You are on page 1of 36

Celest Mech Dyn Astr (2013) 117:349–384

DOI 10.1007/s10569-013-9515-6

ORIGINAL ARTICLE

High-order solutions of invariant manifolds associated


with libration point orbits in the elliptic restricted
three-body system

Hanlun Lei · Bo Xu · Xiyun Hou · Yisui Sun

Received: 28 October 2012 / Revised: 9 July 2013 / Accepted: 20 August 2013 /


Published online: 16 October 2013
© Springer Science+Business Media Dordrecht 2013

Abstract High-order analytical solutions of invariant manifolds, associated with Lissajous


and halo orbits in the elliptic restricted three-body problem (ERTBP), are constructed in
this paper. The equations of motion of ERTBP in the pulsating synodic coordinate system
have five equilibrium points, and the three collinear libration points as well as the associ-
ated center manifolds are unstable. In our calculation, the general solutions of the invari-
ant manifolds associated with Lissajous and halo orbits around collinear libration points
are expressed as power series of five parameters: the orbital eccentricity, two amplitudes
corresponding to the hyperbolic manifolds, and two amplitudes corresponding to the cen-
ter manifolds. The analytical solutions up to arbitrary order are constructed by means of
Lindstedt–Poincaré method, and then the center and invariant manifolds, transit and non-
transit trajectories in ERTBP are all parameterized. Since the circular restricted three-body
problem (CRTBP) is a particular case of ERTBP when the eccentricity is zero, the gen-
eral solutions constructed in this paper can be reduced to describe the dynamics around the
collinear libration points in CRTBP naturally. In order to check the validity of the series
expansions constructed, the practical convergence of the series expansions up to different
orders is studied.

Electronic supplementary material The online version of this article (doi:10.1007/s10569-013-9515-6)


contains supplementary material, which is available to authorized users.

H. Lei · B. Xu (B) · X. Hou · Y. Sun


School of Astronomy and Space Science, Nanjing University, Nanjing 210093, China
e-mail: xubo@nju.edu.cn
H. Lei
e-mail: hanlunlei@sina.com
X. Hou
e-mail: silence@nju.edu.cn
Y. Sun
e-mail: sunys@nju.edu.cn

123
350 H. Lei et al.

Keywords Elliptic restricted three-body problem · Invariant manifold · Lindstedt–Poincaré


method · Lissajous orbit · Halo orbit · Collinear libration point

1 Introduction

Libration points correspond to the dynamical equilibrium solutions of the restricted three-
body problem, and rotate around the barycenter of system with the same angular veloc-
ity as that of the primaries. The libration point orbits, and low-energy transfers based on
dynamical system theory have played an important role in deep space exploration and
attracted much attention recently, due to their potential to generate new kinds of mission
options.
The bounded orbits (for instance, Lissajous and halo orbits) around the L 1 point of
the Sun–Earth system can provide good observation sites of the Sun, due to their priv-
ileged configurations with respect to the primaries, and are usually taken as working
orbits of this kind of missions, such as ISEE–3 (1978), SOHO (1995), Genesis (2001)
etc. (Canalias et al. 2004). The bounded orbits around the L 2 point of Sun–Earth system
are very suitable to place the astronomy telescopes, considering the stable thermal envi-
ronment, without space debris, gravity gradient, and magnetic field from the Earth, etc.
A number of successful missions within the last decade have taken advantage of the spe-
cial location and environment around the L 2 point of Sun–Earth system, such as MAP
(2001), PLANCK (2007), GAIA (2012) etc. (Canalias et al. 2004). In addition, the infor-
mation of the Moon’s opposite side is always unknown to us, and the bounded orbits
around the L 2 point of Earth–Moon system can provide nominal orbits to place scien-
tific spacecrafts (Farquhar 1971). The recent mission, ARTEMIS, is the first mission that
took advantage of the Earth–Moon libration point orbits (Folta et al. 2012; Sweetser et al.
2011).
The unstable dynamical properties of the collinear libration point orbits can be uti-
lized to realize low-energy transfers between different three-body systems (Howell and
Kakoi 2006). The Japanese lunar satellite, Hiten, adopted the low-energy transfer tra-
jectory based on weak stability boundary (WSB) theory (Belbruno 2004), and exploited
the weak stability property of Sun–Earth libration point region to reduce the excess
velocity, so that less impulse maneuver was needed to insert the spacecraft into a sta-
ble, normal circumlunar orbit. The ballistic capture mechanism of WSB trajectory has
been understood with the aid of dynamical system theory. Koon et al. (2001) approxi-
mated the restricted four-body problem (including the Sun, Earth, Moon and spacecraft)
as two circular restricted three-body problems, and a systematic methodology based on
the intersection of invariant manifolds, associated with libration point orbits of Sun–Earth
and Earth–Moon systems, has been developed. Recently, the WSB trajectory has been
adopted by the Gravity Recovery and Interior Laboratory (GRAIL) mission, which is a
part of NASA’s discovery program, to transfer two orbiters to the low lunar orbit (Chung
2010). The other kind of Earth–Moon low energy trajectory corresponding to the interior
capture around the Moon is systematically investigated by Lei et al. (2013), who con-
structed the low energy trajectories in the circular restricted three-body problem (CRTBP)
firstly, and then searched the similar low energy trajectories in the real system by using
evolutionary algorithm, with the initial information provided by the results obtained in
CRTBP.
The invariant manifolds associated with libration point orbits can provide the global under-
standing about the dynamics around collinear libration points from the viewpoint of phase

123
High-order solutions of invariant manifolds in ERTBP 351

space. The low-thrust, low-energy transfers, incorporating the natural dynamics of three-
body system, also have been investigated widely. When the invariant manifolds of libration
point orbits of two three-body systems can not intersect in space, the application of low thrust
propulsion to interconnecting ballistic trajectories on invariant manifolds has been investi-
gated in Pergola et al. (2009). Dellnitz et al. (2006) constructed the Earth–Venus low-thrust,
low-energy transfers incorporating the invariant manifolds of three-body system. The attain-
able sets, playing the same role as invariant manifolds in trajectory design, have been applied
to the design of low-energy, low-thrust transfers to the Moon (Mingotti et al. 2009), and the
design of interplanetary low-thrust transfers (Mingotti et al. 2011).
Usually, the dynamics around the collinear libration points are studied by numerical meth-
ods, such as numerical integration, differential correction, optimization methods, and so
on. However, the numerical method is a trial and error process and depends on the expe-
rience of the designer. Analytical solutions could provide deep insights of the dynamics
around the libration points, and become more and more important. There are two kinds
of methodologies to analytically describe the dynamics around equilibrium points, that
is, the Lindstedt–Poincaré method (L–P) and normal form scheme. Richardson (1980)
used the L–P method to analytically construct a third-order solution for halo-type periodic
motion around the collinear libration points of CRTBP. In Jorba and Masdemont (1999), the
L–P method as well as normal form scheme are adopted to semi-analytically construct the
high-order solutions about the dynamics in the center manifolds of the collinear libration
points in CRTBP. Lei and Xu (2013a) analytically constructed the high-order analytical solu-
tions around triangular libration points in CRTBP, and discussed the practical convergence
in detail. Considering the hyperbolic behaviors together with the center behaviors around
the collinear libration points in CRTBP, Masdemont (2005) expanded the invariant mani-
folds as power series of hyperbolic and center amplitudes. These series expansions could
explicitly describe the general dynamics around collinear libration points of CRTBP, and
have been used to study the two-maneuver transfer problem between LEOs and Lissajous
orbits in the Earth–Moon system (Alessi et al. 2010), to look for rescue trajectories that leave
the surface of the Moon by means of invariant manifolds of libration point orbits (Alessi et
al. 2009), and to construct the low-thrust transfers to the libration point orbits of Sun–Mars
system from the low Earth orbit (Lei and Xu 2013b). For the Hill problem (referring to
the equation of relative motion), which corresponds to the reduced case of CRTBP when
μ = 0, Gómez and Marcote (2006) computed the bounded orbits by using the L–P method.
Taking into account the perturbation of the Solar gravity and lunar eccentricity, Farquhar
and Kamel (1973) analytically developed the third order solutions of quasi-periodic orbits
around the collinear libration point L 2 of Earth-Moon system and discussed the relationship
between the frequency and amplitude for large halo orbits. For a spacecraft moving along a
libration point orbit of Earth–Moon system, Farquhar (1968, 1971) presented detailed stud-
ies about the flight mechanics, practical applications and station-keeping problems. In the
real Earth–Moon system, defined by the JPL ephemeris, the quasi-periodic motions around
the triangular and collinear libration points are analytically studied by Hou and Liu (2010,
2011b).
Compared to CRTBP, the elliptic restricted three-body problem (ERTBP) could approx-
imate the Solar system better. Due to the existence of eccentricity of the primaries, the
equations of motion in ERTBP are non-autonomous. Fortunately, the equations of motion of
ERTBP in the pulsating synodic reference frame have the same symmetries as the ones of
CRTBP, meanwhile, the dynamical properties are similar to those of CRTBP. For example,
libration points and corresponding bounded orbits (Lissajous and halo orbits) also exist, and
are unstable in nature. However, the investigation about the dynamics around the collinear

123
352 H. Lei et al.

libration points in ERTBP is more complicated by analytical method. Hou and Liu (2011a)
constructed the high-order analytical solutions of the center manifolds, such as Lissajous and
halo orbits in ERTBP, by means of semi-analytical method, and discussed the applications in
the Earth–Moon and Sun–Earth + Moon system. For the elliptic equations of relative motion
(or elliptic Hill equations), corresponding to the reduced case of ERTBP when μ = 0, Ren
et al. (2012) obtained a third-order expression of the bounded orbits around the collinear
libration points, and conceptually presented the process of constructing high-order analytical
solutions by using the L–P method.
Considering the unstable dynamics of the collinear libration points and associated center
manifolds in ERTBP, the general solutions of the equations of motion around the collinear
libration points in ERTBP consist of the hyperbolic component (saddle behavior) and cen-
ter component (center behavior). Therefore, in this work, the solutions of invariant man-
ifolds associated with libration point orbits in ERTBP are expanded as formal series of
the orbital eccentricity and four amplitudes, thereinto, two amplitudes correspond to the
hyperbolic manifolds, and the remaining two amplitudes correspond to the center mani-
folds. The series expansions constructed in this paper can describe the general dynamics
around the collinear libration points of ERTBP, and can be considered as an extension
of the ones discussed in Jorba and Masdemont (1999), Masdemont (2005), and Hou and
Liu (2011a). In order to check the validity of the analytical solutions, numerical simula-
tions with the same initial states are also implemented to investigate the practical conver-
gence.
The remainder of this paper is structured as follows. In Sect. 2, the basic dynamical model
about ERTBP is briefly described. Sects. 3 and 4 present the construction process of high-
order solutions of invariant manifolds, associated with Lissajous and halo orbits, around the
collinear libration points in ERTBP, respectively, and the results are presented in Sect. 5. At
last, the conclusions together with discussion are drawn in Sect. 6.

2 Dynamical model

An infinitesimal particle, such as a spacecraft, is placed in the gravitational field generated


by two massive bodies, such as the Sun and Earth, moving around their common center of
mass in Kepler orbits. In this dynamical system, the mass of the spacecraft is much less
than that of any primary, thus the attraction of the spacecraft on the primaries is neglected.
Such a system is called restricted three-body problem (RTBP). In particular, if the primaries
move around each other in circular orbits, such a system is the well-known CRTBP, in
which a motion integral and five equilibrium points exist. Generally, the orbital eccentric-
ities of the primaries are not zero and such a system is the ERTBP, in which the motion
of the spacecraft in the barycentric synodic coordinate system is governed by (Szebehely
1967)
⎧ 1 ∂Ω

⎪ X  − 2Y  = ,

⎪ 1 + e cos f ∂X


⎨ 1 ∂Ω
 
Y + 2X = , (1)

⎪ 1 + e cos f ∂Y



⎪ 1 ∂Ω
⎩ Z  + Z = ,
1 + e cos f ∂Z

with

123
High-order solutions of invariant manifolds in ERTBP 353

1 2  1−μ μ
Ω= X + Y 2 + Z 2 + μ(1 − μ) + + , (2)
2 R1 R2

where μ = m 2 /(m 1 + m 2 ), m 1 and m 2 are the masses of the two primaries. R1 and R2
are the distances of the spacecraft from the massive and secondary primaries, respectively.
To formulate the equations of motion in ERTBP, dimensionless units are adopted, that is,
the total mass of the primaries, the instantaneous distance between the two primaries, and
their angular velocity of the primaries are all taken as unity, such that the values of the
gravitational constant and the period of the secondary are 1 and 2π, respectively. Equation
(1) is formulated in the barycentric synodic system, in which the X -axis is directed from
the massive primary toward the secondary primary, the Z -axis is aligned with the momen-
tum of the secondary and the Y -axis is determined by the right-hand coordinate system.
In Eq. (1), the time-like independent variable is f , which is the true anomaly of the sec-
ondary on the elliptic orbit. The first and second derivatives of coordinate are defined as
follows:

dX d2 X
X = , X  = . (3)
df df2

The Y and Z components have similar expressions. Since the component 1 + e1cos f exists
in Eq. (1), the equations of motion in ERTBP are non-autonomous. However, the equations of
motion in ERTBP have the same symmetric properties as those of CRTBP. To be consistent,
let S1 , S2 and S3 denote the three kinds of symmetries:

S1 : ( f, X, Y, Z , X  , Y  , Z  ) ↔ (− f, X, −Y, Z , −X  , Y  , −Z  ), (4)
S2 : ( f, X, Y, Z , X  , Y  , Z  ) ↔ (− f, X, −Y, −Z , −X  , Y  , Z  ), (5)
S3 : ( f, X, Y, Z , X , Y , Z ) ↔ ( f, X, Y, −Z , X  , Y  , −Z  ). (6)

All the above three symmetries are important for the construction of high-order solutions of
invariant manifolds in ERTBP.
In order to investigate the motion around the collinear libration point L 1 or L 2 conveniently,
it is necessary to move the origin of the coordinate system from the barycenter of system
to the interested libration point, and take the instantaneous distance between the libration
point and its closest primary as length unit. Denote the new reference frame as the L 1 or
L 2 -centered synodic reference frame, and the axes of this new coordinate system are aligned
with the corresponding ones of the barycentric synodic reference frame. Let (x, y, z, x  , y  , z  )
represent the state variables in this new reference frame, and γ denotes the instantaneous
distance between the libration point and its closest primary. The transformation of coordinates
between the original barycentric synodic frame and the L 1 or L 2 -centered synodic system
is,

X = γ (x ∓ 1) + 1 − μ, Y = γ y, Z = γ z, (7)

where the upper sign refers to the L 1 case and the lower one refers to the L 2 case. The
equations of motion in the L 1 or L 2 -centered synodic system can be formulated as follows
(Hou and Liu 2011a):

123
354 H. Lei et al.

