PHY411-AST233 Lecture notes

Alice Quillen

March 23, 2020

Contents
1 Infinitesimal Transformations and Lie Groups 2
1.1 Lie Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Rotation Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Example: Locomotion of Deformable Jello Cubes . . . . . . . . . . . . . . . 7
1.4 The Jello cube! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Mechanical Connection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 A fibre bundle and a Connection . . . . . . . . . . . . . . . . . . . . . . . . 16

2 Rigid Bodies 17
2.1 Configuration space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Generators of Rotation Matrices . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 The kinetic energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Example for the Moment of Inertia tensor . . . . . . . . . . . . . . . . . . . 22
2.5 Stability of a Freely Rotating Body . . . . . . . . . . . . . . . . . . . . . . . 22
2.6 Rigid Body motion from a Lagrangian . . . . . . . . . . . . . . . . . . . . . 24
2.7 A Hamiltonian, a moment map . . . . . . . . . . . . . . . . . . . . . . . . . 26

3 Rotational Dynamics 29
3.1 Euler angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.1 Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Lagrangian and Hamiltonian in terms of Euler angles . . . . . . . . . . . . . 31
3.2.1 Axisymmetric bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3 Torque free rotation of the axi-symmetric body . . . . . . . . . . . . . . . . 34
3.4 Euler angles of the torque free axi-symmetric body . . . . . . . . . . . . . . 35
3.5 Frequencies seen by an external viewer - asteroid rotation . . . . . . . . . . 37
3.6 Long axis and short axis modes . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.7 Comments on Damping of Wobbling . . . . . . . . . . . . . . . . . . . . . . 40
3.8 Andoyer-Deprit variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.8.1 Principal axis rotation . . . . . . . . . . . . . . . . . . . . . . . . . . 45

1
 3.9 Potential energy of a rigid body from an external force field . . . . . . . . . 45
3.10 MacCullagh’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.10.1 For an axisymmetric body . . . . . . . . . . . . . . . . . . . . . . . . 49
3.11 Precession of an axisymmetric body in a circular orbit . . . . . . . . . . . . 49
3.12 Quaternions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.13 A Lie-Poisson numerical method for Solid Body rotation with Quaturnions 57

4 Tidal Evolution 58
4.1 Gravitational Potential Expansion in Legendre Polynomials . . . . . . . . . 58
4.2 Expansion of the gravitational potential of a point mass in spherical harmonics 59
4.3 Tidal deformation of a spherical body . . . . . . . . . . . . . . . . . . . . . 60
4.4 Love numbers for fluid bodies . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4.1 Love numbers for an incompressible homogeneous elastic body . . . 63
4.5 Rotational Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.6 Tidal spin down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.6.1 Conventions for delay . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.6.2 Associated orbital drift rate during spin down . . . . . . . . . . . . . 68
4.7 Other timescales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

5 Spin Resonances 68
5.1 Types of Spin Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 Spin-Orbit resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.3 The dissipative spin-orbit resonance problem . . . . . . . . . . . . . . . . . 71

6 Problems 72

1 Infinitesimal Transformations and Lie Groups

1.1 Lie Groups
A Lie Group is a group that is also a finite dimensional differentiable manifold. The
group operations are smooth functions. A continuous group is a group where continuity is
imposed on the elements of the group in the sense that a small change in one of the factors
of a product produces a correspondingly small change in their product.
Lie algebra operates on the tangent space of the Lie group at the identity element and
this algebra captures the local structure of the Lie Group. Elements of the Lie algebra are
elements of the group that are “infinitesimally close” to the identity, and the Lie bracket
of the Lie algebra is related to the commutator of two such infinitesimal elements.
An r parameter group on an n dimensional space (x) gives transformations

x0 = f (x, a)

2
where a is the r parameters. We desire that for a a vector of zeros, we have the identity

x = f (x, 0, 0, 0....)

Infinitesimal generators can be computed as


∂f 
Li =
∂ai ai =0

where evaluate at aj = 0 for j 6= i. The above f is a matrix or an operator and L has the
dimension of f . Taking into account that f gives a vector we can also write this as

∂fj  ∂
Li = (1)
∂ai ai =0 ∂xj


(using summation notation for j) where ∂x∂ j gives a basis for the tangent space, and these
can be used to operate on any function.
Above we have written Li as a differential operator. It operates on a function of x.
With g(x) a function, then Li g(x) is a number so Li g(x) is a function of x. Using two
different differential operators Li , Lj we can operate sequentially on a function Li (Lj g(x)).
Let us consider two vector operators and write them more simply as
∂
A = ai
∂xi
∂
B = bj
∂xj
where ai , bi are functions of x and we have used summation notation. Our generators Li , Lj
are in this form.
∂ ∂
ABg = ai bj g
∂xi ∂xj
∂2g ∂bj ∂g
= ai bj + ai
∂xi xj ∂xi ∂xj
∂2 ∂bj ∂
AB = ai bj + ai
∂xi ∂xj ∂xi ∂xj

If we take the commutator

∂2 ∂bj ∂ ∂2 ∂ai ∂
[A, B] = ai bj + ai − ai bj − bj
∂xi ∂xj ∂xi ∂xj ∂xi ∂xj ∂xj ∂xi
 
∂bj ∂aj ∂
= ai − bi
∂xi ∂xi ∂xj

3
We notice that the second derivatives cancel. That means that using the commutator our
infinitesimal generators can operate on each other and give us another generator.
The infinitesimal operators form a Lie algebra with commutators (called Lie brackets)
that satisfy
[Li , Lj ] = ckij Lk
with ckij called structure constants. The commutator is a binary operator that satisfies the
Jacobi identity, is antisymmetric and is bilinear w.r.t to multiplication with elements of a
field (typically real or complex numbers).
On a matrix Lie group, we can define an exponential map for any operator in the Lie
algebra, giving us a map between elements of the Lie algebra and the Lie Group. For u in
the Lie algebra
2 3
eu = 1 + u + u2 + u3 ....
2 3!
and this gives an element in the Lie group. In this sense, elements of the Lie algebra are
infinitesimal generators for the group.

u2 v2
eu ev = (1 + u + ....)(1 + v + ....)
2 2
1 1
e(u+v) = 1 + u + v + (u + v)2 .... = 1 + u + v + (u2 + v 2 + uv + vu)....
2 2
We notice that
eu ev 6= eu+v
Locally  
u v 1 1 1
e e = exp u + v + [u, v] + [[u, v], v] − [[u, v], u]...
2 12 12
known as the Baker-Campbell-Hausdorff formula.
Consider two near identity operators A, B and alternating the order that the transfor-
mations are done (see Figure 1 and 7). To second order

A2 B2 A2 B2
    
−A −B A B
e e e e ≈ I −A+ I −B+ I +A+ I +B+
2 2 2 2
A 2 A2 B 2 B 2
= I −A−B+A+B+ + + + + AB − AB − BA + AB − A2 − B 2
2 2 2 2
= I + [A, B]
∼ e[A,B]

The commutator helps us compute what happens when infinitesimal transformation don’t
commute.

4
 Figure 1: Alternating infinitesimal transformations need not commute.

1.2 Rotation Example

Rotation matrices in 3 dimensions form a Lie group. We define the special orthogonal
group SO(3) as
A ∈ R3 , such that detA = 1, AT A = AAT = I
Matrices A operate on x a 3D Cartesian coordinate. Rotation by angle φ about the z axis
 
cos φ − sin φ 0
R(φ) =  sin φ cos φ 0  (2)
0 0 1

Near the identity

   
 − sin φ − cos φ 0  0 −1 0
dR(φ) 

=  cos φ − sin φ 0  = 1 0 0  (3)
dφ φ→0
0 0 0 
φ=0
0 0 0

is the generator for R(φ). Expanding R(φ) for small φ (to first order in φ)
 
1 −φ 0
R(φ) ∼  φ 1 0  = I + φL̂z (4)
0 0 1

with  
0 −1 0
L̂z =  1 0 0 
0 0 0

5
a matrix describing the infinitesimal rotation about the z axis. We can similarly find
 
0 0 1
L̂y =  0 0 0 
−1 0 0
 
0 0 0
L̂x =  0 0 −1 
0 1 0
for rotations about x and y axes. For a small rotation φ about z

x0 = (I + φL̂z )x
 0        
x 0 −1 0 x x − φy
 y 0  = I + φ  1 0 0   y  =  y + φx 
z 0 0 0 0 z z
Suppose we have a function
F (x, y, z)
and we transform it for a small φ.
∂F ∂F
F (x0 , y 0 , z 0 ) = F (x − yφ, y + xφ, z) = F (x, y, z) − y φ+x φ
∂x ∂y
The change in F  
∂F ∂F
∆F = −y +x φ
∂x ∂y
We associate the operator
Lz = −y∂x + x∂y (5)
with our infinitesimal transformation. And this is a vector corresponding to the z compo-
nent of angular momentum. So if the coordinate system is rotated by a small amount φ
about the z axis we can compute the change to any function of the coordinates with

∆F = Lz F φ

where Lz is the differential operator as defined in equation 5

Now let us find the operator using equation 1. Our function
 0   
x x
f (x, y, z) =  y 0  = R(φ)  y 
z 0 z
 
x cos φ − y sin φ
=  x sin φ + y cos φ 
z

6
  
−x sin φ − y cos φ
∂f
=  x cos φ − y sin φ 
∂φ
0
 
 −y
∂f 
=  x 
∂φ φ=0
0
 
∂fx  ∂fy 
Lz = ∂x + ∂y
∂φ φ=0 ∂φ φ=0
= −y∂x + x∂y
consistent with what we have above in equation 5.
This example illustrates two way of describing the infinitesimal operators. One way
with matrices. The other way with differential operators. In both cases we can think of
the infinitesimal operators as lying in the tangent space of the manifold. These operators
also describe the local structure of the Lie group of symmetry transformations.

1.3 Example: Locomotion of Deformable Jello Cubes

Locomotion strategies of small biological organisms might inspire strategies for achiev-
ing locomotion in small artificial mechanisms. Self-propulsion forces can arise from the
mechanical dissipative interactions of the locomotor with a surrounding fluid, or with a
solid substrate. The most common motility modes at microscopic scales are crawling and
swimming. Series of motions of minimal complexity that can give net displacements often
involve periodic shape changes. Discrete-mass crawling strategies require either many de-
grees of freedom (Zimmerman et al., 2004, 2007, 2009; Bolotnik et al., 2011), anisotropic
friction coefficients (Zimmerman et al., 2004), or achieve motion by alternately sticking
and sliding (Chernous’ko, 2002 ). Locomotion at low Reynolds number is not possible if
the body can only deform with one degree of freedom and this has to do with time reversal
symmetry (Purcell 1977).
There are a bunch of possible mechanical mechanisms that give locomotion with a single
degree of freedom but these involve coupling motion mechanically and asymmetry in the
body motion is achieved through the coupling. Another way to do this is have asymmetric
friction coefficients on two masses and in this case motion is achieved with non-sinusoidal
oscillations in time (Wagner and Lauga 2013). For the Zhou paper, the asymmetry is caused
by angle of the spring between two legs. For systems with a single degree of freedom, we
cannot study it with infinitesimal transformations! Only for the two degree of freedom
system with non-commuting motions, are the infinitesimal transformations relevant?

7
Figure 2: Two masses separated by an actuator that can change its length. If the two
masses have bases with different frictional coefficients and the actuator’s length varies
periodically but not sinusoidally, then the scallop can inch forward. A model explored by
Wagner and Lauga 2013.

Figure 3: The green spring represents an actuator and the green arc an angular spring. By
varying the length of the actuator the horse can be made to jump forwards. Based on the
model by Zhou et al. 2014.

Figure 4: A bristle crawler by Noselli and Desimone. The two legs are flexible and their
angle changes with respect to the rough surface.

8
 Figure 5: Variables describing body state for the jello parallelogram.

1.4 The Jello cube!

As an example, we consider a jello cube that allows deformations while maintaining its
density. We assume the body is incompressible but easy to deform. We consider two types
of deformations, shear and stretching the cube vertically, see Figure 6. Each correspond to
a linear deformation gradient and no body rotation. We place the jello cube on a flat surface
and allow some friction between the jello cube and the surface. We allow transformations
only in x, y so the system is effectively two-dimensional. The direction x is along the flat
surface and y is vertical height above the surface. When under no stress or strain, the
jello is a cube with length L0 and area A = L20 . We also allow body translations in the x
direction.
When sheared or stretched the jello surface deforms from a cube into a parallelogram.
The configuration space for our jello cube has three dimensions, one for the x coordinate
of the center of mass, one for the angle of the lower leftmost corner of the parallelogram,
and the ratio of length of its sides. We describe configuration spaces with, the x position of
the jello’s center of mass, and we refer to this as x̄, the length of the base L and f = cot θ
with θ the angle between two edges (see Figure 6).
Stretches we describe with the transformation Dα

Dα :
L0 = L(1 + α)
x0 = x
f0 = f (6)

9
The linear transformation of the body itself is chosen to maintain the parallelogram angles.
We assume the transformation does not shift the center of mass. With friction on the base,
this is equivalent to assuming that the transformation is done quickly enough that friction
at the base does not significantly move the center of mass.
Slow shear transformations we describe with

Sβ :
L0 = L
βA
x0 = x +
2L
f0 = f +β (7)

Here A = L20 is the area of the parallelogram face. The length coordinate is preserved but
the angle θ changes. We assume that the transformation is done slowly enough that the
base does not shift during the deformation and as a consequence the transformation causes
a shift in the the horizontal location of the center of mass.
Translations are described with

Tc :
L0 = L
x0 = x + c
f0 = f (8)

With α = 0, β = 0 and c = 0 we have identity transformations.

Let total configuration space Q is described with coordinates

q = (x, L, f )

and our array of parameters for coordinate transformations

a = (c, α, β)

Our three transformations together generate a Lie group of transformations

q0 = G(q, a)

Our transformation functions (the three expressions 6,7, 8) are elements in the group G.
Infinitesimal generators (as operators) or elements in the Lie algebra of G at point q can
be computed using 

X ∂Gj  ∂
Li = 
∂ai  ∂qj
j
ai =0

10
Figure 6: Volume preserving deformations of a jello cube with a frictional base. We assume
that the center of mass stays fixed when the jello is stretched vertically. We assume that
the base is held fixed by friction when the jello is is sheared.

11
Figure 7: The deformations don’t commute and their consecutive application translates
the center of mass. A slow shear followed by a stretch, followed by the opposite slow shear
and the opposite stretch causes a translation. The jello cube marches forward.

12
We compute the derivatives
 
 L
∂Dα (q) 
= 0  (9)
∂α α=0
0

 
 0
∂Sβ (q)  A 
=  2L (10)
∂β β=0
1

 
 0
∂Tc (q)   1 
= (11)
∂c c=0
0

For the stretches, shears and translations, respectively, the infinitesimal operators at posi-
tion q = (x, L, f ) are

LD = L∂L (12)
A
LS = ∂x + ∂f (13)
2L
LT = ∂x (14)

The space of translations along x is a sub-algebra. The infinitesimal operators are also
vectors in the tangent space of the manifold Q at point q. From the infinitesimal operators
we compute the commutator
A A
[LD , LS ] = LD LS − LS LD = − ∂x = − LT (15)
2L 2L
The operators don’t commute. The commutator gives translation of the center of mass
caused by a series of body transformations. If the operations are done in the order stretch,
shear, inverse stretch, inverse shear, our jello parallelogram should march forward in the
horizontal direction.
The other commutators
[LD , LT ] = [LS , LT ] = 0
The Lie algebra is known as the three-dimensional Heisenberg algebra and it has Bianchi
classification type II.

