3 Kinematics of Rotational Motion

But though the professed aim of all scientific work is to unravel the secrets of nature,
it has another effect, not less valuable, on the mind of the worker. It leaves him in
possession of methods which nothing but scientific work could have led him to invent.
James Clerk Maxwell, The Theory of Molecules

The dynamics of rigid bodies is a chapter of classical mechanics that deserves to be

highlighted not only owing to its intrinsic physical interest but also because it involves
important mathematical techniques. Before, however, embarking on the study of dynamics,
it is necessary to formulate efficacious methods to describe the motion of rigid bodies.
A considerable space will be dedicated to the study of rotational kinematics in the
perspective that several of the mathematical tools to be developed are of great generality,
finding wide application in other domains of theoretical physics.

3.1 Orthogonal Transformations

A rigid body has, in general, six degrees of freedom. Obviously, three of them correspond
to translations of the body as a whole, whereas the other three degrees of freedom describe
the orientations of the body relative to a system of axes fixed in space. A simple way
to specify the orientation of the rigid body consists in setting up a Cartesian system of
axes fixed in the body, which move along with it, and consider the angles that these axes
make with axes parallel to those that remain fixed in space, represented by dashed lines in
Fig. 3.1.

Direction Cosines
Let  be a Cartesian coordinate system (x1 , x2 , x3 ) with corresponding unit vectors
ê1 , ê2 , ê3 representing axes fixed in space, and let   be a Cartesian coordinate system
(x1 , x2 , x3 ) with unit vectors ê1 , ê2 , ê3 whose axes remain attached to the rigid body, as in
Fig. 3.2. An arbitrary vector g can be expressed in terms of the basis  or in terms of the
basis   :
3
g = g1 ê1 + g2 ê2 + g3 ê3 = gj êj , (3.1)
j=1

3
g = g1 ê1 + g2 ê2 + g3 ê3 = gj êj . (3.2)
j=1
86
 87 Orthogonal Transformations

x3
x2´

x3´

x1´

x2

x1
Fig. 3.1 Axes fixed in space and primed axes attached to a rigid body.

x3

x3´
x2´
e^ 3
e^2´
e^ 3´
x2
e^ 2
e^ 1
e^ 1´

x1
x1´
Fig. 3.2 Unit vectors ê1 , ê2 , ê3 fixed in space and unit vectors ê1 , ê2 , ê3 attached to the rigid body.

Of course, the components of g in the bases  and   are related to one another. Indeed,

3 
3
gi = êi · g = gj êi · êj ≡ aij gj , i = 1, 2, 3 , (3.3)
j=1 j=1

where
aij = êi · êj = cosine of the angle between axes xi and xj . (3.4)
The quantities aij – nine in all – are called direction cosines of   relative to . Therefore,
the components of g in   are expressed in terms of its components in  by means of a
linear transformation whose coefficients are the direction cosines.
88 Kinematics of Rotational Motion

The nine quantities aij cannot be mutually independent, for three quantities suffice to
define the orientation of   relative to . In order to obtain the conditions to which the
direction cosines must be subject, all one has to do is take into account that the magnitude
of a vector does not depend on the basis used to express it:

3 
3
g·g= gi gi = gi gi . (3.5)
i=1 i=1

Making use of Eq. (3.3) we find

⎛ ⎞   3 
 3 3 3 3 
3 
 
gi gi = ⎝ aij gj ⎠ aik gk = aij aik gj gk . (3.6)
i=1 i=1 j=1 k=1 k,j=1 i=1

With the help of the Kronecker delta symbol δjk introduced in Appendix A, it is possible
to write
3 
3
gi gi ≡ g21 + g22 + g23 = δjk gj gk . (3.7)
i=1 k,j=1

Comparing equations (3.6) and (3.7), using (3.5) and the arbitrariness of g1 , g2 , g3 , one
infers
 3
aij aik = δjk . (3.8)
i=1

These are the conditions obeyed by the direction cosines, showing that not all of them are
independent.

Orthogonal Matrices
Equation (3.3) is a compact representation of the three equations
g1 = a11 g1 + a12 g2 + a13 g3 ,
g2 = a21 g1 + a22 g2 + a23 g3 , (3.9)
g3 = a31 g1 + a32 g2 + a33 g3 .
Equations of this type can be made much more concise by means of matrix notation.
Definining the matrices1
⎛ ⎞ ⎛  ⎞ ⎛ ⎞
g1 g1 a11 a12 a13
g = ⎝ g2 ⎠ , g  = ⎝ g2 ⎠ , A = (aij ) = ⎝ a21 a22 a23 ⎠ , (3.10)
g3 g3 a31 a32 a33
Eqs. (3.9) can be written in the form
g  = A g . (3.11)

1 In this chapter square matrices will be typically denoted by bold capital calligraphic letters such as A. A vector
written as a column matrix will be denoted by a bold small letter. Whenever necessary, a subscript will be added
to indicate the basis in which the vector is expressed, such as in g .
89 Orthogonal Transformations

On the other hand, noting that aij = (AT )ji , where AT denotes the transpose of A,

Eq. (3.8) is equivalent to 3i=1 (AT )ji (A)ik = δjk . Thus, taking into account the definition
of matrix product, Eq. (3.8) can be rewritten in the matrix form
AT A = I , (3.12)
where
⎛ ⎞
1 0 0
I = (δjk ) = ⎝ 0 1 0 ⎠ (3.13)
0 0 1
is the 3 × 3 identity matrix.
Exercise 3.1.1 Prove that the inverse transformation of (3.3) is

3
gi = aji gj (3.14)
j=1

and, using the invariance of g · g, prove that


3
aji aki = δjk . (3.15)
i=1

Show that, in matrix notation, this equations takes the form2

AAT = I . (3.16)
Prove, further, the direct and inverse transformation laws

3 
3
êj = ajl êl , êj = alj êl . (3.17)
l=1 l=1

Bases {ê1 , ê2 , ê3 } and {ê1 , ê2 , ê3 } are made up of mutually orthogonal unit vectors, so
the following equations hold:
êj · êk = êj · êk = δjk . (3.18)
Derive (3.8) and (3.15) from (3.17) and (3.18).
The transformation matrix between Cartesian systems obeys equations (3.12) and (3.16),
which are equivalent to
A−1 = AT . (3.19)
Any matrix that satisfies this equation is said to be an orthogonal matrix, and the
associated linear transformation (3.11) is called an orthogonal transformation.

2 For finite matrices equations (3.12) and (3.16) are equivalent – that is, one is true if and only if the other is true
(see Appendix D). This property does not extend to infinite matrices.
 90 Kinematics of Rotational Motion

x2
x2´

r
x 1´
f

f
x1
Fig. 3.3 Rotation in the plane.