⎧ 

⎪  

⎪ x − 2y − (1 + 2c 2 )x = (−e)i
cos i
f (1 + 2c 2 )x



⎪ i≥1

⎪   



⎪ + (−e) i
cos i
f c (n + 1)T (x, y, z) ,


n+1 n



⎪ 
i≥0 n≥2




⎪ y  + 2x  − (1 − c2 )y = (−e)i cosi f (1 − c2 )y


⎨ i≥1
⎪    (8)

⎪ + (−e) i i
(x, ,


cos f y c R
n+1 n−1 y, z)




i≥0 n≥2

⎪ 


⎪ z  + c2 z = (−e)i cosi f (1 − c2 )z





⎪ i≥1



⎪   

⎪ + (−e) cos f z
i i
cn+1 Rn−1 (x, y, z) ,

i≥0 n≥2

in which the coefficient cn (μ) is constant and only dependent on the mass parameter of the
three-body system, and is given by
 
1 (1 − μ)γin+1
cn (μ) = (±1)
n
μ + (−1)n , (9)
γi3 (1 ∓ γi )n+1

where γi (i = 1, 2) is the instantaneous distance between L i and its closest primary. In


Eq. (9), the upper sign refers to the L 1 point, and the lower sign refers to the L 2 point. The
case corresponding to the L 3 point is not discussed here since it is not the focus of this paper.
In Eq. (8), Tn and Rn are the homogeneous polynomials of degree n and can be computed
by the following recurrence relations:

2n − 1 n−1 2
Tn = x Tn−1 − (x + y 2 + z 2 )Tn−2 , (10)
n n

which starts with T0 = 1 and T1 = x, and

2n + 3 2n + 2 n+1 2
Rn = x Rn−1 − Tn − (x + y 2 + z 2 )Rn−2 , (11)
n+2 n+2 n+2

which starts with R0 = −1 and R1 = −3x.

3 High-order solutions of invariant manifolds associated with Lissajous orbits


in ERTBP

3.1 Series expansions of the invariant manifolds

The L–P method is adopted to construct the high-order analytical solutions about the invariant
manifolds associated with libration point orbits in ERTBP. The L–P procedure is a process
of recursion, and the unknown coefficients of high-order solutions are calculated from the
lower-order solutions. Linearize the equations of motion in ERTBP as follows:

123
High-order solutions of invariant manifolds in ERTBP 355

⎧ 

⎪  

⎪ x − 2y − (1 + 2c 2 )x = (−e) i
cosi
f (1 + 2c 2 )x ,



⎪ i≥1

⎨  

y + 2x  − (1 − c2 )y = (−e)i cosi f (1 − c2 )y , (12)





⎪ 
i≥1


⎪ 

⎪ z + c2 z = (−e) cosi f (1 − c2 )z ,
i

i≥1

which is a non-autonomous system, and the general solutions cannot be obtained easily.
However, the L–P method only needs a starting point. We take the solution with first-order
amplitude and zero-order eccentricity as the starting point, and the solutions corresponding
to high-order amplitude and eccentricity are calculated by means of the L–P method. The
starting solution we adopt is the general solution of the following equations:
⎧  
⎨ x − 2y − (1 + 2c2 )x = 0,

y + 2x  + (c2 − 1)y = 0,

(13)

⎩ 
z + c2 z = 0,

in which, the motion in the z direction is uncoupled from that in the x–y plane. In fact,
Eq. (13) is the linearized form of the equations of motion in CRTBP, which is a particular
case of ERTBP when the eccentricity is zero. The conjugate characteristic roots of Eq. (13)
are ±λ0 , ±iω0 and ±iν0 ,
   
 
 c − 2 + 9c2 − 8c  2 − c + 9c2 − 8c
 2 2 2  2 2 2 √
λ0 = , ω0 = , ν0 = c2 . (14)
2 2

Thus the general solution of Eq. (13) consists in a harmonic motion in the in-plane component,
an un-coupled oscillation in the out-of-plane component (linear center manifold) and an
exponential part (linear hyperbolic manifold), given by (Masdemont 2005)

⎨ x( f ) = α1 exp(λ0 f ) + α2 exp(−λ0 f ) + α3 cos(ω0 f + φ1 ),

y( f ) = κ̄2 α1 exp(λ0 f ) − κ̄2 α2 exp(−λ0 f ) + κ̄1 α3 sin(ω0 f + φ1 ), (15)


z( f ) = α4 cos(ν0 f + φ2 ),

where α1 and α2 refer to the unstable and stable amplitudes corresponding to the hyperbolic
manifolds, respectively, and the remaining amplitudes α3 and α4 refer to the in-plane and
out-of-plane amplitudes corresponding to the center manifolds, respectively. κ̄1 and κ̄2 are
constants and only dependent on the mass parameter of the three-body system,

ω02 + 2c2 + 1 λ2 − 2c2 − 1


κ¯1 = − , κ̄2 = 0 . (16)
2ω0 2λ0

When considering the nonlinear terms and the perturbation of orbital eccentricity of Eq. (8),
the invariant manifolds associated with Lissajous orbits, around the collinear libration points
in ERTBP, can be expanded as power series of the orbital eccentricity and four amplitude
parameters,

123
356 H. Lei et al.

⎧  r st 
⎪  x pi jkm cos(r f + sθ1 + tθ2 )+ p i j k m


⎪ x( f ) =


exp [(i − j)θ3 ] r st
x̄ pi jkm sin(r f + sθ1 + tθ2 )
e α1 α2 α3 α4 ,



⎪  r st 
⎨  y pi jkm cos(r f + sθ1 + tθ2 )+ p i j k m
y( f ) = exp [(i − j)θ3 ] r st e α1 α2 α3 α4 , (17)

⎪ ȳ pi jkm sin(r f + sθ1 + tθ2 )

⎪  r st 



⎪  z pi jkm cos(r f + sθ1 + tθ2 )+ p i j k m


⎩ z( f ) = exp [(i − j)θ3 ] r st e α1 α2 α3 α4 ,
z̄ pi jkm sin(r f + sθ1 + tθ2 )
in which θ1 = ω f + φ1 , θ2 = ν f + φ2 and θ3 = λ f , here φ1 and φ2 are arbitrarily
initial phase angles corresponding to the in-plane and out-of-plane motion, respectively.
x rpistjkm , y rpistjkm and z rpistjkm refer to the coefficients of cosines, and x̄ rpistjkm , ȳ rpistjkm and z̄ rpistjkm
are the coefficients of sines corresponding to the coordinate series x, y and z to be determined.
Due to the nonlinear terms of the equations of motion and the perturbation of eccentricity,
the frequencies of motion are not constant, and should also be expanded as formal series:
⎧  j

⎪ω = ω pi jkm e p α1i α2 α3k α4m ,

⎨  j
ν= ν pi jkm e p α1i α2 α3k α4m , (18)

⎪ 

⎩λ= j
λ pi jkm e p α1i α2 α3k α4m .
In Eqs. (17) and (18), p, i, j, k, m ∈ N, and r, s, t ∈ Z. When α1 = 0 and α2  = 0, Eq. (17)
describes the stable manifolds associated with Lissajous orbits around the collinear libration
points, whereas, if α1  = 0 and α2 = 0, it describes the unstable manifolds associated with
Lissajous orbits in ERTBP. When time tends to infinity, a spacecraft approaches the Lissajous
orbit exponentially, following the stable manifold. Whereas, the spacecraft departs from the
Lissajous orbit exponentially, following the unstable manifold. Moreover, when α1 · α2 < 0,
Eq. (17) can describe the transit trajectories, following which the spacecraft could transfer
from one side to the other side of the libration point freely. When α1 · α2 > 0, Eq. (17) can
describe the non-transit trajectories, along which the spacecraft can only move in one side
of the libration point for a certain time, but, can’t move in one side forever due to the Arnold
diffusion phenomenon in ERTBP (Xia 1993). In particular, when α1 = α2 = 0, the general
solutions represented by Eq. (17) will be reduced to describe the center manifolds in ERTBP,
such as the Lissajous, plane and vertical Lyapunov orbits. This reduced case is equivalent
to the series expansions discussed in Hou and Liu (2011a). When the orbital eccentricity
satisfies e = 0, it is easy to get f = t, analytical solutions represented by Eq. (17) can be
used to describe the dynamics around the collinear libration points in CRTBP. This reduced
case is equivalent to the series expansions of the center manifolds discussed in Jorba and
Masdemont (1999), and the invariant manifolds discussed in Masdemont (2005).
In our computation, three orders are defined in the following manner. N1 refers to the order
of the orbital eccentricity, N2 represents the order of hyperbolic manifold, N3 stands for the
order of center manifold, and the total order is N = N1 + N2 + N3 . Denote the order of the
analytical solutions as (N1 , N2 , N3 ), and N2 ≤ N3 is required. For any p, i, j, k, m, r, s and
t, it requires that 0 ≤ p ≤ N1 , 0 ≤ i + j ≤ N2 , 0 ≤ k + m ≤ N3 , 0 ≤ r ≤ p (taking into
account the symmetries of cosine and sine functions), −k ≤ s ≤ k and −m ≤ t ≤ m. In
Eq. (17), r, s and t have the same parity of p, k and m, due to the symmetries of Eq. (1). As
stated, Eq. (15) is the general solution of the linearized equations of motion in ERTBP when
the eccentricity is set to zero, and the order is denoted as (0, n 2 , n 3 ), where n 2 + n 3 = 1,
which means that the order of the orbital eccentricity is zero, and the total order of center
and hyperbolic manifolds is one. In detail, for the coordinate series x and y,

123
High-order solutions of invariant manifolds in ERTBP 357

000
x01000 = 1, x00100
000
= 1, x00010
010
= 1,
000
y01000 = κ̄2 , 000
y00100 = −κ̄2 , 010
y00010 = κ̄1 ,
and for the coordinate series z,
001
z 00001 = 1.
For the linear solution described by Eq. (15), the order of frequency is (0, 0, 0), that is,
ω00000 = ω0 , ν00000 = ν0 , λ00000 = λ0 .
Due to the symmetries of the equations of motion in ERTBP, many coefficients of coor-
dinate and frequency are zero. When m is odd, the coefficients x rpistjkm , x̄ rpistjkm , y rpistjkm and
ȳ rpistjkm are zero. When m is even, the coefficients z rpistjkm and z̄ rpistjkm are zero. In addition, the
symmetries of the coefficients also exist, that is,
x rpistjkm = x rpjikm
st
, y rpistjkm = −y rpjikm
st
, z rpistjkm = z rpjikm
st
,
x̄ rpistjkm = −x̄ rpjikm
st
, ȳ rpistjkm = ȳ rpjikm
st
, z̄ rpistjkm = −z̄ rpjikm
st
.

It is easy to verify that when i = j, one gets x̄ rpistjkm = y rpistjkm = z̄ rpistjkm = 0. The coefficients
corresponding to the frequencies are not zero only if both k and m are even, besides satisfying
i = j. These facts could be adopted to save computer storage and help us to check the validity
of the procedure of construction.
In the process of computation, the first and second derivatives of the coordinates x, y and
z with respect to the true anomaly f , appearing in the left-hand side of the equations of
motion, can be computed as follows (taking x as an example):
∂x ∂x ∂x ∂x
x = +ω +ν +λ , (19)
∂f ∂θ1 ∂θ2 ∂θ3
and
∂2x 2∂ x
2
2∂ x
2
2∂ x
2 ∂2x ∂2x
x  = + ω + ν + λ + 2ω + 2ν
∂ f2 ∂θ12 ∂θ22 ∂θ32 ∂ f ∂θ1 ∂ f ∂θ2
∂2x ∂2x ∂2x ∂2x
+ 2λ + 2ων + 2ωλ + 2νλ . (20)
∂ f ∂θ3 ∂θ1 ∂θ2 ∂θ1 ∂θ3 ∂θ2 ∂θ3
The fact of L–P method lies in that high-order solutions of the invariant manifolds around
the collinear libration points are constructed from lower-order solutions. If the solution of
the order (n 1 , n 2 , n 3 ) is to be solved, the unknown coefficients corresponding to coordinates
include (x rpistjkm , x̄ rpistjkm ), (y rpistjkm , ȳ rpistjkm ) and (z rpistjkm , z̄ rpistjkm ), where p = n 1 , i + j = n 2 and
k +m = n 3 , and the unknown coefficients corresponding to frequency include ω pi jkm , ν pi jkm
and λ pi jkm , where p + i + j + k + m = n 1 + n 2 + n 3 − 1. The solution corresponding to
order (n 1 , n 2 , n 3 ) can be constructed only if all terms of order (n̄ 1 , n̄ 2 , n̄ 3 ), n̄ 1 ≤ n 1 , n̄ 2 ≤
n 2 , n̄ 3 ≤ n 3 and n̄ 1 + n̄ 2 + n̄ 3 < n 1 + n 2 + n 3 have been known. In terms of the construction
of analytical solutions up to order (n 1 , n 2 , n 3 ), the process of computation is carried out in
the following manner. Firstly, we construct the terms (:, n̄ 2 , n̄ 3 ), n̄ 2 < n 2 , n̄ 3 < n 3 from
the order (n̄ 1 , n̄ 2 , n̄ 3 ) to the order (n 1 , n̄ 2 , n̄ 3 ), where if n̄ 2 + n̄ 3 = 1, then n̄ 1 = 1, else
n̄ 1 = 0. Secondly, construct the terms (n 1 , :, n̄ 3 ), n̄ 3 < n 3 from the order (n 1 , n̄ 2 , n̄ 3 ) to
the order (n 1 , n 2 , n̄ 3 ), in which if n̄ 3 = 0, then n̄ 2 = 1, else n̄ 2 = 0. At last, we construct
the terms (n 1 , n 2 , :) from the order (n 1 , n 2 , n̄ 3 ) to the order (n 1 , n 2 , n 3 ), where if n 2 = 0,
then n̄ 3 = 1, else n̄ 3 = 0. Substituting the known components up to order (n 1 − 1, n 2 , n 3 ),
(n 1 , n 2 − 1, n 3 ) and (n 1 , n 2 , n 3 − 1) into Eq. (8), the unknown coefficients for the x and y