13
1.5 Mechanical Connection
We can think of configuration space as a principal bundle with full space Q = H × M
and H, the fibre that is also the group of translations of the center of mass, and M the
manifold of body shapes (R2 with coordinates (L, f )). A principal bundle also requires a
projection map π : Q → M that is probably (x, L, f ) → (L, f ).
Body motions are constrained by friction to lie in a horizontal subspace of T Q that is
spanned by the two vectors LD , LS .
A
LD = L∂L LS = ∂x + ∂f (16)
2L
from equation 14. Locally the tangent space of Q can be decomposed into horizontal and
vertical subspaces.
A mechanical connection, Γ is a 1 form (operating on a vector in the tangent space of
Q) that is also a function of elements in the Lie algebra of H. The mechanical connection
sends to zero (projects away) anything in the horizontal subspace. The horizontal sub-
space is that of the body deformation. The vertical subspace is that associated with body
transformation. A mechanical connection is a one form that gives us elements in the Lie
algebra of H, does not change vertical vectors in Q and is zero when applied to vectors in
the horizontal subspace (here spanned by equation 16). We require that the connection Γ
applied to LT give us LT = ∂x and Γ applied to LD and LS give us zero. The mechanical
connection that achieves this is
 
A
Γ = dx − df ∂x (17)
2L

We can check this with

 
A
ΓLT = dx − df ∂x on ∂x
2L
 
A
= dx∂x − df ∂x ∂x
2L
= (1 − 0) ∂x
= ∂x

 
A
ΓLD = dx − df ∂x on L∂L
2L
A
= −(df L∂L )∂x
2L
= 0

14
  
A A
ΓLS = dx − df ∂x on ∂x + ∂f
2L 2L
 
A A
= − ∂x
2L 2L
= 0

Changes in the jello cube body space happen along curves that have tangents specified
by LS and LD . This means they lie in a plane spanned by vectors that are sent to zero by
Γ. A curve that remains in the horizontal space can return to the same f, L coordinates.
But integrating around this loop gives
I
∆x = Γds

Using Stokes’s law we can write the same integral as

Z
∆x = dΓdA

where dΓ is a curvature.
The curvature of the connection is

Ω(u, v) = dΓ(hu, hv)

where hu is the horizontal lift of vector u and hv are the horizontal lifts of vector u, v ∈ Tq Q.
Here h essential projects a vector in Tq Q to the tangent space of M (horizontal projection!).
The curvature can be computed from the exterior derivative of the connection. Inte-
grating the connection about a loop is equivalent to integrating the curvature over the area
inside the loop.
With a trajectory given in the horizontal space, the connection can be integrated to
find out how big a change takes place in the vertical space. A loop in the horizontal space
may not close in the vertical space. Integrating the curvature over the area within the loop
in the horizontal space also gives the change that takes place in the vertical space.
More specifically, we write our connection again, (equation 17)
 
A
Γ = dx − df ∂x (18)
2L
We take the exterior derivative of Γ
 
A A
dΓ = −( + 1)dx ∧ df + dL ∧ dx + dL ∧ df (19)
2L 2L2
Restricting to the horizontal subspace for L, f we get a curvature
A
Ω= dL ∧ df
2L2

15
We consider a loop in L, f space and use Stokes theorem
I Z Z
Γds = dΓ = Ω
A A
Integrating the connection over the loop is equivalent to integrating the curvature over the
area inside the loop.

1.6 A fibre bundle and a Connection

A fibre bundle is a space Q that is locally a product of a base space B and a fibre. A fibre
bundle must have a projection map π : Q → B that shows how to take a point in Q and
project it to the base space. The inverse of the map π −1 (b) gives points in the fibre that
project to a point b ∈ B. The fibre has a structure group G giving homeomorphisms of
the fibre. A principal bundle has fibres homeomorphic to the structure group G. A vector
bundle has fibres that are homeomorphic to a vector space. There is also a condition on
coverings of the base space B by open sets.
For our jello cube, the fibre is the background space giving the center of mass position.
The base space describes the body stretches and tilts.
An Ehresman connection A is a vector valued one form on Q.
Aq : Tq Q → Vq
for q ∈ Q and vertical space Vq the fibre, here and x value. The map is linear and gives a
projection
A(vq ) = vq
for all vq ∈ Vq . If we locally define coordinates q i = (rα , sa ) for base and fiber then
∂
A = ωa for ω a (q) = dsa + Aaα (r, s)dsα .
∂sa
It is common to divide the tangent space Tq Q into horizontal and vertical spaces, where
the horizontal space is projected to zero by the connection and the vertical space is not
changed by the connection.
The mechanical connection above Γ is in the form of A, but it is not affine as it is not
linear in L.
With point q ∈ Q we can project it into the base space r = π(q) ∈ B. We now consider
a vector in the base space at point r, or vr ∈ Tr B. We can find a unique vector in the
tangent space of the whole space Tq Q which we call vrh ∈ Tq Q, that projects to vr when
projected by π. We call this new vector the horizontal lift of vector vr ∈ Tr B. We will use
the connection to determine vrh . Suppose we take any vector X ∈ Tq Q and compute
horXq = Xq − A(q)Xq
The connection removes the part that is in the tangent space of the base. What is left is
the part that is in the tangent space of the fibre.

16
2 Rigid Bodies

Figure 8: Velocity of a particle at position r if spinning about an axis Ω.

We take our rigid body to be a distribution of particles each with coordinate r with
respect to the center of mass of the body. We let the body spin about an axis Ω. The
velocity of a single particle with mass mi , position ri and velocity vi
vi = Ω × r i
The angular momentum of this particle (measured from the origin) is
Li = mi ri × vi = mi ri × (Ω × ri )
We recall a vector identity
A × (B × C) = (A · C)B − (A · B)C
we can write the angular momentum of the particle
Li = mi ri2 Ω − (ri · Ω)ri

Integrating this up over the entire body the entire angular momentum
Z
L = ρ(r) r2 Ω − (r · Ω)r dV

Using summation notation (and now the indices represent coordinate directions, not a
single particle)
Z
Lj = ρ(r) (xi xi Ωj − xj xk Ωk ) dV
Z
= ρ(r (xi xi δjk − xj xk ) dV Ωk
= Ijk Ωk

17
with moment of Inertia tensor
Z
Ijk ≡ ρ(r) (xi xi δjk − xj xk ) dV

Question Suppose we have a particle distribution and pick a spin value and axis Ω. Each
particle has coordinate xi with respect to the center of mass of the body. We set the initial
velocity vi of each particle i at position xi in the distribution equal to

vi = Ω × xi

(see Figure 8). This question is relevant for generating initial conditions for spinning body
simulations.
Q: Is the initial angular momentum parallel to Ω?
A. The answer is usually no. Angular momentum for a particle is xi ×vi = xi ×(Ω×xi )
and so in general contains a component that is perpendicular to Ω. After summing over
all particles the parts perpendicular to Ω do not cancel unless the particle distribution has
a lot of symmetry.
Q: How do you set the velocity vector of each particle to ensure that the total angular
momentum is a chosen vector L?
A. Compute the moment of inertia tensor from the particle distribution and then com-
pute its inverse. Compute Ω = I−1 L. Use this Ω to set the particle velocities. Because
I is not necessarily diagonal and L not necessarily along a principal axis, L need not be
parallel to Ω.

Rigid body rotation is confusing because two coordinate frames are involved, the inertial
frame and the body frame. The motion of the rigid body is observed in the inertial frame
whereas it is simpler to calculate the equations of motion in the body-fixed principal axis
frame. The body rotes about the angular velocity vector Ω but this is not in the same
direction as the angular momentum vector L. Euler’s equations are given in the body-fixed
frame in which the moment of inertial tensor is diagonal and is constant. The solution then
must be rotated back into the inertial frame to describe the rotational motion as seen by
an external observer.

2.1 Configuration space

A rigid body can be viewed as a group of point masses all constrained so that their positions
with respect to each other does not vary. In a center of mass frame the only thing that varies
is the orientation of the body. The configuration space is an element of the rotation
group or SO(3). We can think of rigid body motion as a trajectory in the Lie group SO(3)
or a trajectory in the group of rotations. If we describe rotations with a rotation matrix

18
R the trajectory of a rigid body (w.r.t to its center of mass) is R(t). Neglecting external
forces the total energy is equal to the kinetic energy. A Lagrangian description depends on
the kinetic energy with coordinate describing the body orientation at a particular time and
time derivative depending on the time derivative of body orientation which is equivalent
to its spin.

2.2 Generators of Rotation Matrices

The position of a particle x in a rotating frame is

x = Rx0

where R is a rotation matrix relating two reference frames. We can take x0 to be in the
body’s frame and x to be in an inertial frame.
Using the infinitesimal generators from the Lie algebra of the rotation group (SO(3))

R = eα·l = eαi li (20)

where l is comprised of three matrices, lx , ly , lz . The vector α gives three components

(numbers) or angles, one for each of the matrices lx , ly , lz . The matrices lx , ly , lz are the
infinitesimal generators. I am using lower case l so as not to be confused with angular
momentum.
 
0 0 0
lx =  0 0 −1 
0 1 0
 
0 0 1
ly =  0 0 0 
−1 0 0
 
0 −1 0
lz =  1 0 0  (21)
0 0 0

It may be convenient for to write the indices for the matrix li

li,αβ = −iαβ (22)

where i, α, β are coordinate indices. Let us check this

(lx x)α = −xαβ xβ = (0, −z, y)

(ly x)α = −yαβ xβ = (z, 0, −x)
(lz x)α = −zαβ xβ = (−y, x, 0)

19
and these are consistent with equation 21. More generally [lk x] has α component

[lk x]α = −kαβ xβ (23)

We use these generators to compute the velocity of a particle in the rotating frame
d
ẋ = (Rx0 ) = Ṙx0
dt
d αi li d
= e x0 = (αi li )eαj lj x0
dt dt
= (α̇ · l)Rx0
v = Ωi li x = (Ω · l)x

where we associate α̇ with Ω. Writing out the dot product (using equation 23)
 
0 −Ωz Ωy
Ωi li x = (Ω · l)x =  Ωz 0 −Ωx  x = Ω × x
−Ωy Ωx 0

Using slightly different notation

∂T (Rx0 ) = ∂T (eαi li x0 )
= Ωi li eαj lj x0
= Ωi li Rx0
= Ω i li x
∂
with x = Rx0 and ∂T = ∂t . Note that li x has components

(li x)u = −iuv xv

so

∂T (Rx0 )u = −Ωi iuv xv (24)

or

∂T (Rx0 ) = Ω × x = Ω × Rx0 (25)

as expected.
We took a time derivative of coordinate x but we could have done this for any static
vector. Suppose L0 is a static vector.

L = RL0

20
If the vector L0 in one frame is constant then
L̇ = ṘL0 = Ω × L
dL
where L̇ = dt . If our vector L0 is not constant then we would get
L̇ = ṘL0 + RL̇0 = Ω × L + ∂t L (26)
∂ d
where ∂t = ∂t which is not the same as dt .
This looks like an advective derivative. It also
looks a lot like Euler’s equations! In fact if we chose L0 to be in a frame fixed with the
body (for example aligned with the body’s principal axes ) then ∂t L0 = IΩ̇. This follows
as I is constant in this frame. Here L̇ is the torque and we find
I Ω̇ + Ω × (IΩ) = N (27)
with N the applied torque. This gives Euler’s equations.
We take L0 in a frame aligned with a body’s principal axes where the the moment of
inertia matrix is I = diag(Ix , Iy , Iz ),
L0,x = Ix Ωx
L0,y = Iy Ωy
L0,z = Iz Ωz
Equation 27 is
Ix Ω̇x + (Iz − Iy )Ωy Ωz = Nx
Iy Ω̇y + (Ix − Iz )Ωz Ωx = Ny
Iz Ω̇z + (Ix − Iy )Ωy Ωx = Nz

2.3 The kinetic energy

The kinetic energy depends on the square of the velocity
Z Z
dV dV
T = ρ (∂T (Rx)) = ρ (Ṙx)2
2 2
Z
dV
= ρ ((Ωi · li )x)2
2
Z
dV
= ρ (Ω × y)2
2
where x is in the body frame and y = Rx. Writing this out in index form
(Ω × y)2 = ijk Ωj yk ilm Ωl ym
= (δjl δkm − δjm δkl )Ωj yk Ωl ym
= Ω2 y2 − (Ω · y)2
= Ωi δik yj yj Ωk − Ωi yi yk Ωk
= Ωi (y2 δik − yj yk )Ωk

21
Sticking this back into the integral we recognize that
1
T = ΩIΩ
2
where spin Ω is a vector but I is the moment of inertia tensor.

2.4 Example for the Moment of Inertia tensor

An axisymmetric body has moment of inertia tensor with two eigenvalues that are the
same. If the body is prolate (like a cigar or football) then principal moments I1 = I2 > I3 .
If the body is oblate (like a pancake or a disk) then I1 > I2 = I3 .
We compute the moment of inertia tensor for a dumbbell, comprised of two unequal
masses m1 , m2 , separated rigidly by a distance a on the x axis. We ignore the mass of the
bar holding the masses apart. In a coordinate system with the center of mass at the origin,
the coordinate positions are
m2
x1 = a
m1 + m2
m1
x2 = −a
m1 + m2
We compute the moment of inertia tensor
Ixx = m1 (y12 + z12 ) + m2 (y22 + z22 ) = 0
Iyy = m1 (x21 + z12 ) + m2 (x22 + z22 ) = m1 x21 + m2 x22
m22 m21 m1 m2 2
= m1 a2 + m 2 a 2
= a
(m1 + m2 )2 (m1 + m2 )2 m1 + m2
m1 m2
By symmetry Izz = Iyy . Using the reduced mass µ = m1 +m2

Iyy = Izz = µa2

and is independent of anything but the reduced mass.
Useful to know for modeling lopsided rapidly spinning satellites.

2.5 Stability of a Freely Rotating Body

Starting with the Euler equations
∂L
+ω×L=0
∂t
In a coordinate system aligned with principle axes of the body
I1 ω̇1 = (I2 − I3 )ω2 ω3
I2 ω̇2 = (I3 − I1 )ω3 ω1
I3 ω̇3 = (I1 − I2 )ω1 ω2 (28)

22
where I1 , I2 , I3 are the moments of inertia about the principal axes (eigenvalues of moment
of inertia matrix) and ω1 , ω2 , ω3 are the components of the spin vector in the principle axes
directions (eigenvector directions of the moment of inertia matrix). We can label the axes
so that I1 > I2 > I3 and start the body rotating nearly about direction x̂2 with
ω = α1 x̂1 + sx̂2 + α3 x̂3
with s large and α1 , α2  1 very small perturbations. We put this into Euler’s equations
(equations 28)
(I2 − I3 )
ω̇1 = sα3
I1
(I3 − I1 )
ω̇2 = α1 α3 ∼ 0
I2
(I1 − I2 )
ω̇3 = sα1
I3
The middle equation is second order in small quantities so ω̇2 ∼ 0. If we allow α1 , α2 to be
functions of time
(I2 − I3 )
α̇1 = sα3
I1
(I1 − I2 )
α̇3 = sα1
I3
and putting this together
(I2 − I3 ) (I1 − I2 ) 2
α̈1 = s α1
I1 I3
(I2 − I3 ) (I1 − I2 ) 2
α̈3 = s α3
I1 I3
If the factor on the right is positive then the solutions are exponential growing or shrinking.
If the factor on the right is negative then the solutions are oscillating. If the quantity
(I2 − I3 )(I1 − I2 )
is positive then we can have exponentially growing values for ω1 , ω2 . Since I1 > I2 > I3
this quantity is positive and the system is unstable and will quickly start to spin about a
different direction than started.
Redoing this analysis starting with spin about x̂1 the relevant quantity that must
be negative for stability is (I1 − I3 )(I2 − I1 ) and this is negative with our order for the
moments. Redoing the analysis starting with spin about x̂3 the relevant quantity that must
be negative for stability is (I3 − I1 )(I2 − I3 ), again negative with our ordering.
The body can stably spin about the axis of smallest moment or about the axis of the
largest moment but not the intermediate one.

