Vectors

Chapter 2

Vector Analysis and Classical and

Relativistic Mechanics
2.1 Kinematics and Dynamics of a Point Mass

Kinematics is the study of motion without reference to forces; the Greek

word kinema means "motion." Dynamics is the study of motion under the
action of forces; the Greek word dynamis means "force." Also, the Greek
word mechanikos means "machine."
A curve C, defined by a vector-valued function of time r = r(t) provides
the mathematical description of the trajectory in M3 of a particle (point
mass) so that the location of the particle at time t is at the end point
of the vector r(t). The initial point O of the vector is the origin of the
corresponding frame in which the motion is studied. It is clear that the same
motion can be studied in different frames and in different coordinates. The
Prenet trihedron (page 32) is an example of a coordinate system in which
the particle is at rest, but the coordinate system is moving. The objective
of this section is to derive the rules for describing the motion of a point
mass in different coordinate systems.
Unless explicitly mentioned otherwise, we assume that the curve C is
smooth, that is, the unit tangent vector u exists at every point of the
curve; see page 27.
2.1.1 Newton's Laws of Motion and Gravitation

The motion of a point mass m is related to the net force F acting on the
mass. In an inertial frame, this relation is made precise by Newton's three
laws of motion, and conversely, every frame in which these laws hold is
called i n e r t i a l . These laws were first formulated by Newton around 1666,
less than a year after he received his bachelor's degree from Cambridge
39
40 Kinematics and Dynamics of a Point Mass
University.
Newton's First Law: Unless acted upon by a force, a point mass is either
not moving or moves in a straight line with constant speed. This law is also
called the Law of I n e r t i a or G a l i l e o ' s P r i n c i p l e .
Newton's Second Law: The acceleration of the point mass is directly
proportional to the net force exerted and inversely proportional to the mass.
Newton's Third Law: For every action, there is an equal and opposite
reaction.
Mathematically, the Second Law is
o9r(t) _ F
(2.1.1)
dt2 m
where r = r(t) is the position of the point mass at time t.
EXERCISE 2.1.1. c Show that the Second Law implies the First Law. In
other words, show that if (2.1.1) holds, then the point mass m acted on by
zero external force will move with constant velocity v. Hint: find the general
solution of equation (2.1.1) when F = 0.
The notion of momentum provides an alternative formulation of New-
ton's Second Law. Consider a point mass m moving along the path r — r(t)
with velocity r(t) relative to a reference frame with origin at O. The (linear)
momentum p is the vector
p = mr. (2.1-2)
If the reference frame is inertial, then (2.1.1) becomes
p = F, (2.1.3)
and the force is now interpreted as the rate of change of the linear momen-
tum. Incidentally, the Latin word momentum means "motion" or "cause of
motion." One advantage of (2.1.3) over (2.1.1) is the possibility of variable
mass.
Similarly, the study of circular motion suggest the definition of the
angular momentum about the point O as the vector
Lo = mr x r. (2-1-4)
Note that both p and Lo depend on the reference point O, but do not
depend on the coordinate system.
Newton's Laws of Motion and Gravitation 41
Recall that a force F acting on the point mass m has a torque, or

moment, about O equal to
T0 = rxF. (2.1.5)
Applying the rule (1.3.6), page 26, to formula (2.1.4) we find
dLQ
m(rxr + rxr) = mrxf.
dt
If the frame O is inertial, then r — F/m and
dLc
= r x F = T0. (2.1.6)
dt
Relation (2.1.6) describes the rotational motion just as (2.1.3) describes the
translational motion.
As an example, consider the SIMPLE RIGID PENDULUM, which is a
massless thin rigid rod of length £ connected to a point mass m at one
end. The other end of the rod is connected to a frictionless pin-joint at a
point O, which is a zero-diameter bearing that permits rotation in a fixed
plane. We select a cartesian coordinate system (i, j , K) with center at O
and i, j fixed in the plane of the rotation (Figure 2.1.1), and assume that
the corresponding frame is inertial.
Fig. 2.1.1 Simple Rigid Pendulum
The motion of m is a circular rotation in the (i, j) plane, and is best

described using polar coordinates (r,6); see page 35. The rigidity as-
sumption implies r(t) = £ for all t. Denote by 0 = 6{t) the angle from
% to the rod at time t. Let r(t) be the position vector of the point
mass in the frame O. Then r(t) = £r(0(t)), and by (1.3.24) on page 35,
r — £80. Hence, the angular momentum of the point mass about O is
L — mr x r = m(£r x 196) = m£29A, and
^=m£r6k. (2.1.7)
at
The forces acting on the pendulum are the weight W of m, the air resistance
Fa and the force Fp exerted by the pin at O. Clearly, W = mgi, where
g is the acceleration of gravity. Physical considerations suggest that the
force Fa on m may be assumed to act tangentially to the circular path and
to be proportional to the tangential velocity: Fa = —c£9 6, where c is the
damping constant. The total torque T about O exerted by these forces is
T = r xW + r x Fa + Ox Fp, = £rxmgi-£rxc£9 8,
or
T=-(mg£ sin 9 + £2C9)K. (2.1.8)
Since the frame O is inertial, equation (2.1.6) applies, and by (2.1.7) and
(2.1.8) above, we obtain ml? 6 = —mg£sin6 — l2c6, or
m£e + d6 = -mg sin9. (2.1.9)
Equation (2.1.9) is a nonlinear ordinary differential equation and cannot

be integrated in quadratures, that is, its solution cannot be written using
only elementary functions and their anti-derivatives. When c = 0, such a
solution does exist and involves elliptic integrals; see Problem 2.3, page 417,
if you are curious.
The more familiar harmonic oscillator
6 = -{g/£)6 (2.1.10)
is obtained from (2.1.9) when c = 0 and 6 is small so that sin0 w 6; this

equation should be familiar from the basic course in ordinary differential
equations. If 9(0) = 0O and 0(0) = 0, then the solution of (2.1.10) is
0(t) = 0O cos(cjt), where w = (£/g)1/2.
The period of the small undamped oscillations is 2n(£/g)1/2, and the value
of £ can be adjusted to provide a desired ticking rate for a clock mechanism.
The idea to use a pendulum for time-keeping was studied by the Italian
scientist GALILEO GALILEI (1564-1642) during the last years of his life,
but it was only in 1656 that the Dutch scientist CHRISTIAAN HUYGENS
(1629-1695) patented the first pendulum clock.
EXERCISE 2.1.2r Let point mass m move in a planar path C given by r(t),
where r is the position vector with origin O.
(a) Use formulas (1.3.24), Pa9^ 35, and (2.1.4) to express the angular mo-
mentum of the point mass about O in the coordinate system {r,6).
(b) Suppose that the point mass moves in a circular path C with radius R
and center O. Denote the angular speed by w(i). (i) Compute the angular
momentum LQ of the point mass about O and the corresponding torque,
(ii) Find the force F that is required to produce this motion, assuming the
frame O is inertial. (Hi) Write F as a linear combination ofr and 6. (iv)
How will the expressions simplify ifui(t) does not depend on time?
As we saw in Exercise 2.1.1, the Second Law of Newton implies the
First Law. For further discussion of the logic of Newton's Laws see the
book Foundations of Physics by H. Margenau and R. Lindsay, 1957. Re-
garding the First Law, they quote A. S. Eddington's remark from his book
Nature of the Physical World, first published in the 1920s, that the law,
in effect, says that "every particle continues in its state of rest or uniform
motion in a straight line, except insofar as it doesn't." This is a somewhat
facetious commentary on the logical circularity of Newton's original formu-
lation, which depends on the notion of zero force acting, which can only
be observed in terms of the motion being at constant velocity. The same
logical difficulty arises in the definition of an i n e r t i a l frame as a frame
in which the three laws of Newton hold. We do not concentrate on these
questions here and simply assume that the primary i n e r t i a l frame, that
is, a frame attached to far-away, and approximately fixed, stars is a good
approximation of an inertial frame for all motions in the vicinity of the
Earth. The idea of this frame goes back to the Irish bishop and philoso-
pher G. Berkeley. The deep question "What is a force?" is also beyond the
scope of our presentation; for the discussion of this question, see the above-
mentioned book Foundations of Physics by H. Margenau and R. Lindsay,
or else take as given that there are four basic kinds of forces: gravitational,
electromagnetic, strong nuclear, and weak nuclear. In inertial frames, all
other forces result from these four.
Newton discovered the Law of Universal Gravitation by combining his
laws of motion with Kepler's Laws of Planetary Motion. The history behind
this discovery is a lot more complex than the familiar legend about the apple
falling from the tree and hitting Newton on the head. As many similar
stories, this "apple incident" is questioned by modern historians. Below,
we present some of the highlights of the actual development.
The basic ideas of modern astronomy go back to the Polish astronomer

NICOLAUS COPERNICUS (1473-1543) and his heliocentric theory of the solar
system. Copernicus was a canon (in modern terms, senior manager) of the
cathedral at the town of Prauenburg (now Frombork) in northern Poland,
and observed the stars and planets from his home. Around 1530, he came
to the conclusion that planets in our solar system revolve around the Sun.
He was hesitant to publish his ideas, both for fear of being charged with
heresy and because of the numerous problems he could not resolve; his
work, titled De revolutionibus orbium coelestium ("On the Revolutions of
the Celestial Spheres") was finally published in 1543, apparently just a few
weeks before he died.
It took some time to formalize the heliocentric ideas mathematically,
and the key missing element was the empirical data. The main instrument
for astronomical observations, the telescope, was yet to be invented: it was
only in 1609 that Galileo Galilei made the first one. Without a telescope,
collecting the data required a lot of time and patience, but the Danish
scientist TYCHO BRAHE (1546-1601) had both. Brahe was the royal as-
tronomer and mathematician to Rudolf II, the emperor of the Holy Roman
Empire. At the observatory in Prague, the seat of the Holy Roman Empire
at that time, Brahe compiled the world's first truly accurate and complete
set of astronomical tables. His assistant, German scientist JOHANN K E P L E R
(1571-1630), had been a proponent of the heliocentric theory of Coperni-
cus. After inheriting the position and all the astronomical data from Brahe
in 1601, Kepler analyzed the data for the planet Mars and formulated his
first two laws in 1609. Further investigations led him to the discovery of
the third law in 1619.
K e p l e r ' s First Law: The planets have elliptical orbits with the Sun at
one focus.
K e p l e r ' s Second Law: The radius vector from the Sun to a planet
sweeps over equal areas in equal time intervals.
K e p l e r ' s T h i r d Law: For every planet p, the square of its period Tp of
revolution around the Sun is proportional to the cube of the average distance
Rp from the planet to the Sun. In other words, T% = KsRl, where the
number Ks is the same for every planet.
A planetary orbit has a very small eccentricity and so is close to a
circle of some mean radius R. Kepler speculated that a planet is held
in its orbit by a force of attraction between the Sun and the planet, and
Newton quantified Kepler's qualitative idea. In modern terms, we can
recover Newton's argument by combining equation (1.3.29), page 36, with

his second law (2.1.1). Take an inertial frame with origin at the Sun and
assume that a planet of mass m executes a circular motion around the
origin with constant angular speed w. Then (2.1.1) and (1.3.29) result in
F = ma = -moj2r(t) = -mw2Rr{t),
where r = r/\\r\\ is the unit radius vector pointing from the Sun to the
planet. On the other hand, u> — 2n/T, and therefore the magnitude
of F is ||F|| = m(4rr2/T2)R. Applying Kepler's Third Law, we obtain
||F|| = m{Air2/KsR?)R = Cm/R2, where C = 4n2/Ks. In other words,
the gravitational force exerted by the Sun on the planet of mass m at a
distance R is proportional to m and R~2. By Newton's third law, there
must be an equal and opposite force exerted by m on the Sun. By the
same argument, we conclude that the magnitude of the force must also be
proportional to M and R~2, where M is the mass of the Sun. Therefore,
11*11 = ^ , (2.1.11)
where G is a constant. Newton postulated that G is a u n i v e r s a l
g r a v i t a t i o n a l constant , that is, has the same value for any two masses,
and therefor (2.1.11) is a Universal Law of G r a v i t a t i o n . In 1798, the
English scientist HENRY CAVENDISH (1731-1810), in his quest to determine
the mass and density of Earth, verified the relation (2.1.11) experimentally
and determined a numerical value of G: G « 6.67 x 1 0 - 1 1 m 3 /(kg- sec 2 ).
Since then, the Universal Law of Gravitation has been tested and verified
on many occasions. An extremely small discrepancy has been discovered
in the orbit of Mercury that cannot be derived from (2.1.11), and is ex-
plained by Einstein's law of gravitation in the theory of general relativity;
see Problem 2.2 on page 414.
In our derivation of (2.1.11), we implicitly used the equivalence principle

that the two possible values of m, its inertial and gravitational masses, are
equal. A priori, this is not at all obvious. Indeed, the mass m in equa-
tion (2.1.1) of Newton's Second Law, the i n e r t i a l mass, expresses an
object's resistance to external force: the larger the mass, the smaller the
acceleration. The mass m in (2.1.11), the g r a v i t a t i o n a l mass, expresses
something completely different, namely, its gravitational attraction: the
larger the mass, the stronger the gravitational attraction it produces. The
equivalence principle is one of the foundations of Einstein's theory of gen-
eral relativity and can be traced back to Galilei, who was among the first
to study the motion of bodies under Earth's gravity. Even though many
modern historians question whether indeed, around 1590, he was dropping
different objects from the leaning tower of Pisa, in 1604 Galileo did con-
duct related experiments using an inclined plane; in 1608, he formulated
mathematically the basic laws of accelerated motion under the gravitational
force. The conjecture of Galilei that the acceleration due to gravity is es-
sentially the same for all kinds of matter has been verified experimentally.
Between 1905 and 1908, the Hungarian physicist VASAROSNAMENYI BARO
E O T V O S LORAND (1848-1919), also known as ROLAND E O T V O S , measured
a variation of about 5 x 10~ 9 in the Earth's pull on wood and platinum;
somehow, the result was published only in 1922. In the 1950s, the Ameri-
can physicist ROBERT HENRY DiCKE (1916-1997) measured a difference of
(1.3 ± 1.0) x 1 0 - 1 1 for the Sun's attraction of aluminum and gold objects.
Following the historical developments, we derived relation (2.1.11) from

Newton's Second Law of Motion and Kepler's Third Law of Planetary Mo-
tion. Problem 2.1, page 413, presents a deeper insight into the problem. In
particular, more detailed analysis shows that, in our derivation of (2.1.11),
Kepler's Third Law can be replaced by his First Law, along with the as-
sumption that the gravitational force is attracting and c e n t r a l , that is,
acts along the line connecting the Sun and the planet. Moreover, the reader
who completes Problem 2.1 will see that all three Kepler's laws follow from
(2.1.1) and (2.1.11). This illustrates the power of mathematical models in
reasoning about physical laws.
2.1.2 Parallel Translation of Frames

Recall that Newton's laws of motion hold only in inertial frames; see page 4
for the definition of frame. Ignoring the possible logical issues, we therefore
say that an i n e r t i a l frame is a frame in which a point mass maintains
constant velocity in the absence of external forces. The same frame can
be (approximately) inertial in some situations and not inertial in others.
For example, a frame fixed to the surface of the Earth is inertial if the
objective is to study the motion of a billiard ball on a pool table. The same
frame is no longer inertial if the objective is to study the trajectory of an
intercontinental ballistic missile: the inertial frame for this problem should
not rotate with the Earth; the primary inertial frame fixed to the stars is
a possible choice, see page 43. In this and the following two sections, we
investigate how relation (2.1.1) changes if the frame is not inertial. The
Parallel Translation of Frames 47
starting point is the analysis of the relative motion of frames.

