A&A 432, 1101–1113 (2005)

DOI: 10.1051/0004-6361:20041312
Astronomy


&
c ESO 2005 Astrophysics

Free polar motion of a triaxial and elastic body in Hamiltonian

formalism: Application to the Earth and Mars
M. Folgueira1,2 and J. Souchay2

1
Instituto de Astronomía y Geodesia (UCM – CSIC), Facultad de Ciencias Matemáticas, Universidad Complutense de Madrid,
Ciudad Universitaria, 28040 Madrid, Spain
e-mail: [email protected]
2
Observatoire de Paris, SYRTE, UMR 8630 du CNRS, 61 avenue de l’Observatoire, 75014 Paris, France
e-mail: [email protected]

Received 18 May 2004 / Accepted 25 October 2004

Abstract. The purpose of this paper is to show how to solve in Hamiltonian formalism the equations of the polar motion
of any arbitrarily shaped elastic celestial body, i.e. the motion of its rotation axis (or angular momentum) with respect to its
figure axis. With this aim, we deduce from canonical equations related to the rotational Hamiltonian of the body, the analytical
solution for its free polar motion which depends both on the elasticity and on its moments of inertia. In particular, we study
the influence of the phase angle δ, responsible for the dissipation, on the damping of the polar motion. In order to validate
our analytical equations, we show that, to first order, they are in complete agreement with those obtained from the classical
Liouville equations.
Then we adapt our calculations to the real data obtained from the polar motion of the Earth (polhody). For that purpose,
we characterize precisely the differences in radius J − χ and in angle l − θ between the polar coordinates (χ, θ) and (J, l)
representing respectively the motion of the axis of rotation of the Earth and the motion of its angular momentum axis, with
respect to an Earth-fixed reference frame, after showing the influence of the choice of the origin on these coordinates, and on the
determination of the Chandler period as well. Then we show that the phase lag δ responsible for the damping for the selected
time interval, between Feb. 1982 and Apr. 1990, might be of the order of δ ≈ 6◦ , according to a numerical integration starting
from our analytical equations. Moreover, we emphasize the presence in our calculations for both χ and θ, of an oscillation with
a period T Chandler /2, due to the triaxial shape of our planet, and generally not taken into account.
In a last step, we apply our analytical formulation to the polar motion of Mars, thus showing the high dependence of its damping
on the poorly known value of its Love number k. Moreover we emphasize the large oscillations of Mars’ polar motion due to
the triaxiality of this planet.

Key words. methods: analytical – solar system: general

1. Introduction rotational motion, taking into particular account the elastic de-
formation caused by the centrifugal force due to rotation. We
The torque-free rotational motion of a given body, sometimes first show the process for the formulation of such equations
called the Euler-Poinsot problem, corresponds to the case for of motion expressed in terms of Andoyer’s variables and its
which no torque is exerted by any external body. The appli- canonically conjugate variables.
cation of such motion to a rigid Earth is a classical prob- The Hamiltonian formulation for the rotation of an elas-
lem already discussed by several authors. Usually it has been tic body such as the Earth has been discussed in detail by, to
treated from the analytical point of view by solving either the best of our knowledge, Sevilla & Romero (1985), Getino
Euler’s equations of motion or the Hamiltonian equations of the & Ferrándiz (1990, 1991), Kubo (1991) and Stavinschi &
torque-free motion for an axially symmetric Earth – see for in- Souchay (1994). In this study, we have followed Kubo’s pro-
stance Kinoshita (1972a,b, 1992). In a previous paper (Souchay cedure and we have amplified some aspects of his presentation,
et al. 2002), we have applied a numerical integration method considering a triaxial body instead of a body with rotational
based on the Runge-Kutta-Fehlberg algorithm, to compare the symmetry. Moreover, we have shown the equivalence between
rotation of several typical celestial bodies with different shapes, our Hamiltonian formulation, and the Liouville equations ob-
such as the Earth, Mars and Eros, considered as rigid bodies. tained in a quite different manner.
In this paper we consider an arbitrary elastic and triaxial During the thirteen years that have passed since Kubo
body, and we analyze the analytical solution of its torque-free (1991), the level of accuracy in the observational techniques,

Article published by EDP Sciences and available at http://www.aanda.org or http://dx.doi.org/10.1051/0004-6361:20041312

1102 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism

such VLBI (Very Long Baseline Interferometry), GPS (Global Table 1. Notation.
Positioning System) and LLR (Lunar Laser Ranging), has in-
creased enormously. As an application of the theoretical devel- Symbol Definition
opments above, we not only show that at the level of 1 mas {O; X, Y, Z} Inertial reference frame
some variations related to the triaxial shape of the Earth and {O; x, y, z} Arbitrary body-fixed axes
generally not considered have to be taken into account at the (ω1 , ω2 , ω3 )
The components of the angular rotational vector
present date. We also evaluate the combined effects on the po- →
−
ω
lar motion of the elasticity, the triaxiality of the Earth and the
(L, G, H) Andoyer’s variables
poorly known lag coefficient δ which is responsible for damp-
(l, g, h) Canonically conjugate variables to (L, G, H)
ing. For this purpose we choose real data taken from IERS
(International Earth Rotation and Reference System Service). (χ, θ) Polar coordinates of the motion of the axis of
rotation of the body with respect to its figure axis
Our treatment consists of a detailed discussion and mathe- J Angle between angular momentum equator and
matical modelling for the damping of the free rotational motion the equator of figure (O; x, y)
with the purpose of explaining better how the Chandler wobble
T Kinetic energy
is maintained in the presence of such dissipation. The nature
K = T Hamiltonian for the free rotation
of the continuous excitation for the Earth (such as the response →
−
of the oceans, fluidity of the core, atmospheric excitation, elec- M Angular momentum vector
tromagnetic processes, etc.) is not discussed here. For more in- C = C R + c Inertia tensor
formation, the reader is referred to the works of Dehant et al. CR Inertia tensor of an undeformed body whose
(1999), Nastula & Ponte (1999) and McCarthy & Petit (2003), principal axes of inertia coincide with the
for instance. Therefore, our goal is a refinement, at the present body-fixed axes
accuracy, of the mathematical model, to reduce the residual de- c Tensor which takes into account the deviation of
viation of the model from reality. a deformable body from the rigid model
The Hamiltonian variational equations seem to be more A0 , B0 , C 0 Principal moments of inertia of a rigid body
convenient and useful in many respects, since such equations k Love number
enable us not only to separate precisely the different terms ac- ks Secular Love number
cording to the different aspects considered in this paper (triax- ν Intermediary parameter
iality, elasticity and time lag), but also to make practical com- C0 Intermediary parameter
putations for the study of the free polar motion of an arbitrary ε Triaxial parameter
celestial body. By integrating the general variational equations
α, β, γ, ρ, τ Elastic parameters
for the polar coordinates l (angle) and J (amplitude), we show
δ Phase lag
that the corresponding solutions for these two parameters have
secular and periodic parts, this latter caused by three different σE Euler frequency
effects taken into account here. σC Chandler frequency