23
2.6 Rigid Body motion from a Lagrangian
We notice that the kinetic energy is in fact a Lagrangian density
Z
dV 1
L=T = ρ (∂T (Rx))2 = ΩIΩ (29)
2 2
where the kinetic energy is integrated over the body volume. Again x is in the body frame
and Rx is in the inertial frame. The Lagrangian itself is integrated over a path
Z t2
L= L(t)dt
t1

with path R(t) specified by how the rotation varies in time. The rotation matrices are
functions of angles, α that specify body orientation. Each rotation is a point inside SO(3).
The body rotations so α(t), these angles are functions of time. The time derivatives of
these angles are α̇ = Ω a spin vector. The spin vector is a member of the tangent space of
a point in SO(3). Our Lagrangian density is a function

L(α, α̇) = L(α, Ω).

Taking the angles and their time derivatives we can compute Lagrange’s equations.
∂L
= IΩ (30)
∂Ω
which is the angular momentum. Using equation 29
d ∂L
= L̇ = IΩ̇ (31)
dt ∂Ω
Now for the hard part, taking the derivative of the Lagrangian with respect to the
angles for the rotation (the coordinate in SO(3)). Using equation 20 to write the rotation
matrix in terms of α
∂ ∂ αj lj
Rx = e x
∂αk ∂αk
= lk eαj lj x
= lk Rx
= lk y (32)

with y = Rx. Using equation 23 components

   
∂ ∂
Rx = y
∂αk β ∂αk β
= −kβγ yγ (33)

24
so if we dot this with another vector z
∂y
z· = −kβγ yγ zβ
∂α
= y×z

Recall that the velocity

v = ∂T (Rx)
= Ω × Rx
= Ω×y

The kinetic energy at a particular position x

1 1
[∂T (Rx)]2 = (Ω × y)2
2 2
1 2 2
Ω y − (Ω · y)2


=
2
Taking the derivative with respect to α

∂ v2
= Ω2 y · (lk y) − (Ω · y)Ωi lk yi
∂αk 2
∂ v2
= Ω2 y × y − (Ω · y)(y × Ω)
∂α 2
= (Ω · y)(Ω × y)
= (Ω · y)v (34)

Compare this to an expression for angular momentum at a single location

Ω × L(y) = Ω × (y × v)
= (Ω · v)y − (Ω · y)v
= −(Ω · y)v (35)

where I used a vector identity and

Ω · v = Ω · (Ω × y) = 0.

Above I have computed things for a single location x in the body.

Putting equation 34, 35 together and integrating over the volume we get
∂ 1
[∂T (Rx)]2 = −Ω × L(y)
Z ∂α 2
∂ 1
dV [∂T (Rx)]2 = −Ω × L
∂α 2

25
 ∂L
= −Ω × L (36)
∂α
This and equation 31 with Lagrange’s equations give

L̇ + Ω × L = 0.

Using the moment of inertia tensor in a frame fixed with the body we get Euler’s equations.

I Ω̇ + Ω × IΩ = 0.

2.7 A Hamiltonian, a moment map

Looking at equation 30 or ∂∂L
α̇ = L we can construct a Hamiltonian from the Lagrangian
using the angular momentum vector as a set of canonical momentum variables. More
abstractly the action
aM (q, p) = (aq, ap)
where a ∈SO(3) is a 3x3 matrix. We create a symplectic 2-form with ω = dθ where
X
dθ = pi dqi = hp, dqi
i

The action preserves the symplectic structure

aM θ = hap, d(aq)i = hap, adqi = hp, dqi = θ

The moment map

µA (q, p) = hp, Aqi
where A is in the Lie algebra of SO(3). For SO(3) there is a nice bijective correspondence
between elements of the Lie algebra and spin axes A → Ω with

Av = Ω × v

identifying elements of the Lie algebra with spin vectors.

The moment map is essentially the Hamiltonian. This construction could be used for
other Lie groups. The moment map can be used to generate geodesics and a notion of
curvature.

26
Figure 9: Euler angles. The green line is the intersection of the i, j and I, J planes and is
called a line of nodes with direction n̂. It is computed from the vector K × k. The angle
θ is a nutation angle. The angle φ is a spin precession angle (as seen from the external or
fixed reference frame). ψ is sometimes called a precession angle as seen from the body. ψ̇
rotates the ij axes in the ij plane with respect to a fixed line of nodes. φ̇ rotates the line
of nodes n̂.

27
Figure 10: A toy plane! Practical uses for Euler angles include study of plane stability.

28
3 Rotational Dynamics
3.1 Euler angles
Following Celleti’s book:
We start with a fixed point O and a reference frame i, j, k that coincides with a body’s
principle axes (O, i, j, k). Now we chose an inertial frame with the same origin and three
different axes (O, I, J, K). Let n be the line of nodes defined by the intersection of the i, j
and I, J planes. The direction
K×k
n̂ =
|K × k|
The Euler angles θ, φ, ψ are as follows:
• 0 ≤ θ ≤ π is the nutation angle and is that between K and k axes.
• 0 ≤ ψ ≤ 2π is the proper rotation angle and is that between n̂ and i axis.
• 0 ≤ φ ≤ 2π is the precession angle and is that between n̂ and I axis.
The angle ψ is also sometimes called a precession angle as it can describe the precession
of the spin and angular momentum vectors in the body’s frame.
We need to relate variations in the Euler angles to the spin in the body frame.
ωi = φ̇ sin θ sin ψ + θ̇ cos ψ
ωj = φ̇ sin θ cos ψ − θ̇ sin ψ
ωk = φ̇ cos θ + ψ̇ (37)
The three rotations described below allow us to derive these expressions.
Looking at figure 9 we examine the angles. With θ, φ constant ψ̇ gives a rotation of
the i, j plane with respect to the line of nodes. The body is essentially spinning around
the k body axis. The angle φ moves the line of nodes with respect to the inertial frame. If
the body spin is oriented along k we can look at its projection into the I, J plane. If the
projected spin axis advances we say the body spin precesses. Taking k to be the spin axis,
the angle between it and the K axis we refer to as a nutation angle, θ.

• The angle θ is that between k (body principal axis) and K (inertial frame axis).
cos θ = k · K.

• The angle φ is that between line of nodes n̂ and inertial axis I

cos φ = I · n̂.

• The angle ψ is that between the body principal axis i and the line of nodes n̂
cos ψ = i · n̂.

29
3.1.1 Rotations
Starting in the inertial frame (xyz), three rotations transform into the body frame (x0 y 0 z 0 ).
1. A counter-clockwise rotation by φ about K or the current z axis.
2. A counter-clockwise rotation by θ about the current x axis (here the x000 axis).
3. A counter-clockwise rotation by ψ about the current z axis (here the z 0 or z 000 axis).
First rotate by the precession angle φ then tilt by the nutation angle θ then rotate
by ψ.
We start with xyz in inertial frame, rotate by φ giving x00 , y 00 , z 00 , rotate by θ giving
x000 , y 000 , z 000 then rotate by ψ giving x0 , y 0 , z 0 in body frame.
 00    
x cos φ sin φ 0 x
 y 00  =  − sin φ cos φ 0   y 
z 00 0 0 1 z

x000
     00 
1 0 0 x
 y 000  =  0 cos θ sin θ   y 00 
z 000 0 − sin θ cos θ z 00

x0
     000 
cos ψ sin ψ 0 x
 y 0  =  − sin ψ cos ψ 0   y 000 
z0 0 0 1 z 000
Angular rotation rates arise from time derivatives of each rotation;

ωφ = φ̇ẑ
ωθ = θ̇x̂00
ωψ = ψ̇ẑ0 (38)

To evaluate these we need to recall

ẑ = sin ψ sin θx̂0 + cos ψ sin θŷ0 + cos θẑ0

x̂00 = cos ψx̂0 − sin ψŷ0 (39)

Combining these (equations 38 and 39) together we get equations 37 for the body frame
(x0 , y0 , z0 ) repeated here

ωi = ωx0 = φ̇ sin θ sin ψ + θ̇ cos ψ

ωj = ωy0 = φ̇ sin θ cos ψ − θ̇ sin ψ
ωk = ωz 0 = φ̇ cos θ + ψ̇ (40)

30
 All together the three rotations give this transformation

x0 = (cos ψ cos φ − sin ψ sin φ cos θ)x + (cos ψ sin φ + sin ψ cos φ cos θ)y + sin ψ sin θz
y 0 = −(sin ψ cos φ + cos ψ sin φ cos θ)x + (− sin ψ sin φ + cos ψ cos φ cos θ)y + cos ψ sin θz
z 0 = sin θ sin φx − sin θ cos φy + cos θz, (41)

and the inverse transformation is

x = (cos φ cos ψ − sin φ sin ψ cos θ)x0 − (cos φ sin ψ + sin φ cos ψ cos θ)y 0 + sin φ sin θz 0
y = (sin φ cos ψ + cos φ sin ψ cos θ)x0 + (− sin φ sin ψ + cos φ cos ψ cos θ)y 0 − cos φ sin θz 0
z = sin θ sin ψx0 + sin θ cos ψy 0 + cos θz 0 . (42)

• The angle θ gives the angle between z, z 0 axes.

• Projecting inertial z axis into the body x0 , y 0 plane gives ψ, with

ψ = atan2(z · x0 , z · y0 ) = atan2(K · i, K · j).

• Likewise projecting body z 0 axis into the inertial x, y plane gives φ, with

φ = atan2(z0 · x, −z0 · y) = atan2(k · I, −k · J).

We use the function atan2 so that the quadrant for the angle is specified.

3.2 Lagrangian and Hamiltonian in terms of Euler angles

The Lagrangian in terms of the Euler angles
1
I1 ω12 + I2 ω22 + I3 ω13 − V (θ, φ, ψ)


L(θ̇, φ̇, ψ̇, θ, φ, ψ) =
2
1 1
= I1 (φ̇ sin θ sin ψ + θ̇ cos ψ)2 + I2 (φ̇ sin θ cos ψ − θ̇ sin ψ)2 +
2 2
1 2
I3 (φ̇ cos θ + ψ̇) − V (θ, φ, ψ)
2
and this simplifies if two of the moments are the same.
Taking derivatives of the Lagrangian (with respect to φ̇, θ̇, ψ̇) we can compute canonical
momenta;

pφ = (I1 ω1 sin ψ + I2 ω2 cos ψ) sin θ + I3 ω3 cos θ

pθ = I1 ω1 cos ψ − I2 ω2 sin ψ
pψ = I3 ω3 (43)

31
or

pφ = φ̇ (I1 sin2 ψ + I2 cos2 ψ) sin2 θ + I3 cos2 θ

 

+θ̇(I1 − I2 ) cos ψ sin ψ sin θ

+ψ̇I3 cos θ
pθ = φ̇(I1 − I2 ) sin ψ cos ψ sin θ
+θ̇(I1 cos2 ψ + I2 sin2 ψ)
pψ = I3 (φ̇ cos θ + ψ̇) (44)

With these canonical momenta we can construct a Hamiltonian from the kinetic energy
term and using Euler angles and their associated momenta as canonical variables

H(φ, θ, ψ; pφ , pθ , pψ )

With some manipulation of equations 43

(pφ − pψ cos θ)
I1 ω1 = sin ψ + pθ cos ψ
sin θ
(pφ − pψ cos θ)
I2 ω2 = cos ψ − pθ sin ψ
sin θ
I3 ω3 = pψ

giving us formula for ω1 , ω2 , ω3 in terms of our canonical momenta and angles only. These
are also the angular momentum vector in the body frame. We can replace the spins
ω1 , ω2 , ω3 in the Lagrangian with expressions that depend only on momenta (pθ , pφ , pψ )
and angles giving our Hamiltonian;
 2
1 (pφ − pψ cos θ)
H(φ, θ, ψ; pφ , pθ , pψ ) = sin ψ + pθ cos ψ +
2I1 sin θ
 2
1 (pφ − pψ cos θ)
cos ψ − pθ sin ψ +
2I2 sin θ
1 2
p
2I3 ψ
With I1 = I2 this simplifies. We note that the choice of angles and momenta is poor
when θ ∼ 0. In that setting it is a good idea to change coordinate system via canonical
transformation.

3.2.1 Axisymmetric bodies

For axisymmetric systems (I1 = I2 ) the Lagrangian is
I1  2 2  I
3
Lsym (θ̇, φ̇, ψ̇, θ, φ, ψ) = (φ̇) sin θ + (θ̇)2 + (φ̇ cos θ + ψ̇)2 − V
2 2

32
and the Hamiltonian is
1 (pφ − pψ cos θ)2
 
2 1 2
Hsym (φ, θ, ψ; pφ , pθ , pψ ) = + pθ + p +V (45)
2I1 2
sin θ 2I3 ψ
With V independent of φ, ψ, the Hamiltonian is only dependent on a single angle θ so
momenta pφ , pψ are conserved quantities as is H itself.
In the axisymmetric setting
pφ = (I1 sin2 θ + I3 cos2 θ)φ̇ + I3 ψ̇ cos θ (46)
pψ = I3 (φ̇ cos θ + ψ̇) (47)
pθ = I1 θ̇ (48)
and as long as the potential is independent of φ, ψ, these momenta pφ , pψ remain conserved.
It may be useful to write precession rates in terms of pφ , pψ
(pφ − pψ cos θ)
φ̇ = (49)
I1 sin2 θ
pψ pψ (pφ − pψ cos θ)
ψ̇ = − φ̇ cos θ = − cos θ (50)
I3 I3 I1 sin2 θ
With θ fixed we get constant precession rates, φ̇, ψ̇. The conserved quantity pψ is the
angular momentum projection along the body-fixed principal axis, whereas pφ is the angular
momentum about the z inertial axis. This is probably understood from the last equation
for z in the transformation 42 and applying it to the angular momentum as seen in the
body frame.
What is not obvious that we should be able to find solutions with fixed θ in all axisym-
metric situations. This is equivalent to choosing an orientation for using the Euler angles
– in other words choosing an orientation for the inertial coordinate system.
For torque free rotation we can choose our inertial frame to be with z aligned with the
angular momentum so L = Lẑ that is fixed in the inertial frame. Using our transformation
in equations 41 we find that in the body frame the angular momentum is
Lbody = L(sin ψ sin θx̂0 + cos ψ sin θŷ0 + cos θẑ0 )
and using our relations for ω1 , ω2 , ω3 (equations 37)
L sin ψ sin θ = I1 ω1 = I1 (φ̇ sin θ sin ψ + θ̇ cos ψ)
L cos ψ sin θ = I1 ω2 = I1 (φ̇ sin θ cos ψ − θ̇ sin ψ)
L cos θ = I3 ω3 = I3 (φ̇ cos θ + ψ̇) (51)
Taking the first two equations
L sin θ(sin ψ cos ψ − cos ψ sin ψ) = 0
= I1 ω1 cos ψ − I1 ω2 sin φ = pθ
= I1 θ̇

33
So our choice of coordinate orientation, assuming an axisymmetric body and a constant
angular momentum gave us θ̇ = 0.

