Chapter 7. Rotation in 3 dimensions.

1 Concept of rigid body

A rigid body consists of an arbitrary number of particles such that their relative
distances remain constant. Therefore the entire mass distribution in space of a
solid rigid will remain constant.
A rigid body can be composed of a finite number N of individual particles of
mass (mi , i = 1, . . . , N ). In such case, we refer to the rigid body as a (discrete
mass distribution), and its total mass M can be expressed as:
N
X
M= mi (1)
i=1

Alternatively, a rigid body can consist of continuous medium composed of

infinitesimal mass elements, dm occupying infinitesimal volumes dV , so that we
can define a density function ρ (not necessarily constant throughout the body),
that follows the relation ρ = dm
dV . in this case we will refer to a continuous mass
distribution, and its total mass can be expressed as:
Z
M = ρ dV (2)

2 Angular momentum of a system of particles

The angular momentum is a physical magnitude which represents the rotational
momentum of a rotating body. It plays a role in rotating systems analogous to
that of linear momentum in systems undergoing translations.
Angular momentum must always be referred to a certain point O in space
(typically, the origin of a frame of reference, or the centre of mass of the system).
If a particle of mass mi has velocity ~vi , its linear momentum will of course be
p~ = m~v . If the position of mi with respect to O is ~r, , the angular momentum
of mi will be the cross product:

~ = ~r ∧ p~ = m · ~r ∧ ~v
L (3)
The angular momentum of an entire system composed of N point-like masses
will be:
N
X N
X
L~ = mi · r~i ∧ v~i = mi · r~i ∧ (~
ω ∧ r~i ) (4)
i=1 i=1

where ω (rad/s) is the angular velocity of the system in the considered frame
of reference.

1
Figure 1: Angular momentum with respect to the origin of a point particle of
mass m, located at a distance ~r from the origin and moving with velocity ~v .

~ ∧ (B
Using the vector calculus expression: A ~ ∧ A)
~ = (A
~ · A)
~ ·B
~ − (A
~ · B)
~ · A,
~
we can rewrite:
N
X
~ =
L mi · [(~
ri · r~i ) · ω
~ − (~
ω · r~i ) · r~i ] (5)
i=1
The angular momentum vector for the system in Cartesian coordinates
(Lx , Ly , Lz ) is expressed as:
~ = Lx~i + Ly~j + Lz~k
L (6)
Using that:
~r = x~i + ‘y~j + z~k

~ = ωx~i + ‘ωy~j + ωz~k

ω

=⇒ ~r2 = x2i + yi2 + zi2 , ~ · ~r = ωx xi ‘ωy yi + ωz zi

ω
the general expression for the angular momentum for the axis x is:
N
X
Lx = mi · [(x2i + yi2 + zi2 ) · ωx − (ωx · xi + ωy · yi + ωz · zi ) · xi )]
i=1
(7)
N
X
=⇒ mi · [(yi2 + zi2 ) · ωx − xi · yi · ωy − xi · zi · ωz ]
i=1

Analogously:
N
X
Ly = mi · [(x2i + zi2 ) · ωy − xi · yi · ωx − yi · zi · ωz ] (8)
i=1

2
 N
X
Lz = mi · [(x2i + yi2 ) · ωz − yi · zi · ωy − xi · zi · ωx ] (9)
i=1

Which can be written as:

   PN 2 2
PN PN   
Lx i=1 mi · (yi + zi ) − i=1 mi · xi · yi − i=1 mi · xi · yi ωx
N PN 2 2
PN
− i=1 mi · yi · zi  · ωy 
Ly  =  − P
mi · xi · yi i=1 mi · (xi + zi )
 
Pi=1
N N PN
Lz 2 2 ωz
P
− i=1 mi · xi · zi − i=1 mi · yi · zi i=1 mi · (xi + yi )
| {z }
tensor of inertia (I)
(10)
Where:
(i) The diagonal elements of the tensor of inertia (Iii ) are the moments of
inertia with respect to axis ’i’, e.g. (Ixx ) would be the moment of inertia
about X.
(ii) The off-diagonal elements of the tensor of inertia, Iij with i 6= j are the
products of inertia. They represent cross contributions from two axes and
somehow reflect the symmetry of the system with respect to the axes of
the frame of reference with respect to which {I} is calculated. If symmetry
is perfect with respect to the 3 axes of inertia, then the products of inertia
will be zero and the inertia tensor will be diagonal.