For the case of Mars’ free polar motion, we have analyzed

the extended analytical solutions coming from the Hamiltonian
formalism to show how the phase shift, δ, also plays an impor- 3. Equations of the torque-free rotational motion
tant role in its motion. A study of the normal modes of a triaxial of an elastic body in terms of Andoyer
planet such as the Earth and Mars using the Liouville method variables: Extension of Kubo’s procedure
can be seen in Van Hoolst & Dehant (2002).
3.1. Expression of the Hamiltonian of the torque-free
rotational motion.
In this section, we will generalize the equations of the
torque-free rotational motion of a rigid body in terms of the
2. Notation and definitions Andoyer variables as discussed in Souchay et al. (2002), to
the case of a deformable body and also consider two kinds
For later use, we list in Table 1 the fundamental arguments of shape-configuration, a quasi axisymmetric body and a body
which allow us to write the equations of the torque-free ro- with a triaxial form. For this purpose, we will also summa-
tational motion of an elastic body. We also present in the same rize here some basic facts and formulas which are needed to
table other useful quantities, such as constants and geodetic pa- describe the equations. For details, we refer the reader to text-
rameters, considered in the present paper. For this purpose, we books such as Munk & MacDonald (1975), Lambeck (1980)
have chosen the same notation as in Kinoshita (1977), Moritz & and Moritz & Mueller (1987).
Mueller (1987), Kubo (1991) and Souchay et al. (2002) in order The basic equation
to establish homogeneity of notation. The interested reader is
referred to the above mentioned articles and textbook for more 1→− T −1 →
−
details. K =T = M C M (1)
2
 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism 1103

expresses the Hamiltonian of the torque-free rotational mo- D, E and F are considered small quantities whose products,
tion (equal to the kinetic energy, in this case) in terms of the squares and higher powers can be neglected. Equation (6) then
→
− reduces to:
angular momentum M and the inertia tensor C. It holds for
the rotation of an arbitrary body, rigid or elastic, whenever 1  
K = BC sin2 l + AC cos2 l G2 − L2 + ABL2
we chose as the body-fixed axes {x, y, z}, the Tisserand axes 2 ABC 
(Munk & MacDonald 1975, p.10). By taking these
axes, the −CF G2 − L2 sin 2l
→
−
relative, or deformational, angular momentum h = T → −x ×→
−u – √ 
caused by the existence of a residual deformation velocity → −u , −2L G2 − L2 [BE sin l + AD cos l] . (7)
see for instance (Moritz & Mueller 1987, pp. 110–115) – can Considering rotational symmetry (A = B), Eq. (7) reduces to
→
−
be made zero and the angular momentum vector M takes the the simple form:
same form as for a rigid body. In the following, we will always
1  2 1 2 F
use Tisserand axes. Notice also that, in the case of a deformable K = G − L2 + L − 2 (G2 − L2 ) sin 2l
body, the inertia tensor C will be a function of time since the 2A 2C 2A
1 √ 2
shape of the body will, in general, change with the time. Then, − L G − L [E sin l + D cos l] .
2 (8)
the inertia tensor C can be written as: AC
 
 A F E  Rotational deformation:
 
C = C R + c =  F B D  (2)
  If the elastic body rotates about an axis which deviates from
E D C
the axis of symmetry of the body, then centrifugal forces tend
where, to distort it and therefore this distortion causes variations in the
tensor of inertia. This effect is known as rotational deforma-
 
 A0 0 0  tion, which is the only one to take into account in the study of
 
C R =  0 B0 0  (3) the torque-free rotational motion of an elastic body. Periodic
 variations in the tensor arising from rotational deformation are
0 0 C0
given by (Munk & MacDonald 1975, p. 25):
corresponds to an rigid body whose principal axes of inertia co- 2 k
incide with the body-fixed axes and whose principal moments c11 = c22 = − (C0 − A0 ) m3
3 ks
of inertia A0 , B0 and C0 are constant in time (Souchay et al.
4 k
2002), and c33 = (C0 − A0 ) m3
3 ks
 
 c11 c12 c13  k
D = (C0 − A0 ) m2
 
c =  c12 c22 c23  (4) ks

c13 c23 c33 k
E = (C0 − A0 ) m1
ks
takes into account the deviation of a deformable body from the F=0 (9)
rigid model, so that the elements ci j are small quantities. From
Eqs. (2) and (4), we have: A = A0 + c11 , B = B0 + c22 , C = where, k is a Love number, ks is the secular Love number equal
to 3κ a(C5 Ω0 −A
2
0)
C0 + c33 , D = c23 , E = c13 and F = c12 . 2 (κ2 denotes the gravitational constant and a is the
Taking into account the above considerations, the angular Earth’s equatorial radius), m1 and m2 express deviations of
→
− the rotation axis from the z-axis and m3 express variations of
momentum vector M is written as (Souchay et al. 2002):
the rotational speed. That is to say, the rotation vector → −
ω is
→
− T   √ √ written, in the case of an elastic body, as:
M = L x , Ly , Lz = G2 − L2 sin l, G2 − L2 cos l, L . (5)
→
− −→
ω =→ −
ω 0 + δω (10)
We substitute Eqs. (5) and (2) into Eq. (1). The result is, after
some developments: with
→
− T
ω 0 = (0, 0, Ω), (11)
1
  
K = BC − D sin l + AC − E cos l G − L
2 2 2 2 2 2
which corresponds to a rotation with constant angular veloc-
2∆  ity Ω around the z-axis, and
+ AB − F L − (CF − DE) G − L sin 2l
2 2 2 2

√  −→T
δω = (m1 , m2 , m3 ) Ω. (12)
− 2L G − L (BE − FD) sin l+(AD − EF) cos l
2 2 (6)
For Tisserand axes, we can also use the following relations for
∆ being the determinant of inertia tensor C, that is to say, the components of the angular-momentum vector (Kinoshita
1977):
√
∆ = ABC − AD2 − BE − CF 2 + 2DEF. G2 − L2 sin l = Aω1
√
Note: Expression (6) has the same form as that obtained by G2 − L2 cos l = Bω2
(Kubo 1991) using a different procedure. L = Cω3 . (13)
1104 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism
−→
The variations c11 , c22 , c33 and δω are small quantities whose Numerical values of these parameters (for the Earth and Mars)
products and squares are too small to be of interest in the are presented in Tables 2 and 3. Note that for an Earth with
−→ rotational symmetry (A0 = B0 ): α = β.
present study. Then the components of δω can be written, in
terms of Andoyer variables, as: Following Kubo (1991), D and E should be considered only
as functions of time. So, we will denote J and l in Eq. (18) as
1 √ 2 J  and l . Numerically, they are equal to J and l, respectively.
m1 = G − L2 sin l
A0 Ω The existence of non-elasticity causes a time lag δ between the
1 √ 2 rotational axis and the pole of the equatorial bulge due to the
m2 = G − L2 cos l
B0 Ω centrifugal force. Then, l should be equal to l+δ (δ > 0) and D
L − C0 Ω and E should be written as:
m3 =   · (14)
C0 + 3 ks (C0 − A0 ) Ω D = β C0 sin J  cos l = β C0 sin J  cos (l + δ)
4 k