3.3 Torque free rotation of the axi-symmetric body

In Figure 9 we have two reference frames, an inertial one and a body one. Let us orient
the inertial one with the angular momentum L because lacking any external torques, it is
fixed. We consider a rigid axi-symmetric body. In a coordinate frame aligned with body’s
axis of symmetry the moment of inertia matrix has 3 eigenvalues Ik , I⊥ , I⊥ and two of the
them are the same. If the axis of symmetry lies along the z axis then Ik = Iz . The body is
spinning but with angular momentum and spin not parallel to z or in the xy plane. The
body is not spinning about a principal body axis. Euler’s equations are

I⊥ ω̇x + (Ik − I⊥ )ωz ωy = 0 (52)

I⊥ ω̇y − (Ik − I⊥ )ωz ωx = 0 (53)
Ik ω̇z = 0 (54)

for spin ω in the body’s reference frame. We notice that ωz is fixed. Set
Ik − I⊥
Ω= ωz
I⊥
and
ωz = ω cos α.
The angle α is the angle between spin vector ω and body symmetry vector ẑ. The equations
of motion become

ω̇x + Ωωy = 0 (55)

ω̇y − Ωωx = 0. (56)

A solution is

ωx = ω sin α cos(Ωt) (57)

ωy = ω sin α sin(Ωt) (58)
ωz = ω cos α (59)
Ik − I⊥
 
Ω = ω cos α . (60)
I⊥
Using L = Iω, the angular momentum vector (in the body’s frame)

Lx = I⊥ ω sin α cos(Ωt) (61)

Ly = I⊥ ω sin α sin(Ωt) (62)
Lz = Ik ω cos α (63)

34
It is convenient to define an angle θ between L and body principal axis ẑ that satisfies
q
L2x + L2y I⊥
tan θ = = tan α (64)
Lz Ik

We notice that ω, L and ẑ are all in the same plane.

The above solution describes L and ω in the body frame. To transform a vector into
an inertial frame
d d
= + ω×
dt inertial dt body
We can relate our spin or angular moment in the body frame to those in an inertial frame.
We expect the angular momentum in the inertia frame to be fixed and the spin vector
should precess.

3.4 Euler angles of the torque free axi-symmetric body

Now back to our Euler angles. Take K parallel to L. We associate xyz with i, j, k in Figure
9. The body we orient with axis of symmetry along z. The angle between the body axis
of symmetry and the angular momentum is θ so Lz = L cos θ and as this is fixed θ̇ = 0.
The angle α is that between the spin vector ω and k so ωz = ω cos α. The body does not
nutate but it does precess. Without nutation equations 37 are

ωx = sin ψ sin θφ̇

ωy = cos ψ sin θφ̇
ωz = cos θφ̇ + ψ̇ (65)

Consider a moment when ω and L are in the yz plane and so all have zero x components.
In order for ωx = 0 at this time, ψ = 0 and

ωy = sin θφ̇ = ω sin α. (66)

The first relation follows from equation 65 and the second one from the length of ω with
ωx = 0. Equation 66 gives a relation between θ, α, ω, φ̇ that is satisfied at all times.

35
 Combined with equation 64 we find
ω sin α
φ̇ =
sin θ
!1
Ik2 2
2
= ω sin α 2 cot α + 1
I⊥
#1
Ik2
" !
2
2
= ω 1+ 2 − 1 cos α (67)
I⊥
" !#− 1
2
I⊥ 2

= ω 1 + cos2 θ −1 (68)
Ik2

Using ωz = w cos α and the last of equation 65, some further manipulation using equation
64 gives

ψ̇ = ω cos α − cos θφ̇

= ω(cos α − cot θ sin α)
Ik
 
= ω cos α 1 − = −Ω (69)
I⊥

The last step we have used equation 60. Here we associate ψ̇ with the precession rate of
the angular momentum and spin in the body frame.
For small α (and simultaneously small θ)
Ik
lim φ̇ → ω (70)
α→0 I⊥
Ik
 
ψ̇ → ω 1 − (71)
I⊥
Ik
 
ψ̇ I⊥
→ 1− (72)
φ̇ Ik I⊥

Whereas −ψ̇ is the precession rate in the body frame, φ̇ is the spin precession rate in
the inertial frame. The two are not equal. For a prolate body Ik < I⊥ and vice versa for an
oblate body. Ω < 0 for elongated (prolate) bodies but Ω > 0 for flattened (oblate) bodies,
whereas φ̇ is always positive. Precession in the inertial frame is in the same sense as Lz ,
the vertical component of angular momentum. For a prolate body and small α, the rate
φ̇ φ̇
ω < 1, whereas for an oblate body ω >√1.
For a thin disk Ik = 2I⊥ and φ̇ = ω 1 + 3 cos2 α. For small α, the precession rate φ̇ ∼
2ω and Ω ∼ ω and giving the famous 2:1 ratio discussed by Feynman (though apparently
he had which frequency was smaller backwards).

36
 We have so far explored the torque-free axisymmetric top starting from Euler’s equa-
tions. We should be able to find the same solution using Hamilton’s equations or Lagrange’s
equations. A difficulty is that we need to chose the inertial frame so that the Euler angle θ
is fixed. Once that is done conserved momenta give the precession rates. I think choosing
the good inertial frame is equivalent to computing the angular momentum direction and
choosing the z axis (for the Euler angles) to align with it. Another inconvenience with the
Euler angles is that we are left without a nice expression for the angular momentum in the
inertial frame.

3.5 Frequencies seen by an external viewer - asteroid rotation

How do the body axes move, as seen in the inertial frame? Assuming that we had axis of
symmetry along z 0 (body frame) or ez 0 equation 42 gives us its coordinates in the inertial
frame as a function of the Euler angles

ez 0 (t) = sin φ sin θex − cos φ sin θey + cos θez . (73)

Let us take it’s x coordinate to be 0 at time t = 0, then

ez 0 (t) = sin(φ̇t) sin θex − cos(φ̇t) sin θey + cos θez (74)

as φ̇ (equation 67) is constant for a body of revolution. Likewise we can use equation 42
to determine the other two body axes as a function of time as seen in the inertial frame

ex0 (t) = [cos(φ̇t) cos(Ωt) + sin(φ̇t) sin(Ωt) cos θ]ex

+ [sin(φ̇t) cos(Ωt) − cos(φ̇t) sin(Ωt) cos θ]ey
− sin θ sin(Ωt)ez (75)
ey0 (t) = [cos(φ̇t) sin(Ωt) − sin(φ̇t) cos(Ωt) cos θ]ex
+ [sin(φ̇t) sin(Ωt) + cos(φ̇t) cos(Ωt) cos θ]ey
+ sin θ cos(Ωt)ez (76)

where we have used ψ̇ = −Ω.

Ignoring gravitational torques and those due to radiation forces or outgassing we can
think of an asteroid or a comet as torque free rotating body. Some of them (including
the exotic ‘Oumuamua) do not rotate solely about a principal axis. These we say are
undergoing non-principal axis rotation (NPA). If they are prolate or oblate they would be
described by the precessing torque free top equations considered here. An external viewer
would view the object by monitoring its light curve. The brightness of the asteroid depends
on its illumination angle and its projected surface area (as seen by the viewer). The object
could also have uneven albedo or color variations on its surface. Equations 74, 76 tell us
how the axes of a body of revolution move as seen by the external viewer. An external

37
viewer will see two frequencies in the motion of the precessing body, that set by Ω and
that set by φ̇. Two frequencies present in the light curve gives evidence of non-principal
axis rotation.
It is not easy to determine the body axis ratios from the light curve because the bright-
ness depends on illumination angle. In other words, how much sun light is reflected depends
on the angles between the sun to object line, object to viewer line and surface normal vec-
tors. The surface could also have color or albedo variations (spots!).
If the body is triaxial rather than a body of revolution, then θ is not fixed and there
will be an additional frequency present in the motion, called nutation. And there may be
aperiodicity as well in φ̇. For triaxial bodies the spins and some periods for the Euler angles
can be written in terms of elliptic functions. These give two periods and an associated
precession rate that could be evident in a light curve.
Here we defined the Euler angles with respect to the body symmetry axis. In general
you could define them with respect to a long or short principal body axis. This gives two
conventions for specifying them.
To interpret asteroid rotation, you start with spin and angular momentum vectors and
three moments of inertia. Solve Euler’s equations in the body frame. Then compute the
frequencies associated with the Euler angles and compare them to frequencies that are seen
in a periodogram of a light curve.

3.6 Long axis and short axis modes

Following papers by Nalin Samarasinha and collaborators.....
We take a triaxial body with three moments of inertia, Il , Ii , Is for short, intermedi-
ate and long axes, and defining Euler angles with respect to the short axis (short axis
convention). We define a constant
L2
U≡ (77)
2E
which depends on angular momentum L and rotation kinetic energy E. Long axis modes
(LAMs) have
Il ≤ U < I i (78)
we define a time variable τ and a constant of motion k
s
2E(Ii − Il )(Is − U )
τ =t (79)
Il Ii Is
(Is − Ii ) (U − Il )
k2 = (80)
(Ii − Il ) (Is − U )

38
 s
2E(Is − U )
Ωl = dnτ (81)
Il (Is − Il )
s
2E(U − Il )
Ωi = snτ (82)
Ii (Ii − Il )
s
2E(U − Il )
Ωs = cnτ (83)
Is (Is − Il )

The Euler angles obey

s s 
I (I
l s − U ) I i (U − I l )
ψs = atan2  dnτ, snτ  (84)
Is − Il Ii − Il
s !
−1 Is (U − Il )
θs = cos cnτ (85)
U (Is − Il )
Is − U + (U − Il )sn2 τ
 
φ̇s = L (86)
Il (Is − U ) + Is (U − Il )sn2 τ

It turns out that two of the angles are periodic in τ , but the last one is not, though the
rate φ̇s is. The period
s Z π/2
Il Ii Is du
PψS = 4 p (87)
2E(Ii − Il )(Is − U ) 0 1 − k 2 sin2 u

Short axis modes (SAMs) have

L2
Ii < ≤ Is (88)
2E

s
2E(Is − Ii )(U − Il )
τ =t (89)
Il Ii Is
(Ii − Il ) (Is − U )
k2 = (90)
(Is − Ii ) (U − Ii )

Another set of formulas.

And then you can redo everything using the long axis convention instead of the short
axis convention!
Samarasinha and collaborators place limits on periods and maximum angles .... that
might be useful to classify actual systems from light curves.

39
3.7 Comments on Damping of Wobbling
If an asteroid is not rotating about a principal axis, then Ω is time dependent. Let a0 , v0 , r0
be the acceleration, velocity and position in the body frame and a, v, r be those in the
inertial frame. To relate the time derivative of a position in the inertial frame to that in
the body frame
d ∂
= +Ω×. (91)
dt ∂t
Apply this to position r0 in the body

dr ∂r0
= + Ω × r0
dt ∂t
v = v 0 + Ω × r0

Now we apply our time derivative to v

dv ∂v
= +Ω×v
dt ∂t
∂
v0 + Ω × r0 + Ω × (v0 + Ω × r0 )


a=
∂t
∂v0
= + Ω̇ × r0 + Ω × v0 + Ω × v0 + Ω × (Ω × r0 )
∂t
= a0 + Ω̇ × r0 + 2Ω × v0 + Ω × (Ω × r0 )

If the body does not flex much then a0 and v0 (in the body frame) are small and

a ≈ Ω̇ × r0 + Ω × (Ω × r0 ) (92)

The time dependent stress due to the spin variations can be estimated using this equation.
For a prolate or oblate body, Ω depends on a single frequency and this sets the timescale
of elastic stress variations in the body. If the body is triaxial then an entire spectrum of
frequencies is involved. This is relevant as internal friction due to the elastic deformation
of the body may be frequency dependent.
While angular momentum is conserved, the total energy is not due to dissipation or
internal friction inside the elastic body which goes into heat that can be radiated away.
The rate of energy loss is computing by integrating up over the volume of the body the
stress times the strain rate.

3.8 Andoyer-Deprit variables

Dynamics might be simpler if we keep track of a spin frame, not just an inertial reference
frame and a body frame. Following the nice introduction in Celletti’s book.

40
Figure 11: Andoyer-Deprit angles. The green lines show lines of nodes. The body’s spin
angular momentum is aligned with i3 . The p̂ line of nodes is the intersection of the i, j
plane (body) with the i1 , i2 (spin) plane. The m̂ line of nodes is the intersection of the
I, J plane (inertial) with the i1 , i2 (spin) plane. The n̂ line of nodes is the intersection
of the i, j plane (body) with the I, J plane (inertial). The angles θ, φ, ψ are Euler angles.
The angles g, h, l are Andoyer-Deprit angles. The angles J, K are used in the definition
of Andoyer-Deprit momenta. Under principal axis rotation J = 0, K = θ, m̂ = n̂ = p̂,
φ = h, g = 0, and l = ψ.

41
 • (O, I, J, K) is an inertial reference frame (for example defined with respect to a
Laplace plane or the elliptic). Here O refers to the center of mass of our spinning
body.

• (O, i, j, k) is a body frame with respect to principle axes of moment of inertia.

• (O, i1 , i2 , i3 ) is a spin frame, and i3 is aligned with the angular momentum vector.

Because we have three frames we need more than one line of nodes (see Figure 11) .

• The n̂ line of nodes is the intersection of the i, j plane (body) with the I, J plane
(inertial).
n̂ = K × k/|K × k|
We used this line of nodes when defining the Euler angles.

• The m̂ line of nodes is the intersection of the I, J plane (inertial) with the i1 , i2 plane
(body).
m̂ = i3 × K/|i3 × K|

• The p̂ line of nodes is the intersection of the i, j (body) plane with the i1 , i2 plane
(spin).
p̂ = k × i3 /|k × i3 |

The signs on the cross products need to be checked for convention here! I have adopted
the order in Celleti’s book.
Euler angle lists
• (θ, φ, ψ) are Euler angles of the body frame with respect to the inertial frame.

• (J, g, l) are Euler angles of the body frame with respect to the spin frame.

• (K, h, 0) are Euler angles of the inertial frame with respect to the spin frame, but
assuming that i3 lies on line of nodes m̂.
What are i1 , i2 aligned with? Their directions are not needed to define the Andoyer-
Deprit variables. The angles depend on the lines of nodes rather than the orientation of i1
or i2 .
Andoyer-Deprit variables are a canonical set (G, g), (L, l), (H, h) of momenta paired
with angles.
If M0 is the angular momentum vector M0 = M0 i3 .

G ≡ M0 · i3 = M0 (93)
L ≡ M0 · k = G cos J (94)
H ≡ M0 · K = G cos K (95)

42
 The angle K is an obliquity angle between angular momentum vector and inertial
frame axis. It is not one of the Andoyer-Deprit variables. The angle K is that between i3
(angular momentum vector) and K (inertial frame axis).

cos K = i3 · K

The angle J is called the non-principal rotation angle. When it is zero, J = 0, the
body is rotating about a principal body axis. It is not one of the Andoyer-Deprit variables.
The angle J is that between i3 (angular momentum vector) and k (body principal axis).

cos J = i3 · k

• The Andoyer-Deprit momentum variable G is the total angular momentum.

• The Andoyer-Deprit momentum variable L is the component of the angular momen-

tum in the direction of the principal body axis k. If the body is rotating around this
principal axis then L = G.

• The Andoyer-Deprit momentum variable H is the component of the angular momen-

tum in the direction of the inertial axis K.