The easiest motion is parallel translation, where the corresponding basis
vectors in the frames stay parallel. Consider two such frames with origins
O and Oi respectively. Denote by roi(t) the position vector of 0\ with
respect to O. Let the position vectors of a point mass in frames O and 0\
be r0{t) and ri(i) respectively (Figure 2.1.2).
Fig. 2.1.2 Translation of Frames
Clearly, ro(t) — roi(i) + Vi(t), and the absence of relative rotation

allows us to consider this equality in the frame O for each t We can identify
the parallel vectors that have the same direction and length. Since position
vectors in O and 0\ maintain their relative orientation when there is no
relative rotation of the frames, the coordinate systems in the frames O and
0\ are the same. Then we can apply the rule for differentiating a sum
(1.3.3) to obtain simple relations between the velocities and accelerations
in the frames O and 0\\
r o ( * ) = r 0 i ( t ) + fi(t), r(t)=roi(*) + ri(t). (2.1.12)
If the frame O is inertial, then, by Newton's Second Law, mr(t) = F(t),

where m is the mass of the point, and F is the sum of all forces acting on
the point. The second equality in (2.1.12) then implies
mri(t) = F(t) - mroi(t) (2.1.13)
In effect, there are two forces acting on m in the frame 0\. One is the force
F. The other, — mroi, is called a t r a n s l a t i o n a l a c c e l e r a t i o n force.
It is an example of an apparent, or i n e r t i a l , force, that is, a force that
appears because of the relative motion of frames and is not of any of the
four types described on page 43. If roi(t) is constant, then roi(t) = 0 and
the Second Law of Newton holds in 0\, that is, Oi is also an inertial frame.
Thus, all frames moving with constant velocity relative to an inertial frame
are also inertial frames.
As an EXAMPLE, consider a golf ball in a moving elevator. We fix the

frame O on the ground, and 0\, on the elevator, and select the usual carte-
sian coordinate systems in both frames so that the corresponding coordinate
vectors are parallel. Assume that the elevator is falling down with the grav-
itational acceleration g, so that roi(t) = —gk, and the ball is falling down
inside the elevator, also with the gravitational acceleration g so that ro{t) —
-gk. Then (2.1.12) shows that r\(t) = ro(t) — roi(t) = —gk + gk — 0.
Therefore, f*i(t) is constant, and if r\(to) = 0, then r[(t) = 0 for all t > to
(or until the elevator hits the ground). An observer in the elevator would
see the ball as fixed in the elevator frame 0\: the translational acceleration
force - m f o i ( i ) compensates the gravitational force m?o(t), and the ball
behaves as weightless in the elevator frame.
2.1.3 Uniform Rotation of Frames

Note that it is the absence of rotation that allowed us to use the differenti-
ation rule (1.3.3) in the derivation of relation (2.1.12). This and other rules
of differentiation no longer apply if the frames are rotating relative to each
other, and relation (2.1.12) must be modified.
We start with the analysis of uniform r o t a t i o n , that is, rotation with
constant angular speed around a fixed axis. As a motivational example,
consider a car driving with constant angular speed u>o in a circle with radius
R and center at O. Consider an object (a point mass m) moving inside the
car with constant radial speed VQ relative to O, and rotating together with
the car with constant angular speed UQ around O. Introduce a new (non-
inertial) rotating frame with origin 0\ inside the car and the coordinate
basis vectors i\ = r, j 1 = 0; see Figure 2.1.3. As before, let ro and n be
the position vectors of the point mass in the frames O and Oi, respectively,
and denote 00\ by roi- Assume that T*I(0) = 0.
' * — • -
0 0[ r
Fig. 2.1.3 Rotation of Frames
By construction, roi = Rf. For a passenger riding in the car, the

Uniform Rotation of Frames 49
path r\(t) — votr of the point mass relative to the car is a straight line,
since there is no angular displacement of m relative to the car. Thus,
r
o(t) = (R + vot)r(t). Note that the polar coordinates of m in frame
O are (r(t),9(t)), where r(t) = R + v0t. Hence r(t) = v0 and f(t) = 0.
Also, Q(t) = OJQ and 6(t) = 0. Formulas (1.3.26) on page 36 provide the
acceleration a = ar + a# of the mass in the frame O, where, with f — Q
and 6 = 0,
ar = -(R + v0t)Jl r, ae = 2v0wo §• (2.1.14)
By (1.3.27) on page 36, r 0 i = -Rw$r. In frame 01: r^t) = 0. Then

(2.1.14) shows that the acceleration fo(t) is
r 0 ( 0 = roi(t) + r\(t) - v0ttj$r + 2v0uj0d. (2.1.15)
We see that the simple relation (2.1.12) between the accelerations in trans-
lated frames does not correctly describe acceleration of the point mass in
the frame O in terms of the acceleration in the frame 0\.
If O is an inertial frame and J 1 is a force acting on the point mass in
O to produce the motion, then, by the Newton's Second Law, m ro = F .
According to (2.1.15),
mf\ = F + (mRul + mvotwl) r — 2mv0ujo&- (2.1.16)
Thus, frame 0\ is not inertial. Similar to (2.1.13), inertial forces appear as

correctors to Newton's Second Law: the centrifugal force Fc = {rnRiS^ +
mvoiwg) r and the Coriolis force Fcor = — 2mvou>0 6. The centrifugal force
prevents the mass from flying off at a tangent because of the rotational
motion of the car and the mass. One component of this force, mRwQ r =
—mroi(t), is related to the motion of the car causing the rotation of the
origin of the frame Oi; the other, mvotuiQ r, takes into account the outward
radial motion of the point. Note that the direction of the centrifugal force
is in the direction of r, and is therefore away from the center and opposite
to the direction of the centripetal acceleration (cf. page 35). Incidentally,
the Latin verb fug ere means "to run away."
The Coriolis force — 2mvocoo Q is somewhat less expected. This force is
perpendicular to the linear path in Oi and ensures that the trajectory of
the point in the rotating frame is a straight radial line despite the rotation
of the frame. This force was first described in 1835 by the French scien-
tist GASPARD-GUSTAVE DE CORIOLIS (1792-1843). His motivation for the
study came from the problems of the early 19th-century industry, such as
the design of water-wheels. More familiar effects of the Coriolis force, such
as rotation of the swing plane of the Poucault pendulum and the special
directions of atmospheric winds, were discovered in the 1850s and will be
discussed in the next section.
Recall that in our example r\ = 0. From (2.1.16) we conclude that
F + Fc + Fcor = 0. The real (as opposite to inertial) force F must balance
the effects of the inertial forces to ensure the required motion of the object in
the rotating frame. For a passenger sliding outward with constant velocity
VQ r in a turning car, this real force is the reaction of the seat in the form
of friction and forward pressure of the back of the seat. .
EXERCISE 2.1.3? Find the vector function describing the trajectory ofm in
the O frame. What is the shape of this trajectory? Verify your conclusion
using a computer algebra system.
EXERCISE 2 . 1 . 4 . A
Suppose a point mass m is fixed at a point P in the O
frame, that is, m remains at P in the O frame for all times. Find the vector
function describing the trajectory ofm in the 0\ frame. What is the shape
of this trajectory? Verify your conclusion using a computer algebra system.
Hint. This is the path of m relative to the car seen by a passenger riding in the
car. Show that O is fixed in Ox.
Coming back to Figure 2.1.3, note that the coordinate vectors in the
frame 0\ spin around 0\ with constant angular speed WQ, while the origin
0\ rotates around O with the same angular speed WQ. This observation
leads to further generalization by allowing different speeds of spinning and
rotation.
EXERCISE 2.I.5.' 4 Suppose the origin 0% rotates around the point O with
angular speed U>Q, while the coordinate vectors (r, 9) spin around 0\ with
constant angular speed 2WQ. Suppose a point mass is fixed at a point P in
the 0 frame. Find the vector function describing the trajectory of m in
the 0\ frame. What is the shape of this trajectory? Verify your conclusion
using a computer algebra system.
Our motivational example with the car illustrated some of the main
effects that arise in rotating frames. The example was two-dimensional in
nature, and now we move on to uniform rotations in space. There are many
different ways to describe rotations in R 3 . We present an approach using
vectors and linear algebra.
We start with a simple problem. Consider a point P moving around a
circle of radius R with uniform angular velocity u (Figure 2.1.4). Denote by

r(t) the position vector of the point at time t and assume that the origin O
of the frame is chosen so that ||r|| does not change in time. How to express
r(t) in terms of r(t) and w?
Fig. 2.1.4 Rotating Point
To solve this problem, consider the plane that contains the circle of
rotation and define the r o t a t i o n v e c t o r u> as follows (see Figure 2.1.4).
The vector u> is perpendicular to the plane of the rotation; the direction
of the vector u> is such that the rotation is counterclockwise as seen from
the tip of the vector (alternatively, the rotation is clockwise as seen in the
direction of the vector); the length of the vector u: is w, the angular speed
of the rotation. As seen from Figure 2.1.4, r(t) = r(t) + r* and the vector
r* does not change in time, so that r(t) = r'(t). Consider a cartesian
coordinate system (?, j) with the origin at the center O of the circle in the
plane of rotation so that r(t) = Rcoscjti + Rsinwtj. Direct computations
show that
r'(t) = ( i x j ) x («f(<)) = iv x r{t) = w x r(t), (2.1.17)

r{t) = LJ x r[t). (2.1.18)
EXERCISE 2.1.6.c Verify all equalities in (2.1.17).

We will now use (2.1.18) to derive the relation between the velocities
and accelerations of a point mass m relative to two frames O and Oi,
when the frame 0\ is rotating with respect to O. We assume that the
two frames have the same origins: O = 0\. We also choose cartesian
coordinate systems (i, J, k) and (£i, j l t ki) in the frames; see Figure 2.1.9
on page 62. Let the frame 0\ rotate relative to the frame O so that the
corresponding rotation vector u> is fixed in the frame O. Because of this
rotation, the basis vectors in Oi depend on time when considered in the
frame O: ti = »i(t), Jx =J 1 (*), *i = M * ) - % (2-1.18),
dii/dt = u> x ii, djx/dt = u> x j 1 , dki/dt = u> x k\. (2.1.19)
Denote by ro(t) and ri(t) the position vectors of the point mass in O and
Oi, respectively. If P is the position of the point mass, then ro(t) = OP,
r\(t) = 0\P, and, with O = 0\, we have ro(t) — r\(t) for all t. Still, the
time derivatives of the vectors are different: ro(t) ^ ri(t) because of the
rotation of the frames. Indeed,
ri {t) = xx (t) §i + yi (t) j x + zx (t) ku
r0(t) = x1(t)i1{t)+y1(t)j1{t)+zl(t)K1{t);
recall that the vectors i%, j x , k\ are fixed relative to Oi, but are moving
relative to O. Let us differentiate both equalities in (2.1.20) with respect to
time t. For the computations of ri(t), the basis vectors are constants. For
the computations of ro(t), we use the product rule (1.3.4) and the relations
(2.1.19). The result is
r 0 (t) = r i ( t ) + u ; x r 1 ( t ) . (2.1.21)
EXERCISE 2.1.1? Verify (2.1.21).
There is nothing in the derivation of (2.1.21) that requires us to treat ro
as a position vector of a point. Accordingly, an alternative form of (2.1.21)
can be stated as follows. Introduce the notations Do and D\ for the time
derivatives in the frames O and 0\, respectively. Then, for every vector
function R = R(t), the derivation of (2.1.21) yields
D0R{t) = DiR(t) +UJX R(t). (2.1.22)

Relation (2.1.21) is a particular case of (2.1.22), when R is the position
vector of the point. We now use (2.1.22) with R = r0, the velocity of the
point in the fixed frame, to get the relation between the accelerations. Then
(i) Z V o = r 0 ; (ii) by (2.1.21), Dxr0 = f 1 + w x t , i ; (hi) also by (2.1.21)
w x r 0 = w x (ri + w x n ) . Collecting the terms in (2.1.22),
ro = r i + 2 u x r i + w x ( w x j ' i ) . (2.1.23)
Therefore, the acceleration in the fixed frame has three components: the
acceleration fi in the moving frame, the C o r i o l i s a c c e l e r a t i o n acor =
2u> x r i , and the centripetal acceleration o,c — u} x [u) x f i ) . Note that o,c
is orthogonal to both u> and w x r i .
Assume that the fixed frame O is inertial, and let F be the force acting
on the point mass m in O. By Newton's Second Law, we have mro = F in
the inertial frame O and, by (2.1.23),
mfi = F - 2mw x n - m w x f w x r i ) (2.1.24)
in the rotating frame 0\. Similar to (2.1.16), inertial forces appear as

corrections to Newton's Second Law in the non-inertial frame 0\. There
are two such forces in (2.1.24): the Coriolis force Fcor = - 2 m « x r\ and
the centrifugal force Fc = — mw x (a; x n ) .
As an example illustrating the relation (2.1.23), consider a point mass
m moving on the surface of the Earth along a meridian (great circle through
the poles) with constant angular speed 7; the axis of rotation goes through
the North and South Poles. We place the origins O and 0\ of the fixed and
rotating frames at the center of the Earth and assume that the frame 0\
is rotating with the Earth. (Figure 2.1.5).
Fig. 2.1.5 Motion Along a Meridian
Denote by P the current position of the point, and consider the plane
(NOP) Relative to the Earth, that is, in the frame Ox, the plane (NOP)
is fixed, and the motion of m is a simple circular rotation in this plane
with constant angular speed 7 so that 6(t) = jt. Relative to the fixed
frame O, the plane (NOP) is rotating, and the rotation vector is u>. We
will determine the three components of the acceleration of the point in the
frame O according to (2.1.23).
Introduce the polar coordinate vectors r, 6 in the plane (NOP). By
(1.3.27), page 36, we find ag = 0 and so
rl=ar = -j2Rr, (2.1.25)
where R is the radius of the Earth.