Example 3.1 In the case of rotations in the plane (Fig. 3.3), we have
x1 = x1 cos φ + x2 sin φ , (3.20a)
x2 = −x1 sin φ + x2 cos φ . (3.20b)
The matrix associated with this orthogonal transformation is

cos φ sin φ
A= , (3.21)
− sin φ cos φ
and a single parameter – the angle φ – completely specifies the transformation.

Exercise 3.1.2 Using definition (3.4) and Eq. (3.3) with g = r, derive (3.20a) and (3.20b).
By computing the product AAT , check that the matrix A given by (3.21) is
orthogonal.

The matrix A is associated with a linear transformation which, according to Eq. (3.11),
acts on the components of a vector in  and transforms them into the components of the
same vector in   . This interpretation of Eq. (3.11), in which the vector stays the same and
what changes is the coordinate system, is called the passive point of view. Alternatively,
A may be thought of as an operator which transforms the vector g into another vector g ,
both considered expressed in the same coordinate system. This way of interpreting the
transformation performed by matrix A is known as the active point of view, and in several
occasions, as will be seen, turns out to be more convenient than the passive point of
view. Of course, the specific operation represented by A will depend on the interpretation
adopted. A rotation of the coordinate axes keeping the vector fixed has the same effect
as keeping the axes fixed and rotating the vector by the same angle in the opposite sense.
In Example 3.1, since A represents a counterclockwise rotation of the coordinate axes by
angle φ, it will correspond to a clockwise rotation of any vector by the same angle φ.
91 Orthogonal Transformations

Group of Orthogonal Matrices

It is worthwhile to examine sucessive orthogonal transformations. Let B = (bij ) be the
transformation matrix from  to   , and A = (aij ) be the transformation matrix from  
to   , so that

gi = bij gj , (3.22)
j

gk = aki gi . (3.23)
i

Consequently,
 
   
gk = aki bij gj = aki bij gj ≡ ckj gj , (3.24)
ij j i j

where the ckj defined by


ckj = aki bij (3.25)
i

are the elements of the matrix C which performs the direct transformation from  to   .
According to the definition of matrix product, Eq. (3.25) is equivalent to

C = AB . (3.26)

It is essential to observe the order of factors in this last equation because matrix
multiplication is not commutative. The matrix associated with the first transformation
appears at the rightmost position in the product, immediately to the left comes the matrix
of the second transformation and so on. It is intuitively clear that two successive rotations
are equivalent to a single rotation. So, if A and B are orthogonal matrices, their product
must also be orthogonal. In order to prove that C is orthogonal, it suffices to show that C
satisfies (3.16). As the first step, note that

CC T = AB(AB)T = ABBT AT , (3.27)

where we have used the identity (AB)T = BT AT . With the use of the orthogonality of
B and A we can write
CC T = AIAT = AAT = I , (3.28)

which verifies the orthogonality of C.

For any orthogonal matrix its unique inverse is just its transpose, and it is immediate
that the transpose of an orthogonal matrix is also orthogonal because we can write
(AT )T AT = AAT = I, which shows that (3.12) is satisfied with A replaced by AT .
The set of orthogonal matrices enjoys the following properties:

(O1) If A and B are elements of the set, so is the product AB.

(O2) The product is associative – that is, A(BC) = (AB)C holds true for any elements
A, B, C of the set.
92 Kinematics of Rotational Motion

(O3) The set contains an element I, called the identity, such that for any element A of the
set, AI = IA = A.
(O4) For each element A of the set there exists a unique element B of the set such that
AB = BA = I. The element B is called the inverse of A and denoted by A−1 .

The first and fourth properties have already been proved, the second is satisfied because
matrix multiplication is associative and the third is obviously true.
A set of elements equipped with a composition law − also called multiplication or
product − having the four properties above make up a group (Hamermesh, 1962; Tung,
1985). The group of 3 × 3 orthogonal matrices is denoted by O(3). Group theory plays an
important role in various branches of contemporary physics, most notedly in elementary
particle physics.

3.2 Possible Displacements of a Rigid Body

The determinant of an orthogonal matrix can only assume certains restricted values.
Indeed, taking the determinant of (3.16) there results

(det A)(det AT ) = 1 ⇒ (det A)2 = 1 , (3.29)

where we have used the following well-known properties of the determinant:

det(AB) = (det A)(det B) , det AT = det A . (3.30)

Therefore, the determinant of an orthogonal matrix can only be +1 or −1. A simple

orthogonal matrix with determinant −1 is
⎛ ⎞
−1 0 0
I = −I = ⎝ 0 −1 0 ⎠. (3.31)
0 0 −1
The transformation effected by I is

g1 = −g1 , g2 = −g2 , g3 = −g3 , (3.32)

which corresponds to an inversion of the coordinate axes.

An inversion transforms a right-handed Cartesian system into a left-handed one. It is
clear, therefore, that an inversion does not represent a possible displacement of a rigid
body: once a right-handed Cartesian system of axes is attached to the body, no physical
displacement of the body can change the mutual orientation of those axes unless the body is
deformable, which is not the case by hypothesis. On the other hand, any orthogonal matrix
with determinant −1 can be written in the form A = (−I)(−A) ≡ IB where det B ≡
det(−A) = +1 – that is, A contains an inversion of the axes and does not represent
a possible displacement of the rigid body. Orthogonal matrices with determinant +1 are
said to be proper, whereas those with determinant −1 are said to be improper. Thus, we
conclude that only proper orthogonal transformations correspond to possible displacements
93 Possible Displacements of a Rigid Body

of a rigid body. It is easily shown that the set of proper orthogonal transformations is itself
a group called the rotation group, denoted SO(3).

Example 3.2 The scalar product and the vector product are invariant under rotations.
This means that if R = (Rij ) is a rotation then
(Ra) · (Rb) = a · b (3.33)
and also
(Ra) × (Rb) = R(a × b) . (3.34)
Proving invariance of the scalar product is very simple with the use of the Einstein
convention of sum over repeated indices (Appendix A):
(Ra) · (Rb) = (Ra)i (Rb)i = Rij aj Rik bk
= (RT R)jk aj bk = δjk aj bk = ak bk = a · b . (3.35)
As to the vector product, we have (see Appendix A)
[(Ra) × (Rb)]i = ijk (Ra)j (Rb)k
= ijk Rjm am Rkn bn = ijk Rjm Rkn am bn . (3.36)
According to Appendix A,
ijk Ril Rjm Rkn = (det R)lmn = lmn , (3.37)
where we used det R = 1 because R is a rotation matrix. Multiplying this last equation
by Rrl , which implies summation over the repeated index l, we find
Rrl lmn = ijk Rrl Ril Rjm Rkn = ijk (RRT )ri Rjm Rkn
= ijk δri Rjm Rkn = rjk Rjm Rkn . (3.38)
Finally, substituting this result into (3.36) we obtain
[(Ra) × (Rb)]i = Ril lmn am bn = Ril (a × b)l = [R(a × b)]i . (3.39)
The proof of Eq. (3.34) is complete.