The Andoyer-Deprit momenta in terms of those associated with the Euler angles are

pφ = H
pψ = L
pθ = G sin J sin(l − ψ) (96)

Here the Euler angles are defined with respect to the inertial reference frame I, J, K.
The variables G, J, L, l are related to the angular momentum orientation as seen in the
body frame

I1 ω1 = G sin J sin l
I2 ω2 = G sin J cos l
I3 ω3 = L (97)

How are the Andoyer-Deprit angles g, h, l defined?

• The angle g is that between line of nodes m̂ and p̂.

cos g = m̂ · p̂

This angle is measured in the i1 , i2 plane.

43
 • The angle h is that between I and line of nodes m̂.

cos h = I · m̂

This angle is measured in the I, J plane.

• The angle l is that between i (body principal axis) and line of nodes p̂.

cos l = i · p̂

This angle is measured in the i, j plane.

Using the relation between Euler angles and associated momenta and the Andoyer-
Deprit variables, the Hamiltonian can be written in terms of the Andoyer-Deprit variables
 2
sin l cos2 l 1 L2

1 2 2
H(G, L, H; g, l, h) = (G − L ) + + (98)
2 I1 I2 2 I3
It is independent of H, g, h, so their canonical counter parts h, G, H must be conserved.
A useful relation

cos K = cos J cos θ + sin J sin θ cos(ψ − l)

Celletti’s book shows that the transformation to Andoyer-Deprit variables is canonical.

If the body is axisymmetric

G2 − L2 L2
Haxi = +
2I1 2I3
and  
G 1 1
ġ = l˙ = − L ḣ = 0 (99)
I1 I3 I1
The body rotates around the symmetry axis with constant velocity and that it precedes
uniformly.
With Euler angles, the solution of precessing axisymmetric body is not obvious (you
must establish that you can find a coordinate system with θ̇ = 0) using Andoyer-Deprit
variables, it is clearer that a rigid body rotates about the symmetry axis with a constant
velocity and precessing with a constant velocity.
The Andoyer-Deprit variables are not well defined if J = 0 or K = 0. However one can
transform (canonically) to these variables

λ1 = l + g + h Λ1 = G
λ2 = −l Λ2 = G − L = G(1 − cos J)
λ3 = −h Λ3 = G − H = G(1 − cos K)

44
The new Hamiltonian is
(Λ1 − Λ2 )2
H(Λ1 , Λ2 , Λ3 ; λ1 , λ2 , λ3 ) =
2I3

sin2 λ2 cos2 λ2
 
1 2 2
+ Λ1 − (Λ1 − Λ2 ) + (100)
2 I1 I2
It is independent of Λ3 , λ1 , λ3 so their canonical counterparts λ3 , Λ1 , Λ3 must be conserved.

3.8.1 Principal axis rotation

Under principal axis rotation J = 0, m̂ = n̂, θ = K, φ = h = −λ3 , Λ2 = 0, pθ = 0,
and G = L. The Hamiltonian H(Λ; λ) = Λ21 /(2I3 ). The line of nodes p̂ is defined by an
intersection of two planes that are the same so I think we can chose p̂ = m̂, g = 0 and
l = ψ = −λ2 . Also cos θ = H/G = 1 − Λ3 /Λ1 .

3.9 Potential energy of a rigid body from an external force field

Suppose we have a solid body of mass m subject to an external force. What is the torque
on it? We can ask the related question, what is the potential energy on a rigid body from
an external force field?
The external force field we describe with a potential Φ(x) that we expand about the
center of mass of our body x0
1
Φ(x) = Φ(x0 ) + Φ,i (x0 )(x − x0 )i + Φ,ij (x0 )(x − x0 )i (x − x0 )j + ... (101)
2
In a coordinate system centered about the center of mass of our body y = x − x0
1
Φ(y) = Φ(x0 ) + Φ,i (x0 )y i + Φ,ij (x0 )y i y j + ... (102)
2
In the center of mass coordinate system
Z
d3 yρ(y)y i = 0 (103)

Because the force field is external ∇2 Φ = 0 so Φ,ii (x0 ) = 0. The trace of the potential’s
Hessian is zero because the external potential must satisfy Laplace’s equation.
The total potential energy we find by integrating over body coordinates
Z  
3 i 1 i j
U = d yρ(y) Φ(x0 ) + Φ,i (x0 )y + Φ,ij (x0 )y y + ...
2
Z Z
1
= Φ(x0 )m + Φ,i (x0 ) dy ρ(y)yi + Φ,ij (x0 ) dy 3 ρ(y)y i y j
3
2
Z
1
= Φ(x0 )m + Φ,ij (x0 ) dy 3 ρ(y)y i y j
2

45
The first order term drops out because we are using the center of mass coordinate frame.
The zero-th order term is that for a point mass. The interesting term is the second order
term. Because the trace of the Hessian is zero
Z
0 = Φ,ii (x0 ) dy 3 ρ(y)y k y k
Z
= Φ,ij (x0 )δij dy 3 ρ(y)y k y k

We will use this to rewrite the potential energy with the moment of inertial tensor
Z
1
U2 = Φ,ij (x0 ) dy 3 ρ(y)y i y j
2
Z
1 h i
= Φ,ij (x0 ) dy 3 ρ(y) y i y j − δij y k y k
2
1
= − Φ,ij (x0 )Iij
2
Rewriting this the potential term associated with body torque is
1
U2 = − Φ,ij (x0 )Iij (104)
2
If we require more accuracy we we could add in higher order terms, depending upon higher
order moments of both body and potential.
The external force field could be from an oblate planet, a binary or a point mass. In
all these settings we could compute the derivatives of the external potential and from that
the potential energy integrated over the body. The potential energy of the body can be
put immediately into a Lagrangian or a Hamiltonian using one of the choices above for
canonical variables. By taking derivatives of the potential energy function the torque on
the body can be computed.

3.10 MacCullagh’s Formula

We consider the external perturbation from a point mass of mass M at a position r. This
will external perturbation acts on a small mass of mass m that has moment of inertia
matrix I. We expand the potential from M at the location of m and then use equation 104
to compute the potential perturbation from the moment of inertia matrix. At a position r0
(position of m) the potential from the point mass M can be expanded in terms of Legendre
polynomials
GM GM X r0l
Φ(r0 ) = − = − P l (cos ψ) (105)
|r − r0 | r rl
l=0
where
r·r
cos ψ =
|r||r0 |

46
The second order term is the one that gives us our Hessian matrix.
1
3 cos2 ψ − 1


P2 (cos ψ) =
2
1 02 02 02 02


= 3z − x − y − z
2r02
1
2z 02 − x02 − y 02


= 02
2r
where I have oriented the prime coordinate system so that the point mass M lies along z
axis. In this coordinate system
GM
Φ,xx = Φ,yy =
r3
2GM
Φ,zz = − = −2Φ,xx
r3
Taking only second order terms
GM
Φ2,ij = (δij − 3δiz δjz )
r3
Putting this into equation 104 the left hand part will give us the trace of the moment of
Inertia matrix.
GM
U2 = − 3 (trI − 3Izz ) (106)
2r
where Izz is the moment of inertia of m about the axis between our body m and the
external point mass M . This is equivalent to MacCullagh’s formula.
A small non-round body in the gravitational field of a distant point source has a po-
tential energy perturbation given by this (remember to add it to the zeroth order part
−GM m/r). This can be used to compute the torque on m by M which is the same but
opposite to the torque on M by m. A useful formula for tidal computations.
MacCullagh’s formula is often written

Gm G(Ixx + Iyy + Izz − 3I)

U =− − (107)
r 2r3
where I is the moment of inertia of the oblate body about the axis extending between the
center of the body and the point mass perturber. The moment of inertia about an axis
with direction n̂, defining the direction between center of mass of our extended body and
our point mass, is
I = n̂In̂ = nα Iαβ nβ (108)
Previously we expanded the potential of a point mass, but we should recognize that
terms in a potential are interaction terms so we can just as easily consider them caused by

47
the non-round body on a point mass, rather than integrating the effect of a point mass on
an extended body.
By taking derivatives of the potential function we can compute the torque. Taking
i, j, k to be unit vectors lying along principal axis of the extended spinning body that have
moments of inertia A < B < C and r is the vector from the center of mass of the extended
spinning body to a point mass M . We can write r in a basis aligned with our principal
body axes
r = r((r̂ · i)i + (r̂ · j)j + (r̂ · k)k
and equation 108 as
I = (r̂ · i)2 A + (r̂ · j)2 B + (r̂ · j)2 C
and equation 107 as

Gm G(A + B + C − 3 (r̂ · i)2 A + (r̂ · j)2 B + (r̂ · j)2 C)

 
U =− − (109)
r 2r3
When computing the torque, first ones computes the force by taking the derivative of
the potential. Then r × F gives the torque. The computation is faster if you notice that
terms only containing factors of radius r will drop out because their gradients are parallel
to r. The non-radial part of equation 109 we can write as
3G 
(r · i)2 A + (r · j)2 B + (r · j)2 C)

Unr = − (110)
2r
with derivative (keeping on only non-radial parts)
∂Unr 3G
= − [A(r · i)ii + B(r · j)ji + C(r · k)ki ]
∂ri r
∂Unr 3G
= − [A(r · i)i + B(r · j)j + C(r · k)k]
∂r r
The torque
3GM
T=− [A(r · i)(r × i) + B(r · j)(r × j) + C(r · k)(r × k)] (111)
r
Using
r × i = [(r · j)j + (r · k)k] × i
and similarly for the other terms, the torque can be written
3GM
T = [(C − B)(r̂ · j)(r̂ · k)i + (A − C)(r̂ · k)(r̂ · i)j
r3
+ (B − A)(r̂ · j)(r̂ · i)k] (112)

as is found in Colombo’s 1966 paper on Cassini states.

48
3.10.1 For an axisymmetric body
MacCullagh’s formula is usually described as the gravitational potential of an oblate body,
like the Earth. For an oblate body oriented with z along the pole, Ixx = Iyy = I⊥ . This is
often written with I⊥ as the moment of inertia about an axis perpendicular to the body’s
axis (x or y) of symmetry and Izz = Ik is that about an the axis parallel to the axis of
symmetry (here z).
In a coordinate system with I diagonal

I = n2x Ix + n2y Iy + n2z Iz (113)

The moment I only depends on a latitude angle θ. Let us use a convention θ = 0 gives
the axis of symmetry and θ ∈ [0, π] with θ = π/2 at the equator with z = r cos θ. Taking
n̂ = (cos φ sin θ, sin φ sin θ, cos θ)

I = sin2 θI⊥ + cos2 θIk = I⊥ + cos2 θ(Ik − I⊥ )

Inserting this into equation 107 the gravitational potential

Gm G(2I⊥ + Ik − 3I)
U (r, θ) = − −
r 2r3
Gm G(Ik − I⊥ )(1 − 3 cos2 θ)
= − −
r 2r3
Gm G(I k − I ⊥ )
= − + P2 (cos θ) (114)
r r3
The total potential energy of an axisymmetric extended body described with Ik , I⊥ inter-
acting with a point mass M is

GM m GM (Ik − I⊥ )
E(r, θ) = − + P2 (cos θ) (115)
r r3
where r is the distance between M and the center of mass of m and θ gives the angle
between the vector joining the two masses and the axis of symmetry of the axisymmetric
body. In other words cos θ = r̂ · k with k along the body’s axis of symmetry.
For moments of inertia A = B, the torque in equation 112 simplifies to
3GM
T= (C − A)(r · k)(r × k). (116)
r3

3.11 Precession of an axisymmetric body in a circular orbit

If we assume principal axis rotation (a gyroscopic approximation) then the kinetic energy
is fixed and is Cω 2 /2 with C the principal moment of inertia. However if the object is
not round and is orbit about a point mass then it can precess. If the orbit is circular

49
we can chose a coordinate system oriented with orbit angular momentum aligned in the z
direction,
r = r(cos λt, sin λt, 0)
with λ̇ = n the mean motion. It is convenient to notice that GM/r3 = n2 is the mean
motion as long as the point mass is much larger than that of the precessing body.
For principal axis rotation the body’s spin ŝ = k and with θ the obliquity and φ the
precession angle projected onto the orbital plane,

ŝ = (sin θ cos φ, sin θ sin φ, cos θ).

The spin angular momentum L = Cωŝ. We compute

r̂ · ŝ = sin θ cos(λt − φ) = sin θ(cos(λt) cos φ + sin(λt) sin φ)

r̂ × ŝ = (sin(λt) cos θ, − cos(λt) cos θ, sin θ sin(φ − λt)
Z 2π
1 1
dλ (r̂ · ŝ)(r̂ × ŝ) = sin θ cos θ(sin φ, − cos φ, 0)
2π 0 2
where on the last line I have take the time average. The torque averaged over the orbit

3n2 (C − A)
hTi = sin θ cos θ (sin φ, − cos φ, 0)
2
is in a direction consistent with φ but not θ varying.
dL
= Cω sin θφ̇ (− sin φ, cos φ, 0).
dt
Equating these two expressions we find a precession rate,

3 (C − A)n2
φ̇ = − cos θ (117)
2 Cω
The equations of motion are consistent with a Hamiltonian system with canonical momen-
tum p = 1 − cos θ and canonical coordinate φ
αs αs
H(p, φ) = cos2 θ = (p − 1)2 (118)
2 2
and
3 (C − A)n2
αs ≡ . (119)
2 Cω
A planet can experience torques (and associated obliquity variations) due to perturba-
tions on its orbit. In this case the torque from the Sun is modulated due to orbit orientation
variations, typically described by variations in the longitude of the ascending node. Bill
Ward and collaborators wrote a series of papers accounting for obliquity variations with this

50
type of model, which is usually called ‘secular spin resonance’. The dynamics is considered
secular as it can be approximated by averaging over both orbit and spin angles.
How does a planet precess if it has moons? Moons can orbit a planet quickly compared
to the precession rate orbit of the planet. If the satellites nodal recession due planetary
oblateness is faster than the planet’s spin precession rate, then the moon’s orbit can be
considered a ring that slowly moves along with the planet and its orbit stays oriented
about the equator of the planet. An adiabatic invariant is preserved. The planet and its
moons are considered to precess together as if they were a single object. Moons affect the
planet’s precession rate as they add to the total spin angular momentum and make the
planet behave as if it were flatter. Saturn would precess more slowly if it did not have a
ring and moon satellite system.
The Moon’s orbit nodal recession period is 19 years and is much faster than the Earth’s
spin precession period which is 26,000 years. However precession of the Moon’s orbit is due
to the torque from the Sun, rather from Earth’s oblateness and the Moon’s orbit precesses
about the normal of the Earth’s orbit rather than about the spin axis of the Earth. The
transition between precession due to Earth’s oblateness and about the Earth’s spin axis
and precession due to the Sun and about the Earth’s orbital normal is called the Laplace
transition. A lock between the orbit nodal precession rate and the spin precession rate is
called a Cassini state. Tidally locked moons tend to be in a Cassini state.
Does the Moon’s inclination with respect to the ecliptic or with respect to the Earth’s
equator remain constant as the Earth precesses? The Moon’s orbit w.r.t. ecliptic has
inclination 5.15 degrees. The Moon’s obliquity (angle between orbit planet and spin axis
is 6.688 degrees). Both angles remain fixed as Earth’s spin axis precesses. The Moon does
affect the precession rate of the Earth, but the oblateness of the Earth no longer affects
the Moon’s orbit because the Moon is too far away.