Similarly, we use (1.3.28), page 36, to find the velocity r j , of the point
in the frame 0\\ ti — "fR9, and then the Coriolis acceleration in O is
acor = 2CJ x r\ — 2^Ru> x 9.
To evaluate the product w x 9, we need some additional constructions.
Note that both u> and 9 are in the plane (NOP). Let b be the unit vector
that lies both in the (NOP) and the equator planes as shown in Figure
(2.1.5). This vector is rotating with the plane (NOP) and therefore changes
in time; by (2.1.18), b'(t) = u> x b(t). With this construction, the vector
LJ x 9 is in the plane of the equator and has the same direction as — b'(t).
By definition, 9 is orthogonal to r. By construction, the vectors CJ and b
are also orthogonal, and so the angle 6 between the vectors b and r is equal
to the angle between the vectors u> and 9. Since ||b'(£)|| — u = ||u>|| and
||0|| = 1, we find UJ x 9 — -sinOb', and therefore
acor(t) = -27-Rsin(7i)b'(i). (2.1.26)
Finally, we use 9 and b to write the centripetal acceleration ac = w x ( u x r )

of the point. We have
u>xr = Rsm(n/2-e)b' = RcosQb', u xb'=-uj2b. (2.1.27)
Hence,
ac = -(<Jj2Rcos'yt)b. (2.1.28)
Combining (2.1.25), (2.1.26), and (2.1.28) in (2.1.23), we find the total

acceleration r(t) of the point mass in the fixed frame O:
r(t) = --y2Rr(t) -2^Rsia{jt)b'(t) - cj2Rcos(-ft)b(t). (2.1.29)
EXERCISE 2.1.8. c Verify all the equalities in (2.1.27).

We now use (2.1.24) to study the dynamics of the point mass m moving
on (or close to) the surface of the Earth. Let r\ = r\(t) be the trajectory
of the point in the frame 0\. Unlike the discussion leading to (2.1.29), we
will no longer assume that the point moves along a meridian. Suppose that
O is an inertial frame. If F is the force acting on m in the frame O, then
Newton's Second Law implies mr = F. By (2.1.24),
mf\ = F - 2mu> x f i - mu> x (u> x n). (2.1.30)
The force F is the sum of the Earth's gravitational force Fa =

— {Km/\\r\\\2)ri and the net "propulsive" force Fp(t) that ensures the
motion of the point. If we prescribe the trajectory r\(t) of m in the frame
0\, then the force Fp necessary to produce this trajectory is given by
FP(t) = mri + 2mu>u x r i + rnu x ( w x n) — Fa- (2.1.31)
Conversely, if the force Fp — Fp(t) is specified, then the resulting trajec-

tory is determined by solving (2.1.30) with the corresponding initial condi-
tions ri(0), ri(0) and with F = FG + FP.
Note that (2.1.30) can be written as
mri = F — m acor — mac = FQ + Fp + Fcor + Fc.
Both forces FQ and Fc act in the meridian plane (NOP). Indeed, by the
Law of Universal Gravitation, the gravitational force FQ acts along the
line OP, where P is the current location of the point. The centrifugal force
Fc = mijj x (u> x n ) acts in the direction of the vector b, as follows from
the properties of the cross product; see Figure 2.1.5.
To analyze the effects of the Coriolis force Fcor = —macor = - 2 m u x
f i on the motion of the point mass, we again assume that the point is in
the Northern Hemisphere and moves north along a meridian with constant
angular speed. According to (2.1.26), the force Fcor is perpendicular to
the meridian plane and is acting in the eastward direction. The magnitude
of the force is proportional to sin#, with the angle 6 measured from the
equator; see Figure 2.1.5. In particular, the force is the strongest on the
North pole, and the force is zero on the equator. By (2.1.25) and (2.1.31), to
maintain the motion along a meridian, the force Fp must have a westward
component to balance Fcor. Thus, to move due North, the mass must be
subject to a propulsive force Fp having a westward component. The other
component of Fp is in the meridian plane.
The following exercise analyzes the Coriolis force when the motion is
parallel to the equator.
EXERCISE 2.1.9. Let ii, be the northward vector along the axis through
the North and South poles. We assume that the Earth rotates around this
axis, and denote by u> k the corresponding rotation vector. Let O be the
fixed inertial frame and 0\, the frame rotating with the Earth; the origins
of both frames are at the Earth center. Suppose that a point mass m is in
the Northern hemisphere and moves East along a p a r a l l e l ( a circle cut on
the surface of the Earth by a plane perpendicular to the line through the poles).
Denote by fi the velocity of the point relative to the frame 0\ so that r x
is perpendicular to the meridian plane; see Figure 2.1.6. By (2.1.24), the
Coriolis force acting on the point mass is Fcor = —2mu)h x f\.
Fig. 2.1.6 Motion Along a Parallel
(a)c Assume that both ||ri|| and 9 stay (approximately) constant during
the travel and that there is zero propulsion force Fp, as after a missile has
been fired. Ignore air resistance.
(i) Show that the point mass is deflected to the South, and the magnitude of
the deflection isu\\ri\\ sin9t2, wheret is the time of travel. Hint: verify that
Fcor • 9 = — 2mw||ri||sin0 ) and so 2mw||ri||sin6 | is the force pushing the point
mass to the South, (ii) Suppose that 9 = 41°, ||ri || = 1000 meters per second,
and the point mass travels 1000 kilometers. Verify that the point mass will
be deflected by about 50 kilometers to the South: if the target is due East,
the missile will miss the target if aimed due East. Hint: 2o>sin0 « 10~4.
(b)A Compute the deflection of the point mass taking into account the
change of \\ri\\ and 9 caused by the Coriolis force.
We now summarize the effects of the Coriolis force on the motion of a
point mass near the Earth.
• The force is equal to zero on the equator and is the strongest on the poles.
• For motion in the Northern Hemisphere:
Direction of Motion Deflection of trajectory

North East
South West
East South
West North
• For motion in the Southern Hemisphere
Direction of Motion Deflection of trajectory

North West
South East
East North
West South
EXERCISE 2 . 1 . 1 0 . B Verify the above properties of the Coriolis force. Hint:

The particular shape of the trajectory does not matter; all you need is the direction
ofr\, similar to Exercise 2.1.9.
EXERCISE 2.1.11. A Because of the Coriolis force, an object dropped down

from a high building does not fall along a straight vertical line and lands to
the side. Disregarding the air resistance, compute the direction and mag-
nitude of this deviation for an object dropped from the top of the Empire
State Building (or from another tall structure of your choice).
One of the most famous illustrations of the Coriolis force is the Fou-
CAULT PENDULUM, named after its creator, the French scientist JEAN
BERNARD LEON FOUCAULT (1819-1869). Foucault was looking for an
easy demonstration of the Earth's rotation around its axis, and around
1850 came up with the idea of a pendulum. He started with a small weight
on a 6 feet long wire in his cellar, and gradually increased both the weight
and the length of the wire. He also found a better location to conduct his
experiment. The culmination was the year 1851, when he built a pendulum
consisting of a 67 meter-long wire and a 28 kg weight swinging through a
three-meter arc. The other end of the wire was attached to the dome of the
Paris Pantheon and kept swinging via a special mechanism to compensate
for the air resistance and to allow the swing in any vertical plane. Because
of the rotation of the Earth around its axis, the Coriolis force was turning
the plane of the swing by about 270 degrees every 24 hours, in the clockwise
direction as seen from above.
EXERCISE 2.\.Yl.c Draw a picture and convince yourself that, if the pen-
dulum starts to swing in the Northern Hemisphere in the meridian plane,

then the Coriolis force will tend to turn the plane of the swing clockwise as
seen from above.
Let us perform a simplified analysis of the motion of the Foucault pen-
dulum in the Northern Hemisphere, away from the North Pole and the
equator. Assume that the pendulum starts to swing in a meridian plane.
Figure 2.1.7 presents a (grossly out-of-scale) illustration, with the Earth's
surface represented by the semi-circle. In reality, the length of the sup-
port and the height of the supporting point 0\ are much smaller than the
radius R of the Earth, so that \00\\ « |0.Po| ^ R- Also, the ampli-
tude of the swing is small compared to the length of the support, so that
|OiPo| « \0\PN\ = |OiPs|, and the linear distance \PNPS\ is approxi-
mately equal to the length of the corresponding circular swing arc.
Fig. 2.1.7 Foucault Pendulum
The pendulum is suspended at the point 0\, and the points PN, PS are
the two extreme positions of the weight. The angle 9 is the latitude of the
support point 0\. Denote the distance \PNPS\ by 2r. As seen from the
picture, the point PN is closer to the axis ON of the rotation of the Earth
than the point 0\, and the amount of this difference is \QNPQ\ = rsm9.
Similarly, the point Ps is farther from the axis than 0\ by the same amount.
Since the Earth is rotating around the axis ON with angular speed u, the
points PN, 0\, and Ps will all move in the direction perpendicular to the
meridian plane. The point PN will move slower than Oi, and the point Ps,
faster, causing the plane of the swing to turn. The speed of Ps relative to
0\ and of 0\ relative to PN is ruj sin 9. If we assume that these relative
speeds stay the same throughout the revolution of the swing plane, then
the weight will be rotating around the point Po with the speed rusinO,
and we can find the time T of one complete turn. In time T, the weight
will move the full circle of radius r, covering the distance 2nr. The speed
of this motion is ru sin#, and so
T = - ^ = ^ , (2.1.32)
ru sin 8 sin f)
where To = 27r/o; « 24 hours is the period of Earth's revolution around its
axis. The plane of the swing will rotate 2n sin 0 radians every 24 hours. For
the original Foucault pendulum in Paris, we have 6 « 48.6°, which results
in T = 32 hours, or a 270° turn every 24 hours.
Note that the result (2.1.32) is true, at least formally, on the poles and
on the equator. Still,
• On the poles, T = 24 hours as the Earth is turning under the pendulum,

making a full turn every 24 hours.
• On the equator, where there is no Coriolis force, the points OI,PN,PS
are at the same distance from the axis ON (this is only approximately
true if the swinging is not in the plane of the equator). As a result, the
plane of the swing does not change: T = +oo.
EXERCISE 2.1.13. B Find the period T for the Foucault pendulum in your
home town.
The Coriolis force due to the Earth's rotation has greater effects on the
motion than might be deduced from an intuitive approach based on the
relative velocities of the moving object and the Earth. In particular, these
effects must be taken into account when computing trajectories of long-
range missiles. With all that, we must keep in mind that the effects of the
Coriolis force due to the Earth rotation are noticeable only for large-scale
motions. In particular, the Coriolis force contributes to the erosion of the
river banks, but has nothing to do with the direction of water swirling in
the toilet bowl.
The Coriolis force also influences the direction of the ATMOSPHERIC
WINDS. This was first theorized in 1856 by the American meteorologist
WILLIAM F E R R E L (1817-1891) and formalized in 1857 by the Dutch me-
teorologist CHRISTOPH HEINRICH DIEDRICH BUYS B A L L O T (1817-1890).
/• / Polar Easterlies
/
60° /
//y'Temperate Westerlies/ /
30°/ Calms of Cancer
/ / Easterly Trade Winds / /
Doldrums
XÊasterly Trade Winds \ ^ \ ,
Calms of Capricorn
\ \ \Temperate Westerlies\\/
\
Polar Easterlies '
Fig. 2.1.8 Atmospheric Winds
T h e general wind p a t t e r n on the E a r t h is as follows (see Figure 2.1.8).

W a r m air rises vertically from the surface and is deflected by the Corio-
lis force, resulting in easterly trade winds, temperate westerlies, and polar
easterlies. T h e deflection is to the right in the Northern Hemisphere and to
the left in the Southern Hemisphere, so t h a t the p a t t e r n s in the two hemi-
spheres are mirror images of each other. Three regions of relative calm
form: the doldrums around the equator, calms of Cancer around the 30°
parallel in the Northern Hemisphere, and calms of Capricorn around the
30° parallel in the Southern Hemisphere.
Let us discuss the formation of the easterly trade winds in the Northern
Hemisphere. T h e Sun heats the surface of the E a r t h near the equator.
T h e air near the equator also gets warm, becomes lighter, and moves up,
creating the area of low pressure near the equator and causing the cooler
air from the north to flow south. T h e flow of the cooler air from the North
creates the area of low pressure at high altitudes, deflecting the rising warm
air from the equator to the north. T h e Coriolis force deflects this flow to the
East. At higher altitudes, the air cools down. Cooler, denser air descends
around t h e 30° parallel and flows South back to the lower pressure area
around the equator. T h e Coriolis force deflects this southward flow to the
West. In the stationary regime, this circulation produces a steady wind
from the North-East, the easterly trade winds.
General Accelerating Frames 61
EXERCISE 2.1.14. Explain the formation of the temperate westerlies and

the polar easterlies.
EXERCISE 2.1.15.'4 The flight time from Los Angeles to Boston is usually
different from the flight time from Boston to Los Angeles. Which flight
takes longer? Which of the following factors contributes the most to this
difference, and how: (a) The Earth's rotation under the airplane; (b) The
Coriolis force acting on the airplane; (c) The atmospheric winds? Hint: If
in doubt, check the schedules of direct flights between the two cities.
The complete mathematical model of atmospheric physics is vastly
more complicated and is outside the scope of this book; possible refer-
ence on the subject is the book An Introduction to Dynamic Meteorology
by J. R. Holton, 2004, and some partial differential equations appearing in
the modelling of flows of gases and liquids are discussed below in Section
6.3.5.
In 1963, while studying the differential equations of fluid convection, the
American mathematician and meteorologist EDWARD NORTON LORENZ (b.
1917) discovered a chaotic behavior of the solution and a strange attrac-
tor. These Lorentz differential equations are a prime example of a chaotic
flow. They also illustrate the intrinsic difficulty of accurate weather predic-
tion. His book The Nature And Theory of The General Circulation of The
Atmosphere 1967, is another standard reference in atmospheric physics.
2.1.4 General Accelerating Frames

The analysis in the previous section essentially relied on the equation
(2.1.18) on page 51, which was derived for uniformly rotating frames. In
what follows, we will use linear algebra to show that, with a proper def-
inition of the vector u>, relation (2.1.18) continues to hold for arbitrary
rotating frames.
Consider two cartesian coordinate systems: (i, j , k) with origin O, and
(£i, j l 7 K I ) with origin 0\. We assume that O = 0\\ see Figure 2.1.9.
Consider a point P in R 3 . This point has coordinates (x, y, z) in
(z, j , k) and {x\, j/i, z{) in (fj, j l t ki). Then
x i + y j + z k = xi ii + yi j 1 + zi ki. (2.1.33)
We now take the dot product of both sides of (2.1.33) with i to get
x = xi(ii -i) + yi(j1 •i) + zi(ki -i). (2.1.34)