E = α C0 sin J  sin l = α C0 sin J  sin (l + δ). (20)

We now substitute Eqs. (14) into (9) to get the variations of
the inertia tensor due to the rotational deformation in terms of When these last expressions are substituted in the general ex-
Andoyer variables. The result is: pression of the Hamiltonian (7) one obtains, considering the
approximation for the determinant ∆ ≈ A0 B0C0 and neglecting
2 ν
c11 = c22 = − (L − C0 Ω) terms of second and higher order:
3 C0
L √ 2
4 ν K = KR − G − L2 sin J 
c33 = (L − C0 Ω) A0 B0
3 C0  
ν √ 2 × αB0 sin l sin l + βA0 cos l cos l (21)
D= G − L2 cos l
B0 where
ν √ 2  
E= G − L2 sin l 1 sin2 l cos2 l  2 1 L2
A0 KR = + G − L2 + (22)
2 A0 B0 2 C0
F=0 (15)
is the Hamiltonian of the torque-free motion for a rigid body
where with a triaxial form (Souchay et al. 2003).
k 1 4 When α ≈ β, that is to say the body is quasi-triaxial, we
ν= (C0 − A0 ) and C0 = C0 + νΩ. (16) can do the approximation α ≈ β ≈ γ = α+β 2 and introduce this
ks Ω 3
last value, γ, into the Hamiltonian (21). This substitution leads
Taking into account the relation between the variable L and the to the following expression for the Hamiltonian:
angle J: L = G cos J, we can express (15) as a function of J as γ √
follows: K = KR − L G2 − L2 sin J 
A0 B0
 
2 ν × B0 sin l sin l + A0 cos l cos l . (23)
c11 = c22 = − (G cos J − C0 Ω)
3 C0
Adopting the following notation:
4 ν
c33 = (G cos J − C0 Ω) A0 + B0 B0 − A0
3 C0 Ā = and ε= (24)
ν 2 A0 + B0
D= G sin J cos l we can rewrite Eq. (23) as:
B0
ν γ √
E= G sin J sin l K = K R − L G2 − L2 sin J 
A0
 Ā 
F = 0. (17) × (1 + ε) sin l sin l + (1 − ε) cos l cos l (25)
To first order, we can consider the approximations G ≈ L ≈ and the Hamiltonian of the torque-free motion for a rigid
C0 Ω and cos J ≈ 1. Then the periodic additions to the moments body (22) as:
and products of inertia with respect to the average principal 1  1 L2
axes become: KR = (1 − ε cos 2l) G2 − L2 + · (26)
2 Ā 2 C0
c11 = c22 = c33 ≈ 0 Notice also that in the expressions (25) and (26) ε has a small
D = β C0 sin J cos l value.
For the case where it is not possible to make the above
E = α C0 sin J sin l
approximation: α ≈ β ≈ γ = α+β 2 , the Hamiltonian (21) can
F=0 (18) be rewritten as, in terms of Ā and ε:
L √ 2
where the following notation was adopted: K = KR − G − L2 sin J 
 Ā 
k C 0 − A0 k C 0 − A0
α= and β= · (19) × α(1 + ε) sin l sin l + β(1 − ε) cos l cos l . (27)
ks A 0 ks B 0
 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism 1105

3.2. Variational equations dl  γ
 = L [cos δ − ε cos (2l + δ)]
dt d Ā

Once we have obtained the Hamiltonian corresponding to the dg  γ
torque-free rotational motion of an elastic body in terms of  = − L [cos δ − ε cos (2l + δ)]
dt d Ā
Andoyer variables (L, G, H, l, g, h), we can use the general 
dh 
Hamilton’s equations of motion to establish the variational  =0 (31)
equations of the problem considered here (Kinoshita 1977): dt d

d ∂K and the corresponding pure contribution due to the deformation

(L, G, H) = − for the angle J is given by:
dt ∂(l, g, h)

d ∂K dJ  γ
(l, g, h) =  = − L sin J  [sin δ + ε sin (2l + δ)] (32)
dt d
(28)
dt ∂(L, G, H) Ā
which express the time variations of Andoyer variables as func-
tions of partial derivatives of the Hamiltonian. Variational equations. General case:

α+β The substitution of the Hamiltonian (27) into the general

Variational equations for α ≈ β ≈ γ = 2 :
variational Eqs. (28) gives us the following expressions for the
The substitution of the Hamiltonian (25) in the previous temporal variations of Andoyer variables and the angle J:
expressions gives us: dL ε
= − G2 sin2 J sin 2l
dL ε dt Ā
= − G2 sin2 J sin 2l
dt Ā GL  
γ + sin J sin J  ρ sin δ + τ sin (2l + δ) (a)
+ GL sin J sin J  [sin δ + ε sin (2l + δ)] Ā
Ā dG
dG =0 (b)
=0 dt
dt dH
dH =0 (c)
=0 dt
dt  
  dl L C0 − Ā ε
dl L C0 − Ā =− + L cos 2l
=− dt Ā C0 Ā
dt Ā C0
ε γ L 
+ L cos 2l + L [cos δ − ε cos (2l + δ)] + ρ cos δ − τ cos (2l + δ) (d)
Ā Ā Ā
dg G ε γ dg G ε
= − G cos 2l − L [cos δ − ε cos (2l + δ)] = − G cos 2l
dt Ā Ā Ā dt Ā Ā
dh L 
=0 (29) − ρ cos δ − τ cos (2l + δ) (e)
dt Ā
dh
and taking into account that dJ dt = − G sin J dt , the variational
1 dL
=0 (f)
equation for the angle J is given by: dt
dJ ε
dJ ε γ = G sin J sin 2l
= G sin J sin 2l − L sin J  [sin δ + ε sin (2l + δ)] . (30) dt Ā
dt Ā Ā L  
For later use it is convenient to write the time variations of − sin J  ρ sin δ + τ sin (2l + δ) (g)
Ā
Andoyer variables (z) in the form: (33)
 
dz  dz 
 + 
dz
= where
dt dt r dt d
  1  1 
 dz  ρ= α(1+ε)+β(1−ε) and τ = α(1+ε)−β(1−ε) . (34)
where dz dt r is the corresponding part of a rigid body and dt d 2 2
is the new contribution due to the deformation. The later
Notice that in these equations we have introduced two new
forms (29) and (30) show very well that the terms without the
parameters: ρ and τ, which are related to the elastic parame-
factor γ correspond to the expressions for the torque-free mo-
ters α and β and the triaxial parameter ε, therefore characterize
tion of a rigid body (Souchay et al. 2003). Thus, the pure con-
the shape-configuration and elastic properties of the body. In
tribution due to the deformation is given by:
 the case of rotational symmetry (A = B), we can remark that
dL  γ α = β, then: ρ = α and τ = 0, which is Kubo’s approximation.
 = GL sin J sin J  [sin δ + ε sin (2l + δ)]
dt d Ā According to Eq. (19), we have αβ = AB00 . Then, in the treatment

dG  of the torque-free polar motion of a body with a marked tri-
 =0
dt d axial shape, α and β should be significantly different, so that

dH  a consideration of the terms depending of the parameter τ is
 =0
dt d necessary.
1106 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism

The solution of this system of equations and its correspond- The solution of the first two equations of Eq. (37) gives
ing applications to the Earth’s and Mars’ free polar motion will us the components m1 and m2 that characterize a deviation of
be presented in the next sections. the Earth’s rotation axis from the z-axis for an arbitrary non-
rigid body, that is to say, the polar motion. Similarly, the third
3.3. Comparison of the Hamitonian solution equation can be solved for m3 to get variations of the speed of
with the linearized Liouville equations rotation, or variations of the length of day.
The goal of this section is the derivation of the variational Inserting the relations (14) and its derivatives into the first
equations for l and J, as formulated in the previous section, two Liouville Eqs. (37) and also taking into account that L =
by means of linearized Liouville equations, to demonstrate the G cos J and considering the case when α ≈ β ≈ γ = α+β 2 ,
equivalence between the two different formalisms used in the we get:
literature. A different comparison between these two formula-
dJ dl C0 − B0
tions for a purely elastic Earth model can also be seen in Sevilla cos J sin l + sin J cos l + Ω sin J cos l
& Romero (1985). Nevertheless, we derive here in detail the dt dt B0
equation corresponding to the angle J since its variation has not − γΩ sin J  cos (l + δ) = 0 (40)
been deeply studied till now. A simple presentation of Liouville
equations is offered first. dJ dl C0 − A0
For Tisserand axes, the Liouville equation is the fundamen- cos J cos l − sin J sin l − Ω sin J sin l
dt dt A0
tal equation for the mathematical description of free polar mo- + γΩ sin J  sin (l + δ) = 0. (41)
tion for a non-rigid body. It can be written as follows (Munk &
MacDonald 1975, p.10): By introducing the relation:
∂ → →−
(C −
ω) + →
−
ω × (C→
−
ω) = 0 . (35) C 0 − B0 C 0 − A0
− = −2
εC0
∂t B0 A0 Ā
Here ∂t∂ denotes the time derivative in the body-fixed system
and multiplying Eq. (40) by sin l and Eq. (41) by cos l, and then
(which rotates with the body) and the cross (×) denotes the
adding the two equations, we get the result:
vector product of two vectors.
We will now linearize the above equation. Taking into dJ ε γ
account the expressions (2)–(4), (10)–(12) and that D, E, F = G sin J sin 2l − L sin J  sin δ. (42)
dt Ā Ā
−→
and δω are considered small quantities, and only retaining lin-
In the same way, by multiplying Eq. (40) by cos l and Eq. (41)
ear terms, the Liouville equations become:
by sin l, and after combination, we find:
A0 ṁ1 + (C0 − B0 )Ωm2 + Ė − DΩ = 0  
dl L C0 − Ā ε γ
B0 ṁ2 − (C0 − A0 )Ωm1 + Ḋ + EΩ = 0 =− + L cos 2l + L cos δ. (43)
dt Ā C0 Ā Ā
C0 ṁ3 + Ċ = 0 (36)
∂ Notice that the two equations above give the same expres-
the dot denoting a time derivative ∂t . sions as those for dl dJ
dt and dt in Eqs. (29) and (30). Thus,
The substitution of the derivatives for E, D and c33 yields:
the equivalence between the two formulations, that is to say
ṁ1 + ᾱm2 = 0 the Hamiltonian formulation and that obtained from Liouville
ṁ2 − β̄m1 = 0 equations, has been demonstrated.
ṁ3 = 0 (37)
3.4. Transformation between the axis of angular
with
momentum and the axis of rotation
(C0 − B0 − νΩ)Ω (C0 − A0 − νΩ)Ω
ᾱ = and β̄ = · (38) All the analytical expressions obtained in the preceding sec-
A0 + νΩ B0 + νΩ
tions concerned the axis of angular momentum of the celes-
In terms of the principal moments of inertia and Love numbers: tial body. Indeed, the polar motion which could be really ob-
 
served is that of the axis of rotation of the body. Therefore, it is
ks A 0 − B 0 + 1 − ks C 0 Ω
k k
ᾱ =  necessary to recapitulate and to give with precision how these
1 − kks A0 + kks C0 two motions are related one to the other. Let us represent the

1 − kks (C0 − A0 ) Ω motion of the axis of rotation of the body with respect to its
β̄ = · (39) axis of figure in polar coordinates (χ, θ), where χ is the radius
− kks A0 + B0 + kks C0 of the motion and θ the polar angle. They satisfy the following
Note. For rotational symmetry (A0 = B0 ): equations, from Eq. (14):

1− k 1
ks m1 = χ sin θ = G sin l sin J
ᾱ = β̄ = σC = σE k σE
Chandler frequency A0 Ω
1+ ks Ω 1
(σE : Euler frequency). m2 = χ cos θ = G cos l sin J. (44)
B0 Ω
 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism 1107

Thus, for very small J (J  1) and with the approximation The negative exponential coefficient −J1 characterizes the
G ≈ C0 ω3 ≈ C0 Ω, we get: damping of the polar motion, which depends directly on the
C0 phase lag δ. Therefore, the variations of J are characterized
χ sin θ ≈ J sin l by a main component, decreasing with respect to the time, and
A0
an oscillation with amplitude J0 J2 exp {− J1 t} decreasing in the
C0
χ cos θ ≈ J cos l (45) same proportion as this main component, and a frequency twice
B0 the frequency l1 of the Chandler wobble. Notice that the coef-
which gives the direct relationships: ficient J2 , and also the argument of the exponential function
A0 in J0 , can be written as:
tan l = tan θ (46)  
B0 C0 ε 1
J0 = 0. 5 × exp
and, after transformation: 2 C0 (1 − ρ cos δ) − Ā
C0 ε 1
l ≈ θ − ε sin 2θ. (47) J2 = (53)
2 C0 (1 − ρ cos δ) − Ā
Moreover, by combining the two equations of Eq. (44), we get which depend on both the set of parameters (ρ, δ, ε) and the
to the first order the relationships between J and χ: Chandler frequency. In the following a detailed practical appli-
A0 cation is made to the Earth and Mars.
J≈ χ(1 + ε cos 2θ). (48)
C0
4. Application to the Earth
3.5. Solution of the free polar motion equations Our purpose in this section is to show precisely what are the
with τ  1 various influences acting on the true polar motion of the Earth
Neglecting second order terms, we have solved the variational and which are clearly represented by each of the components in
Eqs. (33) for the variables l and J. In the solution of these the right-hand side of Eqs. (33) giving the expressions of dl/dt
equations, we have considered that τ is negligible (see Tables 3 and dJ/dt. These are the influence of the triaxiality of the Earth,
and 4, for the case of the Earth and Mars) and that J has a very through the parameter ε, of the Earth’s elasticity, through ρ,
small amplitude so that we can write: sin J ≈ sin J  ≈ J. Then, and of the time lag δ due to the inelastic response of the planet.
on performing the integrations with respect to the time t, we For that we use actual polar motion data of the Earth, and as
have found the following explicit form of the solutions: a first step we explain how the angles J and l which are fun-
damental in our equations, are determined starting from their
l = −l1 t + l2 sin (2l1 t) (49) counterparts for the axis of rotation, respectively χ and θ, as
J = J0 × exp {− J1 t − J2 cos (2l1 t)} (50) determined from up-to-date observational data.