123
358 H. Lei et al.

series corresponding to order (n 1 , n 2 , n 3 ) can be determined by solving the following linear


system of algebraic equations:
⎡ ⎤ ⎡ r st ⎤ ⎡ c ⎤ ⎡ r st ⎤
A1 A2 A3 A4 x pi jkm δx X pi jkm
⎢ −A2 A1 −A4 A3 ⎥ ⎢ x̄ rpistjkm ⎥ ⎢ δxs ⎥ ⎢ r st ⎥
⎢ ⎥ ⎢ r st ⎥ + ⎢ c ⎥ = ⎢ ⎢
X̄ pi jkm ⎥
⎥, (21)
⎣ B1 B4 ⎦ ⎣ y pi jkm ⎦ ⎣ δ y ⎦ ⎣ Y pi r st
B2 B3 jkm ⎦
−B2 B1 −B4 B3 ȳ rpistjkm δys r st
Ȳ pi jkm

with

⎪ A1 = −(r + ω0 s + ν0 t)2 + λ20 (i − j)2 − (1 + 2c2 ),


⎨A = 2λ0 (i − j)(r + ω0 s + ν0 t),
2
(22)

⎪ A3 = −2λ0 (i − j),


A4 = −2(r + ω0 s + ν0 t),
and


⎪ B1 = 2λ0 (i − j),


⎨ B2 = 2(r + ω0 s + ν0 t),
(23)

⎪ B3 = −(r + ω0 s + ν0 t)2 + λ20 (i − j)2 − (1 − c2 ),



B4 = 2λ0 (i − j)(r + ω0 s + ν0 t),
and
⎧ c
⎪ δx = − 2(ω0 + κ̄1 )ω pi jk−1m δr 0 δs1 δt0 δi j



⎪ + 2(λ0 − κ̄2 )λ pi−1 jkm δr 0 δs0 δt0 δi−1 j





⎪ + 2(λ0 − κ̄2 )λ pi j−1km δr 0 δs0 δt0 δi j−1 ,

δxs = 0, (24)



⎪ δy
c
= 2(κ̄2 λ0 + 1)λ pi−1 jkm δr 0 δs0 δt0 δi−1 j





⎪ − 2(κ̄2 λ0 + 1)λ pi j−1km δr 0 δs0 δt0 δi j−1 ,

⎩ s
δy = − 2(κ̄1 ω0 + 1)ω pi jk−1m δr 0 δs1 δt0 δi j .
The unknown coefficients of order (n 1 , n 2 , n 3 ) in z are determined by the set of algebraic
equations:
r st c  r st 
C1 C2 z pi jkm δ Z pi jkm
+ zs = , (25)
−C2 C1 z̄ rpistjkm δz Z̄ rpistjkm

with

C1 = −(r + ω0 s + ν0 t)2 + λ20 (i − j)2 + c2 ,
(26)
C2 = 2λ0 (i − j)(r + ω0 s + ν0 t),
and

δzc = −2ν0 ν pi jkm−1 δr 0 δs0 δt1 δi j ,
(27)
δzs = 0.

In Eqs. (21) and (25), X rpistjkm , X̄ rpistjkm , Y pi


r st , Ȳ r st , Z r st
jkm pi jkm
r st
pi jkm and Z̄ pi jkm represent the
known components of the equations of motion corresponding to order (n 1 , n 2 , n 3 ). The
symbol δi j is the Kronecker function, equals zero when i  = j and equals one when i = j.

123
High-order solutions of invariant manifolds in ERTBP 359

We have to mention that the system of notation adopted in this paper is similar to that in
Masdemont (2005).

3.2 Solving the undetermined coefficients

In Eqs. (21) and (25), nine unknown coefficients are required to be computed, including six
coefficients corresponding to coordinate and three coefficients corresponding to frequency. In
the process of computation, we will meet the difficult situation that the number of coefficients
to be computed is not equal to the number of equations. Thus, some special procedures should
be carried out. For clarity, we will discuss how to determine the unknown coefficients in
different situations.
Case 1 r  = 0
Case 1.1 i  = j
In this situation, the undetermined coefficients only include the coefficients of coordinates:
x rpistjkm , x̄ rpistjkm , y rpistjkm , ȳ rpistjkm , z rpistjkm and z̄ rpistjkm . These unknown coefficients satisfy,

⎡ ⎤ ⎡ r st ⎤ ⎡ r st ⎤
A1 A2 A3 A4 x pi jkm X pi jkm
⎢ −A2 ⎥ ⎢ r st ⎥ ⎢ X̄ r st ⎥
⎢ A1 −A4 A3 ⎥ ⎢ pi jkm ⎥ ⎢ pi jkm ⎥

⎣ B1 = ⎢ r st ⎥ , (28)
B2 B3 B4 ⎦ ⎣ y rpistjkm ⎦ ⎣ Y pi jkm ⎦
−B2 B1 −B4 B3 ȳ rpistjkm r st
Ȳ pi jkm

for the x and y components, and

 
C1 C2 z rpistjkm Z rpistjkm
= , (29)
−C2 C1 z̄ rpistjkm Z̄ rpistjkm

for the z component. The elements A1 , A2 , A3 , A4 , B1 , B2 , B3 , B4 , C1 and C2 are the same


as those in Sect. 3.1. These undetermined coefficients can be obtained by solving Eqs. (28)
and (29).
Case 1.2 i = j
As stated in Sect. 3.1, when i = j, one gets x̄ rpistjkm = y rpistjkm = z̄ rpistjkm = 0. In this situation,
the unknown coefficients only include x rpistjkm , ȳ rpistjkm and z rpistjkm , and can be computed by
solving the following algebraic equations:

 
A1 A4 x rpistjkm X rpistjkm
= , (30)
−B2 B3 ȳ rpistjkm r st
Ȳ pi jkm

for the x–y components, and

C1 · z rpistjkm = Z rpistjkm , (31)

for the z component.

123
360 H. Lei et al.

Case 2 r = 0
Case 2.1 s = 0, t = 0 and |i − j| = 1
In this case, the unknown coefficients satisfy,
⎡ ⎤ ⎡ r st ⎤ ⎡ c ⎤ ⎡ r st ⎤
A1 0 A3 0 x pi jkm δx X pi jkm
⎢0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ A1 0 A3 ⎥ ⎢ x̄ pi jkm ⎥ ⎢ 0 ⎥ ⎢ X̄ rpistjkm ⎥
r st
⎣ B1 ⎦ ⎣ ⎦ + ⎣ c ⎦ = ⎢ r st ⎥ , (32)
0 B3 0 r st
y pi jkm δy ⎣ Y pi jkm ⎦
0 B1 0 B3 ȳ rpistjkm 0 r st
Ȳ pi jkm

for the x–y components. In this situation, the coefficient matrix is singular. However, x̄ rpistjkm ,
ȳ rpistjkm , X̄ rpistjkm and Ȳ pi
r st
jkm are the coefficients of sin(0), the same way as Masdemont (2005)
is adopted here, that is, taking x̄ rpistjkm = 0, ȳ rpistjkm = 0, X̄ rpistjkm = 0 and Ȳ pi jkm = 0. For the z
r st

component, the similar analysis is considered so that we can set z̄ pi jkm = 0 and Z̄ rpistjkm = 0.
r st

Then the remaining coefficients satisfy,



A1 A3 x rpistjkm δxc X rpistjkm
+ = , (33)
B1 B3 y rpistjkm δ cy r st
Y pi jkm

where A1 = λ20 − (1 + 2c2 ), A3 = −2(i − j)λ0 , B1 = 2(i − j)λ0 , B3 = λ20 − (1 − c2 ),


δxc = 2(λ0 − κ̄2 )λ pnnkm and δ cy = 2(i − j)(κ̄2 λ0 + 1)λ pnnkm , where n = min(i, j). There
are three undetermined coefficients in Eq. (33), which can be solved by taking y rpistjkm = 0
(or x rpistjkm = 0),

A1 2(λ0 − κ̄2 ) x rpistjkm X rpistjkm
= , (34)
B1 2(i − j)(κ̄2 λ0 + 1) λ pnnkm r st
Y pi jkm

then x rpistjkm and λ pnnkm can be computed. For the z component, z rpistjkm can be obtained by
solving the following equation:

(λ20 + c2 )z rpistjkm = Z rpistjkm . (35)

It is easy to get z rpistjkm = Z rpistjkm /(λ20 + c2 ).


Case 2.2 s = 1, t = 0 and i = j
In this situation, the undetermined coefficients satisfy
⎡ ⎤ ⎡ r st ⎤ ⎡ c ⎤ ⎡ r st ⎤
A1 0 0 A4 x pi jkm δx X pi jkm
⎢0 −A4 0 ⎥ ⎢ x̄ rpistjkm ⎥ ⎢ 0 ⎥ ⎢ r st ⎥
⎢ A1 ⎢
⎥ ⎢ r st ⎥ + ⎢ ⎥ = ⎢ X̄ rpistjkm ⎥ ⎥, (36)
⎣0 B2 B3 0 ⎦ ⎣ y pi jkm ⎦ ⎣ 0 ⎦ ⎣ Y pi jkm ⎦
−B2 0 0 B3 r st
ȳ pi jkm δys r st
Ȳ pi jkm

for the x − y components, where A1 = −ω02 − (1 + 2c2 ), A4 = −2ω0 , B2 = 2ω0 ,


B3 = −ω02 − (1 − c2 ), δxc = −2(ω0 + κ̄1 )ω pi jk−1m and δ sy = −2(1 + κ̄1 ω0 )ω pi jk−1m . For
the z component,
r st  r st 
C1 0 z pi jkm Z pi jkm
= , (37)
0 C1 z̄ rpistjkm Z̄ rpistjkm

123
High-order solutions of invariant manifolds in ERTBP 361

where C1 = −ω02 + c2 . Due to i = j, one gets x̄ rpistjkm = y rpistjkm = z̄ rpistjkm = 0, then the
remaining unknown coefficients satisfy
⎡ ⎤  
x r st
A1 A4 −2(ω0 + κ̄1 ) ⎣ rpistjkm ⎦ X rpistjkm
ȳ pi jkm = , (38)
−B2 B3 −2(1 + κ̄1 ω0 ) r st
Ȳ pi
ω pi jk−1m jkm

for the x–y components, and

z rpistjkm = Z rpistjkm /C1 , (39)

for the z component. From Eq. (38), we can compute x rpistjkm and ω pi jk−1m by taking ȳ rpistjkm =
0, or ȳ rpistjkm and ω pi jk−1m by taking x rpistjkm = 0 (the latter case is adopted in our computation).
Case 2.3 s = 0, t = 1 and i = j
In this case, we can get
⎡ ⎤ ⎡ r st ⎤ ⎡ r st ⎤
A1 0 0 A4 x pi jkm X pi jkm
⎢0 ⎥ ⎢ r st ⎥ ⎢ X̄ r st ⎥
⎢ A1 −A4 0 ⎥ ⎢ pi jkm ⎥ ⎢ pi jkm ⎥

⎣0 = ⎢ r st ⎥ , (40)
B2 B3 0 ⎦ ⎣ y rpistjkm ⎦ ⎣ Y pi jkm ⎦
−B2 0 0 B3 ȳ rpistjkm r st
Ȳ pi jkm

for the x–y components, where A1 = −ν02 − (1 + 2c2 ), A4 = −2ν0 , B2 = 2ν0 and
B3 = −ν02 − (1 − c2 ). For the z component,
 
C1 0 z rpistjkm δc Z rpistjkm
+ z = , (41)
0 C1 z̄ rpistjkm 0 Z̄ rpistjkm

where C1 = −ν02 + c2 = 0 and δzc = −2ν0 ν pi jkm−1 . Due to i = j, one gets x̄ rpistjkm =
y rpistjkm = z̄ rpistjkm = 0, then x rpistjkm and ȳ rpistjkm can be obtained by solving

 
A1 A4 x rpistjkm X rpistjkm
= , (42)
−B2 B3 ȳ rpistjkm r st
Ȳ pi jkm

and ν pi jkm−1 can be obtained by taking z rpistjkm = 0, that is,



ν pi jkm−1 = −Z rpistjkm (2ν0 ). (43)

Case 2.4 s = 0, t = −1 and i = j


The coefficients x rpistjkm and ȳ rpistjkm can be computed by solving

 
A1 A4 x rpistjkm X rpistjkm
= , (44)
−B2 B3 ȳ rpistjkm r st
Ȳ pi jkm

in which A1 = −ν02 − (1 + 2c2 ), A4 = −2ν0 , B2 = 2ν0 and B3 = −ν02 − (1 − c2 ). While


for the z component, take z rpistjkm = 0 and z̄ rpistjkm = 0 due to C1 = −ν02 + c2 = 0 in this case.