Fig. 2.1.9 3-D Rotation of Frames
Similarly, we take the dot product of both sides of (2.1.33) with j :
V = xi(ii • 3) + 2/i(Ji • 3) + *i(«i • 3) (2.1.35)

and with k:
z = xi{i\ • k) + j / i & • k) + zi(ki • k). (2.1.36)
The three equations (2.1.34)-(2.1.36) can be written as a single matrix -

vector equation,
*i • 3 3\-3 K-i-3 (2.1.37)

Kii • k 3i ' £ ki • k
Consider the matrix

l
*i ' 3i ' % Ki • *
U = I 5i • 3 3i3 Ki • 3 (2.1.38)
Ji • k j1-k KI-K,
EXERCISE 2.1.16. ( a ) c Verify that the matrix U is orthogonal, that

T
is UU = U U = I. Hint: 1 = h • fc = (f • d ) 2 + (i • ti) 2 + (A • i"i)2,
T
0 = *i • Ji = (*i • *)(ii • «) + (*i • i)(ji • i) + (ii • K)C?I • «)• WA Veri/j/ i/iai

A
£/ie determinant of the matrix U is equal to 1. (c) Verify that the matrix
U is a representation, in the basis (i\, J 1 ; k\), of an orthogonal transfor-
mation (see Exercise 8.1.4, page 453, in Appendix). This transformation
rotates the frame 0\ so that {i\, j \ , «i) moves into (i, 3, k). (d)A Verify
that the matrix UT is a representation, in the basis (z, j , k), of an or-

thogonal transformation that rotates the space so that (z, j , k) moves into
Now assume that the picture on Figure 2.1.9 is changing in time as

follows:
• Frame O and its coordinate system (i, j , k) are fixed (not moving).
• Oi(t) = 0 for alii.
• The coordinate system (?i, j x , k{) in frame 0\ is moving (rotating)
relative to (i, j , k).
• The point P is fixed in frame 0\ relative to (ii, j l y k\).
Thus, xi,yi,z\ are constants and P is rotating in the (?, j , k) frame. Then
(2.1.33) becomes
x(t) i + y(t) j + z(t) k = xi h(t) +yi3i(t) +ziki(t),
Define the matrix U = U(t) according to (2.1.38), and assume that the
entries of the matrix U are differentiable functions of time; it is a reasonable
assumption if the rotation is without jerking. Since U(t)UT(t) = I for all t,
it follows that d/dt(UUT) = 0, the zero matrix. The product rule applies
to matrix differentiation and therefore
i/uT + ui/T = iiuT + (iiuT)T = o,

which means that fi(£) = U(t)UT(t) is antisymmetric, that is, has the form
/ 0 -w 3 (i) W2(t) \
fi(t) = w 3 (t) 0 -wi(t)
We use the entries of the matrix Q(t) to define mathematically the

i n s t a n t a n e o u s r o t a t i o n v e c t o r in the fixed frame O:
u(t) = u>i (t) i + W2(t)j+ w3 (t) k. (2.1.39)
EXERCISE 2.1.17? (a) Verify that, for every vector R = Rii + R^j+R3k
and each t,
Q.{t) R = w(t) x R. (2.1.40)

Hint: direct computation, (h) Consider the vector OP = r*o(i) = x(t) i +

y(t)j + z(t) k rotating in the frame O. Verify that, with the above definition
of u, we have
r0(t) = w(t) x r 0 ( i ) . (2.1.41)
Hint: write relation (2.1.37) as ro(t) = U{t)f\, where f\ = x\ i + j/i j + z\ k.

Then rx = UT{t)r0(t) and r0{t) = U{t)ri.
Note that (2.1.41) agrees with (2.1.18) on page 51 when u(t) is constant
and justifies the above definition of u>(t) as a rotation vector.
Denote by ro(t) the position vector of a point P in the frame O, and
by T*I(£), the position of the same point in the frame 0\. Since the frames
have the same origin, we have ro(t) = ri(t) for all t. On the other hand,
because of the relative rotation of the frames, the values of ro(t) and ri(i)
are different. As we did earlier on page 52, denote by Do and D\ the
derivatives with respect to time in the frames O and 0\, respectively. If the
point P is fixed in the rotating frame 0\ and 0\P = r\ = x\ £1+2/1 Ji+zi ki
is the position vector of P in 0\, then Dir0(t) = r\(t) = 0, and (2.1.41)
implies Doro(t) = ro(t) — UJ x ro(t).
Similar to the derivation of (2.1.21), we can show that if the point P
moves relative to the frame 0\, then
r0(t) = ri(t)+u>xr0(t). (2.1.42)
Therefore, for every vector R — R(t), expressed as functions in frames O

and Oi, both denoted by R
D0R(t) = DiR(t) + u(t) x R(t). (2.1.43)
EXERCISE 2.1.18? Verify (2.1.43). Hint: write R(t) = x(t)i1(t)+y(t)j1(t) +

z(t) ki(t) and differentiate this equality using the product rule. Since the vectors
*i> J i ! ki are fixed in the frame 0\, you can use equality (2.1.41) to compute
the time derivatives of these vectors. Also, by definition, D\R(t) = x(t)i\(t) +
» ( * ) 5 i ( t ) + * ( * ) « i (*)•
Remark 2.1 Let us stress that the vector ro(t) = x(t)i+y(t)j+z(t) k is

the same as the vector ri(t) = zi(£) ?i(£) +2/1 (£)&(*) +21 (£) ki(t): both are
equal to OP even if the point P moves relative to the frame 0\. As a result,
ro(i) =/= U(t)ri(t). There is no contradiction with (2.1.37), because the
components of the vectors TQ, r\ are defined in different frames and cannot
be related by a matrix-vector product. What does follow from (2.1.37) is
the equality ro(t) = U{t)r\{t), where f\{t) = x\(t)i + y\{t) j + z\{t) k.

Also keep in mind that, despite the equality of the vectors ro(t) = ri(t), the
curve C defined by ro in the frame O is different from the curve C\ defined
by 7*1 in the frame 0\. For example, if P is fixed in 0\, then C\ is jus a
single point.
EXERCISE 2.1.19r Consider the special case of a uniform rotation of frame
0\ relative to frame O so that the origins of the frames coincide, k = ki,
and the rotation vector is u> = W3K. Calculate U(t) and U{t). Show that
the matrix fl = U(t)UT(t) has the form
/O - w 3 0 \
n = w3 0 0 .
\o 0 0/
For the point P fixed in the rotating frame and having the position vector
in the fixed frame ro(t) = x(t) i + y(t) j + z(t) k\ show that
r0{t) = Cl,ro{t) = -u3y{t)i + u>3x(t)3 = w x r0(t).
As a result, you recover relation (2.1.18) we derived geometrically on page
51.
To continue our analysis of rotation, assume that the vector function
w = u){t) is differentiable in t. Then we can set R = ro = A)f*o in (2.1.43)
and use (2.1.42) to derive the relation between the accelerations of the point
in the two frames:
ro(*) = ri(t) + 2 w ( t ) x f i ( t ) + w(t)x ( u ( t ) x r 0 ( t ) ) + w ( t ) x r 0 ( t ) ; (2.1.44)

as before, ro and r i are the position vectors of the point in the frames O
and 0\, respectively.
EXERCISE 2.1.20? (a) Verify (2.1.44). Hint: r0(t) = D0r0(t) = D 0 (fi +u> x
r 0 ) = D\r\ + u x r i + u x r o + u x ( f i + u x ro). Note that both u) and ro are
defined in the same frame O, so the product rule (1.3.6), page 26, applies, (b)
Verify that (2.1.44) can &e written as
Mt)=ri(t) + 2u(t)xri(t)+w(t)x(u(t)xri(t))+w(t)xr1(t). (2.1.45)
Finally, assume that the point 0\ is moving relative to O so that the

function roi(t) = 00\ is twice continuously differentiable. Then we have
r(t) = roi(t) + ri(t). Consider the parallel translation of the frame O
with the origin O' at 0\, and define the rotation vector u to describe the
66 Systems of Point Masses
rotation of the frame 0\ relative to this translated frame O'. We can now
combine relation (2.1.44) for rotation with relation (2.1.12) on page 47 for
parallel translation to get
ro(*) = r o i ( * ) + r i ( t )
+ 2 u ( t ) x fi(t) +w(f) x (u(t) x r i ( t ) ) + u ( J ) x n ( t ) .
EXERCISE 2.1.21? Verify (2.146). Hint: Apply (2.1.45) ton, replacing frame
O with O'.
Suppose that the frame O is inertial, and a force F is acting on the
point mass m. Then, by Newton's Second Law, mro{t) = F; to simplify
the notations we will no longer write the time dependence explicitly. By
(2.1.46),
mf\ = F - mr-Qi - 2mw x r\ - mu x (u x n ) -mCo x T\. (2.1.47)
As before in (2.1.13), page 47, and in (2.1.24), page 53, we have several
corrections to Newton's Second Law in the non-inertial frame 0\. These
corrections are the t r a n s l a t i o n a l a c c e l e r a t i o n force Fta = —inôi,
the C o r i o l i s force Fcor = —2mu> x r i , the c e n t r i f u g a l force Fc =
—rawx ( w x r i ) , and the angular a c c e l e r a t i o n f o r c e F a a = - m w x r i .
2.2 S y s t e m s of Point Masses
The motion of a system of point masses can be decomposed into the motion
of one point, the center of mass, and the rotational motion of the system
around the center of mass. In what follows, we study this decomposition,
first for a finite collection of point masses, and then for certain infinite
collections, namely, rigid bodies.
2.2.1 Non-Rigid Systems of Points

n
Let 5 be a system of n point masses, m i , . . . , mn, and M — J^ rrij, the total
mass of S. We assume that the system is non-rigid, that is, the distances
between the points can change. Denote by rj the position vector of rrij in
some frame 0. By definition, the c e n t e r of mass (CM) of S is the point
Non-Rigid Systems of Points 67
with position vector

n
1
We will see that some information about the motion of S can be obtained
by considering a single point mass M with position vector TCM-
EXERCISE 2.2.1. c Verify that a change of the reference point O does
not change either the location in space of the center of mass or formula
(2.2.1) for determining the location: if O' is any other frame and fj is
the position vector of rtij in O', then the position vector of CM in O' is
n > ^
m
rcM = (1/-W) 5Z j Tj. Hint: fj = Vj + O'O and so TCM = TCM + O'O,
which is the same point in space.
EXERCISE 2.2.2. (a)B Show that the center of mass for three equal
masses not on the same line is at the intersection of the medians of the
corresponding triangle. (b)A Four equal masses are at the vertices of a
regular tetrahedron. Locate the center of mass.
The velocity and acceleration of the center of mass are TCM and fcM,
respectively. Differentiating (2.2.1) with respect to t, we obtain the relations
n
1
m i
rcM = jjYl i 'i> (2-2-2)
J=I
1 "
fCM = jf^Zmjfj. (2.2.3)
3= 1
To study the motion of the center of mass, suppose t h a t t h e reference

frame O is inertial a n d denote by Fj, 1 < j < n, the force acting on the
point mass rrij. T h e n Fj = rrij fj, and, multiplying (2.2.3) by M, we get
the relation
n m
MfCM = Yl 3 *3 =J2Fi = F
m
(2-2-4)
3=1 3=1
Equality (2.2.4) suggests t h a t the total force F can be assumed to act on

a point mass M at the position TCM- As a next step, we will study the
structure of the force F.
Typically, each Fj is a sum of an external force F\ ' from outside of S

and an internal system force F], exerted on rrij by the other n — 1 point
masses. Thus, Fj = Ylk^tj Fjk> where F V is the force exerted by rrik
on rrij. Hence,
F
= itFJ = Y, F<JE) + £ FT = p{B)+F{1) •
j=l j=l j=l
By Newton's Third Law, F$ = -F$. It follows that F ( 7 ) = £ " = 1 Ff] =

(E) {E)
0 and therefore, F = ^ ) " = 1 F = F . By (2.2.4), the motion of the
center of mass is then determined by
MrCM = F{E). (2.2.5)
The (linear) momentum PQM °f the center of mass is, by definition,
MV
PCM = CM-
With this definition, equation (2.2.5) becomes pCM = FÊ\ and if the net
external force F^ ' is zero, then PCM ls constant. By (2.2.2),
n n
m
PCM = J2 ^ =
12PJ> (2-2-6)
where Pj = rrij Tj is the momentum of rrij. Thus, if FÊ' = 0, then the

total linear momentum ps — Yll=i Pj °f ^ e system is conserved.
Next, we consider the rotational motion of the system. By definition
(see (2.1.4) on page 40), the angular momentum LQ-J of the point mass
rrij about the reference point O is given by Loj = Tj X rrij Tj = Vj x Pj.
Accordingly, we define the angular momentum Lo of the system S about
O as the sum of the Lo,j'-
n n n
j=i j=i j=\
For the purpose of the definition, it is not necessary to assume that the
frame O is inertial. Note that, unlike the relation (2.2.6) for the linear
momentum, in general LQ ^ rCM x Mr CM-
From (2.2.7) it follows that

n n
AT
3= 1 j=l 3= 1
since rj x rj = 0. Now, we again assume that the frame O is inertial. Then