where the coefficients li and Ji are:

  4.1. Use of real polar motion data: The problem
C0 Ω C0 − Ā
l1 = − ρ cos δ of determination of the origin
Ā C0
C0 Ωε In the equations of the preceding section the motion of the
l2 = angular momentum axis is given by the two polar coordi-
2 Āl1
  nates l and J, with respect to figure axis. As we have shown
 C0 Ωε
J0 = 0. 5 × exp in Sect. 3.4, the rotation axis is very close to the angular mo-
2Āl1 mentum axis, so that we make an a priori assumption that there
C0 Ωρ is no angular difference between the two axes, and we will as-
J1 = sin δ
Ā sume that J and l in our equations also characterize the motion
C0 Ωε of the pole of rotation. In the following we justify this approxi-
J2 = · (51)
2 Āl1 mation by calculating the quantities l − θ and J − χ as given by
Eqs. (47) and (48).
Equations (49) and (50), together with the set of integration
A preliminary step consists in determining the real val-
constants (51) give the features of the polar motion (l, J) as
ues of χ and θ, which represent the polar coordinates of the
deduced from our Hamiltonian equations and clearly show the
axis of rotation of the Earth with respect to the axis of figure.
effect of elasticity (ρ), time lag (δ) and triaxiality (ε).
For that we use the data of the IERS series C-04 in Cartesian
From Eq. (49), we can observe that the angle l not only
coordinates (x, y) of the pole of rotation with respect to an
undergoes the retrograde linear variation with frequency −l1 ,
arbitrary point Oc adopted conventionally and internationally
corresponding to the Chandler motion, but also a superposed
(IERS Annual Report, 2003). Notice that we restrict our study
oscillation with frequency twice that of the Chandler motion
to the period ranging from 1971, Feb. 18th to 2004, Feb. 19th.
and an amplitude which depends directly on the triaxial coef-
It is obvious that the conventional point Oc is not appropriate
ficient ε. The amplitude of the angle J from Eq. (50) can be
for defining the (χ, θ) data set, for Oc designates a point signif-
re-written, to first order, as:
 icantly different from the point where the axis of figure pierces
J = J0 exp {− J1 t} × (1 − J2 cos (2l1 t)) + O J 2 . (52) the Earth’s surface.
1108 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism
0.50
Table 2. The choice of origin O and its influence on the period of the
Chandler wobble.
0.40
∆x ∆y Rate T Chandler Relative
( ) ( ) (rd/d) (d) uncertainty

Radius chi (arcseconds)

−6
0.05 0.27 −0.01450623 × 10 433.138 9.63 × 10−5 0.30

0.07 0.27 −0.01479483 × 10−6 433.176 1.01 × 10−4

0.09 0.27 −0.01450440 × 10−6 433.191 1.12 × 10−4
0.20
0.05 0.29 −0.01450667 × 10−6 433.126 9.42 × 10−5
0.07 0.29 −0.01450483 × 10−6 433.179 1.00 × 10−4
0.09 0.29 −0.01450279 × 10−6 433.242 1.13 × 10−4 0.10

0.00
4000.0 5000.0 6000.0 7000
Days
For our present study, we will assume that it is represented
by an origin O (intersection of the figure axis with the Earth’s (a)
crust) so that the real path of the rotation axis should satisfy 2.0
symmetry around O. Therefore, the first step determines a suit-
able offset Oc O from Oc such that the polar angle θ has a regular 1.5

behavior, as close as possible linear.

1.0
To do that, we tested a set of different origins O correspond-
ing to ∆x = {0. 05, 0. 07, 0.09} and ∆y = {0. 27, 0.29} which
Angle theta (radians)

0.5
seem at first sight to be very close to the best estimation. We
then made a linear fit of θ and we calculated the slope from 0.0

which we deduce the value of the period of the Chandler wob- −0.5
ble, T Chandler . Our results are shown in Table 2. The relative un-
certainty is equal to the uncertainty of the slope. The period of −1.0

the Chandler wobble determined by the linear regression pro-

−1.5
cedure outlined above varies between 433.126 d and 433.242 d.
It can be shown that a significant drift of O from the best −2.0
4000.0 5000.0 6000.0 7000.0
location of the figure axis will result both in more significant Days
variations of the amplitude of χ and θ (once the linear com- (b)
ponent for this last angle has been removed). Moreover, the
signal for these two parameters should then be characterized Fig. 1. Variation of the radius χ and the angle θ of the polar motion
by the presence, as an artificial effect, of a sinusoidal oscilla- of the rotation axis with respect to the origin O such that Oc O =
tion with a period equal to that of the polar motion (Chandler (0. 05; 0. 27) (thin curve) and Oc O = (0. 07; 0. 29) (thick curve). The
term). The larger this oscillation, the larger the departure of O corresponding time span covers the range between 1982, Jan. 30th and
from its optimal choice. To illustrate this we show respectively 1990, Apr. 19th.
in Figs. 1a, for χ, and 1b, for θ, the two curves correspond-
ing to a different choice of the origin O. In the two figures the
bold curve corresponds to Oc O = (0. 07, 0. 29), whereas the In Fig. 2 we show the classical (x, y) two-dimensional rep-
thin curve corresponds to Oc O = (0. 05, 0. 27). The time t is resentation of the polar motion centered on the selected ori-
measured in days from the original date 1971, Feb. 20th, so gin Os , in the selected time span. We can remark that this curve
that the 3000-day time interval ranges between 1982, Jan. 30th has a circular shape and that the point O, represented by the
and 1990, Apr. 19th. We can note in the two figures that in black point at the center, looks like a suitable point of symme-
the second case (Oc O = (0 . 07, 0 . 29)), an oscillation with try of the curve, which confirms its choice.
period close to T Chandler appears clearly, as an artificial effect
explained above, and due to an inappropriate choice of O, 4.2. Study of specific contributions to the polar motion
whereas this oscillation is not present in the first case. Then
we can admit that the choice of the origin Os such that the We will now analyze the solutions (d) and (g) of Eq. (33) for
offset Oc Os = (0. 05, 0. 29) is a suitable one. In the following the Earth’s free polar motion (J, l). Notice that we are not in-
we will adopt this value. Notice that the time span considered terested here in the question of how the Chandler wobble is
(4000 d < t < 7000 d) has been selected such that the signal maintained in the presence of dissipation, or in the exact nature
does not present an artificial oscillation due to an inappropriate of the excitation. This is the subject of abundant research at the
choice of O, as explained before. We also observed that the sec- present date. For a review of this topic, see e.g. Höpfner (2003).
ular motion of the pole (Höpfner 2003) outside the time span We will focus in particular on showing how the time lag,
above noticeably increases this oscillation. δ, plays an important role in the damping of this motion, and
 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism 1109
0.40 0.0000