A classic result, which reveals one of the fundamental features of a rigid body’s motion,
was established by Euler in 1776.

Theorem 3.2.1 (Euler) The most general displacement of a rigid body with one point fixed is
a rotation about some axis through the fixed point.

Proof Instead of the traditional geometric reasoning, we will resort to the previously
developed algebraic techniques, which will confer the proof a rather instructive character.
But, first, it is necessary to clarify the content of the theorem. Suppose the rigid body is
in an initial configuration and, after moving in any way whatsoever about the fixed point,
it reaches a final configuration. As already argued, the final configuration results from the
94 Kinematics of Rotational Motion

application of a proper orthogonal transformation to the initial configuration. According

to the statement of Euler’s theorem, the final configuration can be attained by means of a
single rotation about some axis through the fixed point – that is, a rotation about a point
is always equivalent to a rotation about an axis containing the point. But a vector along
the rotation axis remains unaltered, its components before and after the rotation are the
same. If A is the proper orthogonal matrix that takes the body from its initial to its final
configuration, Euler’s theorem will be proved as soon as one proves the existence of a
non-zero vector n such that
An = n (3.40)
or, equivalently,
(A − I) n = 0 . (3.41)
If the matrix A − I is invertible, multiplying (3.41) on the left by (A − I)−1 one infers that
n = 0. Therefore, there exists a vector n = 0 satisfying (3.41) if and only if A − I has no
inverse – that is, if and only if
det(A − I) = 0 . (3.42)
In short, proving Euler’s theorem boils down to showing that (3.42) holds. From the
identity
A − I = A − AAT = A(I − AT ) = −A(A − I)T = (−I)A(A − I)T (3.43)
one deduces
det(A − I) = det(−I) det A det(A − I) = − det A det(A − I) . (3.44)
Using det A = 1, this last equation implies
det(A − I) = − det(A − I) , (3.45)
from which one infers (3.42), completing the proof of the theorem. 
The traditional geometric proof obscures the role played by the dimensionality of space,
which crucially enters the above proof. In a space of even dimension det(−I) = 1 and the
previous argument fails. For example, in a two-dimensional space no vector is left unal-
tered by a rotation. Indeed, for a rotation in the plane the axis of rotation is perpendicular
to the plane, with the consequence that any vector along the rotation axis does not belong to
the plane. In summary, Euler’s theorem does not hold in even-dimensional spaces. Euler’s
theorem was generalised by Chasles, in 1830, in the terms below.
Theorem 3.2.2 (Chasles) The most general displacement of a rigid body is a translation
together with a rotation. The rotation axis can be so chosen that the translation is parallel
to it.
The first part of Chasles’s theorem is obvious, since the removal of the constraint of
having one point fixed endows the body with three translational degrees of freedom.
A geometric proof of the second part of Chasles’s theorem can be found in Whittaker
(1944) or Pars (1965). For a modern proof, with the use of purely algebraic methods, see
Corben and Stehle (1960).
 95 Euler Angles

3.3 Euler Angles

In a Lagrangian formulation of rigid body dynamics, the nine direction cosines aij are
not the most convenient coordinates for describing the instantaneous orientation of a rigid
body because they are not mutually independent. The nine equations (3.8) impose only six
conditions on the direction cosines, so they can be expressed in terms of three independent
parameters. Indeed, there are three distinct conditions corresponding to the diagonal part of
Eqs. (3.8), but the six conditions corresponding to the off-diagonal part (i = j) are pairwise
identical. For example, Eq. (3.8) for j = 1, k = 2 is the same as the one for j = 2, k = 1.
From the practical point of view, a convenient way to parameterise the rotation matrix
is by means of the Euler angles. The transformation from the Cartesian system (x, y, z) to
the   (x , y , z ) system is accomplished in three successive stages, each serving to define
one of the Euler angles (Fig. 3.4).
D
(a) Rotation of axes (x, y, z) about the z-axis by angle φ : (x, y, z) −→ (ξ , η, ζ ).
The transformation equations are the same as equations (3.20) with x1 = ξ and x2 = η
supplemented by equation x3 = ζ = z. Therefore, the rotation matrix D is written
⎛ ⎞
cos φ sin φ 0
D = ⎝ − sin φ cos φ 0 ⎠. (3.46)
0 0 1

C
(b) Rotation of axes (ξ , η, ζ ) about the ξ -axis by angle θ : (ξ , η, ζ ) −→ (ξ  , η , ζ  ).

z z z z
h´
q
z´
h qh
f f
y y
f f
x x x x x´
(a) y´ (b)
z z
h´
z´ q
z´ y
qh
f

x´ y
f y

x x x´
(c)

Fig. 3.4 Euler angles.

96 Kinematics of Rotational Motion

The ξ  -axis direction is called the line of nodes and, in analogy with (3.20), the
transformation equations take the form

ξ = ξ ,
η = η cos θ + ζ sin θ , (3.47)

ζ = −η sin θ + ζ cos θ ,

and it follows that

⎛ ⎞
1 0 0
C = ⎝ 0 cos θ sin θ ⎠ . (3.48)
0 − sin θ cos θ

B
(c) Rotation of axes (ξ  , η , ζ  ) about the ζ  -axis by angle ψ : (ξ  , η , ζ  ) −→ (x , y , z ).
The matrix B represents a rotation about the third axis, thus has the same form as D:
⎛ ⎞
cos ψ sin ψ 0
B = ⎝ − sin ψ cos ψ 0 ⎠ . (3.49)
0 0 1
The rotation (x, y, z) → (x , y , z ) is performed by AE = BCD. A direct calculation yields
 
cos ψ cos φ − cos θ sin φ sin ψ cos ψ sin φ + cos θ cos φ sin ψ sin ψ sin θ
AE = − sin ψ cos φ − cos θ sin φ cos ψ − sin ψ sin φ + cos θ cos φ cos ψ cos ψ sin θ .
sin θ sin φ − sin θ cos φ cos θ
(3.50)
The inverse transformation (x , y , z ) → (x, y, z) is effected by A−1
E
which is just the
transpose of AE – that is,
 
cos ψ cos φ − cos θ sin φ sin ψ − sin ψ cos φ − cos θ sin φ cos ψ sin θ sin φ
A−1
E
= cos ψ sin φ + cos θ cos φ sin ψ − sin ψ sin φ + cos θ cos φ cos ψ − sin θ cos φ .
sin ψ sin θ cos ψ sin θ cos θ
(3.51)
In virtue of their definition, the Euler angles take values in the following ranges:

0 ≤ φ ≤ 2π , 0 ≤θ ≤π, 0 ≤ ψ ≤ 2π . (3.52)

3.4 Infinitesimal Rotations and Angular Velocity

The rigid body equations of motion are differential equations, and their formulation
requires the study of the body’s rotation during an infinitesimal time interval.