F j = rrij Tj and dLo/dt = Y^j=i rj x Fj = !Cj=i Tj, where Tj = Tj X FJ
is the torque about O of the force Fj acting on rrij. We define the total
torque To = 2 j = i Tj and conclude that
^ = r0. (2.2.8)
Equation (2.2.8) is an extension of (2.1.6), page 41, to finite systems of
point masses. If To = 0, then Lo is constant, that is, angular momentum
is conserved.
In general, unlike the equation for the linear momentum (2.2.5), the
torque To in (2.2.8) includes both the internal and external forces. If
the internal forces are central, then only external forces appear in (2.2.8).
Indeed, let us compute the total torque in the case of CENTRAL INTERNAL
FORCES. The internal force FJ J ) acting on particle j is FJ 7 ) = f^ F^. By
fc=i
k?j
Newton's Third Law, we have F$ = -F$. Then
, j- n n n n
-^ = J2rjX F> = £ - , x Ff> + J > x X>g>
j=i j=i j ' = i k=i
k?j
n n . .
The terms in the product Yl rj x 2 -^k can
^ e arranged as a sum of
J'=I fc=i
pairs r j x FJj/ + r^ x FJy for each (j, k) with j ^ /c. Also, TJ X FJ fc ' +
rk x FJL- = (r^ — rk) x Fjk- The vector rk — T"j is on the line joining rrij
and rrik- If the forces F-k are central, as in the cases of gravitational and
electrostatic forces, then the vector F L is parallel to the vector (rj —rk),
and (rj — rk) x F k ' = 0. In other words, the internal forces do not
contribute to the torque, and (2.2.8) becomes
d
^ = ±rjxFf=±T^=T^, (2.2.9)
3= 1 .7 = 1
where T J E ) = rj x Ff] is t h e external torque of F J £ ) a n d T{0E) is t h e

total torque on the system S by the external forces.
N e x t , we look a t t h e A N G U L A R M O M E N T U M O F A S Y S T E M RELATIVE
TO THE CENTER OP MASS. Again, let O b e a n arbitrary frame of reference,
let r j (t) be t h e position of mass rrij, 1 < j < n, a n d let TCM (*) D e t h e
position of the center of mass. Define by Xj the position of rrij relative t o
the center of mass:
tj = ^ - rCM. (2.2.10)
T h e n r-j = feu + ij and, according t o (2.2.7),
n n
3=1 3=1
n n n
= ^2 rrij rCM x rcM + ^ rrij Xj x rCM + ^ rrij TCM X ij (2.2.11)
3= 1 3= 1 3=1
n
+ ^mjr j xij.
3=1
E X E R C I S E 2.2.3. c (a) Verify that
n
Y^mjXj = 0, (2.2.12)
j=i
where Xj is the position vector of the point mass rrij relative to the center
n n n n
of mass. Hint: ]>2 rrij tj = ^2 rrijrj — ^2 rrij rcM = ^2 mi ri ~ Mr CM — 0.
3=1 3=1 3=1 3=1
(b) Use (2.2.11) and (2.2.12) to conclude that
n
L0 = M rCM x rc'M + ^ rrij Xj x ij. (2.2.13)
3=1
If we select t h e origin O of the frame at t h e center of mass of the sys-

tem, then TQM = 0, and we get the expression for the a n g u l a r momentum
a r o u n d t h e c e n t e r of mass:
n
LCM = J^2mjXj xij, (2.2.14)
3=1
where Vj(t) = rj(t) - rcM(t). Hence, (2.2.13) becomes
L0 = M rCM x rCM + LCM, (2.2.15)
that is, the angular momentum of the system relative to a point O is equal to
the angular momentum of the center of mass relative to that point plus the
angular momentum of the system relative to the center of mass. Once again,
we see that the center of mass plays a very special role in the description
of the motion of a system of points.
We emphasize that LQ ^ Mr CM X rcM as long as LCM ^ 0. Note
also that the vector functions Xj (t) and ij (t) depend on the choice of the
reference frame.
Let us now compute the time derivative of LcM{t) using the differen-
tiation rules of vector calculus (1.3.3), (1.3.4), (1.3.6) (see page 26). Since
ij x ij = 0, we have
j n n n
rn x x m x =
—LCM = y~} i i *•»' "*" 5 Z J *•» ^' 5 Z m J xi x X
J' (2-2.16)
3= 1 j=l j= l
To express dLcM/dt in terms of the forces Fj acting on the rrij, we would

like to use Newton's Second Law (2.1.1), and then we need the frame to be
inertial. The frame at the center of mass is usually not inertial, because
the center of mass can have a non-zero acceleration relative to an inertial
frame. Accordingly, we choose a convenient inertial frame O and apply
(2.2.4), page 67, in that frame:
ij = jr-j - TCM = Fj/rrij - F/M,

rrij Xj x ij = Xj x Fj - (rrij/M) Xj x F.
By (2.2.16) above,
dLcM n 1 / " \
dt j=i \j=i j
and then (2.2.12) implies
dLcM
= Y,XjXFj. (2.2.17)
dt
If the internal forces i^L are central, then, by (2.2.9), these forces do not
contribute to the total torque. Since Xj —Xk = Tj — TCM — (>*fc — I"CM) =
Tj — rfe, we therefore find
^ f = t*J x Ff> = £ T & , = T & , (2.2.18)
where TCM, is the external torque of Fj ' about the center of mass and
TQM ^S t n e total torque by the external forces. Using (2.2.16) above, we
find
^ = ±mjvj,ij=TcSM. (2.2.19)
Equations (2.2.5) and (2.2.19) provide a complete description of the motion

of the system of point masses in an inertial frame.
As an EXAMPLE illustrating (2.2.5) and (2.2.19), let us consider BI-
NARY STARS. A binary, or double, star is a system of two relatively close
stars bound to each other by mutual gravitational attraction. Mathemati-
cally, a binary star is a system of n = 2 masses mi and mi that are close
enough for the mutual gravitational attraction to be much stronger than
the gravitational attraction from the other stars. In other words, we have
F$ = -F^ and F[B) = F(2E) = 0. Using the equation for the linear
momentum (2.2.5), page 68, we conclude that rcM = 0 and TCM is con-
stant relative to every inertial frame O. We can therefore choose an inertial
frame with origin at the center of mass of the two stars. Applying (2.2.19)
in this frame, we find dLcAi/dt — 0, and by (2.2.16),
mi t i x t i + m.212 x x-2 = 0. (2.2.20)
EXERCISE 2.2.4. B Assume that mi = m^. Show that the two stars move
in a circular orbit around their center of mass. Hint: you can complete the
following argument. By (2.2.1), rcM = (1/2)(T*I -\-T2). SO CM is the midpoint
between m\ and m.2. Hence, t i = —12, i i = —12, t i = —12- By (2.2.20) above,
2ri x Vi = 0 . This implies that ti and vi are parallel (assuming ti ^= 0). Since
dLcM/dt = 0, LCM is constant. By (2.2.14) on page 70, 2m\X\ x ti is constant
as well. Together with X\ x ti = 0, this is consistent with equations (1.3.27),
(1.3.28), and (1.3.29), page 36, for uniform circular motion.
Binary stars provide one of the primary settings in which astronomers
can directly measure the mass. It is estimated that about half of the fifty
stars nearest to the Sun are actually binary stars. The term "binary star"
was suggested in 1802 by the British astronomer Sir WILLIAM HERSCHEL
Rigid Systems of Points 73
(1738-1822), who also discovered the planet Uranus (1781) and infra-red
radiation (around 1800).
2.2.2 Rigid Systems of Points

A system 5 of point masses rrij, 1 < j < n, is called r i g i d if the distance
between every two points rrii, rrij never changes. Let O be a reference frame
and let Vj{t) be the position in that frame of rrij at time t. The rigidity
condition can be stated as
\\rj{t)-ri{t)\\=dii foralU, i,j = l,...,n, (2.2.21)
where the dij are constants.

EXERCISE 2.2.5? Verify that condition (2.2.21) is independent of the choice
of the frame.
EXERCISE 2.2.6.° Let S be a rigid system in motion. Prove that the norm
\\rj — rCM|| and the dot product Tj • TCM remain constant over time for all
j = 1 , . . . , n, that is, the center of mass of a rigid system is fixed relative to
all rrij. Hence, the augmented system mi,... ,mn, M, with M located at
the center of mass, is also a rigid system.
We will now derive the equations of motion for a rigid system. If we con-
sider a motion as a linear transformation of space, then condition (2.2.21)
implies that the motion of a rigid system is an isometry. The physical real-
ity also suggests that this motion is orientation-preserving, that is, if three
vectors in a rigid system form a right-handed triad at the beginning of the
motion, they will be a right-handed triad throughout the motion.
EXERCISE 2.2.7. (a)B Show that an orientation-preserving orthogonal
transformation is necessarily a rotation. (b)c Using the result of part (a)
and the result of Problem 1.9 on page 412 conclude that the only possible
motions of a rigid system are shifts (parallel translations) and rotations.
Let O be a frame with a Cartesian coordinate system (z, j , k). Let S
be a rigid system moving relative to O. Denote by rcM(t) the position of
the center of mass of S in the frame O. We start by introducing two frames
connected with the system. Let OCM be the parallel translation of the
frame O to the center of mass. Thus, OCM moves with the center of mass
of the system S but does not rotate relative to O. Let 0\ be the frame with
00\ — rcM{t) and with the cartesian basis (?i, J j , k\) rotating with the
system. Define the corresponding rotation vector u> according to (2.1.39)

on page 63. Let us apply equation (2.1.46), page 66, to the motions of the
point mass rrij relative to frames O and 0\. Denote by rj the position
vector of rrij in the frame O (draw a picture!) Then Xj = rj — TCM is
the position vector of rrij in OCM • Because of the rigidity condition, the
position vector rij of rrij in the frame Oi does not change in time, and so
f\j = 0 and rij = 0. By (2.1.42) on page 64, rj = rcM + w x Xj, and
then, by (2.1.46), 'fj = TTCM + u x (u> XXJ) + U; x Xj. Since the frame OCM
is not rotating relative to O, we have x = f j — rcM, and
Xj=uxxj, (2.2.22)
Xj = u> x (u> x Xj) + u> x Xj. (2.2.23)
Next, we use identity (1.2.27) on page 22 for the cross product:
w x (u> x Xj) = (w • Xj) OJ - (a; • u>) Xj = (u • Xj) LJ - J1 Xj, (2.2.24)
where u) = ||u>||. To compute the rate of change of the angular momentum,

we will need the cross product Xj x Xj. Applying (1.2.27) one more time,
Xj X(UXXJ) = \\xj\\2u-{xj-uj)xj. (2.2.25)
Putting everything together, we get
Xj x ij = (u • Xj) XjXu> + \\XJf cj - (XJ • u) Xj. (2.2.26)
After summation over all j , the right-hand side of the last equality does
not look very promising, and to proceed we need some new ideas. Let
us look at (2.2.26) in the most simple yet non-trivial situation, when the
rotation axis is fixed in space, and all the point masses rrij are in the plane
perpendicular to that axis. Then w • tj = 0 for all j , and <ii = Co Q, where
Q is the unit vector in the direction of u>; note that since the rotation axis
is fixed, the vectors w and w are parallel. Accordingly, equality (2.2.24)
becomes U}X(LJ xxj) = —co2Xj, and since d(toCj)/dt = u u), equality (2.2.25)
becomes Xj X(UJXXJ) = ||tj || 2 wQ. Summing over all j in (2.2.26) and taking
into account these simplifications, we find from (2.2.16) on page 71
U1U),
m
and it is therefore natural to introduce the quantity ICM = 2 jll*j|| 2 !
which is called the moment of inertia of S around the line that passes
through the center of mass and is parallel to Q. Then
dLcM T . ~ ,„ „ „_*
dt = ICMUU. (2.2.27)
Our main goal is to extend equality (2.2.27) to a more general situation;

the moment of inertia thus becomes the main object to investigate. To
carry out this investigation, we backtrack a bit and look closely at the
angular momentum of the rigid system around the center of mass LCM =
Z3"=i mjVj x ij. By (2.2.22), we have Xj x ij = Xj x (w x Xj). Similar to
(2.2.25) we find Xj x (w x Xj) = ||tj|| 2 w — (XJ • U>)XJ and
VCM = \ J2m^x^2 u
-J2mi(xi ' w ) r J- (2.2.28)
^'=1 / 3=1
As written, equality (2.2.28) does not depend on the basis in the frame
OCM- To calculate LCM, we now choose a cartesian coordinate system
(i, j , k) in the frame OCM- Let Xj(t) = Xj(t)i + Vj(t)j+ Zj(t)k and
w(t) = wx(t) i + ojy(t)j + LJz(t) k. From (2.2.28) above,
LCM = ( J2mi(x2i +
$ + **) I w - I ^2mixi v
i I w*
Vj=1
* I (2.2.29)
rrijZjXj wz,
u=i / \j=i
where we omitted the time dependence notation to simplify the formula.

Since LCM = LCMX * + LCMV 3 + LCMZ «•, to compute the ^-component
LCMX of LCM, we replace the vector Xj in (2.2.29) with Xj, and the vector
u>, with wx:
n n n
LCMX =^x^2 rrij [y] + zf) - uy ^ rrijXjyj - wz ^ rrijXjZj;
3=1 3=1 3=1
similar representations hold for LcMy and LCMZ-

It is therefore natural to introduce the following notations:

n n n
j=i j=\ j=\

n n
= = == =
•*xy -*yx / j ^'j'^jUji ^xz *zx / j TTljXjZj,
lm
•iyz — *zy — / J jVizj-
(2.2.30)
With these notations, LCMx = uxIxx - ujyIxy - wzIxz.

EXERCISE 2.2.8. C Verify that
LcMy — —WxIyx + COylyy — U!zIyz, LQMZ — ~^XIZX — UJyIzy + WZJZZ.
EXERCISE 2.2.9. C ' Assume that all the point masses are in the (i, j) plane.
Show that Ixz = Iyz = 0 and Ixx + Iyy = Izz.
We can easily rewrite (2.2.28) in the matrix-vector form:
CCM(t) = IcM(t)Ct(t), (2.2.31)
where CcM(t) is the column vector (LCMx{t), LCMv(t), LCMz{t))T, Cl(t)

is the column vector {wx{t), u)y(t), ivz(t))T, and
(2.2.32)
The matrix ICM is called the moment of i n e r t i a matrix, or t e n s o r of

i n e r t i a , of the system 5 around the center of mass in the basis (i, j , K).
The Latin word tensor means "the one that stretches," and, in mathematics,
refers to abstract objects that change in a certain way from one coordinate
system to another. All matrices are particular cases of tensors. For a
summary of tensors, see page 457 in Appendix.
As much as we would like it, equality (2.2.31) is not the end of our
investigation, and there two main reasons for that:
(1) It is not at all clear how to compute the entries of the matrix ICM-
(2) Because the system S is rotating relative to the frame O, the entries of
the matrix ICM depend on time.
Remembering that our goal is an equation of the type (2.2.27), we have to