−0.0001
0.30
−0.0002

0.20 −0.0003

Amplitude (arcsecond)
−0.0004
0.10
Y, in arcsecond

−0.0005

0.00
−0.0006

−0.0007
−0.10
−0.0008
−0.20
−0.0009

−0.30 −0.0010
4000.0 5000.0 6000.0 7000.0
Days

−0.40
−0.50 −0.40 −0.30 −0.20 −0.10 0.00 0.10
X, in arcsecond
0.20 0.30 0.40 0.50
Fig. 3. Difference J − χ between the radius J of the path described by
the angular momentum axis and the radius χ of the corresponding ro-
Fig. 2. (x, y) coordinates of the pole of rotation of the Earth for the tation axis with respect to the origin Os such that Oc Os = (0. 05; 0. 29).
3000-day time interval between 1982, Jan. 30th and 1990, Apr. 19th.

Table 3. The Earth parameters. and the anelasticity of the mantle (dispersion) +8.5 days (Smith
& Dahlen 1981).
Parameter Value Unit Origin Notice that our Eq. (d) of Eq. (33) enables us to deduce
A0 8.0101 × 1037 kg m2 Groten (2000) directly from l˙ the analytical determination of the Chandler pe-
B0 8.0103 × 1037 kg m2 Groten (2000) riod taking into account only the elasticity of the mantle:
C0 8.0365 × 1037 kg m2 Groten (2000)  
k 0.29 ˙l = C0 ω × − C0 − Ā + γ cos δ (54)
ks 0.9383 Groten (2000) Ā C0
Ω 7292115 × 10−11 rad s−1 Groten (2000)
which gives T Chandler = 2π/l˙ = −442.2 days. Moreover, this
α 1.01864 × 10−3 This paper
same value when considering the Earth as a rigid body, and cor-
β 1.01861 × 10−3 This paper
responding to the Eulerian motion, is obtained by setting γ = 0
γ 1.01862 × 10−3 This paper
in the equation above. We thus find: T Euler = −304.6 days.
Ā 8.0102 × 1037 kg m2 This paper
Therefore the contribution of the Earth’s elasticity coming from
ε 1.24841 × 10−5 This paper
our calculations is 137.6 days which has to be compared with
ρ 1.01863 × 10−3 This paper
the 143 days above.
τ 2.77167 × 10−8 This paper
Pole offset between the angular momentum axis
and the rotation axis
the influence of the polar motion amplitude itself on the irreg-
ular nature of the damping. But we want also to make specific Figures 3 and 4 show respectively the differences of ra-
conclusions based on the results established in the preceding dius J − χ and of polar angle l − θ between the axis of angular
sections, and which can be applied to the Earth. One is the momentum and the axis of rotation, as deduced from Eqs. (47)
evaluation of the pole offset between the angular momentum and (48), and where χ and θ are taken from the real data.
axis and the rotation axis. Another is the determination of the The amplitude of J is always smaller than the amplitude of χ,
periodic components of the rotation axis of half the Chandler with differences varying between −0.25 and −1.0 milliarcsec
period whose existence and amplitudes can be deduced directly (Fig. 3). In comparison, the variations of l − θ (Fig. 4) are quite
from Eqs. (d) and (g) of Eq. (33). Table 3 presents the Earth pa- regular, with a quasi-sinusoidal behavior, with amplitude a little
rameters that we have used in our computations. The numerical larger than 1◦ and a period that is slightly variable and roughly
values for α, β, γ, Ā, ε, ρ and τ for the Earth were determined equal to T Chandler , as can be easily predicted from Eq. (47).
from the expressions (19), (24) and (34). Notice that the separation between the two axes is not negli-
gible when compared with the present observational precision
Period of the Chandler wobble of GPS and VLBI techniques (better than 0. 001), although it is
small relative to the total amplitude of the polar motion.
Our fitted value for the period of the Chandler wobble as
determined in the preceding section was T Chandler ≈ −433.1 d Determination of the phase lag δ responsible
(in this subsection we use the sign “–” to show that the mo- for the damping
tion is retrograde). Various studies have shown that this period
should range between 410 and 442 days (Höpfner 2004), and The real value of δ has been subject to many investiga-
our value is intermediate between these two extremes. tions and is poorly known. According to the estimate of the
As summarized by Höpfner (2004), the contribution of each quality factor Q−1 , which represents the dissipation inside the
component to the Chandler period has been estimated: the elas- Earth, and the relationship between these damping parameters:
ticity of the mantle brings +143.0 days, the fluidity of the core Q−1 = tan δ ≈ sin δ (see e.g. Lambeck 1988), we can obtain
−50.5 days, the pole tide (effect due to the oceans) +29.8 days, the most probable value for the time lag. Some estimates of
1110 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism
2.0 0.4

0.3

1.0 0.2

Amplitude (arcsecond)
Amplitude (degree)

0.1

0.0 0.0

−0.1

−1.0 −0.2

−0.3

−2.0 −0.4
4000.0 5000.0 6000.0 7000.0 4000.0 5000.0 6000.0 7000.0
Days Days

Fig. 4. Difference l − θ between the angle l describing the path of the Fig. 5. The same curve as Fig. 1 representing the radius χ of the po-
angular momentum axis and the angle θ of the corresponding path lar motion of the rotation axis, fitted by a 7th. order polynomial. The
of the rotation axis with respect to the origin Os such that Oc Os = six curves at the bottom represent the damping effect as calculated by
(0. 05; 0. 29). Eqs. (33), for different values of the phase lag δ, from 1◦ (upper curve)
to 6◦ (lower curve), with 1◦ increment.
0.0010

the specific dissipation function have been computed by Okubo 0.0008

(1982), who showed that: 50 ≤ Q ≤ 100, so that the range of δ 0.0006

should be roughly between 0.◦ 5 and 1.◦ 5. More detailed aspects 0.0004

concerning the Q factor can also be found in Widmer et al. 0.0002

Amplitude (") 0.0000

(1991) and Mathews et al. (2002). The subject is an active one.
−0.0002
We have calculated the theoretical variations of J and l (or −0.0004
χ and θ) given respectively by (d) and (g) of Eq. (33), according −0.0006

