Why do Earth satellites stay up?

Scott Tremaine and Tomer D. Yavetz

Citation: American Journal of Physics 82, 769 (2014); doi: 10.1119/1.4874853

View online: http://dx.doi.org/10.1119/1.4874853
View Table of Contents: http://scitation.aip.org/content/aapt/journal/ajp/82/8?ver=pdfcov
Published by the American Association of Physics Teachers

Articles you may be interested in

The study of gravity gradient effect on attitude of low earth orbit satellite
AIP Conf. Proc. 1522, 636 (2013); 10.1063/1.4801184

EARTH SATELLITES
Phys. Teach. 51, 202 (2013); 10.1119/1.4795356

Why no shear in “Div, grad, curl, and all that”?

Am. J. Phys. 80, 519 (2012); 10.1119/1.3688678

Do not forget to teach about the scientific method

Am. J. Phys. 74, 93 (2006); 10.1119/1.2162550

Simulating Realistic Satellite Orbits in the Undergraduate Classroom

Phys. Teach. 43, 452 (2005); 10.1119/1.2060645

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 Why do Earth satellites stay up?
Scott Tremainea)
Institute for Advanced Study, Princeton, New Jersey 08540
Tomer D. Yavetzb)
Department of Astrophysical Sciences, Peyton Hall, Princeton University, Princeton, New Jersey 08540
(Received 27 September 2013; accepted 22 April 2014)
Satellites in low Earth orbits must accurately conserve their orbital eccentricity, since a decrease in
perigee of only 5–10% would cause them to crash. However, these satellites are subject to
gravitational perturbations from Earth’s multipole moments, the Moon, and the Sun that are not
spherically symmetric and hence do not conserve angular momentum, especially over the tens of
thousands of orbits made by a typical satellite. Why then do satellites not crash? We describe a
vector-based analysis of the long-term behavior of satellite orbits and apply this to several toy
systems containing a single non-Keplerian perturbing potential. If only the quadrupole (or J2)
potential from the Earth’s equatorial bulge is present, all near-circular orbits are stable. If only the
octupole (or J3) potential is present, all such orbits crash. If only the lunar or solar potential is
present, all near-circular orbits with inclinations to the ecliptic exceeding 39 are unstable. We
describe the behavior of satellites in the simultaneous presence of all of these perturbations and
show that almost all low Earth orbits are stable because of an accidental property of the dominant
quadrupole potential. We also relate these results to the phenomenon of Lidov–Kozai oscillations.
C 2014 American Association of Physics Teachers.
V
[http://dx.doi.org/10.1119/1.4874853]

I. INTRODUCTION
The study of the motion of a satellite around its host planet
is one of the oldest problems in dynamics. This subject was
re-invigorated at the dawn of the Space Age in the late 1950s
with a new focus on artificial satellites in low Earth orbits.1,2
Satellites orbiting a spherically symmetric Earth would
have Keplerian orbits, described by their semi-major axis a,
eccentricity e, and other orbital elements. These orbits con-
serve energy E and angular momentum L per unit mass,
which for a test particle are given by
GM丣
E¼
2a
and
jLj ¼ ½GM丣 að1  e2 Þ1=2 ;
where G is the gravitational constant and M丣 is the mass of
the Earth. The perigee of the satellite orbit is að1  eÞ, so if
energy and angular momentum are conserved the perigee is
conserved as well. This is fortunate because the typical peri-
gee of a satellite in low Earth orbit is only 800 km above
the surface of Earth, much smaller than Earth’s radius of
R丣 ¼ 6 378 km. Orbits with perigees less than 200–300 km
are short-lived because of atmospheric drag, so a decrease in
perigee of less than 10% would cause the satellite orbit to
decay quickly (we call this a “crash”).
However, Earth is not spherically symmetric. A better
approximation is an oblate spheroid, a potential in which
energy is conserved but total angular momentum is not. In
this case, if the polar axis of Earth defines the z-axis, then
only the z-component of angular momentum
Lz ¼ ½GM丣 að1  e2 Þ1=2 cos b
is conserved. Here, b is the inclination of the orbit relative to
Earth’s equator. In this case, the perigee is not necessarily
conserved unless the orbit lies in the plane of Earth’s equa-
tor. Why then do satellites in low Earth orbit not crash? A
similar question arises if we treat Earth as spherical and con-
sider the gravitational tides from the Moon and Sun, which
are not spherically symmetric and also do not conserve angu-
lar momentum.
Of course, the answer to why satellites do not crash is
known, in the sense that aerospace engineers design orbits
that are stable over the expected lifetime of the satellite.
Nevertheless, the calculations in textbooks on astrodynamics
and celestial mechanics do not provide a physical explana-
tion for this stability. In asking our colleagues, we have
(1) received a variety of answers, such as “there are only peri-
odic oscillations in the perigee and these are too small to be
important,” or “the orbits are unstable but on timescales
much longer than the satellite lifetime.” We shall show that
(2) neither of these answers captures the most relevant physics.
The primary goal of this paper is to understand physically
the nature of the strongest long-term perturbations to satellite
orbits caused by each non-spherical component that contrib-
utes to the gravitational field around a planet. We will focus
on understanding how these components affect the evolution
of an initially circular orbit. In addition to this primary goal,
we have two secondary goals. First, we derive our results
using a geometric formalism that is simpler and more power-
ful for this purpose than the usual algebraic methods used in
celestial mechanics. Second, we show how our results are
related to Lidov–Kozai oscillations, a remarkable phenom-
enon in orbital mechanics that is relevant to a wide variety of
astrophysical systems.
II. SECULAR DYNAMICS OF EARTH SATELLITES
A. The geometry of Keplerian orbits
(3)
We use a vectorial approach to calculate the behavior of
the satellite orbit. First, we define a coordinate system whose
equatorial plane coincides with the satellite’s orbital plane.