123
362 H. Lei et al.

Case 2.5 otherwise


Case 2.5.1 i  = j
In this situation, x rpistjkm , x̄ rpistjkm , y rpistjkm and ȳ rpistjkm are computed by solving the following
algebraic equations:
⎡ ⎤ ⎡ r st ⎤ ⎡ r st ⎤
A1 A2 A3 A4 x pi jkm X pi jkm
⎢ −A2 A1 −A4 A3 ⎥ ⎢ x̄ rpistjkm ⎥ ⎢ X̄ r st ⎥
⎢ ⎥ ⎢ r st ⎥ = ⎢ ⎥
⎢ rpistjkm ⎥ , (45)
⎣ B1 B2 B3 B4 ⎦ ⎣ y pi jkm ⎦ ⎣ Y pi jkm ⎦
−B2 B1 −B4 B3 ȳ rpistjkm r st
Ȳ pi jkm

for the x–y components, and


 
C1 C2 z rpistjkm Z rpistjkm
= , (46)
−C2 C1 z̄ rpistjkm Z̄ rpistjkm

for the z component. In Eqs. (45) and (46),



⎪ A1 = −(ω0 s + ν0 t)2 + λ20 (i − j)2 − (1 + 2c2 ),


⎨A = 2(i − j)λ0 (ω0 s + ν0 t),
2
(47)

⎪ 3
A = −2(i − j)λ0 ,


A4 = −2(ω0 s + ν0 t),

and


⎪ B1 = 2(i − j)λ0 ,


⎨ B2 = 2(ω0 s + ν0 t),
(48)

⎪ B3 = −(ω0 s + ν0 t)2 + λ20 (i − j)2 − (1 − c2 ),



B4 = 2(i − j)λ0 (ω0 s + ν0 t),

and

C1 = −(ω0 s + ν0 t)2 + λ20 (i − j)2 + c2 ,
(49)
C2 = 2(i − j)λ0 (ω0 s + ν0 t).

Case 2.5.2 i = j
As stated, when i = j, one gets x̄ rpistjkm = 0, y rpistjkm = 0 and z̄ rpistjkm = 0, and the remaining
unknown coefficients are obtained by solving
r st  r st 
A1 A4 x pi jkm X pi jkm
= , (50)
−B2 B3 r st
ȳ pi jkm r st
Ȳ pi jkm

for the x–y components, and



z rpistjkm = Z rpistjkm C1 , (51)

for the z component. In Eqs. (50) and (51), A1 , A4 , B2 , B3 and C1 have the same expressions
as those in Eqs. (47–49).

123
High-order solutions of invariant manifolds in ERTBP 363

4 High-order solutions of invariant manifolds associated with halo orbits in ERTBP

4.1 Series expansions of the invariant manifolds

For large amplitudes, considering the nonlinear terms of the equations of motion in ERTBP,
it is possible to generate ‘halo orbit’ (call ‘halo orbit’ in order to be consistent with the case in
CRTBP) whose in-plane and out-of-plane frequencies of motion are identical. It is worthy to
note that ‘halo orbit’ in ERTBP is quasi-periodic orbit due to the perturbation of eccentricity.
Similar to the halo case in CRTBP discussed in Jorba and Masdemont (1999), modify the
equations of motion in the L i -centered synodic coordinate system as follows:
⎧ 

⎪  

⎪ x − 2y − (1 + 2c 2 )x = (−e)i
cos i
f (1 + 2c 2 )x



⎪ i≥1

⎪   



⎪ + (−e) i
cos i
f c (n + 1)T (x, y, z) ,


n+1 n



⎪ 
i≥0 n≥2




⎪ y 
+ 2x 
− (1 − c )y = (−e)i
cos i
f (1 − c )y


2 2
⎨ i≥1
⎪    (52)

⎪ + (−e) i i
(x, ,


cos f y c n+1 R n−1 y, z)




i≥0 n≥2

⎪ 


⎪ z  + c2 z = (−e)i cosi f (1 − c2 )z + Δz





⎪ i≥1



⎪   

⎪ + (−e) i
cos i
f z c R (x, y, z) ,
⎩ n+1 n−1
i≥0 n≥2

which is different from the form adopted in Hou and Liu (2011a). In Eq. (52), Δ represents the
residual part of the third equation, and Δ = 0 is needed for halo case. Similar to the Lissajous
case discussed in above section, the general solutions of invariant manifolds associated with
halo orbits around the collinear libration points in ERTBP are expanded as
⎧  rs 
⎪  x cos(r f + sθ )+

⎪ x( f ) = exp [(i − j)θ3 ] r s
pi jkm 1 j
e p α1i α2 α3k α4m ,



⎪ x̄ sin(r f + sθ )

⎪  rs
pi jkm 1


⎨  y pi jkm cos(r f + sθ1 )+ p i j k m
y( f ) = exp [(i − j)θ3 ] r s e α1 α2 α3 α4 , (53)

⎪ ȳ pi jkm sin(r f + sθ1 )

⎪  rs 



⎪  z pi jkm cos(r f + sθ1 )+ p i j k m


⎩ z( f ) = exp [(i − j)θ3 ] r s e α1 α2 α3 α4 ,
z̄ pi jkm sin(r f + sθ1 )

where p, i, j, k, m ∈ N and r, s ∈ Z. θ1 = ω f + φ and θ3 = λ f , where φ is the initial phase


angle. The coefficients x rpis jkm , x̄ rpis jkm , y rpis jkm , ȳ rpis jkm , z rpis jkm and z̄ rpis jkm corresponding to
the coordinate will be computed iteratively by using the L–P method. The first and second
derivatives of coordinate with respect to f , appearing in the left-hand side of the equations
of motion, can be computed in the following manner (taking x as an example):

∂x ∂x ∂x
x = +ω +λ , (54)
∂f ∂θ1 ∂θ3

123
364 H. Lei et al.

and
∂2x 2∂ x
2
2∂ x
2 ∂2x ∂2x ∂2x
x  = + ω + λ + 2ω + 2λ + 2ωλ . (55)
∂ f2 ∂θ12 ∂θ32 ∂ f ∂θ1 ∂ f ∂θ3 ∂θ1 ∂θ3
The similar expressions can be derived for y and z.
In the process of construction, the frequencies ω and λ, as well as Δ are not constant,
and also expanded as power series of orbital eccentricity e, the amplitudes corresponding to
the hyperbolic manifolds α1 (unstable component) and α2 (stable component), and ampli-
tudes corresponding to the center manifolds α3 (in-plane component) and α4 (out-of-plane
component) as follows:
⎧  j

⎪ ω= ω pi jkm e p α1i α2 α3k α4m ,

⎨  j
λ= λ pi jkm e p α1i α2 α3k α4m , (56)

⎪ 

⎩Δ = j
d pi jkm e p α1i α2 α3k α4m .

Similar to the Lissajous case in above section, three orders are defined in the following
manner: N1 represents the order of orbital eccentricity, N2 refers to the order of the hyperbolic
manifolds, and N3 corresponds to the order of the center manifolds. The total order of
analytical solution is N = N1 + N2 + N3 . We use (N1 , N2 , N3 ) to stand for the order of
analytical solution to be constructed, and N2 ≤ N3 is required. For any p, i, j, k and m, it
requires 0 ≤ p ≤ N1 , 0 ≤ i + j ≤ N2 and 0 ≤ k + m ≤ N3 . For r and s, it requires
that − p ≤ r ≤ p and −k − m ≤ s ≤ k + m. Moreover, r and s should have the same
parity of p and k + m, respectively. Considering the symmetries of cosine and sine functions,
only the terms corresponding to 0 ≤ s ≤ k + m are computed. It is worthy to note that
Δ(e, α1 , α2 , α3 , α4 ) = 0 is required for halo case, that is, the orbital eccentricity e, and
the amplitudes αi , i = 1, . . . , 4, are not independent. For given parameters e, α1 , α2 , α3
or α4 , the parameter α4 or α3 is computed in order to satisfy Δ = 0. When α1  = 0 and
α2 = 0, Eq. (53) represents the unstable manifolds, whereas if α1 = 0 and α2  = 0, it
represents the stable manifolds associated with halo orbits. In addition, if α1 · α2 < 0,
Eq. (53) stands for the transit trajectories, else if α1 · α2 > 0, it represents the non-transit
trajectories associated with halo orbits in ERTBP. In particular, when α1 = 0 and α2 = 0,
Eq. (53) can be reduced to describe the dynamics in the center manifolds around the collinear
libration points in ERTBP, and this reduced situation is equivalent to the case discussed in
Hou and Liu (2011a). Furthermore, when the orbital eccentricity satisfies e = 0, Eq. (53)
can be reduced to describe the dynamics in the invariant manifolds (including the center
and hyperbolic manifolds) associated with halo orbits around the collinear libration points in
CRTBP. This reduced situation is equivalent to the case discussed in Jorba and Masdemont
(1999) about the center manifolds and the case discussed in Masdemont (2005) about the
invariant manifolds in CRTBP.
For halo case, the starting solution with first-order amplitude and zero-order eccentricity
we adopt for L–P method are modified as

⎨ x( f ) = α1 exp(λ0 f ) + α2 exp(−λ0 f ) + α3 cos(ω0 f + φ),

y( f ) = κ̄2 α1 exp(λ0 f ) − κ̄2 α2 exp(−λ0 f ) + κ̄1 α3 sin(ω0 f + φ), (57)


z( f ) = α4 cos(ω0 f + φ),
where κ̄1 , κ̄2 , ω0 and λ0 have the same forms as Eqs. (14) and (16). Substituting the linear
solution into the linearized equations of Eq. (52) with zero eccentricity, one gets d00000 −

123
High-order solutions of invariant manifolds in ERTBP 365

c2 − ω02  = 0, thus halo orbits don’t exist in the linearized equations of CRTBP. According
to Eq. (57), one gets
00
x01000 = 1, x00100
00
= 1, x00010
01
= 1,

for the x component, and


00
y01000 = κ̄2 , 00
y00100 = −κ̄2 , 01
ȳ00010 = κ̄1 ,

for the y component, and


01
z 00001 = 1.

for the z component. Moreover, for the coefficients of frequency, one gets

ω00000 = ω0 , λ00000 = λ0 , d00000 = c2 − ω02 .

For halo case, some coefficients of the general solution are also zero due to the symmetries of
the equations of motion. When m is odd, the coefficients x rpis jkm , x̄ rpis jkm , y rpis jkm and ȳ rpis jkm
are zero. When m is even, z rpis jkm and z̄ rpis jkm are zero. In addition, for any p, i, j, k and m,
the coefficients of coordinate have the following symmetries:

x rpis jkm = x rpjikm


s
, y rpis jkm = −y rpjikm
s
, z rpis jkm = z rpjikm
s
,
x̄ rpis jkm = −x̄ rpjikm
s
, ȳ rpis jkm = ȳ rpjikm
s
, z̄ rpis jkm = −z̄ rpjikm
s
.

It is easy to verify that when i = j, one gets x̄ rpis jkm = y rpis jkm = z̄ rpis jkm = 0. For the terms of
frequency, the coefficients are not zero only if both k and m are even, besides satisfying i = j.
These properties could also help us to check the validity of the procedure of computation
and save computer storage.
Similar to the Lissajous case, the high-order solutions of invariant manifolds associ-
ated with halo orbits in ERTBP are constructed from lower-order solutions by utilizing
the L–P method. If the general solution is truncated at order (N1 , N2 , N3 ), the terms
of order (n 1 , n 2 , n 3 ), n 1 ≤ N1 , n 2 ≤ N2 and n 3 ≤ N3 , are to be calculated. Taking
the construction of analytical solution corresponding to order (n 1 , n 2 , n 3 ) as an example,
the unknown coefficients include x rpis jkm , x̄ rpis jkm , y rpis jkm , ȳ rpis jkm , z rpis jkm and z̄ rpis jkm , where
p = n 1 , i + j = n 2 and k + m = n 3 , corresponding to the coefficients of the coordi-
nate. The coefficients corresponding to the frequency, ω pi jkm , λ pi jkm and d pi jkm , where
p + i + j + k + m = n 1 + n 2 + n 3 − 1, are also unknown and will be computed iteratively.
The solution corresponding to the order (n 1 , n 2 , n 3 ) can be computed only if the terms of
orders (n̄ 1 , n̄ 2 , n̄ 3 ), where n̄ 1 ≤ n 1 , n̄ 2 ≤ n 2 , n̄ 3 ≤ n 3 and n̄ 1 + n̄ 2 + n̄ 3 < n 1 + n 2 + n 3 , have
been known. The high-order solutions of invariant manifolds up to order (n 1 , n 2 , n 3 ) asso-
ciated with halo orbits in ERTBP are constructed in the following manner. Firstly, construct
the terms (:, n̄ 2 , n̄ 3 ), n̄ 2 < n 2 , n̄ 3 < n 3 from the order (n̄ 1 , n̄ 2 , n̄ 3 ) to the order (n 1 , n̄ 2 , n̄ 3 ),
where if n̄ 2 + n̄ 3 = 1, then n̄ 1 = 1, else n̄ 1 = 0. Then construct the terms (n 1 , :, n̄ 3 ), n̄ 3 < n 3
from the order (n 1 , n̄ 2 , n̄ 3 ) to the order (n 1 , n 2 , n̄ 3 ), in which if n̄ 3 = 0, then n̄ 2 = 1, else
n̄ 2 = 0. Finally, we construct the terms (n 1 , n 2 , :) from the order (n 1 , n 2 , n̄ 3 ) to the order
(n 1 , n 2 , n 3 ), in which if n 2 = 0, then n̄ 3 = 1, else n̄ 3 = 0. Substituting the known terms up
to order (n 1 − 1, n 2 , n 3 ), (n 1 , n 2 − 1, n 3 ) and (n 1 , n 2 , n 3 − 1) into Eq. (52), the unknown
coefficients of analytical solutions corresponding to order (n 1 , n 2 , n 3 ) can be obtained by
solving the following linear system of equations:

123
366 H. Lei et al.

