
the spacecraft on its terminal trajectory.
It remains to determine both the direction and the magnitude of this
disposal maneuver, but the insight obtained from the CR3BP
provides a guide. The Jacobi constant from Eq. (9) could be most

1984

OLIKARA, GMEZ, AND MASDEMONT

Different possibilities exist for determining the maneuver

magnitude. The first possibility is to use the instantaneous rotating
frame state to close the ZVS in the CR3BP at that moment in time.
Unlike the CR3BP, the Jacobi constant can vary significantly in time,
as illustrated in Fig. 15, even increasing above the Jacobi constant of
the libration point. Computing the cost from Eq. (10), the maneuver
distribution shown in Fig. 16 over four revolutions of an L2 Lyapunov
orbit is significantly different than observed in the CR3BP. Because
this is based on the instantaneous value of an oscillating quantity, it

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

efficiently increased by performing a maneuver opposite the

direction of the rotating frame velocity. Because all the propagation in
the ephemeris model is done in an inertial frame, the premaneuver
state is transformed into an instantaneous rotating frame by reversing
the steps from Sec. II.B. Once the rotating frame velocity vector is
obtained, one can transform the parallel, normalized maneuver
direction back to the inertial frame. Note that this sequence of
transformations results in a maneuver direction that is not aligned
with the spacecrafts velocity in the inertial frame.

Fig. 10 Maneuver costs (meters per second) along unstable manifold for L2 Lyapunov orbits.

OLIKARA, GMEZ, AND MASDEMONT

dependency of the departure point phasing, the maximum cost at each

time from the set of all the manifold trajectories is used. In effect, this
is the maximum value along each vertical line from the CR3BP cost
plots. This gives a nominal ephemeris maneuver magnitude v0 as a
function of time along the departure trajectory, as shown in Fig. 17.
Note that this maneuver is not well defined over the first six months
when solutions only exist for certain phasings.
With this nominal maneuver size, one can then verify how well the
disposal schemes goals are met, namely, avoidance of the Earth
bottleneck region. Because there are assumptions and simplifications

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

may not fully capture the orbit characteristics, and this approach is
not used.
The alternative approach that is pursued is to begin by assuming
that some of the structure persists, as was observed in the study of the
natural dynamics in Sec. III, and use costs from the CR3BP for the
ephemeris maneuver. This assumption can then be verified by
checking the orbit lifetimes. The approach begins by associating the
initial ephemeris orbit with the nearby orbit in the CR3BP. The costs
are then obtained at the corresponding time along the unstable
manifold from the CR3BP. To simplify the procedure and remove the

1985

Fig. 11 Maneuver costs (meters per second) along unstable manifold for L2 halo orbits.

1986

OLIKARA, GMEZ, AND MASDEMONT

Fig. 18 for a study of 10,000 ephemeris trajectories. A scaling factor

of zero corresponds to the natural dynamics.
From these representative results, it is apparent that the
incorporation of a maneuver significantly reduces the chance of a
return to the Earth vicinity. When the simulation is rerun, including a
simple cannonball force model for a Gaia-like spacecraft using
parameters from [6,7], the Earth return likelihood is slightly modified
(in fact, returns to Earth are observed to be marginally less likely for
this particular simulation). However, the primary trend persists:

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

built into its computation, a safety factor c is incorporated to scale the

maneuver magnitude: v  cv0 . Note that the maneuver direction
is unchanged. As an illustrative example, consider departure
trajectories leaving over 10 revolutions of a Lyapunov-like ephemeris
orbit with a rotating frame y amplitude of 500,000 km. If one
considers a maneuver after nine months along the unstable manifold,
all nominal maneuvers have a size of 11.6 ms with the direction
opposite the instantaneous rotating frame velocity of each trajectory.
Incorporating various scaling factors gives the results shown in

Fig. 12 Maneuver costs (meters per second) along unstable manifold for 100,000 km L2 Lissajous orbit.

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

OLIKARA, GMEZ, AND MASDEMONT

1987

Fig. 13 Maneuver costs (meters per second) along unstable manifold for 600,000 km L2 Lissajous orbit.

Fig. 14 Maneuver costs (meters per second) along unstable manifold

passing through Earth bottleneck to interior region.

Fig. 15 Instantaneous Jacobi constant along sample manifold

trajectory in ephemeris model (red line corresponds to CR3BP L2
libration point).

1988

OLIKARA, GMEZ, AND MASDEMONT

Adding a disposal maneuver greatly reduces the return probability.

Additional perturbations in this regime are most likely small
compared with the modeled forces, so that the behavior is expected to
continue.
In the computations, an Earth return is defined in a very broad
sense, specifically whether the trajectory passes within 106 km of the

Earth. This corresponds to trajectories that reenter the bottleneck

region where the dynamics are less predictable and where the
assumptions of this disposal scheme are no longer valid. However,
only a limited number of the bottleneck trajectories encounter the
Earth or a protected orbital regime. In particular, the trajectories that
return pass through a small opening in the ZVS. The unstable
manifold in the bottleneck region from a small libration point orbit
approximately bounds this motion, and these manifolds generally
pass far from the Earth. Also note that a whole family of departure
trajectories are being considered. Each individual trajectory could be
studied subject to additional perturbations and with a missionspecific epoch to determine its own likelihood of violating the
disposal constraint. Ultimately, the end-of-life trajectory selected
from the family of options will be one with the most advantageous
characteristics.