Commutativity of Infinitesimal Rotations

Let g be an arbitrary vector submitted to a counterclockwise infinitesimal rotation by an
angle d about an axis defined by the unit vector n̂ (active point of view). According to
(2.136) we have
97 Infinitesimal Rotations and Angular Velocity

g = g + d × g, (3.53)

where
d = d n̂ . (3.54)

The symbols d and d are to be understood as mere names for infinitesimal quantities,
not as differentials of a scalar or of a vector . In fact, generally speaking, there is no
vector  whose differential is equal to d (Problem 3.6).
For successive rotations, with associated vectors d1 and d2 , we write

g = g + d1 × g , (3.55)
  
g = g + d2 × g , (3.56)

whence, neglecting second-order infinitesimals,

g = g + d12 × g (3.57)

with
d12 = d1 + d2 . (3.58)

This last result shows that successive infinitesimal rotations commute (d12 = d21 )
as a consequence of the commutativity of vector addition. Furthermore, the vector
associated with successive infinitesimal rotations is the sum of the vectors associated
with the individual infinitesimal rotations, a property which will be of great value for the
forthcoming developments.
It is rewarding to describe infinitesimal rotations in matrix language. Adopting the active
point of view, Eq. (3.53) can be written in the form

g = A g = (I + ε) g , (3.59)

where ε is an infinitesimal matrix. The commutativity of infinitesimal rotations is easily

checked, since neglecting second-order infinitesimals and taking into account that matrix
sum is commutative,

A1 A2 = (I + ε1 )(I + ε 2 ) = I + ε1 + ε 2 = I + ε2 + ε 1 = A2 A1 . (3.60)

On the other hand, since A = I + ε represents a rotation, A must be an orthogonal matrix.

Then,

I = AAT = (I + ε)(I + εT ) = I + ε + εT ⇒ ε T = −ε . (3.61)

A matrix ε which obeys (3.61) is said to be anti-symmetric. In terms of the elements ij of
ε Eq. (3.61) means ij = −ji . Therefore, the most general form of the matrix ε is
⎛ ⎞
0 α β
ε = ⎝ −α 0 γ ⎠. (3.62)
−β −γ 0
98 Kinematics of Rotational Motion

It follows that the the sum of the diagonal elements of ε is zero: tr ε = 0. Requiring that
the components of (3.59) be identical to those of (3.53), one easily finds
⎛ ⎞
0 −d3 d2
ε = ⎝ d3 0 −d1 ⎠ . (3.63)
−d2 d1 0
Although, as we have just seen, it is possible to associate a vector with each infinitesimal
rotation in such a way that the vector associated with two sucessive rotations is the
sum of the vectors associated with the individual rotations, such a correspondence is not
possible for finite rotations.3 It should be further noted that Eq. (3.63) establishes the
following bijective correspondence between anti-symmetric matrices and vectors in three
dimensions:
⎛ ⎞ ⎛ ⎞
0 −v3 v2 v1
⎝ v3 0 −v1 ⎠ ←→ ⎝ v2 ⎠ . (3.64)
−v2 v1 0 v3

Time Rate of Change of a Vector

In general, the time rate of change of a vector measured in a reference frame attached to
a rotating rigid body will not coincide with that observed in an inertial reference frame
external to the body. For instance, if the vector represents the position of a point fixed in
the body, its time rate of change with respect to the system of axes attached to the body
will be zero, but its time rate of change will not be zero with respect to the inertial system
of axes relative to which the body is rotating. Consider an infinitesimal time interval dt
during which an arbitrary vector g changes. Intuitively, we can write
(dg)inertial = (dg)body + (dg)rot , (3.65)
for the difference between the two systems is due exclusively to the body’s rotation. But
(dg)rot is a consequence of an infinitesimal rotation of g together with the axes attached to
the body – that is,
(dg)rot = d × g , (3.66)
as one infers from Eq. (3.53). The substitution of this last result into (3.65) leads to
 
dg dg
= + ω × g, (3.67)
dt inertial dt body
where
d d
ω== n̂ (3.68)
dt dt
is the instantaneous angular velocity of the rigid body. Note that the magnitude of
ω is the instantaneous rotation speed whereas its direction is along the instantaneous
axis of rotation characterised by the unit vector n̂. We insist that Eq. (3.68) represents
3 A detailed study of finite rotations can be found in Konopinski (1969). See, also, Example 4.1 in the next
chapter.
99 Infinitesimal Rotations and Angular Velocity

symbolically the limit of the ratio of two quantities, d and dt, which jointly tend to zero
as dt → 0, and that ω is not the time derivative of a vector  (Problem 3.6). For a purely
analytic definition of angular velocity, the reader is referred to Problem 3.7.
The important Eq. (3.67) can be derived in a more rigorous manner, which has the
additional virtue of clearing up the meaning of each of its terms. Let {êi }3i=1 be the unit
vectors of the inertial Cartesian system and {êi }3i=1 the unit vectors of the Cartesian system
attached to the body. We can write

3 
3
g= gi êi = gi êi . (3.69)
i=1 i=1

To an inertial observer the êi are constant in time but the êi vary with time because
they rotate together with the body. There are, therefore, two distinct ways to calculate
(dg/dt)inertial , namely
 3
dg dgi
= êi (3.70)
dt inertial dt
i=1
and
  
dg dêi
3 3
dg
= i
êi + gi . (3.71)
dt inertial dt dt
i=1 i=1
Equation (3.70) defines (dg/dt)inertial and, similarly, (dg/dt)body is defined by
  dgi 
3
dg
= ê . (3.72)
dt body dt i
i=1
This last equation is perfectly natural since to an observer attached to the body the unit
vectors êi are constant. But during the time interval dt the mobile system of axes undergoes
an infinitesimal rotation, the same occurring to each unit vector êi – that is,
dêi = d × êi , (3.73)
êi .
where we used (3.66) with g = Substituting (3.72) and (3.73) into (3.71) there results
  
3  3
dg dg  dg
= + 
gi ω × êi = + ω× gi êi , (3.74)
dt inertial dt body dt body
i=1 i=1
which coincides with (3.67).
Equality (3.67) is valid no matter what the vector g is. This allows us to interpret it not
only as a mere equality of vectors, but as a reflection of an equivalence of linear operators
that act on vectors. This operator equality can be conveniently expressed in the symbolic
form  
d d
= + ω×. (3.75)
dt inertial dt body

Exercise 3.4.1 Show that (d/dt)inertial = (d/dt)body when acting on a scalar. Hint: consider
the scalar s = g · h.
 100 Kinematics of Rotational Motion

O2
r2

R P
R2
r
r1

O1
O R1
Fig. 3.5 O1 and O2 are the origins of two-coordinate systems fixed in the body.