3.12 Quaternions
We define a quaternion as
q = q0 + qx i + qy j + qz k
where q0 is called the scalar part and (qx , qy , qz ) is called the vector part of the quaternion.
Multiplication of quaternions is performed using the Hamilton product. Using the rules

i2 = j2 = k2 = −1
ij = k = −ji
jk = i = −kj
ki = j = −ik, (120)

51
the Hamilton product of q1 = a1 + b1 i + c1 j + d1 k and q2 = a2 + b2 i + c2 j + d2 k is

q1 q2 = a1 a2 − b1 b2 − c1 c2 − d1 d2
+ (b1 a2 + a1 b2 + c1 d2 − d1 c2 )i
+ (a1 c2 − b1 d2 + c1 a2 + d1 b2 )j
+ (a1 d2 + b1 c2 − c1 b2 + d1 a2 )k. (121)

A rotation can be described as a rotation by θ about an axis û = (ux , uy , uz ) =

ux i + uy j + uz k where û is a unit vector and we use i, j, k to represent unit vectors for a
Cartesian basis. The associated quaternion
θ
q = e 2 (ux i+uy j+uz k) = cos (θ/2) + sin (θ/2) (ux i + uy j + uz k) (122)

Notice that θ → −θ and u → −u gives the same rotation. This means that q → −q gives
the same rotation. Flipping the sign of θ or u (but not both) gives q−1 .
For a unit quaternion q = q0 + qx i + qy j + qz k with q02 + qx2 + qy2 + qz2 = 1, the axis of
rotation and angle
(qx , qx , qz ) (qx , qx , qz )
u= q = p (123)
qx2 + qy2 + qz2 1 − q02
q 
θ = 2 atan2 qx2 + qy2 + qz2 , q0
q q 
= 2 atan2 qx + qy + qz , 1 − (qx2 + qy2 + qz2 )
2 2 2
q 
= 2 atan2 2
1 − q0 , q0 . (124)

A rotation of vector p = px i + py j + pz k is performed by conjugating the quaternion

p0 = qpq−1 . (125)

To carry out consecutive rotations, quaternions can be multiplied using quaternion multi-
plication.
Taking unit quaternion q = q0 + qx i + qy j + qz k, the rotated vector can be written in
terms of a rotation matrix R with
p0 = Rp
and
1 − 2(qy2 + qz2 ) 2(qx qy − qz q0 ) 2(qx qz + qy q0 )
 

R =  2(qx qy + qz q0 ) 1 − 2(qy2 + qz2 ) 2(qy qz − qx q0 )  . (126)

2(qx qz − qy q0 ) 2(qy qz + qx q0 ) 1 − 2(qx2 + qy2 )
In index notation with indices i, j, k ∈ x, y, z

Rij = δij + 2qi qj − 2qk qk δij − ijk qk q0 (127)

52
To compute RT = R−1 we need only flip the sign of the last term. Rotating a vector p by
q gives

p0i = (1 − qk qk )pi + 2qi qj pj − ijk qk q0 pj

p0 = (1 − qk qk )p + 2q3 (q3 · p) − q0 p × q3

where q3 is the vector part of q.

Equation 126 can be used to relate a quaternion to Euler angles. For example, following
equations 41, 42
Rzz = 1 − 2(qx2 + qz2 ) = cos θ.
If the quaternion transfers to body axes from inertial frame then

ψ = atan2(Rxz , Ryz )
φ = atan2(Rzx , Rzy )

with order reversed if the quaternion transfers from body frame to inertial frame.
Consider two rotations giving a third RC = RB RA
α α
qA = cos + A sin (128)
2 2
β β
qB = cos + B sin (129)
2 2
γ γ
qC = cos + C sin (130)
2 2
with unit vectors A, B, C and angles α, β, γ. The result is known as Rodrigues formula for
composite rotation and is
γ α β α β
cos = cos cos − sin sin B · A (131)
2 2 2 2 2
γ β α α β β α
sin C = sin cos B + sin cos A + sin sin B × A (132)
2 2 2 2 2 2 2
 −1  
γ α β β α α β
tan C = 1 − tan tan B · A tan B + tan A + tan tan B × A . (133)
2 2 2 2 2 2 2
For describing rotations, quaternions have some advantages. Quaternions avoid a phe-
nomenon called gimbal lock which can result in pitch/yaw/roll rotational systems. They
ensure rotations are possible in any orientation. As they depend only on 4 numbers, the
notation is more compact than using rotation matrices. From a quaternion representation
of rotations it is easy to read off the axis and angle of rotation (much easier than when
using Euler angles).
The unit quaternion Lie group (S 3 ; 3D sphere in 4 dimensions) is not exactly the
same as the group of rotations SO(3) (special orthogonal matrices). This is because unit
quaternions q and −q give the same rotation.

53
Question Is there any intuition gained by using canonical variables based on quaternions?

An infinitesimal rotation by angle δθ about axis û gives a quaternion

δθ
qinf = 1 + (ux i + uy j + uz k) (134)
2
We recognize the spin as Ω = θ̇û. Let us consider a time dependent infinitesimal transfor-
mation
θ̇t
q(t) = 1 + (ux i + uy j + uz k)
2
t t
= 1 + (Ωx i + Ωy j + Ωz j) = 1 + Ωi êi
2 2
with êi representing i, j, k. Taking the time derivative
1
q̇ = Ωi êi (135)
2
A position is rotated by
r0 = qrq−1 .
We can differential this w.r.t to time. Multiplying by the quaternions gives us a velocity
v = ṙ = Ω × r, as expected. we can expand
t
q(t)p = p + Ωi pj ijk êk
2
and
d 1
q(t)p = Ωi pj ijk êk
dt 2
d
qpq−1 = Ω × p
dt
To keep track of body orientation, we would need to specify the body’s orientation (as
a function of time) with a quaternion. This could give us the transformation from a body
axis frame (where I is diagonal) to its current orientation with respect to an inertial frame.
We can specify the spin with a different but infinitesimal quaternion. The quaternion
specifying orientation contains the same information as the Euler angles. The infinitesimal
quaternion specifies the instantaneous spin. These two quaternions are everything we need
to describe the dynamics. Using one quaternion for orientation and another for spin are
equivalent to using a rotation matrix R to describe the current orientation and the time
derivative of a rotation matrix Ṙ to describe the current spin.
The total kinetic energy is
Z
1
T = ρ(r)(Ω × r)2 dV
2

54
where r is a vector for each position in the body. Note, that Ω is not equivalent to the
time derivative of a quaternion describing current body orientation. Instead it describes a
quaternion giving the infinitesimal spin rotation associated with rotation.
It may be useful to use the vector identity

(Ω × r)2 = Ω2 r2 − (Ω · r)2
d
(Ω × r)2 = 2Ωi r2 − 2(Ω · r)ri
dΩi
Z
∂T
= Ωj ρ(r)(δij r2 − rj ri )dV = Iij Ωj
∂Ωi
If we decide to adopt a variable whose time derivative is Ω (that would be the rotation
matrices) then our canonical momenta are the angular momentum vectors pi = Iij Ωj and
1
H = pi I(q)−1
ij pj
2
The associated angles are those of the rotations that are hidden in the moment of inertia
matrix.
Perhaps we can find canonical momenta associated with time derivatives of the quater-
nion. Let q describe body orientation and q = q0 + Q with Q the vector part of Q and
q̇ = q̇0 + Q̇. Because the quaturnion must remain a unitpquaternion we can consider only
Q as specifying the entire quaturnion, with q0 (Q) = 1 − Q2 . We perturb q with an
infinitesimal time dependent rotation specified by Ω. Using quaternion multiplication
d 1
q̇ = [(1 + Ωt/2)q] = [−(Ω · Q) + q0 Ω + Ω × Q] (136)
dt 2
is not the same as Ω. The vector part
  
q0 qz −qy Ωx
1 1
Q̇ = [q0 Ω + Ω × Q] = −qz q0 qx   Ω y 
2 2
qy −qx q0 Ωz

By inverting this I could write the Hamiltonian in terms of Q̇. Then I can take derivatives
of the Hamiltonian and find a set of canonical momenta that are conjugate to the vector
part of the quaternion. I invert this and find
 2
(q0 + qx2 )q̇x + (−q0 qz + qx qy )q̇y + (q0 qy + qx qz )q̇z


Ω(q, q̇) = 2q0−1  (q0 qz + qx qy )q̇x + (q02 + qy2 )q̇y + (−q0 qx + qy qz )q̇z 
(qx qz − q0 qy )q̇x + (q0 qx + qy qz )q̇y + (q02 + qz2 )q̇z

Ωi = 2q0−1 [q02 q̇i + q0 ijk qj q̇k + qi qj q̇j ] (137)

Ω = 2q0−1 [q02 Q̇ + q0 Q × Q̇ + Q(Q · Q̇)] (138)

55
 ∂Ωi
= 2q0−1 [q02 δij − q0 ijk qk + qi qj ]
∂ q̇j
The kinetic energy
1
T (Q, Q̇) = ΩIΩ
2
and Ω(Q̇, Q) and I(Q).
∂T
= IΩ = L
∂Ω
in terms of the angular momentum with I(Q) and Ω(Q, Q̇).

∂T ∂T ∂Ωi ∂Ωi
= = Iil Ωl (139)
∂ q̇j ∂Ωi ∂ q̇j ∂ q̇j
πj = 2(Lj q0 + ijk Li qk + Li qi qj /q0 ) (140)
∂T
= π = 2(Lq0 + L × Q + (L · Q)Qq0−1 ) (141)
∂ Q̇

This is linear in L so can also be inverted to solve for L(π).

As a guess  
1 π
L= π×Q+
2 q0
and again as a guess
 
1 πI(Q)π
T (π, Q) = (π × Q)I(Q)(π × Q) +
8 q02

where I am guessing the cross terms go away by symmetry.

Lagrange’s equations would give
d ∂T ∂T
=
dt ∂ Q̇ ∂Q

With T = 21 LI−1 L we can write T (π, Q) and giving canonical coordinates based on
the quaturnion vector Q. However, this Hamiltonian has no obvious conserved quantities,
and we know that angular momentum should be conserved.
In summary it is possible to use Q the vector part of the quaturnion as a canonical
variable and write out Lagrange’s equations in terms of Q, Q̇. It is also possible to derive
a canonical momentum π that is conjugate to Q and an associated Hamiltonian that is
the kinetic energy that is a function of T = H(π, Q). It has the advantage that we
make no assumptions about preferred directions for the angles. However it lacks in obvious
conserved quantities and seems messy, and this probably accounts for its lack of popularity.
Also my choice of Q as variable gives us a potential problem when q0 = 0. This happens

56
when there is a rotation of π/2 about a principal axes. One possible way to redo this might
be to work with 4 degrees of freedom with the quaternion and treat its unit length as a
constraint. Constraints are sometimes ugly to work with in the Hamiltonian view, as you
need to work on a submanifold in phase space.

3.13 A Lie-Poisson numerical method for Solid Body rotation with Quaturnions
We keep track of the body orientation with a quaturnion q(t). The body’s spin vector is
Ω(t). Chose a timestep dt for updating q and Ω(t). In the absence of extra torques the
spin angular momentum vector p is conserved.
Start with orientation defined with quaturnion q(t = 0). The quaturnion should let
you rotate the body from a coordinate with aligned principal axes to the orientation of the
body in an inertial frame. Use initial conditions (current orientation and spin vector) to
compute the initial angular momentum vector p.

• Use the current quaturnion q(t) to compute the inverse of the moment of inertia
matrix I−1 (q) from the quaternion rotation. You don’t need to integrate over the
mass of the body, only rotate or transfer the moment of inertia matrix (tensor) from
the body frame into the current orientation frame.

• Compute the current spin vector Ω = I−1 p.

• Compute unit infinitesimal quaturnion with timestep dt

r
Ω2 dt2 Ωi dt
qdt = 1 − + êi
4 2

• Compute the new quaternion q(t + dt) = qdt q(t) via quaturnion multiplication. This
updates q, describing body orientation at the next timestep.

Repeat these steps.

For torqueless solid body rotation only the quaternion q is changing and angular mo-
mentum p is conserved. If there are external torques then we can compute them using the
body orientation at each timestep. A torque would change only p. So we can alternate
applying torques (involving updating the angular momentum) and updating the orienta-
tion. We can construct a leap-frog integrator in the usual fashion by taking half step of one
operator, full step of the other and then half step of the first operator. Each step would
be Lie-Poisson (or symplectic).
This gives a numerical method, though is not necessarily illuminating or helpful for
finding tumbling or precession frequencies. An advantage is that we keep track of the body
orientation at all times, making it possible to compute additional potential energy terms or
torques in a straightforward manner (for example using MacCullagh’s formula). A disad-
vantage of the Euler angles or Andoyer-Deprit variables is that to include additional forces

57
requires relating the Andoyer-Deprit angles to ordinary coordinates, and this involves a
series of rotations that depends on the current body orientation and the angular momen-
tum vector with respect to an inertial coordinate system. Methods using Euler angles or
Andoyer-Deprit variables suffer from potential gimbal lock and a quaternion method would
not.
A good reference is the 1994 paper by Jihad Touma and J. Wisdom on ‘Lie-Poisson
integrators for Rigid Body dynamics in the Solar System’, AJ 107, 1189. Possibly Sussman
and Wisdom’s book discuss a similar approach to making a quaternion integrator. (Not
sure as I haven’t found the book, but I saw a reference to it).

4 Tidal Evolution
4.1 Gravitational Potential Expansion in Legendre Polynomials
External to a body where the mass density is zero, the gravitational potential satisfies
Laplace’s equation
∇2 Φ = 0
Working in spherical coordinates r, θ, φ, Laplace’s equation becomes

1 ∂2Φ
   
∂ 2 ∂Φ ∂ 2 ∂Φ
r + (1 − µ ) + =0
∂r ∂r ∂µ ∂µ 1 − µ2 ∂φ2
where
µ ≡ cos θ
We assume a trial solution
Vn = rn Sn (µ, φ)
The radial derivative term of the trial solution is this
 
∂ ∂Vn
r2 = n(n + 1)rn Sn
∂r ∂r
Subbing this into Laplace’s equation gives
   
∂ 2 ∂Sn ∂ 2 ∂Φ
(1 − µ ) + (1 − µ ) + n(n + 1)Sn = 0
∂µ ∂µ ∂µ ∂µ
This equation is known as Legendre’s equation. The functions that satisfy it, Sn , are known
as spherical harmonics. We can expand the general solution in spherical harmonics
X 
Φ(r, φ, θ) = An rn + Bn r−(n+1) Sn (µ, θ)
n

where Sn is a spherical harmonic.

58
 If the problem has axial symmetry then we can ignore the dependence on φ. In this
case Sn (µ, φ) = Pn (µ) is a Legendre polynomial satisfying

∂ 2 Pn ∂Pn
(1 − µ2 ) 2
− 2µ + n(n + 1)Pn = 0
∂µ ∂µ
Legendre polynomials can be computed with Rodrigues’s formula
1 dn (µ2 − 1)n
Pn (µ) =
2n n! dµn
and the first few are

P0 (µ) = 1
P1 (µ) = µ = cos θ
1 1
P2 (µ) = (3µ2 − 1) = (3 cos 2θ + 1)
2 4
1 3
P3 (µ) = (5µ − 3µ)
2
1
P4 (µ) = (35µ4 − 30µ2 + 3)
8
The gravitational potential of an axisymmetric body with mass M and radius R and
at r > R (external to the body) is often written as
∞
"  n #
GM X R
Φ(r, θ) = − 1− Jn Pn (cos θ) (142)
r r
n=2

with coefficients Jn dependent on the internal mass distribution and body shape. Oblate
bodies have positive J2 . A comparison of this equation with equation 114 implies that
(Ik − I⊥ )
J2 =
M R2
(and I think the sign is correct).