769 Am. J. Phys. 82 (8), August 2014 http://aapt.org/ajp C 2014 American Association of Physics Teachers
V 769

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 The orthogonal unit vectors of this system are n ^ in the direc- We assume the satellite has unit mass so there is no distinc-
tion of the satellite’s angular-momentum vector, u ^ pointing tion between momentum and velocity.
towards the perigee, and v^ ¼ n ^u ^ . We introduce polar
coordinates ðr; /Þ in the orbital (uv-) plane, with / ¼ 0 coin- B. The multipole expansion of Earth’s gravitational
ciding with the u-axis. The orbit of a test particle with semi- potential
major axis a and eccentricity e is given by
Earth’s gravitational field can be represented by a multi-
2
að1  e Þ pole expansion, of which the first several thousand terms
r¼ ; (4) have been measured.5 Since our goal is physical understand-
1 þ e cos /
ing, for simplicity we shall restrict ourselves to the axisym-
with metric components of this field [the largest non-
axisymmetric components are comparable to the J3 and J4
r ¼ rðcos / u
^ þ sin / v^Þ: (5) axisymmetric components in Eq. (11)]. In this case, the
potential can be written as
The orbital period P ¼ 2pa3=2 =ðGM丣 Þ1=2 and the angular   n 
GM丣 X1
R丣
momentum per unit mass L ¼ r2 d/=dt ¼ ½ðGM丣 aÞ U丣 ðr; hÞ ¼  1 Jn Pn ðcos hÞ : (11)
ð1  e2 Þ1=2 . Using these results, the average of some func- r n¼2
r
tion UðrÞ over one orbit is
ðP ð Here r and h are spherical polar coordinates relative to
1 ðGM丣 Þ1=2 2p dt Earth’s spin axis n ^ 丣 , R丣 is Earth’s radius, Jn is a multipole
hUðrÞi  dt U½rðtÞ ¼ d/ Uðr; /Þ
P 0 2pa 3=2
0 d/ moment, and Pn is a Legendre polynomial. The term with
ð n ¼ 1 is zero if the origin of the coordinates coincides with
ðGM丣 Þ1=2 2p
¼ d/ r2 Uðr; /Þ Earth’s center of mass, as we shall assume. The values of
2pa3=2 L 0 these parameters, and of the first few multipole moments, are
ð given in Table I. The first non-zero moment, J2, is several
ð1  e2 Þ3=2 2p d/
¼ 2
Uðr; /Þ: (6) orders of magnitude larger than the others because of Earth’s
2p 0 ð1 þ e cos /Þ equatorial bulge caused by its spin.
We shall restrict our attention to the first three non-zero
This method of averaging can only be applied in some special moments, J2, J3, and J4, since these are sufficient to illustrate
cases, one of which is the Kepler problem. In the typical inte- how a non-spherical potential affects satellite orbits. The
grable Hamiltonian problem, one can define actions and methods we describe are straightforward to extend to higher
angles such that the actions are constant and the angles circu- multipoles.
late at incommensurable frequencies (except in sub-manifolds Because we are interested in small effects that accumulate
of phase space where the frequencies are resonant). Away over many orbits, we can average the gravitational potential
from such resonances, the averaging is done over all angles, over the Keplerian orbit of a satellite with semi-major axis a
the averaged perturbation is independent of the angles, and and eccentricity e. We write the potential associated with Jn
the actions remain constant even in the presence of the pertur- in Eq. (11) as Un ; then, for example,
bation. However, in the Kepler Hamiltonian, the entire phase
space is resonant in the sense that two of the angles have zero GM丣 J2 R2丣
or resonant frequencies (neither the orbital plane nor the line U2 ðr; hÞ ¼ P2 ðcos hÞ
of apsides precesses). Thus, the averaging is done only over r3  
the one remaining angle, the mean anomaly.3,4 GM丣 J2 R2丣 ðr  n ^ 丣 Þ2 1
¼ 3  3 : (12)
The unit vectors n ^ and u^ are undefined for radial orbits 2 r5 r
(e ¼ 1) and circular orbits (e ¼ 0), respectively. Therefore,
we introduce
Table I. Constants for the Earth’s gravitational potential and the solar and
j  ð1  e2 Þ1=2 n
^ (7) lunar tides.5,6

and Constant Value

e ¼ e^
u; (8) Earth mass GM丣 3:9860  1014 m3 s2
Earth radius R丣 6:3781  106 m
which are well-defined for all bound orbits. Apart from a fac- Multipole moments J2 þ1:0826  103
tor ðGM丣 aÞ1=2 the first of these is the angular-momentum J3 2:5327  106
vector per unit mass, and the second is the eccentricity or J4 1:6196  106
Runge-Lenz vector. In terms of the position and momentum
J5 2:2730  107
vectors,
J6 þ5:4068  107
1 J7 3:5236  107
j¼ rp (9)
ðGM丣 aÞ1=2 J8 2:0480  107
Solar mass GM 1:327  1020 m3 s2
and Earth-Sun semi-major axis a 1:496  1011 m
Lunar mass GMffl 4:903  1012 m3 s2
1 r
e¼ p  ðr  pÞ  : (10) Earth-Moon semi-major axis affl 3:844  108 m
GM丣 r

770 Am. J. Phys., Vol. 82, No. 8, August 2014 S. Tremaine and T. D. Yavetz 770

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 Using Eqs. (4) and (5), the orbit average of U2 is The gravitational potential from the Moon is Uffl ðr; rffl Þ
   ¼ GMffl =jr  rffl j. Since jrj
jrffl j (by a factor of about
GM丣 J2 R2丣 2
2 cos / 60), we expand this potential as a Taylor series to get
hU2 i ¼ un
3ð^ ^丣Þ þ 6ð^ ^丣Þ
un
2 r3 " #
  GMffl r  rffl r2 3ðr  rffl Þ2
cos / sin / Uffl ðr; rffl Þ ¼  1þ 2  2 þ ;
 ð^vn^丣Þ rffl rffl 2r ffl 2rffl4
r3
 2   
2 sin / 1 (21)
þ 3ð^vn^丣Þ 3
 ; (13)
r r3
plus terms of order ðr=rffl Þ3 and higher, which we drop. The
first term is independent of the dynamical variable r and so
and from Eq. (6) we have
can also be dropped. The second term is canceled by the fic-
   2  titious potential due to the acceleration of the Earth’s center
cos2 / sin / 1
3
¼ 3
¼ ; (14) of mass by the Moon. Then averaging over the satellite orbit
r r 2a3 ð1  e2 Þ3=2 using Eqs. (4) and (5), we have
  
cos / sin / GMffl 3ð^u  r ffl Þ2 2 2
¼ 0; (15) hUffl i ¼ 3 hr2 i  hr cos /i
r3 2r ffl r2ffl
and u  rffl Þð^
6ð^ v  rffl Þ 2
 2
hr cos / sin /i
  rffl
1 1 
¼ : (16) v  r ffl Þ2 2 2
3ð^
r3 a3 ð1  e2 Þ3=2  hr sin /i : (22)
rffl2
Finally, since ð^ ^ Þ is an orthonormal triad,
u ; v^; n
2 2 Again, using Eq. (6) gives
ð^
un^ 丣 Þ þ ð^vn^ 丣 Þ þ ðn ^n ^ 丣 Þ2 ¼ 1, and this relation can
be used to eliminate ð^ un ^ 丣 Þ2 and ð^vn^ 丣 Þ2 , giving a2
hr 2 cos2 /i ¼ ð1 þ 4e2 Þ; (23)
GM丣 J2 R2丣 h i 2
2
hU2 i ¼ 1  3ð ^
n 丣  ^
n Þ : (17)
4a3 ð1  e2 Þ3=2 hr 2 cos / sin /i ¼ 0; (24)
a2
In terms of the vectors j and e in Eqs. (7) and (8), we get hr 2 sin2 /i ¼ ð1  e2 Þ; (25)
2
GM丣 J2 R2丣 h 2 2
i
hU2 i ¼ 1  e  3ðj  ^
n 丣 Þ : (18) and
4a3 ð1  e2 Þ5=2
a2
A similar set of calculations leads to hr 2 i ¼ ð2 þ 3e2 Þ: (26)
2
3GM丣 J3 R3丣 h i
2 2 Finally, we eliminate ð^ v  rffl Þ2 using the relation rffl2 ¼
hU3 i ¼ ðe  ^
n 丣 Þ 1  e  5ðj  ^
n 丣 Þ 2 2
^  rffl Þ2 and replace u
8a4 ð1  e2 Þ 7=2 ð^
u  rffl Þ þ ð^
v  r ffl Þ þ ðn ^ with j
^ and n
and e using Eqs. (7) and (8) to obtain
(19)
 