Uniqueness of the Angular Velocity Vector

Intuitively, the angular velocity is expected to be a property of the rigid body as a whole,
independent, therefore, of the point chosen as the origin of the coordinate system attached
to the body. However, a rigorous proof of this fact is advisable (Lemos, 2000b). In Fig. 3.5,
 is an inertial coordinate system with origin at point O, while O1 and O2 are the origins
of two coordinate systems attached to the body, 1 and 2 . Let r be the vector from O to
any point P of the rigid body and let ω1 and ω2 be the angular velocities associated with
1 and 2 , respectively. Given that r = R1 + r1 = R2 + r2 , we have

   
dr dR1 dr1 dR1
= + = + ω1 × r1 (3.76)
dt  dt  dt  dt 

and, similarly,
 
dr dR2
= + ω2 × r2 , (3.77)
dt  dt 

because, P being a point of the body, r1 and r2 are constant vectors with respect to the
coordinate systems 1 and 2 . Putting R = R2 − R1 , from (3.76) and (3.77) one deduces

dR
= ω1 × r1 − ω2 × r2 . (3.78)
dt 

On the other hand,

 
dR dR
= + ω1 × R = ω1 × R , (3.79)
dt  dt 1

because R is a constant vector in the coordinate system 1 attached to the body. Combining
(3.78) and (3.79) we get
ω1 × R = ω1 × r1 − ω2 × r2 (3.80)
101 Infinitesimal Rotations and Angular Velocity

or, inasmuch as R = r1 − r2 ,
(ω2 − ω1 ) × r2 = 0 . (3.81)

Since P is any point of the body, r2 is arbitrary and it follows that ω2 = ω1 , completing
the proof.

Angular Velocity in Terms of the Euler Angles

Some important problems in rigid body dynamics require the angular velocity vector to be
expressed in terms of the Euler angles. An infinitesimal rotation of the rigid body can be
thought of as the result of three sucessive infinitesimal rotations whose angular velocities
have, respectively, magnitudes φ̇, θ̇ , ψ̇. Let ωφ , ωθ and ωψ be the corresponding angular
velocity vectors. The angular velocity vector ω is simply given by

ω = ωφ + ωθ + ωψ (3.82)

in virtue of Eq. (3.58). One can obtain the components of ω either along inertial axes
(x, y, z) or along axes (x , y , z ) attached to the body. Because of its greater utility, we will
consider the latter case.
Clearly, the angular velocity ω depends linearly on φ̇, θ̇ , ψ̇. Therefore, in order to find
the general form taken by ω we can fix a pair of Euler angles at a time, determine the
angular velocity associated with the variation of the third angle and then add the results
(Epstein, 1982). Fixing θ and ψ, the z-axis, which is fixed in space, also becomes fixed in
the body (see Fig. 3.4). So, z is the rotation axis and ωφ is a vector along the z-axis with
component φ̇, and from (3.50) we have
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
(ωφ )x 0 φ̇ sin θ sin ψ
⎝ (ωφ )y ⎠ = AE ⎝ 0 ⎠ = ⎝ φ̇ sin θ cos ψ ⎠ . (3.83)
(ωφ )z φ̇ φ̇ cos θ

Fixing now φ and ψ, the line of nodes (the ξ  -axis direction in Fig. 3.4) becomes fixed
both in space and in the body, so it is the rotation axis. As a consequence, ωθ is a vector
with the single component θ̇ along the ξ  -axis and with the help of (3.49) we find
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
(ωθ )x θ̇ θ̇ cos ψ
⎝ (ωθ )y ⎠ = B ⎝ 0 ⎠ = ⎝ −θ̇ sin ψ ⎠ . (3.84)
(ωθ )z 0 0

Finally, with θ and φ fixed the z -axis, which is fixed in the body, becomes fixed in space.
Therefore, z is the rotation axis and ωψ is a vector along the z -axis with component ψ̇,
there being no need to apply any transformation matrix to ωψ . Collecting the corresponding
components the final result is

ωx = φ̇ sin θ sin ψ + θ̇ cos ψ ,

ωy = φ̇ sin θ cos ψ − θ̇ sin ψ , (3.85)
ωz = ψ̇ + φ̇ cos θ .
102 Kinematics of Rotational Motion

Exercise 3.4.2 Prove that the components of the angular velocity along the axes fixed in
space are
ωx = θ̇ cos φ + ψ̇ sin θ sin φ ,
ωy = θ̇ sin φ − ψ̇ sin θ cos φ , (3.86)
ωz = φ̇ + ψ̇ cos θ .

3.5 Rotation Group and Infinitesimal Generators

With the help of (3.63) we can write (3.59) in the form

g = A g = (I + J1 d1 + J2 d2 + J3 d3 ) g , (3.87)
where
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 0 0 0 0 1 0 −1 0
J1 = ⎝ 0 0 −1 ⎠ , J2 = ⎝ 0 0 0 ⎠, J3 = ⎝ 1 0 0 ⎠. (3.88)
0 1 0 −1 0 0 0 0 0
The matrices Ji are called infinitesimal generators4 of the rotation group SO(3). Note that
tr Ji = 0.
Exercise 3.5.1 Check that the matrices Ji , i = 1, 2, 3, obey the algebra
' (
Ji , Jj = Jk , (3.89)
where (i, j, k) is a cyclic permutation of (1, 2, 3) and [A, B] = AB − BA is the
commutator of matrices A and B. Show that, equivalently, one can write
' (
Ji , Jj = ijk Jk , i, j = 1, 2, 3 , (3.90)
with use of the convention of sum over repeated indices and of the Levi-Civita
symbol ijk defined in Appendix A.
The matrix associated with a finite rotation by an angle α about the direction n̂ (coun-
terclockwise, active point of view) can be found by applying N successive infinitesimal
rotations by the same angle α/N and taking the limit N → ∞:
 α N
g = lim I + n̂ · J g = eα n̂·J g . (3.91)
N→∞ N
Thus, it turns out that a finite rotation is expressed in terms of the exponential function of
the matrix n̂ · J = n1 J1 + n2 J2 + n3 J3 . There’s nothing mysterious about the exponential
of a matrix. By definition,
 Ak ∞
A2 A3
eA = I + A + + + ··· = , (3.92)
2! 3! k!
k=0