continue the investigation of the moment of inertia.
Let us forget for a moment that we are dealing with a rotating system,
and instead concentrate on the matrix (2.2.32). We know from linear al-
gebra that every change of the coordinate system changes the look of the
matrix; see Exercise 8.1.4, page 453 for a brief summary. For many pur-
poses, including ours, the matrix looks the best when diagonal, that is,
has zeros everywhere except on the main diagonal. While not all matrices
can have this look, every symmetric matrix is diagonal in the basis of its
normalized eigenvectors; see Exercise 8.1.5 on page 454.
By (2.2.30) and (2.2.32), the matrix ICM is symmetric, and therefore
there exists a cartesian coordinate system (?*, j * , k*) in which ICM has at
most three non-zero entries 1*^, I22 > -^33 > and all other entries zero. In other
words, there exists an orthogonal matrix U» so that the matrix ICM =
U*ICMUJ is diagonal. The matrix [/* is the representation in the basis
(2, j , k) of a linear transformation (rotation) that moves the vectors i, j , k
to ?*, j * , k*, respectively. The vectors i*, j * , k* are called the p r i n c i p a l
axes of the system S. We will refer to the frame with the origin at the center
of mass and the basis vectors i*, f, k* as the p r i n c i p a l axes frame.
Both the principal axes and the numbers 7 ^ , J£2i ^33 depend only on the
configuration of the rigid system S, that is, the positions of the point masses
rrij relative to the center of mass. A matrix can have only one diagonal look,
but in more than one basis: for example, the identity matrix looks the same
in every basis. Accordingly, the principal axes might not be unique, but
the numbers 1 ^ , I%2, I^3 are uniquely determined by the configuration of the
system, and, in particular, do not depend on time. We know from linear
algebra that the numbers 1^, I^\,I33 are the eigenvalues of the matrix ICM >
and the matrix U* consists of the corresponding eigenvectors.
EXERCISE 2.2.10.A Given an example of a symmetric 3 x 3 matrix whose
entries depend on time, but whose eigenvalues do not. Can you think of a
general method for constructing such a matrix?
Formulas (2.2.30) for the entries of the matrix of inertia are true in
every basis, and therefore can be used to compute the numbers 1^, I22, 7| 3 .
Since these numbers do not depend on time, the coordinates x*-, y*j, z* of
rrij in the principal axes frame should not depend on time either. In other
words, the principal axes frame is not rotating relative to the system S, but
is fixed in S and rotates together with S.
EXERCISE 2.2.11. Consider four identical point masses m at the vertices

of a square with side a. Denote by O the center of the square. Let (i, j , k)
be a cartesian system, at O, with the vectors i and j along the diagonals
of the square (draw a picture!). Verify the following statements: (a) The
center of mass of the system is at O. (b) The vectors i, j , k define the
three principal axis of the system, (c) The diagonal elements of the matric
IQM in {%, j , k) are ma2, ma2, 2ma2. (d) If the system (i, j , k) is rotated
by 7r/4 around k (does not matter clock- or counterclockwise), the result is
again a principal axis frame for the system, and the matrix IQM does not
change.
EXERCISE 2.2.12/1 When are some of the diagonal entries of IQM equal to
zero? Hint: not very often.
We now summarize our excursion into linear algebra: for every rigid
system S of point masses, there exists a special frame, called the p r i n c i p a l
axes frame, in which the matrix of inertia IQM of the system is diagonal
and does not depend on time. This special frame is centered at the center
of mass and is fixed (not rotating) relative to the system S.
Let us go back to the analysis of the motion of the system 5. With
CM = LCMXI + LCMxz +LCMxi, U = OJXZ +u>yj +UJZK ,
and after multiplying by U* on the left and by Uj on the right, equation

(2.2.31) becomes
=
£*CM ^CM^*I (2.2.33)
where CCM = U^CCMUJ is the column vector {L*CMx, L*CMx, L*CMz)T

and ft* = C/*fJ[/J is the column-vector (w*, u>*, w*) T .
By construction, the principal axes frame rotates relative to the frame
O, and the corresponding rotation vector is w. Denoting the time derivative
in the frame O by Do, and in the the principal axes frame, by £>*, and using
the relation (2.1.43) on page 64, we find
D0LCM(t) = D. LCM{t) + u(t) x LCM(t). (2.2.34)
To proceed, let us assume that the underlying frame O is inertial. Then

relation (2.2.17) applies, and we find D0LcM(t) = TcM(t): the change of
LCM (t) in the inertial frame is equal to the torque of all forces about the
center of mass. In the principal axes frame, we have TcM{t) = TQMx{t) i* +
TcMy{t) j * +T ( X Mz (i) k*. Also, since the basis vectors i*, f, k* are fixed in
Rigid Bodies 79
the principal axes frame, we have D*LcM{t) = L*CMx{t) i* 4- L*CMy(t) j * +

L*CMz(t)k*. On the other hand, since the matrix i £ M does not depend
on time, we use (2.2.33) to conclude that the column vector t*CM of the
components of D*LcM{t) satisfies t*CM = IQM^ • Finally, we compute
the cross product in (2.2.34) by writing the vectors in the principal axes
frame and using the relation (2.2.33) for the components C*CM{t) of LCM
in that frame. The result is the three Euler equations describing the
rotation of the rigid system about the center of mass:
< i;yul+u*xul{rxx-rzz) = T*CMy, (2.2.35)
These equations were first published in 1765 by a Swiss mathematician

LEONHARD EULER (1707-1783). Leonhard (or Leonard) Euler was the
most prolific mathematician ever: extensive publication of his works con-
tinued for 50 years after his death and filled 80+ volumes; he also had
13 children. He introduced many modern mathematical notations, such
as e for the base of natural logs (1727), f(x) for a function (1734), E for
summation (1755), and i for the square root of —1 (1777).
EXERCISE 2.2.13? (a) Verify that (2.2.34) is indeed equivalent to (2.2.35).
(b) Write (2.2.27) in an inertial frame and verify that the result is a par-
ticular case of (2.2.35).
With a suitable definition of the numbers I*x, I*y, I*z and the vector
T*CM, equations (2.2.35) also describe the motion of a rigid body. We study
rigid bodies in the following section.
2.2.3 Rigid Bodies

Any collection of points, finite or infinite, can be a rigid system: if two
points in the collection have trajectories rl(t), r2(t) in some frame, then
the rigidity condition ||T*i(i) — T*2(*)|| = ll r i(0) — >*2(0)|| must hold for every
two points in the collection.
Intuitively, a r i g i d body is a rigid system consisting of uncountably
many points, each with infinitesimally small mass. Mathematically, a rigid
body is described in a frame O by a mass d e n s i t y function p = p{r), so
that the volume A y of the body near the point with the position vector ro
has, approximately, the mass A m = p(r) AV. Even more precisely, if the
body occupies the region H in R 3 , then the mass M of the body is given
by the triple (or volume) integral
Af = III p{r)dV.
Without going into the details, let us note that rigid bodies can also be
two-dimensional, for example, a (hard) spherical shell, or one-dimensional,
for example, a piece of hard-to-bend wire. In these cases, we use surface
and line integrals rather then volume integrals. In what follows, we focus
on solid three-dimensional objects.
All the formulas for the motion of a rigid body can be derived from the
corresponding formulas for a finite number of points by replacing rrij with
the mass density function, and summation with integration. For example,
the center of mass of a rigid body is the point with the position vector
VCM = ^ JJJ rp(r)dV. (2.2.36)

n
EXERCISE 2.2.14. C Show that both the mass and the location of the center
of mass of a rigid body are independent of the frame O.
For a rigid body 11 moving in space relative to a frame O, we denote
by lZ(t) the part of the space occupied by the body at time t relative to
that frame O. If the frame O is inertial, then an equation similar to (2.2.5)
connects the trajectory rcM = fCM(t) of the center of mass in the frame
with the external forces per unit mass F1-^ = FÊ\r) acting on the points
of the body:
MfCM(t) = JJJF^'(r(t)) P(r(t))dV, (2.2.37)
The angular momentum of TZ about O is, by definition,
Lo(t) = III r(t) x r{t) P(r(t))dV. (2.2.38)

n{t)
Similar to (2.2.15), page 71, we have
x
L0(t) = MrCM(t) x rCM(t) + [[[*(!) *(*) P(r(t))dV, (2.2.39)
Rigid Bodies 81
where v(t) = r(t) - rcM(t), and is (2.2.14) replaced by
LCM(t) = llf*(t) x «(*) P(r(t)) dV, (2.2.40)

TC(t)
which is the angular momentum relative to the center of mass.

As with finite systems of points, consider the parallel translation of
the frame O to the center of mass of the body. The rotation of the rigid
body relative to this translated frame is described by the rotation vector
u>(t) = u>x{t) i + Wy(t) 3 + ojz(t) k, where (S, j , k) is the cartesian basis in
the translated frame. By analogy with (2.2.30), we write v(t) = x(t)i +
vit)3+ *(*)&, LCM(t) = LCMx(t)i + LCMy{t)3 + LcMz(t)k, and define
Ixx = iy2{t) +Z2{t))P{x{t)) dV Iyy = {x2{t) +Z2{t))
-R(t)
III n(t)
' III
= IJJ(x2(t) + y2(t))p(x(t))dV,
K{t)
•*-xy — -*-yx jJjx(t)y{t)p{x{t))dV,

*xz — Izx —
it(t) n(t)
JJJx(t)z(t)p(x(t))dV,
Tl(t)
lyz — J-zy
= JJJy{t)z{t)p{x{t))dV.
W)
(2.2.41)
Then we have relation (2.2.31), page 76, for rigid bodies:
CcM(t)=ICM(t)n(t), (2.2.42)
where CcM{t) is the column vector (LcMx(t), LcMy{t), LcMz(t))T, ft is
the column vector (wx{t), wy(t), uz(t))T, and
( J-xx
-Iyx
•'zx
~~±xy *xz \
Iyy —IyZ I •
~*zy *zz /
(2.2.43)
The matrix IQM is called the moment of i n e r t i a m a t r i x , or t e n s o r of

i n e r t i a , of the rigid body TZ around the center of mass in the basis
(i, j , k). As in the case of a finite rigid system of points, there exists a
principal axes frame, in which the matrix ICM is diagonal, and the diag-
onal elements I*x,I*y,I*z are uniquely determined by mass density func-
tion p. The principal axes frame, with center at the center of mass and
basis vectors i*, j * , k*, is attached to the body and rotates with it. If
u>(t) = u* i* + LJ*J* + w* k*, then the Euler equations (2.2.35) describe
the rotation of the body about the center of mass. Equations (2.2.37) and
(2.2.35) provide a complete description of the motion of a rigid body.
As an example, consider the DISTRIBUTED RIGID PENDULUM. Recall
that in the simple rigid pendulum (page 41), a point mass is attached to the
end of a weightless rod. In the distributed rigid pendulum, a uniform rod
of mass M and length I is suspended by one end with a pin joint (Figure
2.2.1).
-*-3
Cross-section of the rod
Fig. 2.2.1 Distributed Pendulum
We assume that the cross-section of the rod is a square with side a. Then
the volume of the rod is la2 and the density is constant: p(r) = M/(£a2).
EXERCISE 2.2.15. c Verify that the center of mass CM of the rod is the
mid-points of the axis of the rod.
Consider the cartesian coordinates (z*, j * , k*) with the origin at the
center of mass (Figure 2.2.1). As usual, k* = i* x j * .
Let us compute I*z, as this is the only entry of the matrix 7£ M we will
need:
+ y2)p(r)dV
/ a/2 a/2 1/2 e/2 a/2 a/2 \

M 2 2
dy I dz I x dx + I dx I dz I y dy
la?
\-a/2 -a/2 -1/2 -1/2 -a/2 -a/2
2 3 2 2
= (pa £ /12) + (plo^jYL) = (M/12){1 + a ).
Rigid Bodies 83
EXERCISE 2.2.16. C Verify that the vectors i*, j * , k* define a principal axis
frame. Hint: I*y = I*z = I*z = 0, as seen from (2.2.41) and the symmetry of
the rod.
The motion of the rod is a 2-D rotation around the pin at O, and the
vector k* is not rotating. There are no internal forces in the rod to affect
the motion. As a result, the angular velocity vector is UJ = ui*z k* —Ok*,
and the Euler equations (2.2.35) simplify to I*z —^ = T^MI, or
M{P+a?).. (B)
^ V~1CMz' (2.2.44)
where TQM'Z is the rc*-component of the external torque around the center
of mass.
To simplify the analysis, we ignore air resistance. Then the external
torque T*jM' around the center of mass is produced by two forces: the force
of gravity W and the force Fpin exerted by the pin at O. Since the rod
is uniform, the torque due to gravity is zero around the CM. To compute
Fpin we now assume that the frame fixed at O is inertial and Newton's
Second Law (2.2.5) applies. The external force on mass M at the CM is the
sum of the total gravity force, W = Mgi, and the reaction of the pin FPin.
Hence, Fpin = M rcM — W. Since CM moves around O in a circle of radius
£/2 we use (1.3.27) on page 36 to obtain the components ar, ag of TCM
in polar coordinates with origin at O: ar = {—1/2)UJ2 r, a$ = (^/2)w 6.
The point of application of Fpin relative to CM has the position vector
-rCM = -{i/2)r. Then
T
CM = ~rCM x Fpin + 0xW = --rx Fpin.
Hence,
? X {MV,CM
T*C(M = 4 -W) = ^(^Lbrx0-gvxi^
-IM (I . . „ , .»
2 V2-uj + gsmv IK .
Substituting in (2.2.44, we get (M(£2 + a2)/12)6 = -(£M/2)({£/2)u> +

g sin 6) or {(£2 + a2)/l2)0 + (£2w/4) + ( ^ s i n 0 / 2 ) = 0, or, with w = 0,
({At2 + o?)/l2)e = -{gl/2) sinS. (2.2.45)

If a « 0 and 6 is so small that sin# « #, then equation (2.2.45) becomes
{21/3)0+ g0 = 0. (2.2.46)
Comparing this with a similar approximation for the simple rigid pendu-
lum (2.1.10), page 42, we conclude that a thin uniform stick of length £
suspended at one end oscillates at about the same frequency as a simple
rigid pendulum of length (2/3)1
The objective of the above example was to illustrate how the Euler
equations work. Because of the simple nature of the problem, the system
of three equations (2.2.35) degenerates to one equation (2.2.45). In fact, an
alternative derivation of (2.2.45) is possible by avoiding (2.2.35) altogether;
the details are in Problem 2.4, page 417.
Note that if the frame O is fixed on the Earth, then this frame is not
inertial, and the Coriolis force will act on the pendulum, but a good pin
joint can minimize the effects of this force.
For two more examples of rigid body motion, see Problems 2.7 and 2.8
starting on page 419.
2.3 The Lagrange-Hamilton Method
So far, we used the Newton-Euler method to analyze motion using forces

and the three laws of Newton (it was L. Euler who, around 1737, gave a
precise mathematical description of the method). An alternative method
using energy and work was introduced in 1788 by the French mathe-
matician JOSEPH-LOUIS LAGRANGE (1736-1813) and further developed in
1833 by the Irish mathematician Sir WILLIAM ROWAN HAMILTON (1805-
1865). This Lagrange-Hamilton method is sometimes more efficient than the
Newton-Euler method, especially to study systems with constraints. An ex-
ample of a constraint is the rigidity condition, ensuring that the distance
between any two points is constant.
In what follows, we provide a brief description of the Lagrange-Hamilton
method. The reader is assumed to be familiar with the basic tools of multi-
variable calculus, in particular, the chain rule and line integration.
Lagrange's Equations 85
2.3.1 Lagrange's Equations

We first illustrate Lagrange's method for a point mass moving in the plane.
The modern methods of multi-variable calculus reduce the derivation of the
main result (equation (2.3.17)) to a succession of simple applications of the
chain rule.
Consider an inertial frame in M2 with origin O and basis vectors i, j .
Let r = xi + yjbe the position of point mass m moving under the action
of force F = F2 i + F2 j . By Newton's Second Law (2.1.1),
Fi = m i , F2 = my. (2.3.1)
The state of m at any time t is given by the four-dimensional vector

(x,y,x,y), that is, by the position and velocity. Knowledge of the state
at a given time allows us to determine the state at all future times by
solving equations (2.3.1).
Now consider a different pair ((71,(72) of coordinates in the plane, for
example, for example, polar coordinates, so that
x = x(q1,q2), y = y{qi,q2); q\ = qi(x, y), q2 = q2{x, y), (2.3.2)
and all the functions are sufficiently smooth. Differentiating (2.3.2),
dx . dx . . dy . dy . ,„ „ „,
x = —qi + —q2, y=~-qi + ^-q2. (2.3.3)
dqi dq2 dqi dq2
We call qi,q2,q\,q2 the g e n e r a l i z e d coordinates, since their values de-
termine the state (x,y,x,y) by (2.3.2) and (2.3.3). Note that the partial
derivatives dx/dqi, dy/dqi, i = 1,2, depend only on q\ and q2. We then
differentiate (2.3.3) to find
dx _ dx dx _ dx dy dy dy dy
dq\ dq\' dq2 dq2 ' dq\ dqi' dq2 dq2 '
Next, we apply the chain rule to the function dx(qi(t),q2(t))/dqi to find
_d_ / dx\ _ d2x dq\ d2x dq2 _ d2x . d2x .