GMffl a2 2 3ðj  r ffl Þ2 15ðe  rffl Þ2
and hUffl i ¼  1 þ 6e þ  :
4rffl3 rffl2 r2ffl
3GM丣 J4 R4丣 n
hU4 i ¼ ð6  e2 Þð1  e2 Þ2 (27)
128a5 ð1  e2 Þ11=2
We now carry out a second averaging, this time over the
^ 丣 Þ2
 10ð6 þ e2 Þð1  e2 Þðj  n lunar orbit rather than the satellite orbit. For simplicity we
assume that the Moon’s orbit is circular, with semi-major
^ 丣 Þ4 þ 20ðe  n
þ 35ð2 þ e2 Þðj  n ^ 丣 Þ2 ð1  e2 Þ axis affl and a fixed normal n ^ ffl . Averaging any fixed vector c
h i over the Moon’s orbit, we have hðc  r ffl Þ2 i ¼ 12 a2ffl
2
^
 1  e  7ðj  n 丣 Þ 2
: (20) ^ ffl Þ2 , and thus
½c2  ðc  n
GMffl a2
2 2 2 hhUffl ii ¼ ^ ffl Þ2  3ðj  n
½15ðe  n ^ ffl Þ2 þ 1  6e2  :
Since j þ e ¼ 1, any terms involving e can be replaced by 8a3ffl
1  j2 .
(28)
C. The gravitational potential from the Moon
D. Rotating reference frame
The effects on satellite orbits of the tides from the Sun
and the Moon are qualitatively similar, and since the lunar In some cases, it will be useful to work in a reference
tide is stronger by a factor of about 2.2 we consider only frame that rotates about Earth’s spin axis with angular speed
the Moon. x. In this frame, the Lagrangian for a particle of unit mass in

771 Am. J. Phys., Vol. 82, No. 8, August 2014 S. Tremaine and T. D. Yavetz 771

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 a potential UðrÞ is L ¼ 12 ðr_ þ x n ^ 丣  rÞ2  UðrÞ. The ca- Since the Kepler Hamiltonian HKep is independent of e and j,
nonical momentum is p ¼ @L=@ r_ ¼ r_ þ x n ^ 丣  r, and the its gradients vanish so we can replace H ¼ HKep þ hUi by
Hamiltonian is H ¼ p  r_  L ¼ 12 p2 þ UðrÞ  x p  n ^丣  r hUi in these equations.
¼ 12 p2 þ UðrÞ  x n^ 丣  r  p ¼ 12 p2 þ UðrÞ  xðGM丣 aÞ1=2 j By definition, j and e must satisfy
^ 丣 ; where the last equality follows from Eqs. (9) and (10).
n
Thus, transforming to a rotating frame simply adds the term je¼0 (37)

Urot ¼ xðGM丣 aÞ1=2 j  n

^丣 (29) and

to the Hamiltonian. j2 þ e2 ¼ 1: (38)

E. Equations of motion It is straightforward to show that these constraints are con-

Having derived the orbit-averaged perturbing potentials
from the Earth’s multipole moments and the lunar tidal
field, we must now determine how the satellite orbit
responds to these perturbations. Once again we use a vecto-
rial method.
The evolution of the orbit is determined by a Hamiltonian
H ¼ HKep þ UðrÞ, where HKep ¼ v2 =2  GM丣 =r is the
Kepler Hamiltonian. If UðrÞ ¼ 0 the energy EKep ¼ GM丣 =
2a is conserved, where as usual a is the semi-major axis. If the
perturbing potential is non-zero, dEKep =dt ¼ p  $U. If we
orbit-average, as we did for the potentials, we find that
hp  $Ui ¼ 0, since the orbit is closed. Thus, EKep and the
semi-major axis are conserved. Then the orbit’s shape and orien-
tation are determined entirely by the vectors j and e [Eqs. (7)
and (8)].
From Eqs. (9) and (10), the Poisson brackets of j and e are
served by the equations of motion in Eqs. (35) and (36).
There is a gauge freedom in the Hamiltonian H since it can
be replaced by H þ F, where Fðj; eÞ is any function that is
constant on the manifold (37). Adding such a function, how-
ever, has no effect on the equations of motion.8
Note that if M  e þ j and N  e  j then jMj ¼ jNj ¼ 1
for all time; thus the geometry of the averaged equations of
motion is that of the direct product of two 2-spheres.3,4
III. STABILITY OF SATELLITES ON CIRCULAR
ORBITS
We now investigate the effect of each of the perturbations
we have discussed, the multipole moments J2, J3, and J4 and
the lunar tide, on the evolution of a satellite in a circular low
Earth orbit. We have verified most of these conclusions by
numerical integrations of test-particle orbits.

1 A. Quadrupole (J2 ) potential

fji ; jj g ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ijk jk ; (30)
GM丣 a
If the only perturbation is the quadrupole term of Eq. (18),
1 we have
fei ; ej g ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ijk jk ; (31)
GM丣 a
3GM丣 J2 R2丣
$j hU2 i ¼  ^ 丣 Þn
ðj  n ^丣 (39)
and 2a3 ð1  e2 Þ5=2
1 and
fji ; ej g ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ijk ek ; (32)
GM丣 a
3GM丣 J2 R2丣 h i
where ijk is the antisymmetric or Levi–Civita tensor. The $e hU2 i ¼ ^ 丣 Þ2 e:
1  e2  5ðj  n (40)
4a3 ð1  e2 Þ7=2
time evolution under the Hamiltonian H of any function f of
the phase-space coordinates is Substituting these results into Eqs. (35) and (36) and drop-
df ping all terms higher than linear order in eccentricity gives
¼ f f ; H g: (33)
dt 1=2
dj 3ðGM丣 Þ J2 R2丣
¼ ^丣Þ j  n
ðj  n ^丣 (41)
Then by the chain rule, we have dt 2a7=2
df and
¼ f f ; jg$j H þ f f ; eg$e H; (34)
dt 1=2
de 3ðGM丣 Þ J2 R2丣
where $j is the vector ð@=@j1 ; @=@j2 ; @=@j3 Þ, and similarly ¼ 7=2
dt nh 4a i o
for $e . Replacing f by ji and ei and using Eqs. (30) and (31)
we have7  ^ 丣 Þ2 e  j þ 2ðj  n
1  5ðj  n ^丣 :
^丣Þ e  n
(42)
dj 1
¼  pffiffiffiffiffiffiffiffiffiffiffiffiffi ðj  $j H þ e  $e H Þ; (35)
dt GM丣 a The first of these equations describes precession of the satel-
lite’s orbital angular momentum j around the symmetry axis
and of the Earth n^ 丣 . The angular frequency of the precession is

de 1
¼  pffiffiffiffiffiffiffiffiffiffiffiffiffi ðj  $e H þ e  $j H Þ: (36) 3ðGM丣 Þ1=2 J2 R2丣
dt GM丣 a x¼ ^ 丣 Þ;
ðj  n (43)
2a7=2