4 Strictly speaking, the matrices J are a particular representation of the generators of SO(3) called the
i
fundamental representation.
103 Dynamics in Non-Inertial Reference Frames

where A and its powers are n×n matrices. The series (3.92) is always convergent (Courant
& Hilbert, 1953).
Matrix Jk is the generator of infinitesimal rotations about the kth Cartesian axis and
all matrices in the three-dimensional rotation group SO(3) can be represented in the form
R(α ) = eα ·J where α ≡ α n̂. The matrices R(α) form a Lie group whose generators obey
the Lie algebra associated with the group, Eq. (3.90). The quantities ijk that appear in
Eq. (3.90) are called the structure constants of the Lie algebra of the three-dimensional
rotation group.
Exercise 3.5.2 Making use of the identity
det eA = etr A , (3.93)
verify that all matrices in SO(3) have unit determinant.

3.6 Dynamics in Non-Inertial Reference Frames

Let  and   be reference frames with common origin,  being an inertial frame and  
being a rotating frame with angular velocity ω. Let r be the position vector of a particle of
mass m whose equation of motion in  is
 
d2 r
m = F, (3.94)
dt2

where F is the total force on the particle. Using Eq. (3.67) we have
 
dr dr
vin ≡ = + ω × r ≡ v + ω × r, (3.95)
dt  dt  
where vin is the particle’s velocity in the inertial frame  and v denotes the particle’s
velocity in the rotating frame – owing to the coincidence of origins, r is also the particle’s
position vector in   . From (3.95), with further use of (3.67), one derives
   
d2 r dvin dvin
= = + ω × vin
dt2 dt  dt  


dv dω
= + × r + 2ω × v + ω × (ω × r) . (3.96)
dt   dt
Equation (3.67) with g = ω shows that the angular acceleration is the same in both
frames, hence the lack of subscript in dω/dt in the last equation. Denoting the particle’s
acceleration in the rotating frame by a = (dv/dt)  and using (3.94), the equation of
motion for the particle in the rotating frame is finally obtained:
dω
ma = F + 2mv × ω + mω × (r × ω) + mr × . (3.97)
dt
To an observer in the non-inertial frame   it all happens as if the particle were subject to
an effective force which is the actual force F plus: a first term known as the Coriolis force;
 104 Kinematics of Rotational Motion

S
S´ m .
r

Fig. 3.6 Simultaneous rotation and translation (Exercise 3.6.1).

a second term, quadratic in ω, called the centrifugal force; and a third term, proportional
to the angular acceleration of the non-inertial frame, sometimes called the Euler force.

Exercise 3.6.1 If the origins of  and   are not coincident (Fig. 3.6), show that instead of
(3.97) we have
dω
ma = F − mao + 2mv × ω + mω × (r × ω) + mr × , (3.98)
dt
where ao = (d2 h/dt2 ) is the acceleration of the origin of   relative to .

Dynamics on the Rotating Earth

As one of the most important applications of the previous developments, consider the
motion of a particle near the Earth’s surface. It is convenient to introduce a Cartesian
coordinate system which rotates together with the Earth with origin at the Earth’s surface,
as shown in Fig. 3.7. The z-axis is vertical, that is, it is perpendicular to the Earth’s
surface; the x-axis is tangent to the meridian pointing south; the y-axis is tangent to the
parallel pointing east. The latitude of the place on the Earth’s surface taken as origin of the
Cartesian system xyz is λ. Taking the origin of the inertial frame  at the centre of the Earth,
the position vector of the origin of the rotating frame is R = Rẑ, where R is the radius of
the Earth. Therefore, R is a constant vector in the rotating frame attached to the Earth’s
surface and the repeated use of Eq. (3.67) immediately gives
 
d2 R
ao = m = ω × (ω × R) , (3.99)
dt2

 105 Dynamics in Non-Inertial Reference Frames

w
N z
y

S
Fig. 3.7 Rotating frame xyz fixed on the Earth’s surface.

where we have used dω/dt = 0 because the Earth’s angular velocity is practically constant.
Substitution of this result into (3.98) yields

ma = T + mg + mω × (R × ω) + 2mv × ω + mω × (r × ω) , (3.100)

where F = mg + T with T denoting the sum of all forces on the particle other than its
weight. Equation (3.100) can be written in the form

ma = T + mgeff + 2mv × ω + mω × (r × ω) , (3.101)

where
geff = g + ω × (R × ω) (3.102)

is the effective acceleration of gravity. In order to justify this name, consider a plumb bob,
that is, a particle of mass m suspended from a string under tension T and in equilibrium at
the origin (r ≡ 0). Equation (3.101) yields

T + mgeff = 0 , (3.103)

that is, the direction of the plum line (the measured vertical direction) is determined by the
effective acceleration of gravity vector. The vector geff does not point exactly toward the
centre of the Earth. In the coordinate system of Fig. 3.7 we have g = −gẑ and

ω = −ω cos λ x̂ + ω sin λ ẑ , (3.104)

whence
geff = −(g − Rω2 cos2 λ) ẑ + Rω2 sin λ cos λ x̂ . (3.105)

Therefore, there is a reduction of the magnitude of the gravitational acceleration in the

vertical direction and the appearence of a component in the north-south direction that
106 Kinematics of Rotational Motion

prevents geff from pointing toward the centre of the Earth – the true vertical direction.
The effects are small, however, because
2π rad
ω≈ ≈ 7.3 × 10−5 rad/s (3.106)
24 × 3600 s
and in consequence
Rω2 ≈ (6400 km) × (7.3 × 10−5 s−1 )2 ≈ 3 × 10−2 m/s2 , (3.107)
which is about 300 times smaller that the acceleration of gravity. This centrifugal effect,
although minute, cannot be neglected in studies of the shape of the Earth.
For objects moving near the surface of the Earth, the last term on the right-hand side of
Eq. (3.100) is usually much smaller that the Coriolis term. Essentially, the centrifugal effect
reduces to converting g into geff , and the main modification to the motion brought about by
the rotation of the Earth arises from the Coriolis force. For a supersonic airplane with speed
v ≈ 2000 km/h, the Coriolis acceleration does not exceed 2ωv ≈ 0.08 m/s2 or about
0.008g. Though small, the Coriolis acceleration is important in several circumstances,
as will be seen in forthcoming examples. In the northern hemisphere the Coriolis force
deflects to the right an object that is moving horizontally – that is, parallelly to the
Earth’s surface, the deflection being to the left in the southern hemisphere (Problem
3.11). Presumably this effect explains the behaviour of cyclones, which typically rotate
counterclockwise in the northern hemisphere but clockwise in the southern hemisphere
(Goldstein, 1980).