+
Jt\dTi)^Wi^ d~q~2~bTlltt^Wiqi +
^dTiq2' }
Prom (2.3.3), differentiating with respect to the variable q\,
dx d2x . d2x . d (dx\ ,n „ „,

— = -jr-jgi + T-j^q2 =- £ ; , - • 2.3.6)
dqi dq{ dqidq2 dt \dqi J
86 The Lagrange-Hamilton Method
Similarly,
d / dx \ dx d ( dy\ _ dy
dt\dqi) dqi' dt \dqiJ dqi'
(2.3.7)
d ( dx \ _ dx d f dy\ _ dy
dt \dq2J dq2 dt \dq2J dq2'
EXERCISE 2.3.1. C Verify (2.3.7).
By definition, the kinetic energy £K of m is
£K = j(i2 + y2). (2-3.8)
Replacing x and y by their functions of qi,q2 as given by (2.3.3), we obtain

the function
£if = £/c(9i,92,91,92)- (2.3.9)

From (2.3.8), again by differentiation and using (2.3.4),
d£K 9£K dx Q£K dy . dx . dy
dq\ dx dq\ dy dqi dq\ dqi
Differentiating with respect to t and applying (2.3.7), we get
d (d£ \ ..dx ..dy . dx . dy /0q1f^
TJT \-^r-K )= mx-— + my— + m i - \-my-. (2.3.10)
dt \ dqi J dqi dq\ dqi dqi
From (2.3.8) we also get
9£K . dx .dy ,OQ1l,
—— = mx-^— +my—. (2.3.11)
dqi dqi dqi
Subtracting (2.3.11) from (2.3.10), we find that
d fd£K\ d£K ..dx ...dy ,„,,„>
dt \ dqi J — mx-dqi\-my-—.
dqi dqi (2.3.12)
EXERCISE 2.3.2.c Verify that

d (d£K\ d£K ..dx ..dy ,„,.„,
-77 [-XT- - - 5 — =mx— +my—. (2.3.13)
dt \ dq2 ) dq2 dq2 dq2
To continue our derivation, we use equations (2.3.1) to rewrite (2.3.12)
and (2.3.13) as
±(^)-%«=F1^ + F * j = 1>2. (2.3.14)

dt \ dqj J dqj dqj dqj
Recall that the work W done by the force F is the line integral JcF • dr.
Hence, the differential is dW = F • dr. Since dr — dxi + dyj, we have
dW = Fidx + F2dy. (2.3.15)
By the chain rule,
dx = (dx/dqi)dq1 + {dx/dq2)dq2, dy = (dy/dq^dqi + (dy/dq2)dq2.
Substituting in (2.3.15), we get dW = Q\dq\ + Q2dq2, where
dx
n - p _u F Qy - T? dr
Ql = r\- V t2 — = t • ——,
91 Ql qi
(2 3 16)
Q2 = Flp. + F2^=F.^L.
oq2 aq2 aq2
The functions Q\ and Q2 are called g e n e r a l i z e d forces corresponding to
the generalized coordinates q\,q2. Substituting in (2.3.14), we obtain the
Lagrange equations of motion,
*(w )-%">'• -1'2- <2 317)

'
where Qj is given by (2.3.16).
For the point mass moving in cartesian coordinates under the force
F = F\ i + F2j, we have q\ = x, q2 = y, 8K = m(x2 +y2)/2, and equations
(2.3.17) coincide with (2.3.1), so it might seem that we did not accomplish
much. It is clear, though, that (2.3.17) is more general, and covers mo-
tions not only in cartesian, but also polar and any other coordinates that
might exist in R 2 . In other words, equation (2.3.17) is at a higher level of
abstraction than (2.3.1), and therefore has a higher mathematical value.
EXERCISE 2.3.3. (a)c Verify that for qi = x, q2 = y, equations (2.3.17)
indeed coincide with (2.3.1). (b)A Write (2.3.17) in polar coordinates, with
9i =t, q2= 0.
Next, we consider a system of n point masses in IR3, and assume that
there are k degrees of freedom, k < 3n. In other words, only k out of 3n
coordinates can change independently of one another; the remaining 2>n - k
coordinates are uniquely determined by those k. The state vector is now
{xi,yi,zi, x2,y2,z2,... ,xn,yn,zn,±i,yi,zi,... ,xn,yn,zn)

and the generalized coordinates are
(9i,92,93,.-.,9fc,9i,---,9fc)>
two for each degree of freedom. Equations of the type (2.3.1)-(2.3.8) can
be written for Xi, qi, Xi, qi. The kinetic energy is then given by
n
£K = J2(mi/2)(x2i+yf + zf) or
£K = £ * ( ? i , . • • ,9fc,9i, • • •,«*)• (2-3.18)
i=l
With forces Fi = (•Fi,»,i:2,i,-p3,i), i = l,---,n, equations (2.3.1) become
Fi,j = mjZj, F2,j = rriij/i, F 3 ) i = m , ^ , i = 1 , . . . , n,
equation (2.3.12) becomes
d (d£K\ d£K <A .. dxi ^ ~ <9?/i , v ^ •• ^ /o o m\

T7 "^v-
d* V dqj ) - - 5d—
qj
= V r a j Z i — + > m m - 5 — + > m^t-S—>
j ^ dqj f^ dqj £ > dqj (2.3.19)
equation (2.3.14) becomes
Tf (iT-) - IT = £ (F"P + F^7T + F>.*jr) • <2-3-20)

dt V d?j y dq, f^ \ dqj dqj dqj J
j = 1 , . . . , fc, and equation (2.3.15),
n k
F dr
dW = J2 i- i = J2 Qid(ir (2.3.21)
i=l j=l
The generalized forces are
Qi = £ (F^P +^ +^ ) = £ * • P> (2-3-22)

j — l,...,k. Thus, there are exactly k Lagrange's equations, one for each
degree of freedom:
1 k <2 323)
*(%)-%-*>• ' - - '
Let us now look at the particular case of CONSERVATIVE FORCE FIELDS.
By definition, we say that the force F is c o n s e r v a t i v e or defines a
c o n s e r v a t i v e force f i e l d if the vector F is a gradient of some scalar
function: F = — VV; the negative sign is used by convention. The func-

tion V is called the p o t e n t i a l of the force field F. The work done by a
conservative force along a path r(t), a < t < b, is
rb
pO rb
no
w= / F-rdt
Ja
= /
Ja
-W-rdt
= _ [b (^te + ^dy + dV_dz\dt=_ fbbdV(r(t))

w,„uw M
(2.3.24)
Ja \ dx dt dy dt dz dt J Ja dt
= -{V(r(b))-V(r(a))),
That is, dW = —dV, and the work done by a conservative force equals the
change in potential.
Let us assume that all the forces Fi,...,Fn are conservative, and
denote by V\,...,Vn the corresponding potentials. By construction, we
have Fi = Fi(xuyi,Zi), i = l,...,n and so V* = Vi(xi,yi,Zi). De-
fine V = Vi + • • • + Vn. According to (2.3.21), dW = X)" =1 Ft • dr{ =
— Y17=i dVi = —dV. We now write V as a function of the generalized coordi-
nates: V = V(qi,...,qk), and get dW = -(dV/dqi)dqi (dV/dqk)dqk.
Comparison of the last equality with (2.3.21) results in the conclusion
Qj = —dV/dqj, j — l,...,fc. Thus, for conservative forces, equations
(2.3.23) become
±(0£K\_ dtK=_dV^ j =1 (2325)

dt \ dq\j J dqj dqj'
We will transform (2.3.25) even further by noticing that V does not depend
on the velocities, and therefore is independent of qy. dV/dqj = 0. We then
define the Lagrangian
L = SK-V, (2.3.26)
and re-write equation (2.3.25) as
d fdL\ dL
dtvdqj-wr0- (2 3 27)
--
Equation (2.3.27) has at least three advantages over (2.3.23): (a) a more
compact form; (b) a higher level of abstraction; (c) a possibility to include
constraints.
To conclude this section on Lagrange's equations, let us look briefly
at some NON-CONSERVATIVE FORCES. AS before, q = (q\,...,qk) is the
vector of generalized coordinates and q = (gi, • • • ,<7fc), the vector of the
corresponding time derivatives, so that the vector (q, q) uniquely deter-

mines the state of the system. We derived (2.3.27) under the assumption
Q — (Qii • • • i Qk) = — W , where V = V(q). Now assume that the gener-
alized force Q is no longer conservative, and has a non-conservative compo-
nent that depends only on q; many non-conservative forces, such as friction
and some electromagnetic forces, indeed depend only on the velocity. In
other words, we assume that
Q = Q ( 1 ) + Q ( 2 ) , Q ( 1 ) = - W i , Q ( 2 ) = jVV2 (2.3.28)
for some scalar functions V\ = V\{q) and V% = V2(q). Then
Define V = Vi + V2 so that dVi/dqj = dV/dqj, dV2/dqj = OV/dq,, and
If L = SK — V, then Lagrange's equations (2.3.27) follow after substi-

tuting (2.3.30) into (2.3.17). Alternatively, if we define the conservative
Lagrangian L\ = £K — Vi, then (2.3.17) and (2.3.30) imply
d dL\ dLi ^(2i

* ^ T - ^ = «? ) ' (2 3 31)
--
a modification of (2.3.17).
2.3.2 An Example of Lagrange's Method

Our goal in this section is to verify the Lagrange equations (2.3.17) for
a points mass that moves in space, but has only one degree of freedom.
The example below is adapted from the book Introduction to Analytical
Mechanics by N. M. J. Woodhouse, 1987.
Consider a smooth wire in the shape of an elliptical helix. Choose an
inertial frame with origin O and cartesian basis vectors (?, J, k), so that k
is along the axis of the helix. In this frame, the helix is the set of points
with coordinates (x, y, z) so that
An Example of Lagrange's Method 91
see also Exercise 2.3.4 below. A bead of mass m placed on the wire will slide
down under the force of gravity. Note that the two equations in (2.3.32),
describing the helix, define two constraints and leave the bead with only
one degree of freedom. We assume that the friction force is negligible, so
that the restraining force N = (N\,N2,Ns), exerted by the wire on the
bead, is orthogonal to the velocity vector at every point of the motion. Our
objective is to derive the equations of the motion of the bead along the
wire.
First, let us look at what the Newton-Euler method will produce. By
Newton's Second Law, we get three equations
mx = Nx, my — Ny, m'z = Nz — mg, (2.3.33)
with the six unknowns x,y,z,Ni,N2,Ns. Equations (2.3.32) are the two
constraint equations on x, y, z. The sixth equation comes from the orthog-
onality condition for frictionless motion:
r-N = xN1+yN2 + zN3 = 0. (2.3.34)
To solve this system of six equations, it is natural to parameterize the helix

by setting
x = acosq, y = bsinq, z = cq, (2.3.35)
where q is the generalized coordinate for this problem.