If we now consider a continuous distribution instead of a discrete distribu-

tion, we can rewrite by analogy:

mi −→ dm
N
X Z
−→
i=1 M

Therefore, the moments of inertia will be:

N
X Z
Ix,x = mi · (yi2 + zi2 ) =⇒ Ix,x = (y 2 + z 2 ) · dm (11)
i=1

N
X Z
Iy,y = mi · (x2i + zi2 ) =⇒ Iy,y = (x2 + z 2 ) · dm (12)
i=1

N
X Z
Iz,z = mi · (x2i + yi2 ) =⇒ Iz,z = (x2 + y 2 ) · dm (13)
i=1

3
 And the products of inertia will be:
N
X Z
Ix,y = − mi · xi · yi = Iy,x =⇒ Ix,y = − x · y · dm = Iy,x (14)
i=1

N
X Z
Ix,z = − mi · xi · zi = Iz,x =⇒ Ix,z = − x · z · dm = Iz,x (15)
i=1

N
X Z
Iy,z = − mi · yi · zi = Iz,y =⇒ Iy,z = − y · z · dm = Iz,y (16)
i=1

The kinetic energy due to the rotation of an object expressed in terms of the
tensor of inertia is given by:
1 1 ~
Krotation = ~ T · I|{z}
·ω ·ω
~ =⇒ Krotation = · ω
~ ·L (17)
2 2
~
L

Note that, just as the moment of inertia relevant for rotation about a fixed
axis, was given for a specific object and a specific axis, the tensor of inertia is
given for a specific object and for a specific frame of reference (origin and set of
Cartesian axes which follow the right hand’s rule).

By construction, the tensor of inertia is real and symmetric, and

thus it can always be diagonalized. Mathematically, the results of the
diagonalization of a 3x3 matrix is a set of 3 eigenvalues and 3 eigenvectors.
From the physical point of view, it means that we can always find a set of
perpendicular axes,
- with the same origin as in our initial frame of reference,
- with directions given by the eigenvectors,

and such that the tensor of inertia in that set of axes will be diagonal. The
corresponding moments of inertia of the new diagonal tensor will be the eigen-
values. Note that, after the diagonalization process, the origin of the frame of
reference remains constant, and the set of axes which makes a tensor of inertia
diagonal is referred to as a set of Principal Axes of inertia. It can be naturally
inferred that for an arbitrary object and an arbitrary point in space,
we can always find a set of 3 principal axes of inertia with origin at the
given arbitrary point, and such that the object will be able to rotate
with constant ω around any of these axis without the need of external
torques. Actually, in such case, the angular momentum vector will be parallel
to the angular velocity: ω ~
~ k L.

4
HEADS UP!
• If an object rotates about a principal axis of inertia, its angular velocity
will be parallel to its angular momentum.

• The object can rotate with constant angular velocity about a principal
axis of inertia without the need of external torques.
• If an object is to rotate with constant angular velocity about an axis
which is not a principal axis of inertia, it will require the effect of external
torques.

3 Steiner’s (parallel axis) theorem in matrix form

The theorem in its 1-dimensional formulation proves that, given an axis which
passes through the center of mass of a solid, and another axis parallel to the
first one, the moment of inertia of both axes are related as as follows:

IP = ICM + m · r2 (1D) (18)

Where ICM Is the moment of inertia of the solid given for an axis going through
the center of mass, IP is the moment of inertia of the solid given for the parallel
axis, m is the solid mass and r is the perpendicular distance between the two
axes.
Steiner’s theorem in its 3-dimensional version can be written as:

y2 + z2
 
−x · y −x · z
x2 + z 2
 
IP = ICM + m ·  −y · x −y · z  (3D) (19)
−z · x −z · y x2 + y 2

Where:
(i) ICM is the tensor of inertia of the solid given for a frame of reference
centred at the center of mass of a solid, F RCM .
(ii) IP is the tensor of inertia of the solid given for a frame of reference parallel
to the F RCM , F RP .
(iii) m is the mass of the solid.
(iv) x, y and z are the coordinates of the origin of F RP with respect to
F RCM .

5
 ZP

ZCM YP
XP
OP

~
R

YCM
XCM OCM

Figure 2: Sketch of parallel frames of reference (FRCM and FRP) in Steiner’s

therorem in matrix form.

3.1 Rotation matrices:

Rotation matrices can be active, when they refer to the motion of a vector with
respect to a frame of reference (as below), or passive, when they refer to the
motion of a frame of reference with respect to a fixed vector (inverting the ma-
trices below). An active matrix transform a vector by rotating it. A passive
matrix leaves the vector unchanged and, instead, rotates the frame of reference
accordingly.

An ACTIVE rotation matrix of an angle θ about the X axis is:

 
1 0 0
About x axis: Rx (θ) = 0 cos(θ) − sin(θ) (20)
0 sin(θ) cos(θ)

An ACTIVE rotation matrix of an angle θ about the Y axis is:

 
cos(θ) 0 sin(θ)
About y axis: Ry (θ) =  0 1 0  (21)
− sin(θ) 0 cos(θ)

6
An ACTIVE rotation matrix of an angle θ about the Z axis is:
 
cos(θ) − sin(θ) 0
About z axis: Rz (θ) =  sin(θ) cos(θ) 0 (22)
0 0 1

We can apply an ACTIVE rotation matrix R on a vector ~v as follows:

~vrotated = R~v

We can apply two consecutive ACTIVE rotations (first rotation R1 and then
R2 ) as follows:
~vrotated = R2 R1~v

A PASSIVE rotation matrix of an angle θ about the X axis is:

 
1 0 0
About x axis: Rx (θ) = 0 cos(θ) sin(θ)  (23)
0 − sin(θ) cos(θ)

A PASSIVE rotation matrix of an angle θ about the Y axis is:

 
cos(θ) 0 − sin(θ)
About y axis: Ry (θ) =  0 1 0  (24)
sin(θ) 0 cos(θ)

A PASSIVE rotation matrix of an angle θ about the Z axis is:

 
cos(θ) sin(θ) 0
About z axis: Rz (θ) =  − sin(θ) cos(θ) 0 (25)
0 0 1

Analogously to what happens with active rotation matrices:

- We can apply an PASSIVE rotation matrix R on a vector ~v as follows:

~vrotated = R~v

- We can apply two consecutive PASIVE rotations (first rotation R1 and

then R2 ) as follows:
~vrotated = R2 R1~v

7
 Rotation matrices (ACTIVE AND PASSIVE) are orthogonal, and thus their
inverse is equal to their transposed form: R−1 = RT . Matrices form a non-
Abelian group with the product, therefore the product of matrices is non-
commutative. This means that the order in which we apply matrix rotations
does matter.
~vrotated = R2 R1~v 6= R1 R2~v
Just as with vectors, rotation matrices can be applied
 on other matrices and,
specifically, can be applied to the tensor of inertia I :

I = R I RT
 

8
4 Euler’s equations
Let us recall that the tensor of inertia of a body depends on the mass distribution
of the body itself with respect to the system of reference with which tensor of
inertia has been calculated. If we choose a fixed frame of reference, as the
body moves, its tensor of inertia will change. However, if we consider a system
of reference which is attached to the body, the tensor of inertia will remain
constant, regardless the motion of the body.
We can use this fact and the relation between the time variation of the
angular momentum L ~ (in fixed system of reference SR) and the time variation
of the angular momentum L ~ ∗ (in the rotating system of reference SR):

~
dL ~
d∗ L
= + ω ~
~ ∧L (26)
dt
|{z} dt
|{z}
Time variation in non-rotating SR Time variation in rotating SR

According to Newton’s second law for rotating objects:

~ τ1
dL
= ~τ = τ2 (27)
dt |{z}
sum of all external torques τ3

~ = I ·ω
 
We also know that L ~ because I is constant in the system of
reference that rotates with the body. Besides, it is clever to chose a frame of
reference attached to the body which is also a system of principal axes of inertia,
so that the associated tensor of inertia will be
 diagonal. The unitary vectors
associated to this frame of reference will be e~1 , e~2 , e~3 ), and we will use the
standard notation x↔1, y↔2, z↔3. Then we can express:

~ I1 ω1
d∗ L
~˙ = I1 ω2

= I ·ω (28)
dt
I3 ω3
Therefore we can write:
~
dL
~˙ + ω
 
~τ = = I ·ω ~ ∧( I ·ω
~ ), (29)
dt
with

~i ~j ~k
 

ω∧( I ·ω
]~ ~ ) =  ω1 ω2 ω3  =
I 1 · ω1 I2 · ω2 I3 · ω3

= (I3 − I2 ) · ω2 · ω3 · ~i + (I1 − I3 ) · ω1 · ω3 · ~j + (I2 − I1 ) · ω1 · ω2 · ~k

9
 Therefore:
τ1 = I1 · ω˙1 + (I3 − I2 ) · ω3 · ω2
τ2 = I2 · ω˙2 + (I1 − I3 ) · ω1 · ω2 (30)
τ3 = I3 · ω˙3 + (I2 − I1 ) · ω2 · ω1
The former are the so-called Euler equations for rotation, and are the key
to understand 3-dimensional rotation. Usually if the body rotates about a fixed
point =⇒ The origin of the SR’ is @ that point. If the body is unconstrained
can move freely =⇒ The origin of the SR’ is taken @ the center of mass of the
body.
We can use Euler equations to express the energy variation of the solid rigid:

multiplying by ω
~
~˙ + ω ~˙
  
~τ = I · ω ~ ∧( I ·ω~ ) −−−−−−−−−−→ ω ~ · ~τ = ω
~ · I ·ω

The tensor of inertia I is symmetric, therefore:

1 d0    dK
~˙ = ω
~˙ · I · ω
 
~ · I ·ω
ω ~ = · · ω
~ · I ·ω
~ =
2 dt dt
Therefore:
dK
~ · ~τ
=ω (31)
dt
It is also easy to prove that, unless external torques are applied, the body
can only rotate with constant ω about a principal axis of inertia.

Proof:
~˙ + ω
  
~τ = I · ω ~ ∧( I ·ω ~ is constant ⇒ ~τ = ω
~ ). Therefore, if ω ~ ∧( I ·ω ~)
~ ~
 
Under these conditions ~τ = 0 ⇐⇒ ω ~ ∧( I ·ω ~ ) = 0 ⇐⇒ ω
~ || I · ω~ =L
This will only happen when ω ~ goes along a principal axis of inertia, as we
wanted to prove.

PLEASE SEE MANUSCRIPT NOTES FOR EXAMPLES.

10