to 1◦ incremental values of the phase shift δ from (1◦ to 6◦ ). The −0.0008

influence of δ on the damping of the angle χ is quite remarkable −0.0010

as can be seen in Fig. 5. We have fitted the variations of χ with

4000.0 5000.0 6000.0 7000.0
Days

a seventh order polynomial in order to enhance the trend and to Fig. 6. Short-period oscillations, with period T Chandler /2, calculated
compare with the long periodic damping effect. Notice that, be- from Eq. (33), characterizing the radius χ of the polar motion of the
tween t = 5000 d and t = 6000 d, there seems to be a period of axis of rotation of the Earth, with respect to the origin Os . For com-
damping accompanied by a significantly decreasing amplitude parison, the difference J − χ between the two axes (Fig. 3) is shown as
(from 0. 3 to 0. 1) that appears to be in accordance with a value a dashed line.
of δ ≈ 6◦ , which is significantly larger than the value found
by other authors. For instance, Williams et al. (1978) and Bois
The curve for J (Fig. 6) consists of large oscillations in
et al. (1996), starting from the acceleration of the Moon’s mo-
amplitude of between 0.4 and 1.6 milliarcsec peak-to-peak,
tion due to tides, consider a time lag of 0.041 rd and 0.054 rd,
and for which the 2l˙ frequency appears clearly. Notice that
respectively, corresponding roughly to 2.◦ 4 and 3◦ . Notice also
the amplitudes of these oscillations are significantly larger
that the effect of damping is not linear with respect to time, for
than the amplitudes of the differences of radius between the
its derivative is directly proportional to the amplitude χ of the
two axes. Moreover, they are larger than the level of precision
motion.
of the GPS and VLBI determinations of the polar motion (x, y)
Determination of the amplitude of periodic analytical (a few 0.1 mas). Thus they might a priori be detected, and it
contributions seems instructive to show in parallel the short periodic curve
of the variations of χ obtained by subtracting the 7th. order fit-
In Sect. 3.3 we have shown and defined with precision the ting polynomial from the global variation in Fig. 5. The results
existence of sinusoidal oscillations, with frequency 2l,˙ (or 2θ̇) are shown in Fig. 7. The amplitude of these variations is larger
both for J and l (or respectively for χ and θ). These oscillations by one order of magnitude, so that the detection of the effects
come from the first and third term containing 2l of the right discussed here might not be possible.
hand-side of Eq. (d) in Eq. (33) and the two terms on the right On the other hand, in Fig. 8 the amplitude of the χ oscilla-
hand-side of (g) in Eq. (33). These equations give respectively tions with a frequency 2l˙ (full line) is significantly smaller than
the expressions for dJ/dt and dl/dt. We have integrated these that of the differences J − χ of the angle between the axis of
equations using the real values of J and l for the Earth in the figure and the axis of rotation (dashed line).
3000-day time interval already chosen, and inserted them in the
equations. The results are shown respectively in Figs. 6 and 7
5. Application to Mars
for χ and θ. As we think that it is of some interest to compare
these contributions with the amplitudes of the differences J − χ As one of our main purposes in this paper was to set up in a
and l − θ already shown in Figs. 3 and 4, these are presented straightforward and clear way the equations giving the polar
again in dashed lines. motion (J, l) of any given celestial body with irregular shape
 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism 1111

1.3
1.2
1.1
1.0
0.9
0.8
Amplitude (degree)

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
−0.1
−0.2
−0.3
−0.4
4000.0 5000.0 6000.0 7000.0
Days

Fig. 7. Short-period oscillations, calculated from Eq. (33) with pe-

riod T Chandler /2, characterizing the angle θ of the polar motion of the
axis of rotation of the Earth with respect to the origin Os . For com-
parison, the difference l − θ between the two axes (Fig. 4) is shown in
dashed line. Fig. 9. Global solution of J for Mars with changing k and δ. The units
are in arcseconds.
0.05

0.04

0.03 dependence of the slope of the curve on the Love number

0.02 adopted. Mathematically, this comes from the dependency
Amplitude (arcsecond)

0.01 of J1 on the coefficient ρ which is deduced from kMars itself.

0.00
The periodic parts of the solutions for JMars and lMars are
−0.01
represented respectively in Fig. 10, for δ = 5◦ and k =
−0.02
(0.15, 0.25). Notice that the curve for kMars = 0.15 varies more
−0.03
quickly than that for k = 0.25. This is again due to the presence
−0.04
of l1 in the argument of the sine and cosine functions and also in
−0.05
4000.0 5000.0
Days
6000.0 7000.0
the coefficient J2 , closely related to the damping. Consequently,
when k increases, l1 diminishes slightly and J2 gets larger. The
Fig. 8. Short-period oscillations of χ according to Eq. (33) already corresponding period is lengthened. The damping is more sig-
presented in Fig. 6 (thick line) in comparison with the short-period
nificant for k = 0.25 than that for k = 0.15. Notice also that the
variations of J − χ (dashed line).
amplitudes of these oscillations for JMars and lMars (t = 0) are
respectively 0. 8 peak-to-peak and 2.◦ 3, which is considerably
and elastic properties, according to the set of Eqs. (33), we larger than those corresponding to the Earth. Therefore, in the
show in this section how they can be applied advantageously frame of future space missions, it is necessary to take into ac-
in the case of Mars, whose physical characteristics, compiled count such oscillations, in case excitation still prevails on the
in Table 4 from several works and recent modelings of the ro- planet. Finally, in Fig. 11, we have plotted the mi -components
tation of Mars such as Yoder & Standish (1997), Folker et al. (i = 1, 2) for different values of the Love number k and for
(1997), Bouquillon & Souchay (1999) and Gauchez & Souchay δ = 5◦ .
(2000), are rather well known. The numerical values for the
elastic parameters and Ā and ε for Mars were determined from
6. Concluding remarks
the expressions (19), (24) and (34). Since accurate determina-
tions of the damping parameter δ and the Love number of Mars In this paper, we have adopted a Hamiltonian approach to the
are quite difficult to obtain, we have considered in this paper theory of the rotation of a triaxial and elastic celestial body to
a wide range of values for δMars (between 0.◦ 5 and 10◦ ) and give the equations of the polar motion (l, J). This follows the
two extreme values for kMars (0.15 and 0.25) to show the clear pioneer work of Kubo (1991), by taking into account not only
influence of both quantities on the free polar motion of Mars. elastic characteristics of the body but also the presence of a
As we have no information on the excitation to which the phase shift δ that represents a lag of the readjustment of masses
polar motion of Mars should be subject, we have only con- inside the body, and also the influence of its triaxiality. In order
sidered the damping effect which leads to the exponentially to validate our analytical expressions, we have compared those
decreasing tendency of the polar motion radius JMars shown for J and l with the ones obtained from the linearized Liouville
by the formula (50) depending on the three parameters J0 , equations, thus showing that they are equivalent to first order.
J1 and J2 . In Fig. 9 we have plotted the variations of JMars As the torque-free rotational motion of a celestial body has
(with an initial amplitude of 0. 5) associated with different a direct dependence on its moments and products of inertia, the
values of the time lag δMars and Love number kMars . They comparisons between their analytical solutions and present or
clearly show the linear tendency of such solutions. Sine (or co- future observational data can therefore provide useful material
sine) curves are superposed on these linear components, prin- to understand the internal behavior of this body, as is already
cipally due to the effects (damping, triaxiality and elasticity) the case for the Earth. Some interesting connections may be
considered in the calculations. In particular, we note a large found, for instance with slow changes in the figure of the Earth
1112 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism

Fig. 10. Periodic part of l and J for δ = 5◦ and k = (0.15, 0.25). The units for l are in radians and for J in arcseconds.