⎡ ⎤ ⎡ xrs ⎤ ⎡ ⎤ ⎡ rs
X pi jkm

A1 A2 A3 A4 0 0 pi jkm δxc
⎢ −A2 ⎢ x̄ r s ⎥ ⎢ δ s ⎥ ⎢ X̄ r s ⎥
⎢ A1 −A4 A3 0 0 ⎥⎥⎢ pi jkm ⎥ ⎢ pi jkm ⎥
⎢ B3 ⎥ ⎢ yr s ⎥ ⎢ ⎢
x⎥
⎥ ⎢ r s ⎥
⎢ B4 B1 B2 0 0 ⎥ ⎢ pi jkm ⎥ ⎢ δ y ⎥ ⎢ Y pi jkm ⎥
c
⎢ −B4 ⎥ ⎢ rs ⎥ + ⎢ s ⎥ = ⎢ rs ⎥, (58)
⎢ B3 −B2 B1 0 0 ⎥ ⎢ ȳ pi jkm ⎥ ⎢ δ y ⎥ ⎢ Ȳ pi jkm ⎥
⎣0 ⎢ ⎥ ⎢ ⎥
0 0 0 C1 C2 ⎦ ⎣ z rpis jkm ⎦ ⎣ δzc ⎦ ⎣ Z rpis jkm ⎦
0 0 0 0 −C2 C1 z̄ r s δzs Z̄ r s
pi jkm pi jkm

where X rpis jkm , X̄ rpis jkm , Y pi


rs
jkm , Ȳ pi jkm , Z pi jkm and Z̄ pi jkm are the known components of the
rs rs rs

equations of motion corresponding to order (n 1 , n 2 , n 3 ), and the elements of the constant


coefficient matrix in Eq. (58) are given by

⎪ A1 = −(r + ω0 s)2 + λ20 (i − j)2 − (1 + 2c2 ),


⎨A = 2λ0 (i − j)(r + ω0 s),
2
(59)

⎪ A = −2(i − j)λ0 ,


3
A4 = −2(r + ω0 s),

and

⎪ B1 = −(r + ω0 s)2 + λ20 (i − j)2 − (1 − c2 ),


⎨B = 2λ0 (i − j)(r + ω0 s),
2
(60)

⎪ 3
B = 2(i − j)λ0 ,


B4 = 2(r + ω0 s),

and

C1 = −(r + ω0 s)2 + λ20 (i − j)2 + c2 − d0 ,
(61)
C2 = 2λ0 (i − j)(r + ω0 s),

and
⎧ c
⎪ δx


= − 2(ω0 + κ̄1 )ω pi jk−1m δr 0 δs1 δi j



⎪ + 2(λ0 − κ̄2 )λ pi−1 jkm δr 0 δs0 δi−1 j



⎪ + 2(λ0 − κ̄2 )λ pi j−1km δr 0 δs0 δi j−1 ,




⎪ δxs
⎪ = 0,

δ cy = 2(κ̄2 λ0 + 1)λ pi−1 jkm δr 0 δs0 δi−1 j (62)



⎪ − 2(κ̄2 λ0 + 1)λ pi j−1km δr 0 δs0 δi j−1 ,





⎪ δ sy = − 2(κ̄1 ω0 + 1)ω pi jk−1m δr 0 δs1 δi j ,



⎪ δzc = − (d pi jkm−1 + 2ω0 ω pi jkm−1 )δr 0 δs1 δi j ,



⎩ s
δz = 0.

4.2 Solving the undetermined coefficients

Similar to the Lissajous case, we can solve the undetermined coefficients in different situa-
tions.

123
High-order solutions of invariant manifolds in ERTBP 367

Case 1 r  = 0
Case 1.1 i  = j
In this situation, the coefficients of frequency are zero, and the unknown coefficients includ-
ing x rpis jkm , x̄ rpis jkm , y rpis jkm , ȳ rpis jkm , z rpis jkm and z̄ rpis jkm , satisfy the following linear system of
equations:
⎡ ⎤ ⎡ xrs ⎤ ⎡ rs
X pi jkm

A1 A2 A3 A4 0 0 pi jkm
⎢ −A2 A1 −A4 A3 0 ⎢ x̄ r s ⎥ ⎢ X̄ r s ⎥
⎢ 0 ⎥ ⎥⎢ pi jkm ⎥ ⎢ pi jkm ⎥
⎢ B3 ⎥ ⎢ r s ⎥ ⎢ Y rs ⎥
⎢ B4 B1 B2 0 0 ⎥ ⎢ pi jkm ⎥ ⎢ pi jkm ⎥
y
⎢ −B4 B3 −B2 B1 0 ⎢ ⎥ = ⎢ ⎥. (63)
⎢ 0 ⎥ ⎥⎢ ȳ rpis jkm ⎥ ⎢ Ȳ pirs
jkm ⎥
⎣0 ⎢ ⎥ ⎢ ⎥
0 0 0 C1 C2 ⎦ ⎣ z rpis jkm ⎦ ⎣ Z rpis jkm ⎦
0 0 0 0 −C2 C1 z̄ rpis jkm Z̄ rpis jkm
The x–y components and the z component are uncoupled, and can be solved separatively.
Case 1.2 i = j
As stated in Sect. 4.1, when i = j, the coefficients including x̄ rpis jkm , y rpis jkm and z̄ rpis jkm are
zero, thus the unknown coefficients satisfy
⎡ ⎤ ⎡ rs ⎤ ⎡ rs ⎤
A1 A4 0 x pi jkm X pi jkm
⎣ −B4 B1 0 ⎦ ⎣ ȳ rpis jkm ⎦ = ⎢ rs ⎥
⎣ Ȳ pi jkm ⎦ . (64)
0 0 C1 r s
z pi jkm r s
Z pi jkm
In Eqs. (63) and (64), A1 , A2 , A3 , A4 , B1 , B2 , B3 , B4 , C1 and C2 have the same forms as
Eqs. (59–61). The undetermined coefficients x rpis jkm , ȳ rpis jkm and z rpis jkm can be solved easily.
Case 2 r = 0
Case 2.1 s = 0 and |i − j| = 1
In this case, x̄ rpis jkm , ȳ rpis jkm and z̄ rpis jkm are the coefficients of sin(0), and can be taken as
zero, then the unknown coefficients satisfy
rs c rs
A1 A3 x pi jkm δx X pi jkm
+ = , (65)
B3 B1 y rpis jkm δy
c rs
Y pi jkm

for the x − y components, where A1 = λ20 − (1 + 2c2 ), A3 = −2(i − j)λ0 , B1 = λ20 − (1 −


c2 ), B3 = 2(i − j)λ0 , δxc = 2(λ0 − κ̄2 )λ pnnkm and δ cy = 2(i − j)(κ̄2 λ0 + 1)λ pnnkm , here n =
min(i, j). In Eq. (65), one can compute x rpis jkm and λ pnnkm by taking y rpis jkm = 0, or compute
y rpis jkm and λ pnnkm by taking x rpis jkm = 0 (the first case is adopted in our computation). For
the z component, the unknown coefficients satisfy
C1 · z rpis jkm = Z rpis jkm , (66)

where C1 = λ20 + c2 − d0 .
Case 2.2 s = 1 and i = j
As analyzed similarly, when i = j, the coefficients including x̄ rpis jkm , y rpis jkm and z̄ rpis jkm are
zero, thus the unknown coefficients satisfy
rs c  rs 
A1 A4 x pi jkm δx X pi jkm
+ s = , (67)
−B4 B1 ȳ rpis jkm δy rs
Ȳ pi jkm

corresponding to the x − y components. In Eq. (67), A1 = −ω02 − (1 + 2c2 ), A4 =


−2ω0 , B1 = −ω02 − (1 − c2 ), B4 = 2ω0 , δxc = −2(ω0 + κ̄1 )ω pi jk−1m and δ sy =

123
368 H. Lei et al.

−2(κ̄1 ω0 + 1)ω pi jk−1m . From Eq. (67), one can compute x rpis jkm and ω pi jk−1m by taking
ȳ rpis jkm = 0, or compute ȳ rpis jkm and ω pi jk−1m by taking x rpis jkm = 0 (the latter case is adopted
in our computation). For the z component, the unknown coefficients satisfy
C1 · z rpis jkm − (d pi jkm−1 + 2ω0 ω pi jkm−1 ) = Z rpis jkm , (68)

where C1 = −ω02 + c2 − d0 = 0. From Eq. (68), one can compute d pi jkm−1 by taking
z rpis jkm = 0.

Case 2.3 otherwise

Case 2.3.1 i  = j
In this situation, the coefficients of frequency are all zero, and the unknown coefficients
including x rpis jkm , x̄ rpis jkm , y rpis jkm , ȳ rpis jkm , z rpis jkm and z̄ rpis jkm are computed by solving the fol-
lowing algebraic equations:
⎡ ⎤ ⎡ xrs ⎤ ⎡ rs
X pi jkm

A1 A2 A3 A4 0 0 pi jkm
⎢ −A2 A1 −A4 A3 0 ⎢ x̄ r s ⎥ ⎢ X̄ r s ⎥
⎢ 0 ⎥⎥⎢ pi jkm ⎥ ⎢ pi jkm ⎥
⎢ B3 ⎢ y pi jkm ⎥ ⎢ Y pi jkm ⎥
⎥ ⎢
0 ⎥ r s r s
⎢ B4 B1 B2 0 ⎥⎢⎢ ⎥ = ⎢ rs

⎥. (69)
⎢ −B4 B3 −B2 B1 0 0 ⎥ ȳ rpis jkm ⎥ ⎢ Ȳ pi
⎢ ⎥⎢⎢ ⎥ ⎢ jkm ⎥

⎣0 0 0 0 C1 C2 ⎦ ⎣ z rpis jkm ⎦ ⎣ Z rpis jkm ⎦
0 0 0 0 −C2 C1 z̄ rpis jkm Z̄ rpis jkm
Case 2.3.2 i = j
The coefficients including x̄ rpis jkm , y rpis jkm and z̄ rpis jkm are known as zero, thus the undeter-
mined coefficients satisfy
⎡ ⎤ ⎡ rs ⎤ ⎡ rs ⎤
A1 A4 0 x pi jkm X pi jkm
⎣ −B4 B1 0 ⎦ ⎣ ȳ rpis jkm ⎦ = ⎢ rs ⎥
⎣ Ȳ pi jkm ⎦ . (70)
0 0 C1 r s
z pi jkm r s
Z pi jkm

In Eqs. (69) and (70), the elements of constant coefficient matrix are given by

⎪ A = −ω02 s 2 + λ20 (i − j)2 − (1 + 2c2 ),
⎪ 1

⎨ A = 2ω λ s(i − j),
2 0 0
(71)

⎪ A3 = −2λ0 (i − j),


A4 = −2ω0 s,
and

⎪ B1 = −ω02 s 2 + λ20 (i − j)2 − (1 − c2 ),


⎨ B2 = 2ω0 λ0 s(i − j),
(72)

⎪ B3 = 2λ0 (i − j),


B4 = 2ω0 s,
and

C1 = −ω02 s 2 + λ20 (i − j)2 + c2 − d0 ,
(73)
C2 = 2ω0 λ0 s(i − j).
All undetermined coefficients corresponding to both coordinate and frequency can be easily
computed by solving the corresponding algebraic equations.

123
High-order solutions of invariant manifolds in ERTBP 369

Table 1 CPU time (in seconds, for a PC with a 2.60 GHz Intel (R) Core (TM) i5 vPro CPU), the average
residual acceleration of series expansions defined in the context, and RAM memory (in Kilobytes for frequency,
and Megabytes for coordinate) for the coefficients of analytical solutions of the invariant manifolds associated
with Lissajous and halo orbits around the L 1 point in the Sun–Earth ERTBP

Case Order Freq. (KB) Coord. (MB) CPU time (s) Residual
acceleration

Lissajous (1, 1, 1) 1.60E−01 8.09E−03 1.56E−02 1.05E−04


(2, 2, 2) 1.86 1.02E−01 1.72E−01 3.68E−06
(3, 3, 3) 1.86 6.05E−01 3.03E+00 2.90E−07
(4, 4, 4) 8.38 2.66E+00 3.81E+01 1.55E−08
(4, 4, 5) 8.38 4.79E+00 1.05E+02 1.93E−09
(4, 3, 6) 9.31 5.33E+00 1.17E+02 1.18E−10
(5, 3, 5) 5.58 4.26E+00 7.58E+01 1.93E−09
(4, 2, 7) 9.31 5.02E+00 1.00E+02 1.98E−11
Halo (3, 3, 3) 1.86 4.70E−01 1.42E+00 3.63E−04
(4, 4, 4) 8.38 1.95E+00 1.39E+01 1.10E−04
(5, 5, 5) 8.38 6.00E+00 1.07E+02 4.03E−05
(5, 5, 9) 2.09E+01 2.46E+01 1.42E+03 4.80E−07
(5, 3, 12) 2.60E+01 2.48E+01 1.43E+03 1.84E−08

5 Results

In our computation, the Sun–Earth elliptic restricted three-body problem is taken as an


example. The mass parameter of this three-body system is μ = 3.003480575402412 × 10−6 ,
and the orbital eccentricity of the Earth on the elliptic orbit is e = 0.01671022. The procedures
corresponding to the construction of high-order analytical solutions of invariant manifolds
associated with both Lissajous and halo orbits in ERTBP are completed in FORTRAN 95
language. It is worthy to note that, in the process of construction, nine- and eight-dimensional
arrays for Tn (and Rn ) appearing in the equations of motion are needed to define and compute
for Lissajous and halo cases, respectively. Unfortunately, the Fortran compiler can only
admit an array up to seven dimensions, and we transform the multi-dimensional array into a
single-dimensional array to overcome this technological difficulty. Due to the large storage
of coefficients corresponding to the coordinate and frequency, the analytical solutions of
invariant manifolds associated with libration point orbits in ERTBP can be constructed up to
a limited order. When the general solutions are reduced to different cases, the order of the
solution (N1 , N2 , N3 ) can be adapted as needed for describing the corresponding dynamics
around the collinear libration points more accurately. For example, the series expansions
truncated at order (N1 , 0, N3 ) are suitable for computing Lissajous or halo orbits in ERTBP,
the ones truncated at order (0, N2 , N3 ) are suitable for computing the invariant manifolds in
CRTBP, and the ones constructed up to order (0, 0, N3 ) are suitable for computing Lissajous
or halo orbits in CRTBP.
The CPU time, RAM memory required to generate the coefficients of frequency and
coordinate series expansions truncated at different orders are summarized in Table 1. The
last column of Table 1 is the average residual acceleration, which indicates the accuracy of
the series expansions constructed and is computed in this manner: a spacecraft is assumed to
move along a nominal trajectory obtained by series expansions, at any point of the nominal

123
370 H. Lei et al.