Example 3.3 An object at height h falls from rest. Find the transversal deflection of the
object when it hits the ground caused by the Coriolis force.

Solution
If the object falls vertically, v = −vẑ with v > 0 and the Coriolis force is

FCor = 2mv × ω = −2mvẑ × (−ω cos λ x̂ + ω sin λ ẑ) = 2mωv cos λ ŷ . (3.108)

The deviation is in the west-to-east direction in both hemispheres, since cos(−λ) =

cos λ. In the vertical direction the approximate equation of motion is
1
z̈ = −g ⇒ ż = −gt ⇒ z = h − gt2 , (3.109)
2
where we used the initial conditions ż(0) = 0, z(0) = h. In the west-to-east direction we
have
ÿ = 2ωv cos λ ≈ 2ωgt cos λ . (3.110)

Strictly speaking, only immediately after the beginning of the fall is the velocity vector
vertical. As the object falls, it acquires a transversal velocity that remains much smaller
that the vertical component because the Coriolis acceleration is negligible compared
to the acceleration of gravity. This justifies approximating the speed v by its vertical
 107 Dynamics in Non-Inertial Reference Frames

component alone during the whole fall, as done in (3.110). With the initial conditions
y(0) = ẏ(0) = 0 Eq. (3.110) implies

ωg 3 ω 8(h − z)3
y= t cos λ = cos λ . (3.111)
3 3 g
Upon hitting the ground (z = 0) the transversal displacement is

ω 8h3
y = cos λ . (3.112)
3 g
At the equator, where the effect is biggest, for a fall from a height h = 100 m the
deflection is y ≈ 2.2 cm. Although the deviation is measurable, the experiment is
hard to do because one has to eliminate the influence of winds, air resistance and other
disturbing factors which may mask the Coriolis deflection.

Example 3.4 Determine the effect of the rotation of the Earth on a pendulum whose
suspension point can turn freely, known as the Foucault pendulum.

Solution
Disregarding the quadratic term in ω in (3.101) and taking geff ≈ g, the equation of
motion for a simple pendulum becomes
ma = T + mg + 2mv × ω , (3.113)
where T denotes the tension in the string. It is convenient to choose the origin of the
coordinate system at the pendulum’s equilibrium point (Fig. 3.8). In the case of small
oscillations, the pendulum’s vertical displacement is much smaller that its horizontal
displacement. In fact,
θ2
z = l(1 − cos θ ) ≈ l , (3.114)
2

q l

T
y
m
f r
x
Fig. 3.8 Foucault pendulum.
108 Kinematics of Rotational Motion

so, with ρ = l sin θ ≈ lθ ,

z θ
≈ 1 (3.115)
ρ 2
for small θ . Therefore, in the vertical direction the pendulum remains essentially at rest
with the tension in the string balancing its weight:
T cos θ ≈ mg ⇒ T ≈ mg (3.116)
since cos θ ≈ 1. The Coriolis force is transversal to the instantaneous plane of
oscillation of the pendulum, which prevents it from oscillating in a fixed vertical plane.
The horizontal components of (3.113) are
T T
ẍ = − sin θ cos φ + 2ωz ẏ , ÿ = − sin θ sin φ − 2ωz ẋ , (3.117)
m m
with
ωz = ω sin λ . (3.118)
Taking into account that
ρ
sin θ = , ρ cos φ = x , ρ sin φ = y (3.119)
l
and using (3.116), equations (3.117) take the form
ẍ + ω02 x = 2ωz ẏ , ÿ + ω02 y = −2ωz ẋ , (3.120)
where

g
ω0 = . (3.121)
l
Note that for any pendulum with l < 1 km we have ω0 > 0.1 s−1  ωz . The equations
of motion for the pendulum are more easily solved with the help of the complex function
ζ (t) = x(t) + iy(t) because equations (3.120) are the real and imaginary parts of
ζ̈ + 2iωz ζ̇ + ω02 ζ = 0 . (3.122)
This last equation admits solutions in the form ζ (t) = ζ0 e−ipt with

− p2 + 2ωz p + ω02 = 0 ⇒ p = ωz ± ωz2 + ω02 ≈ ωz ± ω0 . (3.123)
The general solution for ζ (t) is
ζ (t) = Ae−i(ωz −ω0 )t + Be−i(ωz +ω0 )t . (3.124)
Adopting the initial conditions x(0) = a, ẋ(0) = 0, y(0) = ẏ(0) = 0, so that the
pendulum starts oscillating in the xz-plane, we get
ω0 − ωz ω0
A + B = a, B= A≈ A=A ⇒ A = B = a/2 , (3.125)
ω0 + ωz ω0
whence
ζ (t) = a cos ω0 t e−iωz t . (3.126)
109 Problems

Therefore,
x(t) = Re ζ (t) = a cos ωz t cos ω0 t ,
y(t) = Im ζ (t) = −a sin ωz t cos ω0 t . (3.127)
The angle φ that the plane of oscillation makes with the initial plane of oscillation (the
xz-plane) is such that
y
tan φ = = − tan ωz t ⇒ φ = −ωz t = −(ω sin λ)t . (3.128)
x
The plane of oscillation rotates about the vertical with angular velocity ω sin λ, the
sense, viewed from above, being clockwise in the northern hemisphere (λ > 0) and
counterclockwise in the southern hemisphere (λ < 0). Each 24 hours the plane of
oscilllation rotates by 360 sin λ degrees.5 At the latitude of Rio de Janeiro (λ =
−22◦ 54 ), the plane of oscillation turns 140◦ a day counterclockwise. In the North Pole
the rate of rotation is 360◦ a day clockwise, which can be understood as follows: The
plane of oscillation is fixed in an inertial reference frame, but the Earth rotates under
the pendulum by 360◦ a day in the counterclockwise sense. So, to an observer on Earth
the sense of rotation of the plane of oscillation is clockwise. As a final remark, note that
(3.127) shows that the order of magnitude of ẋ and ẏ is ω0 a since ω0  ω > ωz . On
the other hand, the order of magnitude of |ω × (r × ω)| is ω2 a. Thus, the ratio of the
centrifugal and Coriolis terms in (3.101) is ω2 a/2ω0 aω = ω/2ω0  1, which justifies
a posteriori the approximation that led to (3.113).