772 Am. J. Phys., Vol. 82, No. 8, August 2014 S. Tremaine and T. D. Yavetz 772

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 where j  n^ 丣 ¼ cos b and b is the constant inclination of the and
orbit to Earth’s equator. The minus sign indicates retrograde
precession of the angular momentum vector around the sym- de 3ðGM丣 Þ1=2 J3 R3丣 h 2
i
metry axis. ¼ 1  5ðj  ^
n 丣 Þ ^丣:
jn (50)
dt 8a9=2
The second equation is best analyzed by transforming to
the frame rotating with the precession of the orbit. In this Thus, the eccentricity vector grows linearly with time:
frame, j is a constant and the Hamiltonian is modified eðtÞ ¼ e0 þ ct, where
according to Eq. (29). The equation of motion for the eccen-
tricity is modified to 3ðGM丣 Þ1=2 jJ3 jR3丣

jcj ¼ 9=2
sin b 1  5 cos2 b : (51)
de x
 8a
¼ 5cos2 b  1 e  j: (44)
dt 2cos b The behavior of an orbit in the octupole potential is different
from the other potentials considered here. In the other poten-
We now switch to Cartesian coordinates with the positive tials, the sub-manifold in phase space e ¼ 0 is invariant, and
z-axis along n^ 丣 and with j in the xz-plane, so j ¼ ðsin b; we investigate the stability of perturbations around this solu-
0; cos bÞ, and look for a solution of the form e ¼ e0 expðktÞ. tion. However, Eqs. (49) and (50) imply that in the octupole
Solving this eigenvalue problem yields either k ¼ 0 or case e ¼ 0 is no longer invariant. That is, circular orbits do
not even exist in the averaged dynamics.
x
k ¼ 6i ð1  5cos2 bÞ We conclude that the J3 term leads to eccentricity growth
2 cos b in initially circular orbits of all inclinations except b ¼ 0
(equatorial orbit) and b ¼ bcrit . Hence, if only the J3 multi-
3ðGM丣 Þ1=2 J2 R2丣 pole were non-zero and there were no other external forces,
¼ 7i ð1  5 cos2 bÞ: (45) Earth satellites on all orbits except these two would crash.
4a7=2
The survival timescale would be much smaller than
The eigenvalue k ¼ 0 is unphysical because the correspond-  9=2
ing eigenvector does not satisfy the constraints of Eq. (37). 26:9 yr a
Since the eigenvalues are zero or purely imaginary, we jcj1 ¼ : (52)
sin bj1  5 cos2 bj R丣
conclude that nearly circular low Earth orbits of all inclina-
tions are stable under the influence of the quadrupole
moment of the Earth. Note that if the inclination
pffiffiffiffiffiffiffiffi b satisfies
1  5cos2 b ¼ 0, or b ¼ bcrit ¼ cos 1 1=5 ¼ 63:43 , all C. J4 potential
three eigenvalues k are zero in Eq. (45). This is the so-called
The analysis of orbital evolution for higher multipoles
critical inclination, and the motion of satellites in the vicin-
proceeds in the same way as for J2 and J3, although the alge-
ity of this inclination has been the subject of many studies.9
bra rapidly becomes more complicated. For J4, we substitute
In physical units, the timescale of these oscillations is
the averaged potential hU4 i [Eq. (20)] into Eqs. (35) and
 7=2 (49). Keeping terms up to first order in eccentricity on the
11:5 d a
jkj1 ¼ 2
: (46) right-hand side we find
j1  5 cos bj R丣
1=2 h i
dj 15ðGM丣 Þ J4 R4丣 3
¼ 3ðj  ^
n 丣 Þ  7ðj  ^
n 丣 Þ ^丣
jn
dt 16a11=2
B. Octupole (J3 ) potential
(53)
If J3 is the only non-zero multipole moment then Eq. (20)
gives and
1=2
15GM丣 J3 R3丣 de 15ðGM丣 Þ J4 R4丣
$j hU3 i ¼  ^ 丣 Þðe  n
ðj  n ^ 丣 Þn
^丣 (47) ¼ f ½3ðj  n ^ 丣 Þ3 e
^ 丣 Þ  7ðj  n
4a4 ð1  e2 Þ 7=2 dt 16a11=2
^ 丣  ½1  7ðj  n
n ^ 丣 Þ2 ðe  n
^丣Þ j  n^丣
and
^ 丣 Þ2 þ 21ðj  n
 ½1  14ðj  n ^ 丣 Þ4 j  eg: (54)
 h i
3GM丣 J3 R3丣 2 2
$e hU3 i ¼ ^
n 丣 1  e  5ðj  ^
n 丣 Þ The first of these equations describes precession of the sat-
8a4 ð1  e2 Þ7=2 ellite’s orbital angular momentum with angular frequency
h i
e
15ðGM丣 Þ1=2 J4 R4丣 h i
2 2
þ5 ðe  ^
n 丣 Þ 1  e  7ðj  ^
n 丣 Þ :
1  e2 x¼ 3ðj  ^
n 丣 Þ  7ðj  ^
n 丣 Þ 3

(48) 16a11=2
15ðGM丣 Þ1=2 J4 R4丣

As before, we substitute these results into Eqs. (35) and (36) ¼ 11=2
3 cos b  7cos3 b ; (55)
16a
and drop all terms linear or higher in the eccentricity from
the right-hand side to get where the inclination b is a constant. Equation (54) is best
analyzed by transforming to the frame rotating with the pre-
dj cession of the orbit. In this frame j is a constant, so the
¼0 (49)
dt Hamiltonian is modified according to Eq. (29). Thus, the

773 Am. J. Phys., Vol. 82, No. 8, August 2014 S. Tremaine and T. D. Yavetz 773

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 equation of motion for the eccentricity of Eqs. (35) and (36) 3ðGM丣 Þ1=2 a3=2 Mffl
is modified to k¼6 ð3  5 cos2 bffl Þ1=2 : (63)
23=2 a3ffl M丣
de 15ðGM丣 Þ1=2 J4 R4丣