−3 −4
x 10 x 10
2.5 2

1
1.5
y [adim]

z [adim]
1
0
L
1
0.5

0 −1

−0.5

−2
−1
0.9865 0.987 0.9875 0.988 0.9885 0.989 0.9895 0.99 0.9905 0.991 0.987 0.9875 0.988 0.9885 0.989 0.9895 0.99 0.9905

x [adim] x [adim]

−4
x 10
2
−4
x 10
2
1
1
z [adim]
z [adim]

0
0

−1 L
1

−1 −2
20
10 0.99
x 10
−4 0.989
−2 0 0.988
−5 0 5 10 15 20 0.987
−4 y [adim] x [adim]
x 10
y [adim]

Fig. 1 Projections on the coordinate plane and a 3D representation of the analytical solutions, truncated at
order (4, 4, 5), corresponding to the negative branch of the unstable manifold associated with Lissajous orbit
around the collinear libration point L 1 in the Sun–Earth elliptic restricted three-body system. The amplitudes
of the manifold are α1 = −1×10−5 , α2 = 0, α3 = 0.02 and α4 = 0.02. The initial phase angle corresponding
to the out-of-plane component is taken as φ2 = 0, and ten values of φ1 corresponding to the in-plane initial
phase angle in Eq. (17) are taken in [0, 2π ] uniformly. The trajectory specified by (φ1 , φ2 ) in the unstable
manifolds is terminated when the Euclidean norm of the position deviation between the analytical trajectory
and numerically integrated trajectory reaches the given tolerance 1 × 10−5

trajectory, the acceleration of the spacecraft, computed by the equations of motion, is denoted
by a N um , and the second derivatives of series x, y and z represented by Eqs. (17) and (53) are
denoted by a Ana . The average residual acceleration means the average value of the Euclidean
norm of a N um − a Ana for a certain nominal trajectory during π units of dimensionless time.
The dimensionless magnitude of acceleration a = 1.0 is about 5.93 mm/s2 for the Sun–
Earth elliptic restricted three-body system. The similar consideration of residual acceleration
about analytical solutions can be seen in Farquhar and Kamel (1973). For Lissajous case,
the nominal trajectory lies in the positive branch of the unstable manifolds with amplitudes
α3 = α4 = 0.02 and is specified by φ1 = φ2 = 0. For halo case, the nominal trajectory lies
in the positive branch of unstable manifold with α4 = 0.05 and is specified by φ = 0.
The high-order solutions of invariant manifolds associated with Lissajous orbit, up to order
(4, 4, 5), and halo orbit, up to order (5, 5, 9), are constructed following the procedure presented
in Sects. 3 and 4, and the corresponding coefficients of frequencies and coordinates of the
series expansions are given in the form of supplementary material linked to the electronic
version of this paper. In the supplement, the subroutines corresponding to Lissajous and halo
cases written using the FORTRAN language are provided to compute the analytical state
of the invariant manifolds in ERTBP by calling the coefficients provided. Supplementary
material 1 includes a description of the online appended files.

123
High-order solutions of invariant manifolds in ERTBP 371

−4 −4
x 10 x 10
2
5
L1

0
1
y [adim]

z [adim]
−5
0

−10

−1
−15

−20 −2
0.9895 0.99 0.9905 0.991 0.9915 0.992 0.9925 0.993 0.9935 0.9895 0.99 0.9905 0.991 0.9915 0.992 0.9925 0.993 0.9935

x [adim] x [adim]

−4
x 10
2
−4
x 10
2
1
1
z [adim]
z [adim]

0
0
−1 L
1

−1 −2

0
0.993
x 10
−4
−10 0.992
−2 0.991
−20 0.99
−20 −15 −10 −5 0 5 y [adim]
x 10
−4 x [adim]
y [adim]

Fig. 2 Projections on the coordinate plane and a 3D representation of the analytical solutions, truncated at
order (4, 4, 5), corresponding to the positive branch of the unstable manifold associated with Lissajous orbit
around the libration point L 1 in the Sun–Earth elliptic restricted three-body system. The amplitudes of the
manifolds are α1 = 1 × 10−5 , α2 = 0, α3 = 0.02 and α4 = 0.02. The out-of-plane initial phase angle is
taken as φ2 = 0, and ten values of in-plane initial phase angle φ1 in Eq. (17) are taken in [0, 2π ] uniformly.
The terminal condition of these trajectories in the manifolds is the same as that in Fig. 1

Figure 1 shows the negative branch of the unstable manifold with amplitudes α1 = −1 ×
10−5 , α2 = 0, α3 = 0.02 and α4 = 0.02, and Fig. 2 shows the positive branch of the unstable
manifold with amplitudes α1 = 1×10−5 , α2 = 0, α3 = 0.02 and α4 = 0.02. The trajectories
in the invariant manifolds presented in Figs. 1 and 2 are terminated when the Euclidean norm
of the position deviation between the analytical and numerically integrated orbits reaches the
given threshold or tolerance 1 × 10−5 . This terminal condition is applied to the computation
of the invariant manifolds associated with halo orbit in the following tests. The symmetries
of the equations of motion indicate that the stable and unstable manifolds are symmetric
with respect to the x–z plane, and both branches of stable manifolds can be obtained easily
by using this symmetric property. As stated in Sect. 3, Eq. (17) can also describe the transit
(α1 · α2 < 0) or non-transit (α1 · α2 > 0) trajectories around the collinear libration points.
Figure 3 presents the sets of transit and non-transit trajectory. For halo case, the out-of-plane
amplitude of the halo orbit is taken as α4 = 0.05, and the in-plane amplitude is computed to
be α3 = 0.140151260237441 in order to satisfy Δ(e, α1 , α2 , α3 , α4 ) = 0. Figures 4 and 5
show the negative and positive branches of the unstable manifolds associated with halo orbit
in ERTBP, respectively. The transit and non-transit trajectories associated with halo orbit in
ERTBP can be seen in Fig. 6.
Besides describing the hyperbolic manifolds, Eqs. (17) and (53) can be reduced to describe
the dynamics in the center manifolds around the collinear libration points in ERTBP. For

123
372 H. Lei et al.

−3 −3
x 10 x 10
2
2.5
1.5
2
1
1.5
0.5
y [adim]

y [adim]
1
0
0.5
−0.5
0

−0.5 −1
L
L 1
1

−1 −1.5

−1.5 −2
0.988 0.989 0.99 0.991 0.992 0.993 0.9875 0.988 0.9885 0.989 0.9895 0.99 0.9905

x [adim] x [adim]
−4 −4
x 10 x 10
2
2

1 1
z [adim]

z [adim]
0 0

−1 −1

−2 −2
0.987 0.988 0.989 0.99 0.991 0.992 0.993 0.9875 0.988 0.9885 0.989 0.9895 0.99 0.9905

x [adim] x [adim]
−4 −4
x 10 x 10
2 2

1 1
z [adim]

z [adim]

0 0

−1 −1

−2 −2

−5 0 5 10 15 20 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2


−4 −3
x 10 x 10
y [adim] y [adim]

−4
x 10
2

1
L1
z [adim]

L
−4 1
x 10 0
z [adim]

2 0.993
0
−2 0.992 −1
0.991
15 −2
10 0.99
−4 5 2
x 10 0.989
0 1
−5 0.988 x [adim] 0.99
y [adim] −3 0
x 10 0.989
−1
0.988
−2
y [adim] x [adim]

Fig. 3 Projections on the coordinate plane and 3D representation of the sets of transit trajectory (left panel),
and non-transit trajectory (right panel) associated with Lissajous orbit around the L 1 point in the Sun–Earth
ERTBP. The initial phase corresponding to the out-of-plane component is taken as φ2 = 0, and ten values of
the phase angle corresponding to the in-plane motion φ1 in Eq. (17) are taken in [0, 2π ] uniformly

123
High-order solutions of invariant manifolds in ERTBP 373

−3 −4
x 10 x 10

6 5
4
4 3
2
y [adim]

z [adim]
2 1
0
0 −1

L −2
1

−2 −3
−4
−4 −5

0.986 0.987 0.988 0.989 0.99 0.991 0.992 0.986 0.987 0.988 0.989 0.99 0.991 0.992

x [adim] x [adim]
−4
x 10
6

−4
x 10
4 6

4
2
2
z [adim]
z [adim]

0
0 L1
−2

−2 −4

−4 5 0.992
0.99
−3 0
x 10 0.988
0.986
−6 0.984
−6 −4 −2 0 2 4 6 8 −5
−3
y [adim] x [adim]
x 10
y [adim]

Fig. 4 Projections on the coordinate plane and 3D view of the analytical solutions, truncated at order
(5, 5, 9), corresponding to the negative branch of the unstable manifold associated with halo orbit around
the collinear libration point L 1 in the Sun–Earth elliptic restricted three-body system. The amplitudes of
the invariant manifold are α1 = −1 × 10−5 , α2 = 0 and α4 = 0.05, then the in-plane amplitude is
α3 = 0.140151260237441 such that Δ(e, α1 , α2 , α3 , α4 ) = 0. Ten values of the initial phase angle φ in
Eq. (53) are taken in [0, 2π ] uniformly. The terminal condition of the trajectory in the manifold is the same
as that in Fig. 1

example, (i) the Lissajous orbit, corresponding to α1 = α2 = 0, α3  = 0 and α4  = 0 in


Eq. (17); (ii) the plane Lyapunov orbit, corresponding to α1 = α2 = α4 = 0 and α3  = 0 in
Eq. (17); (iii) the vertical Lyapunov orbit, corresponding to α1 = α2 = α3 = 0 and α4  = 0
in Eq. (17); and (iv) the halo orbit, corresponding to α1 = α2 = 0, α3  = 0 and α4  = 0
in Eq. (53). Hou and Liu (2011a) directly expanded the Lissajous and halo orbits around
the collinear libration points in ERTBP as power series of orbital eccentricity e, in-plane
amplitude α3 and out-of-plane amplitude α4 , and constructed the high-order solutions up to
arbitrary order by means of the L–P method. In order to compare with the series expansions
constructed in this paper, the algebraic manipulator for computing the analytical solutions of
Lissajous and halo orbits in ERTBP discussed in Hou and Liu (2011a) is modified in terms
of the following aspects: (a) the instantaneous distance between the collinear libration point
and its closest primary is taken as length unit; (b) the equations of motion represented by
Eq. (52) are taken to construct the high-order solutions of halo orbit, which is different from
the form adopted by Hou and Liu (2011a); and (c) the main construction process of analytical
solutions is modified.
To compute the trajectories in the center manifolds in ERTBP, the high-order solutions
are constructed up to order (7, 0, 9), where the order of the hyperbolic manifolds is zero.
The first three graphs of Figs. 7 and 9 present the projections on the coordinate plane of
Lissajous and halo orbits, respectively, and Fig. 8 shows a plane and vertical Lyapunov orbit

123
374 H. Lei et al.

−3 −4
x 10 x 10
5 5
4 4
3 3
2 2
L1
1 1
y [adim]

z [adim]
0 0
−1 −1
−2 −2
−3 −3
−4 −4
−5 −5
−6
0.988 0.989 0.99 0.991 0.992 0.993 0.994 0.995 0.988 0.989 0.99 0.991 0.992 0.993 0.994 0.995

x [adim] x [adim]
−4
x 10
6

−4
x 10
4 5

2
z [adim]
z [adim]

0
0

L
1
−2 −5

5
−4
0.994
−3
0
x 10 0.992
−6 0.99
−5
−6 −4 −2 0 2 4 6 0.988
−3
y [adim] x [adim]
x 10
y [adim]

Fig. 5 Projections on the coordinate plane and 3D view of the analytical solutions, truncated at order (5, 5, 9),
corresponding to the positive branch of the unstable manifold associated with halo orbit around the collinear
libration point L 1 in the Sun–Earth elliptic restricted three-body system. The amplitudes of the invariant
manifold are α1 = 1 × 10−5 , α2 = 0 and α4 = 0.05. Ten values of the initial phase angle φ in Eq. (53) are
taken in [0, 2π ] uniformly. The terminal condition of the trajectory in the manifold is the same as that in Fig. 1

in ERTBP. The Lissajous and halo orbits with the same amplitudes are also computed by
the modified procedure discussed in Hou and Liu (2011a), and Figs. 7 (bottom-right graph)
and 9 (bottom-right graph) show the Euclidean norm of the position deviation between the
orbits obtained by the two analytical approaches, including the series expansions constructed
in this paper, and the series expansions discussed in Hou and Liu (2011a), up to the same
order within 40π units of dimensionless time. The time histories of the position deviation
indicate that the series expansions represented by Eqs. (17) and (53) can be exactly reduced
to the series expansions discussed in Hou and Liu (2011a).
As is known, when the orbital eccentricity of the primary is zero, ERTBP is reduced to the
case of CRTBP. Thus the analytical solutions of invariant manifold associated with libration
point orbits in ERTBP can also describe the dynamics around the collinear libration points
in CRTBP. That is, when taking e = 0, Eq. (17) can describe Lissajous orbit and its invariant
manifolds, and Eq. (53) can describe halo orbit and its invariant manifolds in CRTBP. In
Jorba and Masdemont (1999), the Lissajous and halo orbits around the collinear libration
points in CRTBP are directly expanded as power series of in-plane amplitude α3 and out-of-
plane amplitude α4 . In Masdemont (2005), the invariant manifolds associated with Lissajous
and halo orbits in CRTBP are directly expressed as power series of hyperbolic amplitudes α1
(unstable amplitude) and α2 (stable amplitude), and center amplitudes α3 (in-plane amplitude)
and α4 (out-of-plane amplitude). In order to compare with the series expansions constructed
in this paper, the algebraic manipulator for computing the analytical solutions of center and