Problems

3.1 In three dimensions Eq. (3.64) establishes a bijective correspondence between

vectors and real antisymmetric matrices. (a) If A = (akl ) is the antisymmetric
matrix associated with the vector v = (v1 , v2 , v3 ), show that akl = kml vm with
sum over repeated indices and kml the Levi-Civita symbol defined in Appendix A.
Show that, conversely, vm = 12 kml akl . (b) Prove that the eigenvalues of the antisym-
metric matrix associated with v are zero and ± i|v|. (c) If A is a real antisym-
metric matrix, show that the matrices I ± A are non-singular and that the matrix
B = (I + A)(I − A)−1 is orthogonal.
3.2 Construct the vector r obtained by reflecting vector r in the plane whose unit normal
vector is n̂. Without any calculations, using only geometric arguments, determine
the eigenvalues and eigenvectors of the corresponding transformation matrix A. If
n̂ = (n1 , n2 , n3 ), show that A has elements aij = δij − 2ni nj and is an improper
orthogonal matrix.

5 Foucault intuited this result in 1851, before the mathematical treatment of the problem taking into account the
Coriolis force as the cause of the rotation (Dugas, 1988).
110 Kinematics of Rotational Motion

3.3 For the matrix D in Eq. (3.46), prove that

⎛ ⎞
cos nφ sin nφ 0
D n = ⎝ − sin nφ cos nφ 0⎠
0 0 1

for any natural number n. Was this result intuitively expected?

3.4 Any proper orthogonal matrix A corresponds to a rotation by some angle about
some direction. The rotation axis is determined by the eigenvector of A with
eigenvalue 1. Picking coordinates such that the z-axis is along the said eigenvector
the matrix A will have the form (3.46). Taking into account that the trace of a matrix
is independent of the coordinate system, show that the rotation angle is given by
cos = (tr A − 1)/2.
3.5 A sphere of radius R rolls without slipping on a plane surface. If (x, y, R) are Cartesian
coordinates of the centre of the sphere, show that, in terms of the Euler angles, the
conditions for rolling without slipping are

ẋ − R(θ̇ sin φ − ψ̇ sin θ cos φ) = 0 , ẏ + R(θ̇ cos φ + ψ̇ sin θ sin φ) = 0 .

Prove that these constraints are not holonomic (see Appendix B).
3.6 The translational displacement of a body can be represented by a vector and the
linear velocity is the time derivative of the position vector. The angular velocity
vector, however, is not, in general, the time derivative of an angular displacement
vector. In order to prove this, note that if there exists such a vector , its components
x , y , z can be expressed in terms of the Euler angles and must be such that their
time derivatives equal the corresponding components of the angular velocity:
∂x ∂x ∂x ∂y ∂y ∂y
ωx = θ̇ + φ̇ + ψ̇ , ωy = θ̇ + φ̇ + ψ̇ ,
∂θ ∂φ ∂ψ ∂θ ∂φ ∂ψ
∂z ∂z ∂z
ωz = θ̇ + φ̇ + ψ̇ .
∂θ ∂φ ∂ψ
Using (3.86), prove that there can be no vector (x , y , z ) such that these equations
are satisfied.
3.7 Equation (3.67) can be derived in a fully analytic fashion without appealing to the
geometric definition of the angular velocity vector. (a) By definition, the derivative
of a matrix is the matrix formed by the derivatives of its elements. If A is a time-
dependent orthogonal matrix, by differentiating the identity ATA = I with respect
T T T
to time prove that the matrix Ȧ A is antisymmetric – that is, Ȧ A = − (Ȧ A)T .
(b) Let r and r be the components of the same position vector r of a particle
relative to the inertial frame  and the rotating frame   , and let A be the rotation
matrix from  to   . If v and v are the velocities relative to the respective
inertial and rotating frames, we have v = dr /dt and v = dr /dt. Starting
from r = AT r derive

v = v + Ȧ A r .
T
111 Problems

Taking into account part (a), the correspondence (3.64) between vectors and
antisymmetric matrices and the fact that an equality of vectors which is true in one
coordinate system is true in all coordinate systems, prove that there exists a vector ω
such that
v = v + ω × r ,
which coincides with Eq. (3.67).
3.8 Let P be the matrix associated with a rotation through 180◦ about an arbitrary
axis. (a) Determine P 2 without calculations, reflecting on its meaning. (b) Letting
A = (I + P)/2 and B = (I − P)/2, prove that A2 = A and B2 = B. (c) Show
that the matrices A and B are singular and compute their product.
3.9 A particle is fired vertically upward from the surface of the Earth with initial speed v0 ,
reaches its maximum height and returns to the ground. Show that the Coriolis
deflection when it hits the ground has the opposite sense and is four times bigger
than the deviation for a particle dropped from the same maximum height.
3.10 Show that the equation of motion (3.98) can be derived from the Lagrangian
mv2 m
L= − V − mr · ao + (ω × r)2 + mv · (ω × r) ,
2 2
where v = ṙ and V is the potential energy corresponding to the force F. Obtain the
Jacobi integral associated with this Lagrangian in the case in which ao and ω are
constant.
3.11 At t = 0 a projectile is fired horizontally near the surface of the Earth with speed v.
(a) Neglecting gravity, show that, to a good approximation, the horizontal deviation
of the projectile due to the Coriolis force is
dH = ωv| sin λ|t2 ,
where λ is the latitude. (b) Show that the projectile is deflected to the right in the
northern hemisphere and to the left in the southern hemisphere. (c) To the same
degree of approximation, express dH in terms of the projectile’s distance D from the
firing point at time t. (d) How would the inclusion of the influence of gravity affect
the result? (e) What changes if the projectile is fired at an angle above the horizontal?
(f) It is told that during a World War I naval battle in the south Atlantic British shells
missed German ships by about 90 meters to the left because the British gunsights
had been corrected to a latitude 50◦ north (for battles in Europe) instead of 50◦ south
where the battle took place. Conclude that the British shells would have missed the
targets by twice the deflection calculated above. (g) Assuming the tale is true6 and
v = 700 m/s, how distant must the German ships have been?

6 It is a pity that such a savoury story is most likely but an urban legend, as convincingly argued at
http://dreadnoughtproject.org/tfs/index.php/Battle of the Falkland Islands.