¼ 11=2
1  7cos2 b ðe  n
^丣Þ j pffiffiffiffiffiffiffiisffi stable if and only if the inclination bffl < bL
The solution
dt 16a

  cos 1 3=5 ¼ 39:23 (Ref. 10) or bffl > 180  bL . In
n^ 丣 þ 1  14cos2 b þ 21cos 4 b j  eg: (56) other words, all nearly circular orbits with an inclination
between 39:23 and 140:77 with respect to the lunar orbit
We assume e ¼ e0 expðktÞ and solve the resulting eigenvalue will have an exponentially increasing eccentricity. The
equation for k and find that either k ¼ 0 or growth rate is
 3=2
15ðGM丣 Þ1=2 J4 R4丣 1 248 yr R丣
k ¼ 6i ð3 þ 6 cos 2b þ 7 cos 4bÞ1=2 jkj ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (64)
64a11=2 5 2
a
1  cos bffl
 ð15 þ 28 cos 2b þ 21 cos 4bÞ1=2 : (57) 3

For nearly all choices of b, the value of k is purely imagi- E. Power-law quadrupole potential
nary, indicating a stable orbit. However, for 34:40
< b < 40:09 , and for 71:10 < b < 73:42 , the orbits are It is striking that both the J2 potential and the lunar poten-
unstable. Taking the inclination 37:04 that maximizes the tial are quadrupoles with angular dependence / P2 ðcos hÞ,
growth rate, the characteristic timescale for this growth is yet circular orbits are stable in one of these perturbing poten-
tials at all inclinations and unstable in the other for a wide
 11=2 range of inclinations. Some insight into this behavior comes
a
jkj1 ¼ 26:8 yr : (58) from examining an artificial axisymmetric perturbing poten-
R丣 tial Uc ¼ cr b P2 ðcos hÞ with symmetry axis n ^ c . In these cases,
the J2 potential corresponds to b ¼ 3 and the lunar poten-
For the second unstable regime (71:10 < b < 73:42 ), tial to b ¼ þ2. After orbit-averaging and dropping all terms
the maximum growth rate occurs at b ¼ 72:26 and the char- of order higher than e2, the potential becomes
acteristic growth time is larger by a factor of 4.08.
cab
D. The lunar potential hUc i ¼ ^ c Þ2 þ 8
f6ð2 þ bÞð3 þ bÞðe  n
32
We substitute the doubly averaged potential hhUffl ii from  ð18 þ 13b þ b2 Þe2
Eq. (28) into Eqs. (35) and (36) to obtain ^ c Þ2 g:
 3½8 þ ð2  3b þ b2 Þe2 ðj  n (65)
dj 3G1=2 M ffl a3=2 We substitute this potential into Eqs. (35) and (36) to find
¼ 1=2
^fflÞ j  n
½ðj  n ^ ffl  5ðe  n
^fflÞ e  n
^ffl
dt 4M a3 that e ¼ e0 expðktÞ with
丣 ffl
(59)
cab
k¼6 pffiffiffiffiffiffiffiffiffiffiffiffiffi  f½18 þ 13b þ b2
and 16 GM丣 a
þ 3ð2  3b þ b2 Þcos 2 b
de 3G1=2 Mffl a3=2
¼ 1=2
^ffl Þ e  n
½ðj  n ^ffl  ½18 þ 17b þ 5b2  3ð14 þ 7b þ 3b2 Þcos2 bg1=2 ;
dt 4M a3 丣 ffl
(66)
^fflÞ j  n
 5ðe  n ^ffl þ 2 j  e: (60)
where cos b ¼ n ^n^ c . The satellite orbit is unstable if the
To first order in the eccentricity the first of these describes quantity in braces is positive.
uniform precession of j around n ^ ffl with frequency Figure 1 shows the unstable inclinations (stippled) as a
function of the inclination b and the exponent b of the poten-
3ðGM丣 Þ1=2 a3=2 Mffl tial. Remarkably, the only exponents for which all inclina-
x¼ ^ ffl Þ;
ðj  n (61)
4a3ffl M丣 tions are stable are b ¼ 3, corresponding to the potential
arising from the planetary quadrupole J2, and b ¼ 2, which
where j  n^ ffl ¼ cos bffl and bffl is the inclination of the satel- is unphysical. We conclude that the stability of orbits of all
lite orbit to the lunar orbit, which is nearly in the ecliptic inclinations circling an oblate planet is an accidental conse-
plane. We then transform to the frame rotating at this fre- quence of the properties of the quadrupole potential.
quency and the equation of motion for the eccentricity sim-
plifies to F. General axisymmetric potential
Now consider the more general case of an arbitrary axi-
de 3ðGM丣 Þ1=2 a3=2 Mffl
¼ ^fflÞ j n
½2 j  e  5ðe  n ^ffl : symmetric perturbation UðrÞ to the Kepler potential. Let ^z
dt 4a3ffl M丣 be the symmetry axis of the potential. The orbit-averaged
(62) potential hUi can only depend on the semi-major axis a and
the dimensionless eccentricity and angular-momentum vec-
Assuming e ¼ e0 expðktÞ and solving the eigenvalue equa- tors e and j. Because the semi-major axis is fixed in the
tion, we find k ¼ 0 or orbit-averaged equations of motion (Sec. III E) we do not

774 Am. J. Phys., Vol. 82, No. 8, August 2014 S. Tremaine and T. D. Yavetz 774

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 2hUi;3
x ¼ pffiffiffiffiffiffiffiffiffiffi ðj  ^z Þ: (71)
GMa

In the frame rotating at x, dj=dt ¼ 0 and the Hamiltonian is

modified according to Eq. (29). Thus, the equation of motion
for the eccentricity becomes
de 1
¼  pffiffiffiffiffiffiffiffiffiffi ½2hUi;1 j  e þ hUi;2 j  ^z
dt GMa
þ hUi;22 ðe  ^z Þ j  ^z : (72)