123
High-order solutions of invariant manifolds in ERTBP 375

−3
x 10
6 0.01

0.008
4
0.006 L1
2
0.004
0
0.002
y [adim]

y [adim]
−2 0

−0.002
−4
−0.004
−6
−0.006
L1
−8
−0.008

−10 −0.01
0.985 0.99 0.995 0.984 0.986 0.988 0.99 0.992 0.994

x [adim] x [adim]
−4 −4
x 10 x 10
6
5
4
4
3
2
2
z [adim]

z [adim]
0 1
0
−2 −1
−2
−4
−3
−4
−6
−5
−8
0.984 0.986 0.988 0.99 0.992 0.994 0.996 0.984 0.985 0.986 0.987 0.988 0.989 0.99 0.991 0.992 0.993

x [adim] x [adim]
−4 −4
x 10 x 10
6 6

4 4

2 2
z [adim]

z [adim]

0 0

−2 −2

−4 −4

−6 −6

−8 −8
−12 −10 −8 −6 −4 −2 0 2 4 6 −0.015 −0.01 −0.005 0 0.005 0.01 0.015
−3
x 10
y [adim] y [adim]

−4 −4
x 10 x 10
6
5
4

2
z [adim]

z [adim]

0
0

−2
−5
−4 L1

5 0.01
L1
0.995 0.005 0.992
0
−3 0 0.99
x 10 −5 0.99 0.988
−0.005
0.986
−10 0.985 −0.01 0.984
y [adim] x [adim] y [adim] x [adim]

Fig. 6 Sets of transit trajectories (left panel) and non-transit trajectories (right panel) associated with halo
orbit around the collinear libration point L 1 in the Sun–Earth elliptic restricted three-body system, computed
by series expansions truncated at order (5, 5, 9). Ten values of the initial phase angle φ in Eq. (53) are taken
in [0, 2π ] uniformly

123
376 H. Lei et al.

−4
−4 x 10
x 10 4
8
3
6

2
4

1
2

z [adim]
y [adim]

0 0
L1

−2 −1

−4 −2

−6 −3

−8 −4
0.9898 0.9898 0.9899 0.9899 0.99 0.99 0.9901 0.9901 0.9902 0.9902 0.9898 0.9898 0.9899 0.9899 0.99 0.99 0.9901 0.9901 0.9902 0.9902

x [adim] x [adim]
−4 −8
x 10 x 10
4 1

0.8
3
0.6
2
0.4
1
Deviation

0.2
z [adim]

0 0

−0.2
−1
−0.4
−2
−0.6
−3
−0.8

−4 −1
−8 −6 −4 −2 0 2 4 6 8 0 20 40 60 80 100 120
−4
x 10
y [adim] Time [adim]

Fig. 7 Projections on the coordinate plane of a Lissajous orbit (the first three graphs) around the collinear
libration point L 1 in the Sun–Earth elliptic restricted three-body system, computed by series expansions up to
order (7, 0, 9), and the Euclidean norm of the position deviation (bottom-right graph) between the Lissajous
orbits obtained by the series expansions constructed in this paper and the series expansions constructed up to
order (7, 9) discussed in Hou and Liu (2011a) during 40π units of dimensionless time. The amplitudes of the
Lissajous orbit are α3 = 0.02 and α4 = 0.03

−4
x 10
8
α = 0.02
3 −4
6 x 10
4
4 α4 = 0.03
2
z [adim]

2
y [adim]

0
0 L1
L
−2 1
−2

−4 −4
1
0.5 0.99
−6 0.99
−6 0 0.99
x 10 0.99
−0.5
−8 0.99
0.9898 0.9898 0.9899 0.9899 0.99 0.99 0.9901 0.9901 0.9902 0.9902 −1 0.99
y [adim] x [adim]
x [adim]

Fig. 8 Plane (left panel) and vertical (right panel) Lyapunov orbits around the collinear libration point L 1
in the Sun–Earth elliptic restricted three-body problem. The amplitudes of the plane Lyapunov orbit are
α3 = 0.02 and α4 = 0, and the ones of the vertical Lyapunov orbit are α3 = 0 and α4 = 0.03

invariant manifolds in CRTBP are developed according to the discussion presented in Jorba
and Masdemont (1999) and Masdemont (2005), respectively.
The high-order analytical solutions of invariant manifolds associated with Lissajous and
halo orbits in CRTBP are constructed up to order (0, 7, 9) and (0, 9, 12), respectively. The first

123
High-order solutions of invariant manifolds in ERTBP 377

−4
−3 x 10
x 10 6
5

4
4
3

2 2

z [adim]
1
y [adim]

0 0

−1
L
1 −2
−2

−3 −4
−4

−5 −6
0.9885 0.989 0.9895 0.99 0.9905 0.991 0.9915 0.992 0.9885 0.989 0.9895 0.99 0.9905 0.991 0.9915 0.992

x [adim] x [adim]
−4 −8
x 10 x 10
6 1

0.8
4
0.6

0.4
2
Deviation
0.2
z [adim]

0 0

−0.2
−2
−0.4

−0.6
−4
−0.8

−6 −1
−5 0 5 0 2 4 6 8 10 12
−3
x 10
y [adim] Time [adim]

Fig. 9 Projections on the coordinate plane of a halo orbit (the first three graphs) with α4 = 0.05 around the
collinear libration point L 1 in the Sun–Earth ERTBP, computed by series expansions up to order (9, 0, 12),
and the Euclidean norm of the position deviation (bottom-right graph) between the halo orbits obtained by
series expansions constructed in this paper and series expansions, up to order (9, 12), discussed in Hou and
Liu (2011a), during 4π units of dimensionless time

three graphs of Figs. 10 and 11 show the projections on the coordinate plane of the positive
branches of the unstable manifolds associated with Lissajous and halo orbits, respectively.
A trajectory specified by φ1 = 1.6π and φ2 = 0 in the manifolds is represented by a red
solid line. At the same time, the manifolds with the same amplitudes are computed by the
algebraic manipulator developed according to Masdemont (2005). For the trajectory specified
by φ1 = 1.6π and φ2 = 0, the Euclidean norm of the position deviation between the analytical
results obtained by the two analytical approaches can be seen in Fig. 10 (bottom-right graph)
corresponding to the Lissajous case and Fig. 11 (bottom-right graph) corresponding to the
halo case. The time histories of the deviation indicate that the reduced case (corresponding to
e = 0) of the series expansions constructed in this paper is equivalent to the series expansions
constructed in Masdemont (2005).
Figures 12 (left panel) and 13 (left panel) present a Lissajous orbit with α3 = 0.05 and
α4 = 0.05, and a halo orbit with α4 = 0.05 obtained by series expansions truncated at order
(0, 0, 15). Figures 12 (right panel) and 13 (right panel) show the Euclidean norm of position
deviation between the orbit computed by series expansions constructed in this paper and the
one computed by the algebraic manipulator developed according to Jorba and Masdemont
(1999), and indicate that the series expansions presented in this paper can be further reduced
to the ones discussed in Jorba and Masdemont (1999).
We utilize the method of numerical integration with the initial states provided by series
expansions to investigate the domain of convergence of the high-order solutions constructed

123
378 H. Lei et al.

−3 −4
x 10 x 10
1 4

3
0.5
2
0
1
y [adim]

z [adim]
−0.5 0

−1
−1
L
1
−2
−1.5
−3

−2 −4
0.9895 0.99 0.9905 0.991 0.9915 0.992 0.9895 0.99 0.9905 0.991 0.9915 0.992

x [adim] x [adim]
−8
−4 x 10
x 10 1
4
0.8
3
0.6
2 0.4
Deviation
1 0.2
z [adim]

0 0

−0.2
−1
−0.4
−2
−0.6
−3 −0.8

−4 −1
−2 −1.5 −1 −0.5 0 0.5 1 0 0.5 1 1.5 2 2.5 3 3.5
−3
x 10
y [adim] Time [adim]

Fig. 10 Projections on the coordinate plane of the analytical solutions corresponding to the positive branch
of the unstable manifold associated with Lissajous orbit around L 1 point in the Sun–Earth CRTBP (the first
three graphs), and the Euclidean norm of the position deviation (bottom-right graph) for a particular trajectory
represented by the red solid line obtained by two analytical approaches, including the series expansions up to
order (0, 7, 9) constructed in this paper and series expansions up to order (7, 9) discussed in Masdemont (2005),
during 3.8 units of dimensionless time. The amplitudes of the invariant manifold are α1 = 1 × 10−5 , α2 =
0, α3 = 0.03 and α4 = 0.03

in this paper. The numerical tool used for integrating the equations of motion is a Runge–
Kutta Fehlberg seventh-order integrator with eighth-order automatic step-size control, called
RKF78 for short (Fehlberg 1968), and the tolerance of error is taken as 10−14 . For Lissajous
case, the trajectory specified by φ1 = 0 and φ2 = 0 in the positive branch of the unstable
manifold corresponding to α1 = 1 × 10−5 and α2 = 0 is selected to test the domain of
convergence of series expansions truncated at different orders in terms of the pair of in-
plane and out-of-plane amplitudes (α3 , α4 ). In the following computations, the maximum
values of α3 and α4 are taken as 0.1 and 0.3, respectively, and 100 × 100 mesh points of
(α3 , α4 ) distributed in [0, 0.1] × [0, 0.3] are evaluated. Other choices can be carried out in
a similar way. For given parameters (e, α1 , α2 , α3 , α4 , φ1 , φ2 ), the initial condition can be
computed by the series expansions constructed. Then, the initial condition is numerically
integrated by integrator RKF78, and the numerical integration is terminated at time π, which
is about the period of libration point orbit. Next, the state of the numerical orbit at time π
is compared against the state obtained by series expansions, and the Euclidean norm of the
position deviation between numerical and analytical orbits at time π is computed. Finally,
the base 10 logarithm of the Euclidean norm of the position deviation is considered as the
index of accuracy, that is, the index n means the deviation 10n . Figure 14 shows the domain
of convergence of the high-order analytical solutions truncated at different orders, such as

123
High-order solutions of invariant manifolds in ERTBP 379

−3 −4
x 10 x 10
5 6

4
4
3

2
2
1
y [adim]

z [adim]
L1
0 0

−1
−2
−2

−3
−4
−4

−5 −6
0.988 0.989 0.99 0.991 0.992 0.993 0.994 0.995 0.988 0.989 0.99 0.991 0.992 0.993 0.994 0.995

x [adim] x [adim]
−8
−4 x 10
x 10 1
6
0.8

4 0.6

0.4
2
Deviation
0.2
z [adim]

0 0

−0.2
−2 −0.4

−0.6
−4
−0.8

−6 −1
−5 0 5 0 0.5 1 1.5 2 2.5 3 3.5 4
−3
x 10
y [adim] Time [adim]

Fig. 11 Projections on the coordinate plane of the analytical solutions corresponding to the positive branch
of the unstable manifold (the first three graphs) associated with halo orbit around L 1 in the Sun–Earth circular
restricted three-body problem, and for a particular trajectory represented by the red solid line, the Euclidean
norm of the position deviation (bottom-right graph) between the orbits obtained by the series expansions
truncated at order (0, 9, 12), constructed in this paper, and the series expansions up to order (9, 12) discussed
in Masdemont (2005). The amplitudes of the manifold are α1 = 1 × 10−5 , α2 = 0 and α4 = 0.05, then
α3 = 0.140156724783764 such that Δ(e = 0, α1 , α2 = 0, α3 , α4 ) = 0

−8
x 10
1

0.8
−3
x 10 e = α1 = α2 = 0
0.6
1 α3 = α4 = 0.05
0.4
0.5
Deviation
z [adim]

0.2
0 0

−0.5 −0.2

−0.4
−1
2 −0.6
1 0.991
L1
−3 0 0.9905 −0.8
x 10
−1 0.99
−1
−2 0.9895 0 20 40 60 80 100 120
y [adim] x [adim]
Time [adim]

Fig. 12 A Lissajous orbit with α3 = 0.05 and α4 = 0.05 (left panel) around the collinear libration point L 1
in the Sun–Earth CRTBP computed by series expansions truncated at order (0, 0, 15), and the Euclidean norm
of the position deviation (right panel) between the Lissajous orbits obtained by series expansions constructed
in this paper and series expansions up to order 15 discussed in Jorba and Masdemont (1999) during 40π units
of dimensionless time

123
380 H. Lei et al.