4.2 Expansion of the gravitational potential of a point mass in spherical

harmonics
We consider a point mass of mass Mp at position r0 . At position r the gravitational
potential is
GMp
Φ(r) = −
|r − r0 |
We would like to expand the gravitational potential near the origin in spherical harmonics.
Assume that we want to know the potential with r < r0 or close to the origin.

59
 p p
|r − r0 | = r2 + r02 − 2r · r0 = r2 + r02 − 2rr0 cos µ
with
r · r0
cos µ ≡
rr0
The potential has axial symmetry. There is an important direction, that connecting the
origin and r0 that is the axis of symmetry.
Using the Legendre polynomials
∞
X rl
Φ(r) = Φ(r, µ) = −GMp Pl (µ)
r0l+1
l=0

and the expansion is good (converges) as long as r < r0 .

Let’s look at the first few terms.
GMp r r2
Φ(r, µ) = − P0 (µ) − GM P
p 02 1 (µ) − GM p 03 P2 (µ) ... (143)
r0 r r
GMp r r2 1
= − 0 − GMp 02 µ − GMp 03 (3µ2 − 1) ... (144)
r r r 2
Consider Mp as an external mass perturbing a body of mass M with center of mass at
r. If we integrate the l = 1 term over the body, the condition that r is at the center of
mass implies that that the total potential would be zero. The l = 2 term is a quadrupole
term and that contributes the strongest non-trivial term to the tidal potential caused by
Mp .

4.3 Tidal deformation of a spherical body

We consider an external perturber of mass Mp on a spherical body of mass M and radius R.
The tidal potential, VT , on the surface has size VT ∼ GMp R2 /D3 where D is the distance
to the perturber. More accurately the tidal potential

GMp r2
VT (r, µ) = P2 (µ)
D3
We assume the surface of the deformed body is an equipotential surface.

gh(µ) ∼ VT (R, µ)

where g = GM/R2 is the surface acceleration and h(µ) is the surface height. We can solve
for the surface height giving
h(µ) Mp R 3
= P2 (µ)
R M D3

60
4.4 Love numbers for fluid bodies
A homogenous density nearly spherical body of mean radius C has surface radius

R(θ, φ) = C(1 + h2 P2 (cos θ) + h3 P3 (cos θ)) (145)

with P2 , P3 Legendre polynomials. This could be expansed to contain additional and

higher order spherical harmonics. Here I have included quadrupole and octupole terms.
The gravitational potential to the octupole term outside the body r > C
 2  3 !
GM C C
V (r, θ, φ) = − 1 + k2 P2 (cos θ) + k3 P3 (cos θ)
r r r
!
3 C 2 3 C 3
   
GM
=− 1+ h2 P2 (cos θ) + h3 P3 (cos θ) (146)
r 5 r 7 r

The coefficients for both expressions can be called Love numbers. They are related by
3
kl = hl (147)
2l + 1
They depend on l which is the multipole index. Above I have only included l = 2, 3 terms
but higher order expansions should be consistent with this expression. This expression
is derived by expanding 1/∆ where ∆ is the distance between a point in the shell and a
distant point θ, φ and then integrating over the mass shell. The integration is easier using
spherical harmonics. This is done by M+D with unclearly defined (vaguely normalized)
spherical harmonics.
To summarize: With a deformed body described by the Love numbers for the surface
deformation, hl , we can find the external gravitational potential of the deformed body and
it is described with Love numbers kl .
The tidal potential from an external point mass body of mass mpert expanded in the
coordinate system at the center of M is
 
Gmpert  r 2  r 3
V (r, θ, φ) = − P2 (cos θ) + P3 (cos θ) (148)
a a a
where the tidal perturber is oriented along θ = 0 and is a distance a away from M . The
surface of a tidally deformed fluid body should be an equipotential surface, so the two
expressions (equation146 and 148) for the potential should be equal on the surface.
On the surface (where r = R(θ, φ)), equation 146 gives
 
GM 3 3
V (θ, φ) = − 1 − h2 P2 (cos θ) − h3 P3 (cos θ) + h2 P2 (cos θ) + h3 P3 (cos θ) (149)
C 5 7
 
GM 2 4
=− 1 + h2 P2 (cos θ) + h3 P3 (cos θ) (150)
C 5 7

61
where I have expanded to first order in the Love numbers. The coefficients are
3 2l − 2
1− = (151)
2l + 1 2l + 1
On the surface equation 148 gives
" 2  3 #
Gmpert C C
V (θ, φ) = − P2 (cos θ) + P3 (cos θ) . (152)
a a a
Here we just set r = C because mpert is small so we only take the zero-th order term. We
match coefficients and find Love numbers
5 mpert C 3
 
h2 = (153)
2 M a
7 mpert C 4
 
h3 = (154)
4 M a
The coefficients are  l+1
2l + 1 m C
hl = (155)
2l − 2 M a
Inserting these back into equation 146 gives
 l+1
3 mpert C
kl = (156)
2(l + 1) M a
and this agrees with expressions by Efroimsky+12, though the factors of mpert /M and C/a
are not included in their coefficients.
What is the external gravitational potential of our tidally perturbed fluid body?
!
3 mpert C 3 C 2 3 mpert C 4 C 3
       
GM
V (r, θ, φ) = − 1+ P2 (cos θ) + P3 (cos θ)
r 5 M a r 7 M a r
(157)
The dependence on distance and mass ratio is usually not included in the Love numbers
so that they are independent of the tidal perturbation. If we follow this convention and
evaluate the potential at r = a,
!
GM X mpert  C 2l+1
V (a, θ, φ) = − 1+ kl Pl (cos θ) (158)
a M a
l
This is the potential from M at the location of mpert due to tidal deformation of M by the
tide caused by mpert . The quadrupolar term is the strongest one
GM mpert C 5
 
V2 ∼ −k2 P2 (cos θ) (159)
a M a
We will use this below when we estimate tidally generated torque on a spinning body.

62
4.4.1 Love numbers for an incompressible homogeneous elastic body
When the body is not a fluid, the deformation is lower (h2 is lower) and the Love number
h2 is inversely proportional to the rigidity or the elastic modulus of the deformed body.
Stiffer bodies deform less than stretchy ones. For an incompressible homogeneous elastic
body the Love number
k2 ∼ 0.038eg /µ
where µ is the rigidity and
GM 2
eg =
R4
and R = C. The other Love number h2 can be computed using equation 147.

Figure 12: Illustration of the rotational deformation of a spinning body.

4.5 Rotational Deformation

We assume we have an axisymmetric body of mass M with axis of symmetry about the z
axis which is also the spin axis. The external gravitational potential for our axisymmetric
spinning body with mass M
∞
"  n #
GM X R
V (r, µ) = − 1− Jn Pn (µ) (160)
r r
n=2

63
(equation 142) where µ = cos θ is the colatitude and R is the equatorial radius. There is
no n = 1 term as we set the coordinate system to have origin at the center of mass. For
an axisymmetric body
C −A
J2 =
M R2
where C, A are body principal moments of inertia with C > A. For an oblate body, C
corresponds to that for the short axis which is also the axis of symmetry. The coefficient
Z R Z 1
1 2+n
Jn = 2πr dr dµ Pn (µ)ρ(r, µ)
M Rn 0 −1

where ρ(r, µ) is the internal density distribution.

If a body is undergoing solid body rotation, the centrifugal acceleration at radius r
and angle θ is acf = Ω2 (xx̂ + yŷ) (see Figure 12). The centrifugal force gives an effective
potential
1
Vcf (r, θ) = Ω2 r2 sin2 θ
2
where θ is the colatitude (which is zero from the +z axis) and goes to π on the -z axis. It
is convenient that we can write the centrifugal potential in terms of P2 (µ) = 12 (3µ2 − 1)
with
1
Vcf (r, θ) = Ω2 r2 (P2 (µ) − 1)
3
Adding together the centrifugal term and the J2 term arising from the internal mass dis-
tribution the external potential including the centrifugal term

GM R2 1 2 2
 
GM
V (r, µ) = − + + Ω r P2 (µ).
r r3 3

We also expand the surface in spherical harmonics with the surface

rs (µ) = R(1 + P2 (µ))

where  is small. We expand the potential on the surface and assume that it is an equipo-
tential and ignore terms that are not ∝ P2 ,
 
GM GM GM 1 2 2
V (rs (µ)) = − + P2 (µ) + J2 + Ω R P2 (µ) = constant. (161)
R R R 3

We solve for 
1 Ω2 R 3 q
− = J2 + = J2 + (162)
3 GM 3
where
Ω2 R 3
q≡
GM

64
is a measure of centrifugal support. If q = 1, the body flies apart.
Using the formula for P2 (µ) = 12 (3µ2 − 1) we compute the polar and equatorial radii,
the surface has at the pole with θ = 0, µ = 1, rs,polar = R(1 + ). At the equator, θ = π/2,
µ = 0, rs,equatorial = R(1 − /2). A measure of the oblateness

requatorial − rpolar 3
f≡ = − .
requatorial 2

We insert this into equation 162 and find

3 q
f = J2 + (163)
2 2
Equation 163 gives a relation between rotation rate (through q), oblateness (through f )
and internal mass distribution (through J2 ).
A variant of this equation that is written using the moment of inertia instead of J2 is
called the Darwin Radau relation. The relation is a bit more complicated as it solves for the
body shape assuming that the rotating and distorted body is in hydrostatic equilibrium.

4.6 Tidal spin down

Figure 13: Illustration of the lag angle. The body on the left tidally perturbs the one on
the right which becomes elongated. But since it is spinning its deformation is not exactly
aligned with the perturber. This gives a torque on the spinning body causing it to spin
down or up. Because angular momentum is conserved, this also caused the semi-major
axis a to drift.

We assume principal axis rotation of a spherical homogeneous moon with mass M that
is perturbed tidally by a body M∗ . We assume that they are in circular orbit with semi-
major axis a and zero obliquity. The radius of M is R. We assume that M is not tidally
locked (is not in a spin synchronous state).
We estimate the torque on M∗ by the quadrupolar potential term from M ’s deformation
assuming there is a slight angular shift in V2 given by equation 159 (but with mpert = M∗

65
and C = R). The potential term (equation 159)
 5
GM∗ R
V2 ∼ −k2 P2 (cos θ) (164)
a a

Using P2 (θ) = 21 (3 cos2 θ − 1) we add a phase shift δ to the orientation giving

 5
GM∗ R
V2 ∼ −k2 P2 (cos(θ + δ)) (165)
a a

The torque on M∗ is T = M∗ r × −∇V2 . We want to evaluate the torque at θ = 0 and we

want only the the z component (that is aligned with the orbit normal and spin axis of M )

∂V2 
T = M∗
∂θ θ=0

GM∗2 R 5 ∂ 1
  
2
= k2 (3 cos (θ + δ) − 1) 

a a ∂θ 2 
 θ=0
2
 5
GM∗ R 
= k2 3 cos(θ + δ) sin(θ + δ)

a a 
θ=0
2
 5
3 GM∗ R
≈ k2 k2 sin(2δ).
2 a a

The secular part of the semi-diurnal (l = 2) term in the Fourier expansion of the
perturbing potential from point mass M∗ , gives a tidally induced torque
 5
3 GM∗2 R
T = k2 sin 2
2 a a

where k2 is the Love number and k2 sin 2 is known as the quality function. The quality
function is often approximated as k2 /Q where Q is a dissipation factor and k2 is a Love
number (but with out factors that depend on mass ratio or distance).
Because the torque is equal and opposite, this torque is exerted on the spinning body
M causing its spin down.
T ∼ I ω̇
where I is the moment of inertia about the spin axis and ω is the spin rate.
So far we have not made any approximations that depend upon whether M is larger or
smaller than M∗ . With M∗ > M the orbital period is P = 2π/n where mean motion
p
n = GM∗ /a3 .

66
Spin down times are often written in terms of P .
ω0
tdespin ≈
ω̇
where ω0 is the initial spin rate. We can approximate our spinning body as a sphere with
moment of inertia I ∼ 25 M R2 . If we chose ω0 = GM/R3 equal to a centrifugal break up
p

speed then the timescale for spin-down

GM 52 M R2
p r
ω0 GM/R3 I
tdespin ≈ = =
ω̇ I ω̇ R3 T
 3  9
P M 2 a 2 Q
=
15π M∗ R k2
Recent estimates for shear modulus and dissipation factor for icy or rocky rubble give
µQ ∼ 1011 Pa (e.g., Pravic et al. 2014). This is consistent with shear modulus µ = 1 GPa
and dissipation factor Q = 100.
Because angular momentum is conserved there is a relation between the spin down
rate and the drift rate
√ in orbital semi-major axis. If M∗ > M then the orbital angular
momentum L ≈ M GM∗ a and
r
M GM∗ M nȧa
T = L̇ = ȧ = .
2 a 2
With this we can estimate the drift rate in orbital semi-major axis ȧ from the torque
associated with spin down (in the case that the spinning object M < M∗ is low mass)

a M na2
tdrif t = =
ȧ  2T
1 M  a 5 Q
=
3n M∗ R k2
  
P M a 5 Q
=
6π M∗ R k2

4.6.1 Conventions for delay

The torque can be written explicitly in terms of a phase delay. The delay can be a function
of the frequency of tidal perturbations on the body which is n − ω. Some popular conven-
tions are Darwin-Kaula-Golreich constant phase delay giving a constant Q = 1/2 . Also
popular are constant time delay (Darwin-Mignard) giving 2 = (ω − n)∆t with constant ∆t
and in this case Q(ω − n) = 1/2 is frequency dependent. If the body is solid, another ap-
proach is to adopt a viscoelastic model for the body’s interior. The time dependent elastic
compliance can be used to predict a frequency dependent quality function (see numerous
papers by Michael Efroimsky).

67
4.6.2 Associated orbital drift rate during spin down
Using conservation of angular momentum
 5
3k2 M R
ȧ = sign(ω − n) na
Q M∗ a

4.7 Other timescales

The timescale for wobble damping or non-principal axis rotation to decay is
 n 4
twobble ∼ tdespin
ω
where ω is the spin of the moon.
Once tidally locked and at zero obliquity the timescale for eccentricity damping
2 M  a 5 Q −1
tedamp ≈ n (166)
21 M∗ R k2
The asphericity parameter p
α≡ 3(B − A)/C
and A < B < C are the body’s principal axis moments of inertia. For a triaxial ellipsoid
with body semi-axis lengths ab > bb > cb the moments of inertia are A = M 2 2
5 (bb + cb ),
B=M 2 2 M 2 2
5 (ab + cb ), C = 5 (ab + bb ) and the asphericity
r r
3(B − A) 3(a2 − b2 )
α= =
C a2 + b2
.
The frequency of free libration in the tidally locked state is

ωlib ∼ αn

If the body is nearly spherically symmetric, the libration frequency is slow.

5 Spin Resonances
Two approaches to spin resonance. You can expand the potential perturbation taking into
account the orientation of the body. Then with some work you force a Hamiltonian to look
like a pendulum model. Alternatively you compute the torque on the body and try to find
the equations of motion for the body spin/orientation. You also try to force the equations
of motion to look like a pendulum model.

68
5.1 Types of Spin Resonance
A number of possible types of resonance are possible. I will list them!

1. Spin-orbit resonance. Spin rate is a half multiple of mean motion giving

jn
w=
2
where n is the mean motion, w is the spin and j is an integer. Mercury is in a
3:2 spin-orbit resonance. If the body is not spherical and on an eccentric orbit, the
spin-orbit resonances overlap and the dynamics is chaotic. Spin orbit resonance can
be altitude unstable so when chaotic the body will start tumbling. Hyperion (very
elongated and in an eccentric orbit) is in such a tumbling state.