All quantities other than e on the right side are

P constants, so
the general solution is of the form e ¼ e0 þ 2i¼1 ci expðki tÞ,
where11
  1=2
2i 1 2
Fig. 1. Stability of nearly circular orbits in a potential of the form k ¼ 6 pffiffiffiffiffiffiffiffiffiffi hUi;1 hUi;1 þ hUi;22 sin b ;
r b P2 ðcoshÞ. The horizontal axis is the exponent b and the vertical axis is the GMa 2
orbit inclination in degrees, relative to the symmetry plane of the potential. (73)
Unstable regions of parameter space are stippled, and bounded by solid
lines. Vacuum solutions of Poisson’s equations require b ¼ 3 or 2 and are and
marked by vertical dashed lines.
2hUi;1 j  e0 þ hUi;22 ðe0  ^z Þ j  ^z ¼ hUi;2 j  ^z ; (74)
need to consider the dependence of hUi on it. We shall
assume that the potential UðrÞ is a smooth function of posi-
tion r, so hUi must be a smooth function of e and j. with cos b ¼ j  ^z for circular orbits.
Moreover, because of axisymmetry, the dependence on these With this formula several of our earlier results become
vectors can only occur through the scalars e2, j2, e  ^z , and straightforward to interpret as follows:
j^z , and since j2 þ e2 ¼ 1 we need only one of the first two (I) The potential hU2 i½e2 ; e  z^; ðj  ^z Þ2  associated with
in this list. Finally, since the shape of the orbit is unchanged Earth’s quadrupole moment [Eq. (18)] has no depend-
if j ! j the potential can only depend on ðj  ^z Þ2 . Thus, the ence on e  n ^ 丣 , so hU2 i;22 ¼ 0 and k is always pure
orbit-averaged potential can be written hUi½e2 ; e  ^z ; ðj  ^z Þ2 . imaginary, which explains why orbits in this perturb-
Denoting the derivative with respect to argument i by U;i , ing potential are always stable.
Eqs. (35) and (36) become (II) The potential hU3 i½e2 ; e  ^z ; ðj  ^z Þ2  associated with the
dj 1   multipole moment J3 is an odd function of e  ^z [Eq.
¼  pffiffiffiffiffiffiffiffiffiffi hUi;2 e  ^z þ 2hUi;3 ðj  ^z Þ j  ^z (67) (20)]. Thus, hU3 i;1 is also odd, so the first and third
dt GMa terms of Eq. (72) are zero at e ¼ 0 and the second is
non-zero. This implies that the eccentricity vector grows
and linearly with time in the octupole potential or any poten-
de 1 tial that is an odd function of e  ^z , so circular orbits do
¼  pffiffiffiffiffiffiffiffiffiffi ½2hUi;1 j  e þ hUi;2 j  ^z not exist in the averaged dynamics of such potentials.
dt GMa (III) Note also that hU3 i is linear in e  z [Eq. (20)], so
þ 2hUi;3 ðj  ^z Þ e  ^z Þ: (68) hU3 i;22 ¼ 0. Thus, for the combined potential pffiffiffiffiffiffiffiffiffiffi
hU2 þ U3 i, Eq. (73) is simply k ¼ 62ihU2 i;1 = GMa,
The first of these shows that j  ^z is constant, i.e., the z-com- independent of U3 . We conclude that any quadrupole
ponent of angular momentum is conserved. potential, no matter how small, will stabilize a circu-
Now assume jej
1 and keep only terms on the right side lar satellite orbit against the octupole potential U3 .
that are independent of e in the first equation and up to linear (IV) The quadrupole potential associated with J2 is the
in e in the second to get strongest non-Keplerian potential for satellites in low
Earth orbit by several orders of magnitude. Let us then
dj 2 write the perturbing potential as hUi ¼ hU2 i þ h/i
¼  pffiffiffiffiffiffiffiffiffiffi hUi;3 ðj  ^z Þ j  ^z (69) where 
1 and /ðrÞ represents the potentials from
dt GMa
higher order multipoles, the Moon and Sun, etc. Since
and hU2 i;22 ¼ 0, we have to first order in ,
de 1 
¼  pffiffiffiffiffiffiffiffiffiffi ½2hUi;1 j  e þ hUi;2 j  ^z 2i
dt GMa k ¼ 6 pffiffiffiffiffiffiffiffiffiffi hU2 i2;1 þ hU2 i;1
GMa
þ hUi;22 ðe  ^z Þ j  ^z þ 2hUi;3 ðj  ^z Þ e  ^z ; (70)  1=2
1 2
 2h/i;1 þ h/i;22 sin b : (75)
where the derivatives of hUi are evaluated at e ¼ 0. Since j  ^z 2
is constant, these derivatives can be taken to be constants. The
first equation describes uniform precession of j around ^z at At most inclinations hU2 i2;1 is much larger than the other
angular frequency term since the latter is multiplied by 
1. Thus, k is pure