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

VI.

Fig. 16 Disposal maneuver cost (meters per second) in ephemeris model

using instantaneous Jacobi constant.

Conclusions

This study provides a look at the dynamics applicable to the endof-life disposal for libration point spacecraft. It is apparent that the
unstable manifold is the primary driver of motion away from the
orbit. Selecting an appropriate departure phase, a trajectory not
returning for a few revolutions about the zero-velocity surface can be
followed without any additional operations. By including a disposal
maneuver on the unstable manifold, however, the bottleneck can be
closed. This is observed to have a relatively low cost (less than
50 ms), though the maneuver must be performed after
approximately eight months. Nevertheless, science objectives may
still be met during part of this time period due to the initial shadowing
of the nominal libration point orbit. Ephemeris results suggest that
with a safety factor this disposal scheme can be transitioned to
realistic mission scenarios. With sufficient budgeting of both
propellant and time, this is a simple option to dispose of a libration
point spacecraft.

Acknowledgments
The authors would like to acknowledge funding from the Marie
Curie Initial Training Network PITN-GA-2011-289240, AstroNetII, and Spanish grants MTM2010-16425, MTM2013-41168-P (G.
Gomez), MTM2012-31714, and 2014SGR504 (J. J. Masdemont).
The authors are also grateful for the insightful comments and
suggestions from the reviewers.

References

Fig. 17 Maximum maneuver costs as function of time for 500,000 km

amplitude L2 Lyapunov orbit.

[1] Space Systems Space Debris Mitigation Requirements, International

Organization for Standardization, ISO-24113:2011, Geneva, 2011.
[2] Bell, M., Gmez, G., and Masdemont, J. J., Invariant Manifolds,
Lagrangian Trajectories and Space Mission Design, Space Manifold
Dynamics: Novel Spaceways for Science and Exploration, edited
by Perozzi, E., and Ferraz-Mello, S., 1st ed., SpringerVerlag, New

Fig. 18 Percentage of trajectories returning to bottleneck region for safety scaling factors (plot on right is an enlargement).

OLIKARA, GMEZ, AND MASDEMONT

[3]
[4]

[5]

Downloaded by RYERSON UNIVERSITY on September 13, 2016 | http://arc.aiaa.org | DOI: 10.2514/1.G000581

[6]

[7]

York, 2010, pp. 196.

doi:10.1007/978-1-4419-0348-8
Garcia, L., Farquhar, R., and Eastman, T., New Opportunities for a
Historic Spacecraft, Space Weather, Vol. 10, No. 8, 2012, pp. 12.
doi:10.1029/2012SW000832
Gmez, G., Llibre, J., Martnez, R., and Sim, C., Station Keeping of
Libration Point Orbits, ESA CR ESOC 5648/83/D/JS(SC), 1985;
reprint, Dynamics and Mission Design Near Libration Points I:
Fundamentals: The Case of Collinear Libration Points, 1st ed., World
Scientific, Singapore, Republic of Singapore, 2001.
Gmez, G., Jorba, ., Masdemont, J. J., and Sim, C., Study
Refinement of Semi-Analytical Halo Orbit Theory, ESA CR ESOC
8625/89/D/MD(SC), 1991; reprint, Dynamics and Mission Design Near
Libration Points III: Advanced Methods for Collinear Points, 1st ed.,
World Scientific, Singapore, Republic of Singapore, 2001.
Colombo, C., Letizia, F., Soldini, S., Lewis, H., Alessi, E. M., Rossi, A.,
Vasile, M., Vetrisano, M., Van der Weg, W. J., and Landgraf, M., Endof-Life Disposal Concepts for Libration Point and Highly Elliptical
Orbit Missions, Second IAA Conference on Dynamics and Control of
Space Systems, International Academy of Astronautics Paper IAAAAS-DyCoSS2-14-03-01, March 2014.
Soldini, S., Colombo, C., Walker, S., and Landgraf, M., LibrationPoint Orbit Missions Disposal at the End-of-Life Through Solar

[8]
[9]
[10]
[11]

[12]
[13]

1989

Radiation Pressure, Second IAA Conference on Dynamics and Control

of Space Systems, International Academy of Astronautics Paper IAAAAS-DyCoSS2-14-11-01, March 2014.
Szebehely, V. G., Theory of Orbits: The Restricted Problem of Three
Bodies, 1st ed., Academic Press, New York, 1967, pp. 741, Chap. 1.
Hartman, P., Ordinary Differential Equations, 1st ed., Wiley, New York,
1964, pp. 228272, Chap. 9.
Hairer, E., Nrsett, S. P., and Wanner, G., Solving Ordinary Differential
Equations I. Nonstiff Problems, 2nd ed., Springer, New York, 1993,
pp. 224240.
Jorba, ., Numerical Computation of the Normal Behavior of Invariant
Curves of n-Dimensional Maps, Nonlinearity, Vol. 14, No. 5, 2001,
pp. 943976.
doi:10.1088/0951-7715/14/5/303
Gmez, G., Masdemont, J. J., and Sim, C., Quasihalo Orbits
Associated with Libration Points, Journal of Astronautical Sciences,
Vol. 42, No. 2, 1998, pp. 135176.
Conley, C. C., Low Energy Transit Orbits in the Restricted Three-Body
Problem, SIAM Journal on Applied Mathematics, Vol. 16, No. 4, 1968,
pp. 732746.
doi:10.1137/0116060