2. Spin-secular resonance. There is a relation between the planet’s spin precession rate
and an orbital precession rate (like Ω̇) induced by other planets. Saturn’s obliquity
is explained with such a model by Bill Ward and Doug Hamilton where the obliquity
is increased by drifting within such a resonance.

3. Spin-Spin resonance. Two non-spherical asteroids in orbit about each other. Their
spins are multiples of each other.

4. Spin-Binary resonance. Pluto-Charon’s binary orbital frequency could affect the spin
states of Pluto and Charon’s minor satellites.
j
(n − nb ) = w − nb
2
See recent work by Alexandre Correia.

5. Mean motion-spin precession. A mean motion resonance between two objects some-
how affects the spin of one of them. We saw this in some of our simulations for Pluto
and Charon’s minor satellites (and we wrote in 2017 about it).

6. Wobble excitation resonance. The same simulations also showed Pluto and Charon’s
binary perturbations exciting wobbling primarily in the minor satellite Kerberos (but
also sometimes in Styx).

There are probably more types of spin resonances!

69
 Figure 14: Angles relevant for spin orbit resonance

5.2 Spin-Orbit resonance

We take a non-spherical body m in a nearly circular orbit about mass Mp . The non-
spherical body is spinning about the body axis giving its smallest principal moment of
inertia C and with orbit normal and spin axis coinciding (undergoing principal axis rota-
tion). The equation of motion for body orientation angle θ (with respect to the inertial
frame)
3 GMp
C θ̈ + (B − A) sin(2(θ − f )) = 0 (167)
2 r3
p
where moments of inertia are A ≥ B ≥ C. Here n = GMp /a3 is the orbit mean motion,
in terms of the semi-major axis a. The angle θ describes body orientation (see Figure 14)
in an inertial frame and f is the orbital true anomaly. For a circular orbit f = nt. The
right term we recognize as torque arising from MacCullagh’s potential term. The left term
is the torque on the spinning body itself. Here we have only torque from the central mass
on the spinning body in orbit. There could be additional torques (such as tidal). This
setting is relevant for study of spin-orbit resonance for Mercury in orbit about the Sun or
Phobos in orbit about Mars.
Expanding f in eccentricity gives a series of terms depending on multiples of the mean
anomaly M . To first order f = M + 2e sin M , cos f = cos M + e(cos 2M − 1) and sin f =
sin M +e sin 2M . Writing sin 2(θ −f ) = sin 2θ cos 2f −cos 2θ sin 2f one can replace the true
anomaly with sums containing sines and cosines with the mean anomaly. We also need to
expand the radius as r−1 = (1+e cos f )
a(1−e2 )
.
After doing the expansion in eccentricity we should get
GMp 2
X
sin(2(θ − f )) = n H(p, e) sin(2(θ − pM ))
r3 p

70
summing over non-zero half integer p and with n the mean motion. The coefficients H(p, e)
are polynomials in eccentricity e. To lowest order in e

H(1, e) = 1
H(1/2, e) = −e/2
H(3/2, e) = 7e/2
H(2, e) = 17e2 /2

The higher the half integer |p|, the higher the power of eccentricity in the lowest term;
H(p, e) = O(e2|p−1| )
It is convenient to transfer to a new angle γ = θ − pM for a single half integer p. Then
the equation of motion (and taking only a single sine term) becomes

3 (B − A) hT i
γ̈ + n2 H(p, e) sin 2γ =
2 C C
Resonance capture can be studied using tidal evolution for the averaged torque hT i. Re-
cent work showing that all terms need to be taken into account to estimate the capture
probability which is strongly dependent on body shape (through the moments of inertia)
and eccentricity.
The resonances are separated by half integers of mean motion. q The spin synchronous
or tidal lock term has p = 1 and has libration width dependent on n B−A C . The 3/2 term
q
has libration width that depends n e B−A C . Using widths and distance between resonance
Jack Wisdom developed a resonance overlap criterion for chaotic behavior that was applied
to Hyperion. Apparently once the system is chaotic it is also attitude unstable and so will
start tumbling.

5.3 The dissipative spin-orbit resonance problem

It is convenient to assume that the torque on the spinning body due to tides is linearly
dependent on the angular velocity giving a modified spin orbit equation of motion (from
equation 167)
3 GMp
C θ̈ + (B − A) sin(2(θ − f )) = −Kd (θ̇ − n) (168)
2 r3
where the functions L(e) and N (e) are averaged over the orbit and the dissipation factor
 2
3n k2 R M∗
Kd = (169)
C Q a M

71
6 Problems
1. Reference frames for a symmetric top
Consider a rigid axi-symmetric body with no external forces. In a coordinate frame
aligned with body’s axis of symmetry the moment of inertia matrix has 3 eigenvalues
Ik , I⊥ , I⊥ and two of the them are the same. If the axis of symmetry lies along the z
axis then Ik = Iz . The body is spinning but with angular momentum and spin not
parallel to z or in the xy plane. The body is not spinning about a principal body
axis. Euler’s equations are

I⊥ ω̇x + (Ik − I⊥ )ωz ωy = 0 (170)

I⊥ ω̇y − (Ik − I⊥ )ωz ωx = 0 (171)
Ik ω̇z = 0 (172)

for spin ω in the body’s reference frame. A solution is

ωx = ω sin α cos(Ωt) (173)

ωy = ω sin α sin(Ωt) (174)
ωz = ω cos α (175)
Ik − I⊥
 
Ω = ω cos α (176)
I⊥

The angular momentum vector (in the body’s frame)

Lx = I⊥ ω sin α cos(Ωt) (177)

Ly = I⊥ ω sin α sin(Ωt) (178)
Lz = Ik ω cos α (179)

a) Show that in the inertial frame that the angular momentum vector is constant.
b) What is the spin vector in the inertial frame?
c) What are the trajectories of the body x, y, z axes as seen in an inertial frame?
Hint:
dL ∂L
= +ω×L
dt ∂t
2. Quadrupole moment and moment of Inertia:
The Quadrupole moment tensor of an extended mass mass distribution with density
ρ(x) is Z
Qαβ ≡ ρ(x)(3xα xβ − δαβ r2 ) d3 x

72
 where r2 = 2
P
i xi . The moment of inertia tensor is
Z
Iαβ ≡ ρ(x)(δαβ r2 − xα xβ ) d3 x.

Consider a point mass of mass M at the origin. And consider an extended body of
mass m at coordinate position x = R. The gravitational potential of m is approxi-
mately
Gm GQαβ Rα Rβ
Φ(R) = − −
R 2R5
using summation notation (every pair of indices is summed over each coordinate) and
only expanding to the quadrupolar term.
Show that the gravitational force on point mass M from extended mass m is F with
components Fi
Fi GmRi GQαi Rα 5 Ri
= − + GQαβ Rα Rβ
M R3 R5 2 R7
Show that the torque on M has components
GRj Qkα Rα
Ti = ijk Rj Fk = −M ijk
R5
where ijk the Levi-Civita symbol.
What is the torque on m?
Show that we can replace Q with I giving
Rj GIkα Rα
Ti = 3M ijk
R5
(hint: use symmetry or antisymmetry to drop terms containing the trace).
Show that this can be derived from MacCullagh’s formula.

3. Tumbling or wobbling decay due to internal dissipation

It is common to assume that celestial bodies are spinning about a principal axis
because the timescale for wobble to decay due to internal dissipation is shorter than
the tidal spin-down time.
Explain why a stiff but elastic body dissipates energy when it is tumbling.
The wobble decay timescale for a spinning asteroid is
3GCQ
twobble ∼
w3 k2 R5
where R is a mean body radius, C is the maximum moment of inertia about a
principal body axis, Q is a unitless energy dissipation parameter, w is the initial spin

73
 rate and k2 is a Love number. The timescale does not depend upon an external tidal
field. This formula I think is from Peale 1977 in a book called Planetary Satellites
that I cannot get an electronic version of, though it might be in the library.
A similar formula is by Burns and Safronov (1973) (Burns, J.A., and Safronov, V.S.
1973. Asteroid nutation angles. Monthly Notices of the Royal Astronomical Society
of London, Vol. 165, pp. 403 - 411). and is

µQ
τ∼ A
ρR2 w3
with constant A ∼ 100. Here µ, Q, ρ, R, w are the mean shear rigidity, quality factor,
density, body radius and spin rate.
The two formulas are consistent with each other if we use C ∼ mR2 and k2 ∼
Gm2 R−4 µ−1 .
Explain the form of this timescale to order of magnitude.
Hints: Estimate variations in stress, σ, and strain  on the body. Recall that stress is
strain times elastic modulus σ ∼ µ. Energy dissipation per unit volume of an elastic
material ė ∼ σ .
˙

4. To a spin synchronous state and dissipation timescales

Suppose one of Saturn’s small moons, such as Pan, suffers an impact, giving it an
obliquity, or exciting libration, or exciting tumbling or increasing or decreasing its
spin. The same impact could increase orbital eccentricity and orbital inclination.
How fast does the moon spin down?
Once in a spin synchronous state (tidally locked) how fast does eccentricity decay?
Once the spin reaches the orbital mean motion, how fast does the obliquity decay?
Would the orbital inclination decay and if so, how fast?
Pan suffers a drift in its semi-major axis, ȧ, primarily due to tides excited in Saturn.
Which decays are set by dissipation in Saturn and which by dissipation in Pan?
Note orbital eccentricity and inclination can also decay due to excitation of waves
into Saturn’s rings.

5. Averaging a binary
Consider a rapidly spinning binary comprised of masses m1 , m2 separated by distance
a.
Compute the moment of inertia matrix of a ring with mass m and radius r about its
center of mass.

74
 If our binary is rapidly spinning (rotation in a plane) while maintaining the separation
a, what is the average value of the moment of inertia matrix (about the center of
mass)? In other words what is
Z P
1
hIi = dt I
P 0

where P is the rotation period.

6. Spin Precession
Consider a non-round body of mass m spinning about its maximum principal axis,
the direction of its maximum moment of inertia, equivalently the body’s minimum
axis.

p mass M with M  m. It is in a circular orbit with radius R, mean

It orbits a point
motion n = G(M + m)/R3 .
The angle between the body’s spin axis and orbit normal can be called the obliquity,
. Assume that the spin w  n.
As a function of , what is the torque on m caused by the point mass M ? What are
the components of the torque vector T? Write these in terms of the orientation of
the spin vector. Consider the component of the torque that is perpendicular to the
spin vector.
Show that the precession rate is
S
Ω̇ = − cos 
w
with
3 C − (A + B)/2
S = n2
2 C
where the moments of inertia (about principal axes) are C > B > A.

7. Tidal evolution
For a two-body system of mass M and m the total orbital angular momentum is
p
Lo = µ G(M + m)a(1 − e2 )

where a and e are the semi-major axis and eccentricity and µ = mM/(m + M ) is the
reduced mass. Here L is not angular momentum per unit mass. The total orbital
energy is
G(M + m)
Eo = −µ
2a

75
 With spin parallel to orbital angular momentum and principal axis rotation, the spin
of m has angular momentum
Lsm = Cm ωm
where Cm is the moment of inertia about the spin axis for mass m and ωm is m’s
spin. The kinetic energy associated with m’s spin is
Cm 2
Esm = ω
2 m
We can describe the spin of body M with moment of inertia CM and spin ωM .
Body deformation associated with tides dissipate energy in m so total energy (orbit
+ spin) is not conserved however the total angular momentum (orbit + spin) is
conserved.

(a) Consider m only dissipating energy (let M be a point mass) and ωm 6= n where
n is the mean motion. When ωm = n the system is tidally locked or in a
spin-synchronous state. What sign is ȧ? Is ė important?
Hint: consider the sign of ωm − n? Is m speeding up or slowing down?

(b) Consider m only dissipating and ωm = n but e > 0. What sign is ȧ? Is ė
important? If so what sign is ė?

(c) Io is drifting outwards (ȧ > 0) but is tidally locked with Jupiter. Why? What
mechanism maintains Io’s non-zero eccentricity? What type of tide is responsible
for Io’s vulcanism? Jupiter is spinning and Io is in an eccentric orbit.

(d) The Moon is drifting away from the Earth. Which type of tide is responsible?

(e) Hyperion never reaches a spin-synchronous state. It is tumbling due to strong

spin-orbit resonances and attitude instability in them. How much higher is the
dissipation rate (by what factor) in Hyperion in the tumbling state compared
to the spin-synchronous state? Hyperion’s eccentricity is 0.123.
Hint: consider energy damping rate during tidal circularization versus spin
down.

8. Radial distribution of tidally generated internal heat

Consider an isotropic, homogeneous visco-elastic spherical spinning body of mass m
that is orbiting mass M . Consider the tidally generated heat per unit volume .
˙ The
interior of the body is heated more than the material near the surface. Why is that?
Can you estimate how the heat per unit volume scales with radius?

76
9. Excitation of Tumbling
Consider an axi-symmetric torque free body initially spinning about a principal body
axis. Recall (equations 67, 69) with precession rate
#1
Ik2
" !
2
2
φ̇ = ω 1 + 2 − 1 cos α (180)
I⊥

where α is the angle between spin vector and body principal axis. Since we are
considering spinning about a principal body axis consider α small. In the limit of
small α and small θ (angle between angular momentum vector and body axis)

φ̇ Ik
≈
ω I⊥
Ik
 
ψ̇
≈ 1−
ω I⊥

Suppose you can perturb the body with an oscillating torque (a torque that averages
to zero). What frequency would excite tumbling?

10. Wobble damping

A torque free, and freely rotating asteroid has constant angular momentum but its
total energy decreases due to elastic dissipation and heat loss.
Consider a body with spin Ω = (Ωl , Ωi , Ωs ) for long, intermediate and short body
axes with moments of inertia Il , Ii , Is . The body initially is not spinning about a
principal axis.
a) What is the initial rotational kinetic energy? What is the angular momentum?
b) What are the possible final spin states after the wobble has damped away?
In wobbling damping computations, we often assume that the moment of inertia is
not time dependent. If normal modes are excited in the asteroid then this is possibly
a bad assumption. Asteroids tend to rotate with periods of a few hours to a days.
c) Why is it reasonable to assume that normal modes are not excited during tumbling
damping of asteroids?
Hint: estimate the frequency of the slowest normal modes using an elastic modulus
and assuming an asteroid size. Compare this frequency to a typical spin period.

11. Accretion on to Pan

Suppose Pan is accreting material through one of its Lagrange points at a rate of Ṁ .
What is the rate of change of Pan’s angular momentum?

77
 What accretion rate would keep Pan spinning and no longer in tidal lock with Saturn?
How fast does Pan’s moment of inertia matrix change due to accretion?

12. Spin orbit resonance

We take a non-spherical body m in a nearly circular orbit about mass M > m that
is spinning about its principal moment of inertia and with orbit normal and spin
axis coinciding (this is principal axis rotation). The equation of motion for body
orientation angle θ
3 GM
C θ̈ + (B − A) sin(2(θ − f )) = 0
2 r3
p
where moments of inertia are A ≥ B ≥ C. Here n = GM/a3 is the orbit mean
motion, in terms of the semi-major axis a. The angle θ describes the orientation of
m in an inertial frame and f is the orbital true anomaly and describes the position
of m in its orbit about M .
Show that the right term arises from MacCullagh’s potential term and is due to the
torque on the non-spherical body by M .
Show that the left term is just the torque on the spinning body itself.

78