775 Am. J. Phys., Vol. 82, No. 8, August 2014 S. Tremaine and T. D. Yavetz 775

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
 imaginary and the orbit is stable. Physically, the rapid pre- orbital distribution is that no satellites are found with incli-
cession of the angular-momentum and eccentricity vectors nations between 55 and 130 (relative to the ecliptic),
averages out the perturbations from other sources and sup- although prograde orbits with smaller inclination and retro-
presses the instabilities that they would otherwise induce. grade orbits with larger inclination are common (20%
However, hU2 i2;1 is proportional to ð1  5cos2 bÞ so near the and 80% of the total population, respectively). This gap
critical inclination bcrit there can be a narrow range of incli- is explained naturally by Lidov-Kozai oscillations: at incli-
nations in which the square brackets in Eq. (75) are negative nations close to 90 the oscillations are so strong that the
so the orbit is unstable. satellite either collides with one of the much larger Jovian
satellites or the planet at periapsis, or escapes from the
planet’s gravitational field at apapsis.17,18
IV. LIDOV-KOZAI OSCILLATIONS • Some extrasolar planets have remarkably high eccentric-
ities. The current record-holder has e ¼ 0.97, and it is likely
The nonlinear trajectories of the linear instabilities we
that some or most of the highest eccentricities have been
have described are known as Kozai, Kozai-Lidov, or Lidov-
excited by Lidov-Kozai oscillations due to a companion
Kozai oscillations. Although Laplace had all of the tools
star.19–21 In fact, the four planets with the largest known
needed to investigate this phenomenon, it was only discov-
eccentricities all orbit host stars with companions.22
ered in the early 1960s by Lidov10 in the Soviet Union and • “Hot Jupiters” are giant planets orbiting within 0.1 AU of
brought to the West by Kozai.12 The simplest and most astro-
their host star, several hundred times closer than the giant
physically relevant examples of Lidov-Kozai oscillations
planets in our own solar system. Such planets cannot form
arise when a distant third body perturbs a binary system, as
in situ. One plausible formation mechanism is “high-
in the discussion of the lunar potential in Sec. III D. The per-
eccentricity migration,” which involves the following
turbing potential hUffl i [Eq. (28)] depends on the orbital ele-
steps. First, the planet forms at 5–10 AU from the host
ments of the satellite through the semi-major axis a, the
star, like Jupiter and Saturn in our own solar system.
eccentricity e, and the projections of the eccentricity and
Second, it is excited to high eccentricity by gravitational
angular-momentum vectors on the lunar orbit axis, e  n ^ffl
scattering off another planet. Third, Lidov-Kozai oscilla-
and j  n^ ffl . The semi-major axis is conserved because we
tions due to a companion star or other giant planets peri-
have orbit-averaged (see discussion in Sec. II E), j  n ^ ffl is
odically bring the planet so close to the host star that it
conserved because the orbit-averaged potential is axisym-
loses orbital energy through tidal friction, and lastly the
metric, and the Hamiltonian hUffl i is conserved because it is
orbit then decays, at a faster and faster rate, until the
autonomous. Given these four variables and three conserved
planet settles into a circular orbit close to the host star.23,24
quantities, the orbit-averaged oscillation has only one degree • Many stars are found in close binary systems, with separations
of freedom and thus is integrable,12,13 although the integra-
of only a few stellar radii. Forming such systems is a chal-
bility disappears when higher-order multipole moments are
lenge because the radius of a star is now much smaller than it
included or the angular momentum in the inner orbit is not
used to be in the past. It is possible that most or all close bi-
small compared to the outer orbit.14–16
nary systems were formed from much wider binaries by the
Some properties of the oscillations are straightforward to
combined effects of Lidov-Kozai oscillations induced by a
determine. Since hUffl i and j  n ^ ffl are conserved, Eq. (28) tells
distant companion star and tidal friction.24 Supporting this hy-
us that the eccentricity e and the normal component of the ec-
pothesis is the remarkable observation that almost all close bi-
centricity e  n ^ ffl must evolve along the track
^ ffl Þ2  2e2 ¼ constant. If the satellite is initially on a nary systems (orbital period less than 3 days) have a tertiary
5ðe  n
^ ffl Þ2 ¼ 2e2 =5. companion star.25 The formation rate of Type-Ia supernovae
circular orbit, the constant must be zero so ðe  n
2 2 2 may be dominated by a similar process in triple systems con-
Since e  j ¼ 0, ðe  n^ ffl Þ =e cannot exceed sin b, where b is
taining a white dwarf-white dwarf binary and a distant com-
the inclination, so sin2 b  2=5 if e > 0, which immediately
panion. Either Lidov-Kozai oscillations plus energy loss
gives the stability criterion for circular orbits b < bL [Eq.
through gravitational radiation26 or Lidov-Kozai oscillations
(63)]. If the inclination b0 of the initial circular orbit exceeds
^ ffl Þ2 ¼ ð1  e2 Þcos2 b ¼ cos2 b0 and we find that excite the binary to sufficiently high eccentricity lead to a
bL , then ðj  n
situation where the two white dwarfs collide.27
e  1  ð5=3Þcos2 b0 , which gives the maximum eccentric-
2
• Most galaxies contain supermassive black holes at the
ity achieved in the Lidov-Kozai oscillation. At the maximum
centers, and when galaxies merge their black holes spiral
eccentricity, the inclination is bL .
towards the center of the merged galaxy through dynami-
As an example, Lidov pointed out that if the inclination of
cal friction.28 However, dynamical friction becomes less
the lunar orbit to the ecliptic were changed to 90 , with all
effective once the black holes have formed a tightly bound
its other orbital elements kept the same, this hypothetical
binary and it is unclear whether the black hole inspiral
“vertical Moon” would collide with the Earth in about four
will “stall” before gravitational radiation becomes effec-
years as a result of a Lidov-Kozai oscillation induced by the
tive. Lidov-Kozai oscillations in the binary, induced either
gravitational field of the Sun. More recent work shows that
by the overall gravitational field of the galaxy or the pres-
Lidov-Kozai oscillations may play a significant role in the
ence of a third black hole, can pump the binary to high ec-
formation and evolution of a wide variety of astrophysical
centricity where gravitational radiation is more efficient,
systems. These include the following.
leading to inspiral and merger of the black holes.29
• The giant planets in the solar system are surrounded by
over 100 small “irregular” satellites, most of them discov-
ered in the last two decades. These satellites, typical- V. DISCUSSION
ly < 100 km in radius, are found at much larger semi-major
axes and have more eccentric and inclined orbits than the Satellites in low Earth orbit are subject to a variety of non-
classical satellites. One of the striking features of their Keplerian perturbations, both from the multipole moments of

776 Am. J. Phys., Vol. 82, No. 8, August 2014 S. Tremaine and T. D. Yavetz 776

This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25

Earth’s gravitational field and from tidal forces from the
Moon and Sun. For practical purposes in astrodynamics, the
effects of these perturbations have been well-understood for
decades. Nevertheless, in generic perturbing potentials (even
axisymmetric ones), many orbits are likely to crash in a short
time. A classic example is Lidov’s vertical lunar orbit,
described in Sec. IV. Thus, it is worthwhile to understand
physically what features the perturbing potential must have
so that low Earth orbits are stable.
Our most general result, for an axisymmetric perturbing
potential, is that the orbit is stable if [Eq. (73)]
! of the critical inclination,” Astron. J. 65, pp. 624–627 (1960); A. G.
@hUi @hUi 1 @ 2 hUi
þ sin2 b > 0; (76)
@e2 @e2 2 @ðe  zÞ2
where z is the symmetry axis of the potential and hUi is the
orbit-averaged potential written as a function of e2, e  ^z , and
z Þ2 ¼ cos2 b with b the inclination. It is an accidental
ðj  ^
property of the potential due to an internal quadrupole
moment that it is independent of e  ^z so the motion is always
stable. This statement is not meant to imply that all other
potentials are unstable. The J4 potential analyzed in Sec. III C