Fig. 11. m1 and m2 -components for δ = 5◦ . The units are in arcseconds.

Table 4. Parameters of Mars. and other celestial bodies, as well as in the relationships be-
tween the pole of figure and rotation.
Parameter Value Unit Origin
We have applied our analytical calculations for the Earth to
C0
A0
1.005741 Bouquillon & Souchay (1999) emphasize small but significant contributions to the polar mo-
tion. At first, as our two-dimensional parametrization of the an-
C0
B0
1.005044 Bouquillon & Souchay (1999) gular momentum pole and of the rotation pole of the Earth with
respect to a given origin is done though polar coordinates – re-
B0 −A0
C0
0.0006896 Bouquillon & Souchay (1999) spectively (J, l) and (χ, θ) – instead of the usual (x, y) rectan-
gular representation, we have studied in detail how the choice
C0
MR2
0.3662 Folkner et al. (1997) of the origin influences both the values of the coordinates and
the determination of the Chandler period T Chandler . We have
k 0.15–0.25 Yoder & Standish (1997) shown that for our selection of real data of polar motion be-
ks 1.203 Gauchez & Souchay (2000) tween 1982, Jan. 30th and 1990, Apr. 19th, this motion is very
Ω 7.08822 × 10−5 rad s−1 Folkner et al. (1997) regular around the point with coordinates Oc Os = (0. 05, 0. 29)
α(k = 0.15) 7.15836 × 10−4 This paper with respect to the conventional CIP pole, and that this point
β(k = 0.15) 7.1534 × 10−4 This paper
seems to be the most appropriate choice for the origin.
γ(k = 0.15) 7.15588 × 10−4 This paper We also thought it was of some interest to evaluate the dif-
α(k = 0.25) 1.19306 × 10−3 This paper ferences in amplitude (J−χ) and in angle (l−θ) between the axis
β(k = 0.25) 1.19223 × 10−3 This paper of angular momentum and the axis of rotation. Moreover we
γ(k = 0.25) 1.192645 × 10−3 This paper
have shown that in addition to the classical Chandlerian com-
Ā 2.68657 × 1021 kg m2 This paper
ponent of the polar motion, dampened by the influence of the
phase lag δ, an oscillation exists for both for J and l, with fre-
ε 0.346625 × 10−3 This paper
quency twice the frequency of the Chandler wobble itself and
ρ(k = 0.15) 0.71559 × 10−3 This paper
an amplitude directly proportional to the coefficient ε charac-
τ(k = 0.15) 0.49604 × 10−6 This paper
terizing the triaxiality.
ρ(k = 0.25) 0.11926 × 10−2 This paper
One of our important results concerns the value of the
τ(k = 0.25) 0.82840 × 10−6 This paper
phase lag δ responsible for the damping. For that we selected a
 M. Folgueira and J. Souchay: Free polar motion of a triaxial and elastic body in Hamiltonian formalism 1113

time interval for which the motion of the pole clearly undergoes Kinoshita, H. 1972a, Publ. Astron. Soc. Japan, 24, 409
damping, and we could show that according to our simulations, Kinoshita, H. 1972b, Publ. Astron. Soc. Japan, 24, 423
a value of δ ≈ 6◦ seems in accordance with the negative slope Kinoshita, H. 1977, Celest. Mech., 15, 277
caused by the decrease of the angle θ; this value is significantly Kinoshita, H. 1992, Celest. Mech., 53, 365
larger than that usually found in the literature. Kubo, Y. 1991, Celest. Mech., 50, 165
Lambeck, K. 1980, The Earth’s variable rotation (Cambridge
Finally, we showed what should be the behavior of the polar
University Press)
motion for the planet Mars, and the dependence on parameters
Lambeck, K. 1988, Geophysical Geodesy. The slow deformations of
which are still poorly know, such as the Love number kMars and the Earth (Oxford: Clarendon Press)
the rotational phase lag δMars . We emphasize the fact that the Mathews, P. M., Herring, T. A., & Buffett, B. A. 2002, J. Geophys.
oscillations with period T Chandler /2 are very large and of the Res., 107, B4
same order as the amplitude of the polar motion itself, because McCarthy, D., & Petit, G. 2003, IERS Conventions 2003, IERS
of a more pronounced triaxial shape than for the Earth. Technical Notes 32
Moritz, H., & Mueller, I. I. 1987, Earth rotation: Theory and
Acknowledgements. This research was carried out in the Department Observation (New York: The Ungar Publishing Company)
of “Systèmes de Référence Temps Espace” (SYRTE), Paris Munk, W. H., & MacDonald, G. J. F. 1975, The rotation of the Earth:
Observatory, under a combination of a financial support of A geophysical discussion (Cambridge University Press)
Paris Observatory (2003) and Descartes Prize Allowance (2004) Nastula, J., & Ponte, R. M. 1999, Geophys. J. Int., 139, 123
(M. Folgueira). Special thanks to Dr. D. McCarthy for several Okubo, S. 1982, J. Geophys. J. R. Astr. Soc., 71, 647
suggestions clarifying the presentation. Sevilla, M. J., & Romero, P. 1985, in Proc. of I Hotine-Marussi
Symposium on Mathematical Geodesy, 703
Smith, M. L., & Dahlen, F. A. 1981, Geophys. J. R. Astr. Soc., 64, 223
References Souchay, J., Folgueira, M., & Bouquillon, S. 2003, Earth, Moon and
Bois, E., Boudin, F., & Journet, A. 1996, A&A, 314, 989 Planets, 93, 107
Bouquillon, S., & Souchay, J. 1999, A&A, 345, 282 Stavinschi, M., & Souchay, J. 1994, in Proc. of Journées, Systèmes de
Dehant, V., Arias, F., Bizouard, Ch., et al. 1999, Celest. Mech., 72(4), Référence spatio-temporels, 81
245 Van Hoolst, T., & Dehant, V. 2002, Physics of the Earth and Planetary
Folkner, W. M., Yoder, C. F., Yuan, D. N., et al. 1997, Science, 278 Interiors, 134, 17
Gauchez, D., & Souchay, J. 2000, Earth, Moon and Planets, 84, 33 Widmer, R., Masters, G., & Gilbert, F. 1991, Geophys. J. Int., 104,
Getino, J., & Ferrándiz, J. M. 1990, Celest. Mech., 49, 303 541
Getino, J., & Ferrándiz, J. M. 1991, Celest. Mech., 51, 17 Williams, J. G., Sinclair, W. S., & Yoder, C. F. 1978, Geophys. Res.
Groten, E. 2000, J. Geodesy, 74-1, 134 Lett. 5, 11, 943
Höpfner, J. 2003, J. Geodesy, 77, 388 Yoder, C. F., & Standish, E. M. 1997, J. Geophys. Res., 102(2), 4065
Höpfner, J. 2004, Surveys Geophys., 25, 1