−8
x 10
1

0.8
−4
x 10
5 e=α =α =0 0.6
1 2
α = 0.140156813432901 0.4
3
α = 0.05

Deviation
0 4
z [adim]

0.2
L1
0
−5
−0.2

−0.4
−10
5 −0.6
0.993
0.992 −0.8
−3 0 0.991
x 10 0.99
0.989 −1
−5 0.988 0 2 4 6 8 10 12
y [adim] x [adim]
Time [adim]

Fig. 13 A halo orbit (left panel) around L 1 in the Sun–Earth CRTBP obtained by series expansions truncated
at order (0, 0, 15), and the Euclidean norm of the position deviation (right panel) between halo orbits computed
by series expansions constructed in this paper and series expansions truncated at order 15 discussed in Jorba and
Masdemont (1999) during 4π units of dimensionless time. The out-of-plane amplitude is taken as α4 = 0.05
and the in-plane amplitude α3 is computed according to Δ(e = 0, α1 = 0, α2 = 0, α3 , α4 ) = 0

(4, 4, 5), (4, 3, 6), (5, 3, 5), (4, 2, 7). The results indicate that, with the same threshold, the
region of convergence shown in Fig. 14 (bottom-right graph) is the largest since its order
corresponding to the center manifolds is the largest.
For high-order solutions of invariant manifold associated with halo orbit, a similar test is
carried out. The trajectory specified by φ = 0 in the positive branch of unstable manifold
corresponding to α1 = 1 × 10−5 and α2 = 0 is chosen to test the domain of convergence.
Due to Δ(e, α1 , α2 , α3 , α4 ) = 0, amplitudes α3 and α4 are no longer independent. In our
computation, α4 is taken as an independent variable, and the maximum value of α4 is 0.25.
A hundred values of α4 are taken in [0, 0.25] uniformly. The initial condition corresponding to
(e, α1 , α2 , α3 , α4 , φ) can be computed by the series expansions directly. Then, the equations
of motion are integrated until the time reaches π units of dimensionless time. The Euclidean
norm of the position deviation between analytical and numerical orbits at time π can be
obtained. The domain of convergence of series expansions up to order (5, 5, 9) and (5, 3,
12) corresponding to invariant manifolds associated with halo orbit in ERTBP is considered.
Figures 15 (left panel) and 16 (left panel) show the relationship between α3 and α4 , and
indicate that the in-plane amplitude α3 has a minimum non-zero value. Figures 15 (right
panel) and 16 (right panel) present the deviation for position as a function of out-of-plane
amplitude α4 . The solid lines on the right panels of Figs. 15 and 16 denote the tolerance
1 × 10−5 . The maximum allowed values of α4 for series expansions truncated at order (5, 5,
9) and (5, 3, 12) are 0.137 and 0.160 adim in terms of the threshold 1×10−5 , respectively. The
domain of convergence corresponding to other orders can be explored in a similar manner.

6 Conclusions and discussion

The high-order solutions of invariant manifolds, associated with Lissajous and halo orbits,
around the collinear libration points in the elliptic restricted three-body problem, are con-
structed in this paper. The general solutions of the invariant manifolds are expanded as
power series of five parameters, including the orbital eccentricity of the secondary e, two
amplitudes corresponding to the hyperbolic manifolds α1 (unstable) and α2 (stable), and two
amplitudes corresponding to the center manifolds α3 (in-plane) and α4 (out-of-plane). For the

123
High-order solutions of invariant manifolds in ERTBP 381

−3 −4 −3

−5
0.2

−5

−3
−4
−4

−5
−4 0.25 −4
−4

−6
−3
0.15 −5 −5
0.2
α4 (adim)

α (adim)
−5 −5

−4
−6 −6

−5 −6
−6

0.1 0.15 −6
−7 −7 −7

4
−7

−5
−6

−4
−8
0.1
−6

−8 −8
0.05 −
8 0.05 −8
−9 −9 −7 −9
−6 −5
−7 −5
0 −10 0 −7 −10
0 0.02 0.04 0.06 0 0.05 0.1
α3 (adim) α (adim)
3
−3 −3
−5

0.2 −3

−6
−5
−4

−4
−4 −4 −4
0.25

−3
−4

−6
−4
−3

0.15 −5 −5

−5
0.2 −5
α4 (adim)

−5 α (adim) −5
−4

−6 −6

−7
−5 −6

−6
−6

−6
0.15

−4
0.1
−7 −7
4
−7

−6

−5
−7
0.1

−3
−6

−8 −7 −8
0.05
−8
−9 0.05 −9
−4
−7 −5 −7 −6 −5
0 −10 0 −10
0 0.02 0.04 0.06 0 0.05 0.1
α3 (adim) α3 (adim)

Fig. 14 Practical convergence of the high-order analytical solutions of invariant manifolds, truncated at
different orders, associated with Lissajous orbits, around L 1 point in the Sun–Earth elliptic restricted three-
body system. The index shown in the color bar represents the base 10 logarithm of the Euclidean norm of the
position deviation between the analytical orbit and numerically integrated orbit at π units of dimensionless
time. In order to display, the deviation is taken as 1 × 10−10 when it is less than 1 × 10−10 , and the deviation
is taken as 1 × 10−3 when it is greater than 1 × 10−3 . Left panel: the analytical solutions are truncated at
order (4, 4, 5) (top graph) and (5, 3, 5) (bottom graph). Right panel: the analytical solutions are truncated at
order (4, 3, 6) (top graph) and (4, 2, 7) (bottom graph)

−5
x 10
0.25 1.8
Truncated at order (5, 5, 9) Truncated at order (5, 5, 9)
1.6
0.2
1.4
Error [adim]

1.2
α4 [adim]

0.15

0.1
0.8

0.6
0.05
0.4

0 0.2
0.135 0.14 0.145 0.15 0.155 0.16 0.165 0 0.05 0.1 0.15 0.2 0.25

α [adim] α [adim]
3 4

Fig. 15 The relationship between α3 and α4 (left panel), and the Euclidean norm of the deviation for position
between analytical orbit and corresponding numerically integrated orbit at π units of dimensionless time (right
panel). The analytical solution is constructed up to order (5, 5, 9). The red solid line denotes the threshold
1 × 10−5

123
382 H. Lei et al.

−4
x 10
0.25 1
Truncated at order (5, 3, 12) Truncated at order (5, 3, 12)
0.9

0.2 0.8

0.7

Error [adim]
α [adim]

0.15 0.6

0.5
4

0.1 0.4

0.3

0.05 0.2

0.1

0 0
0.135 0.14 0.145 0.15 0.155 0.16 0.165 0 0.05 0.1 0.15 0.2 0.25

α3 [adim] α4 [adim]

Fig. 16 The relationship between α3 and α4 (left panel), and the Euclidean norm of the deviation for position
between analytical orbit and corresponding numerically integrated orbit at π units of dimensionless time (right
panel). The analytical solution is constructed up to order (5, 3, 12). The red solid line also denotes the threshold
1 × 10−5

analytical solutions of invariant manifolds associated with halo orbits, these five parameters
are not independent and should satisfy Δ(e, α1 , α2 , α3 , α4 ) = 0. The advantage of the series
expansions proposed and constructed in this paper lies in that it can provide a parametric
description to the general dynamics around the collinear libration points of ERTBP. That is,
the general solutions with e  = 0 can describe the following dynamics around the collinear
libration points in ERTBP: (i) the stable manifolds, corresponding to αi  = 0, i = 2, 3, 4,
and α1 = 0; (ii) the unstable manifolds, corresponding to αi  = 0, i = 1, 3, 4, and α2 = 0;
(iii) the transit trajectories, corresponding to αi  = 0, i = 1, . . . , 4, and α1 · α2 < 0; (iv)
the non-transit trajectories, corresponding to αi  = 0, i = 1, . . . , 4, and α1 · α2 > 0; (v) the
Lissajous and halo orbits, corresponding to α1 = α2 = 0, α3  = 0, and α4  = 0; and (vi)
the plane and vertical Lyapunov orbits (the special case of Lissajous orbits). Furthermore,
the center and hyperbolic manifolds around the collinear libration points in CRTBP which
is the special case of ERTBP can also be described by the general solutions with e = 0.
For the reduced cases, such as the center manifolds in ERTBP, the hyperbolic and center
manifolds in CRTBP, the algebraic manipulator for computing the analytical solutions are
developed following the description presented in Hou and Liu (2011a), Masdemont (2005)
and Jorba and Masdemont (1999) in order to compare with the series expansions constructed
in this paper. The results indicate that the series expansions constructed in this paper can
be exactly reduced to the series expansions discussed in the existing literature. In order to
check the validity of the series expansions constructed, the initial states obtained by series
expansions are numerically integrated by RKF78, and the analytical results are compared
against the numerical results to investigate the domain of convergence of series expansions
up to different orders.
Due to the large storage of the coefficients of coordinate and frequency, the series expan-
sions of the invariant manifolds associated with collinear libration point orbits can be con-
structed up to a limited order. The applications of analytical solutions of invariant manifolds
in ERTBP will be discussed in our future work.

Acknowledgments This work was carried out with financial support from the National Basic Research
Program 973 of China (2013CB834103), the National High Technology Research and Development Program
863 of China (2012AA121602), the National Natural Science Foundation of China (Grant No. 11078001) and
the Research and Innovation Project for College Graduates of Jiangsu Province (Grant No. CXZZ13_0042).

123
High-order solutions of invariant manifolds in ERTBP 383

The authors are much obliged to the reviewers for their many insightful comments that substantially improved
the quality of this paper.

References

Alessi, E.M., Gómez, G., Masdemont, J.J.: Leaving the Moon by means of invariant manifolds of libration
point orbits. Commun. Nonlinear Sci. Numer. Simul. 14, 4153–4167 (2009)
Alessi, E.M., Gómez, G., Masdemont, J.J.: Two-maneuvres transfers between LEOs and Lissajous orbits in
the Earth-Moon system. Adv. Space Res. 45, 1276–1291 (2010)
Belbruno, E.: Capture Dynamics and Chaotic Motions in Celestial Mechanics: With Applications to the
Construction of Low Energy Transfers. Princeton University Press, Princeton (2004)
Canalias, E., Gómez, G., Marcote, M., et al.: Assessment of mission design including utilization of libration
points and weak stability boundaries. Technical report 18142/04/NL/MV (2004)
Chung, M.J.: Trans-Lunar Cruise trajectory design of GRAIL (Gravity Recovery and Interior Laboratory) mis-
sion. In: AIAA/AAS Astrodynamical Specialist Conference, Toroto, Ontario Canada, 2–5 August (2010)
Dellnitz, M., Junge, O., Post, M., et al.: On target for Venus-set oriented computation of energy efficient low
thrust trajectories. Celest. Mech. Dyn. Astron. 95, 357–370 (2006)
Farquhar, R.W.: The control and use of libration point satellites. Technical report TR R346, Stanford University
report SUDAAR-350 (1968)
Farquhar, R.W.: The utilization of halo orbits in advanced lunar operations. Technical report NASA TN D-6365
(1971)
Farquhar, R.W., Kamel, A.A.: Quasi-periodic orbits about the translunar libration point. Celest. Mech. 7,
458–473 (1973)
Fehlberg, E.: Classical fifth-, sixth-, seventh-, and eighth-order Runge-Kutta formulas with stepsize control.
Technical report NASA TR R-287 (1968)
Folta, D.C., Woodard, M., Howell, K.C., et al.: Applications of multi-body dynamical environments: the
ARTEMIS transfer trajectory design. Acta Astronaut. 73, 237–249 (2012)
Gómez, G., Marcote, M.: High-order analytical solutions of Hill’s equations. Celest. Mech. Dyn. Astron. 94,
197–211 (2006)
Hou, X.Y., Liu, L.: On quasi-periodic motions around the triangular libration points of the real Earth-Moon
system. Celest. Mech. Dyn. Astron. 108, 301–313 (2010)
Hou, X.Y., Liu, L.: On motions around the collinear libration points in the elliptic restricted three-body problem.
Mon. Not. R. Astron. Soc. 415, 3552–3560 (2011a)
Hou, X.Y., Liu, L.: On quasi-periodic motions around the collinear libration points in the real Earth-Moon
system. Celest. Mech. Dyn. Astron. 110, 71–98 (2011b)
Howell, K.C., Kakoi, M.: Transfers between the Earth-Moon and Sun-Earth systems using manifolds and
transit orbits. Acta Astronaut. 59, 367–380 (2006)
Jorba, À., Masdemont, J.: Dynamics in the center manifold of the collinear points of the restricted three body
problem. Physica D 132, 189–213 (1999)
Koon, W.S., Lo, M.W., Marsden, J.E., et al.: Low energy transfer to the Moon. Celest. Mech. Dyn. Astron.
81, 63–73 (2001)
Lei, H.L., Xu, B., Sun, Y.S.: Earth-Moon low energy trajectory optimization in the real system. Adv. Space
Res. 51, 917–929 (2013)
Lei, H.L., Xu, B.: High-order analytical solutions around triangular libration points in circular restricted
three-body problem. Mon. Not. R. Astron. Soc. 434, 1376–1386 (2013a)
Lei, H.L., Xu, B.: Low thrust transfer to the libration point orbits of Sun Mars system from the Earth. J.
Astronaut. 34, 763–772 (2013b)
Masdemont, J.J.: High-order expansions of invariant manifolds of libration point orbits with application to
mission design. Dyn. Syst. 20, 59–113 (2005)
Mingotti, G., Topputo, F., Bernelli-Zazzera, F.: Low-energy, low-thrust transfers to the Moon. Celest. Mech.
Dyn. Astron. 105, 61–74 (2009)
Mingotti, G., Topputo, F., Bernelli-Zazzera, F.: Earth-Mars transfers with ballistic escape and low-thrust
capture. Celest. Mech. Dyn. Astron. 110, 169–188 (2011)
Pergola, P., Geurts, K., Casaregola, C., et al.: Earth-Mars halo to halo low thrust manifold transfers. Celest.
Mech. Dyn. Astron. 105, 19–32 (2009)
Ren, Y., Masdemont, J.J., Marcote, M., et al.: Computation of analytical solutions of the relative motion about
a Keplerian elliptic orbit. Acta Astronaut. 81, 186–199 (2012)

123
384 H. Lei et al.

Richardson, D.L.: Analytic construction of periodic orbits about the collinear points. Celest. Mech. 22,
241–253 (1980)
Sweetser, T.H., Broschart, S.B., Angelopoulos, V., et al.: Artemis mission design. Space Sci. Rev. 165, 27–57
(2011)
Szebehely, V.: Theory of Orbits. Academic Press, New York (1967)
Xia, Z.H.: Arnold diffusion in the elliptic restricted three-body problem. J. Dyn. Differ. Equ. 5, 219–240
(1993)

123

You might also like