is unstable only over a total range of inclinations of about 8 ,
and Fig. 1 shows that other potentials can be stable for most
inclinations. But the potential associated with J2 is the only
axisymmetric potential arising from a physically plausible
mass distribution for which all inclinations are stable.
Because the other perturbing potentials are much smaller
than the one due to the Earth’s quadrupole, all low Earth
orbits are stable except perhaps inp a ffiffiffiffiffiffiffi
smallffi region near the
critical inclination bcrit ¼ cos 1 1=5 ¼ 63:43 , where
@hUi=@e2 ¼ 0 for the quadrupole potential.
We chose to investigate only azimuthally symmetric terms
in the perturbing potential, but we strongly suspect that simi-
lar arguments apply to small non-axisymmetric potentials.
The strong quadrupole potential of the Earth guarantees sta-
bility except for small interval(s) in the inclination b. These
intervals will be centered on resonances11,30,31 between the
precession frequencies of the angular-momentum vector
and the eccentricity vector [Eqs. (43) and (45)]; that is, when
1  5cos2 b ¼ 2ncos b for integer n (b ¼ 46:38 ; 63:43 ;
73:14 ; 78:46 ; …).
Finally, we hope that this paper will introduce readers to
the use of vector elements to study the long-term evolution
of Keplerian orbits. These elements are not new,7,8,32 but
they deserve to be more widely known.
ACKNOWLEDGMENTS
The authors thank Boaz Katz and Renu Malhotra for com-
ments and insights.
a)
Electronic mail: tremaine@ias.edu
b)
Electronic mail: tyavetz@princeton.edu
1
D. G. King-Hele, “The effect of the Earth’s oblateness on the orbit of a
near satellite,” Proc. R. Soc. A 247, 49–72 (1958).
2
P. E. El’yasberg, Introduction to the Theory of Flight of Artificial Earth
Satellites. Moscow: Izdatel’stvo Nauka Glavnaya Redaktsiya. Translated
by the Israel Program for Scientific Translations (1965).
3
J. Moser, “Regularization of Kepler’s problem and the averaging method
on a manifold,” Commun. Pure Appl. Math. 23, 609–636 (1970).
4
M. Kummer, “On the three-dimensional lunar problem and other perturba-
tion problems of the Kepler problem,” J. Math. Anal. Appl. 93, 142–194
(1983).
777 Am. J. Phys., Vol. 82, No. 8, August 2014
This article is copyrighted as indicated in the article. Reuse of AAPT content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
131.151.244.7 On: Wed, 20 Aug 2014 21:22:25
5
N. K. Pavlis, S. A. Holmes, S. C. Kenyon, and J. K. Factor, “The develop-
ment and evaluation of the Earth Gravitational Model 2008 (EGM2008),”
J. Geophys. Res.-Sol. Ea. 117, B04406 (2012).
6
B. Luzum, N. Capitaine, A. Fienga, W. Folkner, T. Fukushima, J. Hilton,
C. Hohenkerk, G. Krasinsky, G. Petit, E. Pitjeva, M. Soffel, and P.
Wallace, “The IAU 2009 system of astronomical constants: The report of
the IAU working group on numerical standards for Fundamental
Astronomy,” Celest. Mech. Dyn. Astron. 110, 293–304 (2011).
7
M. Milankovich, “Uber € die verwendung vektorieller bahnelemente in der
st€
orungsrechnung,” Bull. Serb. Acad. Math. Nat. A 6, 1 (1939).
8
S. Tremaine, J. Touma, and F. Namouni, “Satellite dynamics on the
Laplace surface,” Astron. J. 137, 3706–3717 (2009).
9
See, for example, B. Garfinkel, “On the motion of a satellite in the vicinity
Lubowe, “How critical is the critical inclination?,” Celestial Mech. 1,
6–10 (1969).
10
M. L. Lidov, “The approximate analysis of the evolution of the orbits of
artificial satellites,” in Problems of the Motion of Artificial Celestial
Bodies (Academy of Sciences, Moscow, 1961), pp. 119–134.
11
B. Katz and S. Dong, “Exponential growth of eccentricity in secular the-
ory,” e-print arXiv:1105.3953 (2011).
12
Y. Kozai, “Secular perturbations of asteroids with high inclination and
eccentricity,” Astron. J. 67, 591–598 (1962).
13
H. Kinoshita and H. Nakai, “General solution of the Kozai mechanism,”
Celest. Mech. Dyn. Astron. 98, 67–74 (2007).
14
E. B. Ford, B. Kozinsky, and F. A. Rasio, “Secular evolution of hierarchi-
cal triple star systems,” Astrophys. J. 535, 385–401 (2000).
15
B. Katz, S. Dong, and R. Malhotra, “Long-term cycling of Kozai-Lidov
cycles: Extreme eccentricities and inclinations excited by a distant eccen-
tric perturber,” Phys. Rev. Lett. 107, 181101 (2011).
16
S. Naoz, W. M. Farr, Y. Lithwick, F. A. Rasio, and J. Teyssandier,
“Secular dynamics in hierarchical three-body systems,” Mon. Not. Roy.
17
Astron. Soc. 431, 2155–2171 (2013).
V. Carruba, J. A. Burns, P. D. Nicholson, and B. J. Gladman, “On the in-
clination distribution of the Jovian irregular satellites,” Icarus 158,
434–449 (2002).
18
D. Nesvorn y, J. L. A. Alvarellos, L. Dones, and H. F. Levison, “Orbital
and collisional evolution of the irregular satellites,” Astron. J. 126,
398–429 (2003).
19
M. Holman, J. Touma, and S. Tremaine, “Chaotic variations in the eccen-
tricity of the planet orbiting 16 Cygni B,” Nature 386, 254–256 (1997).
20
K. A. Innanen, J. Q. Zheng, S. Mikkola, and M. J. Valtonen, “The Kozai
mechanism and the stability of planetary orbits in binary star systems,”
Astron. J. 113, 1915–1919 (1997).
21
T. Mazeh, Y. Krymolowski, and G. Rosenfeld, “The high eccentricity of
the planet orbiting 16 Cygni B,” Astrophys. J. 477, L103 (1997).
22
J. T. Wright, O. Fakhouri, G. W. Marcy, E. Han, Y. Feng, J. A. Johnson,
A. W. Howard, D. A. Fischer, J. A. Valenti, J. Anderson, and N. Piskunov,
“The exoplanet orbit database,” Publ. Astron. Soc. Pacific 123, 412–422
(2011).
23
Y. Wu and N. Murray, “Planet migration and binary companions: the case
of HD 80606b,” Astrophys. J. 589, 605–614 (2003).
24
D. Fabrycky and S. Tremaine, “Shrinking binary and planetary orbits by
Kozai cycles with tidal friction,” Astrophys. J. 669, 1298–1315 (2007).
25
A. Tokovinin, S. Thomas, M. Sterzik, and S. Udry, “Tertiary companions
to close spectroscopic binaries,” Astronon. Astrophys. 450, 681–693
(2006).
26
T. A. Thompson, “Accelerating compact object mergers in triple systems
with the Kozai resonance: a mechanism for “prompt” Type Ia supernovae,
gamma-ray bursts, and other exotica,” Astrophys. J. 741, 82 (2011).
27
D. Kushnir, B. Katz, S. Dong, E. Livne, and R. Fernandez, “Head-on colli-
sions of white dwarfs in triple systems could explain Type Ia supernova,”
Astrophys. J. 778, L37 (2013).
28
M. C. Begelman, R. D. Blandford, and M. J. Rees, “Massive black hole
binaries in active galactic nuclei,” Nature 287, 307–309 (1980).
29
O. Blaes, M. H. Lee, and A. Socrates, “The Kozai mechanism and the evolu-
tion of binary supermassive black holes,” Astrophys. J. 578, 775–786 (2002).
30
M. A. Vashkov’yak, “Stability of circular satellite orbits for combined
action of perturbations from an external body and from the noncentrality
of the planetary gravitational field,” Cosm. Res. 12, 757 (1974).
31
J. F. Palacian, “Dynamics of a satellite orbiting a planet with an inhomoge-
neous gravitational field,” Celest. Mech. Dyn. Astron. 98, 219–249 (2007).
32
P. Musen, “Special perturbations of the vectorial elements,” Astron. J. 59,
262–267 (1954).
S. Tremaine and T. D. Yavetz 777