EXERCISE 2.3.4-5 Verify that equations (2.3.32) and (2.3.35) define the same
set of points. How should one interpret the inverse cosine in (2.3.32)?
The position vector of the bead is r(q) = x(q) i + y(q) j + z(q) k, and
the motion of the bead is determined by the function q = q(t). Note
that the tangent vector to the helix is dr/dq. Using the six equation, the
unknown reaction force N can be eliminated and a single second-order
ordinary differential equation for q = q(t) can be obtained; the details of
this approach are the subject of Problem 2.9, page 421.
We now look more closely at the alternative approach using the La-
grange method. Let p = m r be the momentum of the sliding bead. Then
equation (2.3.33) in vector form becomes
p = N -mgk. (2.3.36)
Denote by ptan = P • (dr/dq) and Q = —mgk • (dr/dq) the tangential

components of the momentum p and the gravitational force —mg k. This
definition of Q agrees with (2.3.22) on page 88. Differentiating ptan with

respect to t, we get
dr d2r
Ptan = P • — ++PP • "J-J Q-
dq~ 'df
Using (2.3.36) and p = mr = m (dr/dq) q, we find
,.r ... dr dr d?r . 2
Ptan = {N -mgn)- — +m— • J - J <T,
or, since N • (dr/dq) = 0,

(IT (IT
Ptan = Q + m-j--j^q2, (2.3.37)
The kinetic energy is then
^ = -2mvr^----e, (2.3.38)
and by (2.3.35),
£K = ^ ( a 2 sin2 q + b2 cos2 q + (?) q2. (2.3.39)
Hence, from (2.3.38),

8£K dr dr . dr dr
dq dq dq dq dt
From (2.3.39), using standard rules of partial differentiation,
—— = m(a2 sin q cos q — b2 cos q sin q) q2.
Now, r(q) = a cos #2 + b sin qj + cqk, so that

dr d2r ^
— = —asmqt + bcosqj + CK, —5- = —acosqt — bsmqj.
dq dqz
Therefore,
8£K dr d2r . 9
dq dq dq2
and equation (2.3.37) is equivalent to Lagrange's equation of motion
d ( 8£K\ OSK „
1
* *
dt \ dq J dq =Q, (2.3.40)
Hamilton's Equations 93
where Q is the generalized force from (2.3.16), that is,
Q = F • dr/dq = -mg k • dr/dq. (2.3.41)
We therefore accomplished our objective and verified the Lagrange method

for our example by explicitly deriving equation (2.3.40). Once (2.3.39) and
(2.3.41) are taken into account, equation (2.3.40) becomes a second-order
ODE for the function q = q(t). A reader interested in solving this equation
should try Problem 2.9, page 421.
Notice that the unknown, constraint-induced, reaction force N appears
in Newton's equation (2.3.33), but not in the Lagrange equation (2.3.40).
Of course, the Newton-Euler method, requiring elimination of the unknown
force N, results in the same equation of motion; see Problem 2.9 for details.
2.3.3 Hamilton's Equations

Lagrange's equations (2.3.27) are a system of k second-order ODEs. A
change of variables reduces this system to Ik first-order ODEs. While there
are many changes of variables to achieve this reduction, one special change
transforms (2.3.27) into a particularly elegant form, known as Hamilton's
equations.
Given the Lagrange function (2.3.26), introduce new variables p =
iPu • • • ,Pk) b y
(2 3 42)
*=!£• --
and assume that equations (2.3.42) are solvable in the form
q = f(p,q) (2.3.43)
for some vector function / = ( / i , . •-,/&)• For example, if L =

\J2j=i(mj(lj ~ hjQj)i t n e n (2.3.42) becomes pj = rrijqj or qj = Pj/rrij.
In applications to mechanics, <jj usually has the dimension of distance and
rrij is the mass, so that pj has the dimension of momentum. This is why
the variable pj defined by (2.3.42) is called the g e n e r a l i z e d momentum.
Together with (2.3.42), equations (2.3.27) become a system of k first-
order ODEs
To obtain another k equation, define the Hamiltonian

k
H(p, q) = Y<PM ~ L(l> <?)> ( 2 - 3 - 45 )
where, according to (2.3.43), each qj is a function of p and q. Then direct

calculations show that
dqL_d_H_ dpl_ d_H_
? i fc
dt - dp^ dt - dq,' • - "--' - ^6Ab>
Equations (2.3.46) are a system of 2k first-order ODEs, known as

Hamilton's equations.
EXERCISE 2.3.5.B Derive equations (2.3-46) Hint: the first follows directly by
computing dH/dpj and using (2.3.42), (2.3-43), and qj = dqj/dt; for the second,
note that dL/dqj = —dH/dqj and use (2.3-44)-
EXERCISE 2.3.6F Show that if (2.3.31) is used instead of (2.3.27), then the
second equation in (2.3-46) becomes
We conclude the section by showing that, in a typical mechanical prob-

lem, when the vector q is the position and V is the potential energy, the
Hamiltonian is the total energy of the system.
Theorem 2.3.1 Consider a system of n point masses in M3. Assume

that the position vectors rj and the potential V of the system do not depend
on q and depend only on q. Then the Hamiltonian of the system is equal
to the total energy:
H(p,q)=£K + V. (2.3.48)
Proof. By assumption,
ri = ri(q), i = l,...,n. (2.3.49)
Then
ElH-
d<
(2-3-50)
* b
Hamilton's Equations 95
Substituting (2.3.50) into
1 ™
results in
3fe
a
£K = Yl MiQi = QTAq, (2.3.51)
where A = (a,j:e, j,t= 1 , . . . ,3k) is a symmetric matrix and
1 v^ dri dri
2
~i dqj dqe
Clearly, if F(x) — xTBx = Y2i,j bijXjXj is a quadratic form with the

coefficients bij independent of x, then V F = 2Bx and V F • x = 2.F.
Therefore, by (2.3.51),
V f ^ = 2£K. (2.3.52)
Since V does not depend on q, it follows that d£x/dqj = dL/dqj = pj,

and (2.3.52) becomes
3fe
an
Since ff = J2i=i PiQi ~ L d L = £K -V, equality (2.3.48) follows. D
As with most equation related to the Lagrange-Hamilton method, the

main significance of (2.3.46) is its mathematical elegance and high level of
abstraction; it is the level of abstraction that makes (2.3.46) an extremely
interesting object to study in various branches of mathematics. With all
that, equations (2.3.46) simplify the analysis of many concrete problems in
mechanics. Unfortunately, these topics fall outside the scope of this book.
2.4 Elements of the Theory of Relativity
Newtonian mechanics, as discussed in the previous sections, describes the

motion of objects at speeds much smaller than the speed of light; at speeds
96 Theory of Relativity
close to the speed of light the description is provided by the theory of

relativity.
Special relativity mechanics arose from Newtonian mechanics and the
hypothesis that the mathematical equations describing laws of mechanics
should have the same form in every two inertial frames that move with the
constant velocity relative to each other. This hypothesis of mathematical
invariance does not just express a mathematical aesthetic. From a physical
standpoint, the hypothesis requires that two observers, one in a frame O
and the other in a frame 0\ moving with constant velocity v relative to
O, should be able to describe the motion of a particle using the same
law. One such law could be Newton's Second Law (2.1.1) on page 40.
Although Newton understood the position is relative to the observer in
different frames, he assumed time to be an absolute physical quantity that
is the same in different frames. Mathematically, these assumptions are
expressed by the Galilean transformation r i = r + vt, t\ = t. Note
that the Galilean transformation does leave (2.1.1) unchanged: f\ = r,
and so if m r = F, then m'f\ = F.
The assumption of absolute time contradicts the experimental result,
not known at Newton's time, that light has the same speed in O and 0\.
We will see that the constant speed of light implies that time is as rela-
tive as position, and the Galilean transformation must be replaced by that
of Lorentz. Not surprisingly, the Lorentz transformation no longer leaves
(2.1.1) unchanged, and we will see how the law of motion must be modified
to satisfy the invariance hypothesis; see (2.4.15) on page 102 below. This
new law of motion is a more general mathematical model of mechanics than
(2.4.1) and leads to a number of remarkable conclusions, such as the famous
relation £ — mc2 between the mass and energy.
The theory of relativity is universally considered as one of the pinnacles
of human intellectual achievements. Many also believe that understanding
of this theory is beyond the ability of all but very few bright minds. Our
main objective in the following few sections is to demystify the subject and
to demonstrate that one only needs a basic knowledge of linear algebra
and multi-variable calculus to understand the main ideas of the theory, to
follow the computations that lead to the most famous implications of this
theory, and to see how the mathematical models of relativity interact with
the underlying physical phenomena.
We start with a short historical background, mostly related to the de-
velopment of special relativity. Next, we present an elementary derivation
of the Lorentz transformation, and finally discuss the main results of gen-
Historical Background 97
eral relativity. An interested reader can follow up on some of the topics in

Problem 2.2, page 414, and in Section 3.
2.4.1 Historical Background

As one legend puts it, back when he was a little boy, A. Einstein was curious
about two questions: (1) What happens if you run after a ray of light and
try to catch it? (2) What happens if you are in a moving elevator and
suddenly somebody turns off the gravity? The answers to these questions
later became the theory of special and general relativity.
We mentioned earlier that both the "relative" and "absolute" points
of view had proponents from the early years of modern science. In 1707,
the Irish bishop George Berkeley disputed the concept of absolute space
as described in Newton's 1687 book Principia. Newton contended that an
absolute rotation can be determined experimentally by measuring the effect
of a centrifugal force as in his famous experiment with a rotating bucket of
water in which the surface of the rotating water assumed a concave shape.
Hence, Newton argued, there is no need for a second body relative to which
the rotation is measured. Berkeley disputed this on philosophical grounds,
saying that "motion cannot be understood without a determination of its
direction, which in turn cannot be understood except in relation to our or
some other body." Absolute space, he claimed, is unobservable. If a body
is the only one in the universe, it is meaningless to speak of its rotation,
and we need the existence of other bodies, relative to which the rotation
can be measured. We now know mathematical formulas supporting this
argument: for example, the centrifugal force - m w x (a; x r\) in (2.1.47) on
page 66 involves the rotation vector u> that must be defined in an inertial
frame. For more about the dispute between Newton and Berkeley, see the
book The Physical Foundations of General Relativity by D. Sciama, 1969.
The propagation of light was more difficult to understand. Up until the
late 19th century, it was assumed that light propagated through the aether
(or ether), a hypothetical substance believed to be filling all absolute space
and serving as the medium for the propagation of electromagnetic waves.
In 1887 in Ohio, two American physicists, ALBERT ABRAHAM MICHELSON
(1852-1931) and EDWARD WILLIAMS MORLEY (1838-1923) conducted an
experiment aimed at measuring the velocity of the Earth relative to the
aether.
The relative propagation speed of other wave disturbances in a medium,
like sound and water waves, depends on the velocity of the source. If the
aether existed, then the Earth's motion around the Sun, at the speed of
about 30km/sec, would affect the speed of light depending on the direction
of the light beam. Michelson and Morley emitted a beam of light, first in
the direction of the Earth's motion and then perpendicular to it. They used
mirrors and an interferometer to measure the speed of light in each case,
and failed to detect any significant difference; see Problem 3.3 on page 424
for details. As a result, contrary to its initial intent, the Michelson-Morley
experiment provided the first evidence against the existence of the aether,
and in that respect became the most famous failed experiment in history.
The experiment certainly had a profound effect on the further development
of natural sciences, and in 1907 Michelson became the first American to
receive a Nobel prize in physics.
In 1905, Albert Einstein, who was 26 at the time, made the result of the
Michelson-Morley experiment the basis of a new theory of relative motion,
special relativity. He proposed two postulates:
(i) An observer measures the speed of light to be the same in any direction
regardless of the observer's motion relative to other bodies; this is a
consequence of the Michelson-Morley experiment,
(ii) Any two observers moving with constant relative velocity can agree on
all observed data and mathematical relations (laws) between the data;
this is the postulate of mathematical invariance.
According to Postulate (i), all observers measure the same value for the
speed of light in a vacuum: c = 3 x 108 meters/second. It follows that
absolute motion (in some special frame) is unobservable, hence physically
meaningless. In other words, Einstein sided with Berkeley against Newton
on the question of relative motion. The letter c in the notation for the
speed of light is for celerity (swiftness).
According to Postulate (ii), for two observers in frames O and 0\, math-
ematical relations which hold in the frame O are to be transformed into
corresponding relations in the other frame 0\ by a transformation of the
variables (x,y,z,t) to the variables (xi,yi,zi,t\); the experimental study
of light propagation suggests that this transformation should be linear.
The result, called the Lorentz transformation, is a necessary consequence
of Einstein's two postulates.
The Lorentz Transformation 99
2.4.2 The Lorentz Transformation and Special Relativity

Consider two frames, the fixed frame O and the moving frame 0\. We
choose cartesian coordinate systems in both frames and assume that, at
each time moment, the frame 0\ is a parallel translation of the frame O
so that the vectors i, i\ are always on the same line; see Figure 2.4.1. As
usual, we represent the points in the frames O and 0\ with vectors of the
form xi + yj + zk and x\ %\ + y\ j 1 + z\ k\, respectively.
3\
l ^l
O / Ox
R, ki
Fig. 2.4.1 Fixed and Moving Frames
Thus, 0\ is moving to the right relative to O with constant velocity v.

A linear transformation of the space-time coordinates is
x = axi+0ti, y = 3/1, z = zi (2.4.1)
and
t = 7 H + ati (2.4.2)
for some real numbers a, (3, ~f,a. Equation (2.4.2) means that time is no
more absolute than space: each frame has its own time coordinate as well
as space coordinates. Equations (2.4.1) and (2.4.2) also demonstrate that
space and time are bound together in one space-time continuum. We will
see that Postulate (i) cannot hold without relation (2.4.2) with appropriate
7 and a.
Consider point 0\, the origin of the moving frame. It has the f i coordi-
nate x\ = 0 in the 0\ frame at all times t\. In the fixed O frame, by (2.4.1),
the i coordinate of 0\ is x = @t\; by (2.4.2), we also have t = at\. Note
that /3 > 0. If v = ||u|| is the speed of 0\ relative to O, then v = x/t = f3/a.
A light impulse emitted at the instant when O = 0\ will reach, at time t,
a point P, where the coordinates (x, y, z) of P in O satisfy
x2+y2 + z2 = c2t2. (2.4.3)
By Postulate (i), if (xi,yi,zi) are the coordinates of the same point P in

the frame 0\, then
A + y\ + *\ = cH\ (2.4.4)
By Postulate (ii), and because of relations (2.4.1), (2.4.2) above, equation

(2.4.3) becomes
(axi + (3hf + y\ + z\ = c 2 (7Zi + ah)2, or

2 2 2 2 2
{a - c V ) x + yj + z\ + 2(a0 - c V ) z i * i = c V - p /c )t .
Since the last equality must agree with (2.4.4) for all (xi, j/i, z\, t\) values,
it follows that
a2 - c 2 7 2 = 1, (2.4.5)
2
a/3 - c 7<7 = 0, (2.4.6)
2 2 2
a - p /c = 1. (2.4.7)
These three equation together with 0/a — v can be solved easily for
a,P, 7,<T. Prom (2.4.6), (5/a — c2^/a = v. So 7 = va/c2. Substitut-
ing in (2.4.5), we get a = (1 -v2/c2)~* and so 7 = (v/c2)(l — v2/c2)~2. In
(2.4.7), substitute /? = va and solve for a to get a = a. Therefore, (2.4.1)
and (2.4.2) become y — yi, z = z\ and
v
x= . t =—==£=. (2.4.8)
In what follows, we will often use the notations
a = [ l - - } , 7=^«, (2A9)
and write (2.4.8) as
x = ax\ + vati, t = ati+jxi. (2.4.10)
Then the inverse transformation becomes
xi — ax - vat, t\ = at - jx. (2.4.11)
One can verify this directly or use the physical argument that the inverse
transformation must be obtained from (2.4.10) by switching the roles of
{x,t), {x\,t\) and replacing v with —v.
Equations (2.4.9), (2.4.10), and (2.4.11) are the Lorentz

t r a n s f o r m a t i o n equations that govern special relativity kinematics. If v is
very small compared to c, then 1 — (v2/c2) is close to 1 and the Lorentz
equations become (approximately) x = x\ + vt, t = t\. This is the tradi-
tional Galilean transformation relating two frames moving with relative
velocity v in Newtonian kinematics. We always use the Galilean transfor-
mation, especially the t = t\ part, in our investigations of the classical
Newtonian mechanics. Let us stress again that, in Newtonian mechanics,
time is the same in both frames.
Equations (2.4.11) were first proposed in 1904 by the Dutch scientist
HENDRIK ANTOON LORENTZ (1853-1928) as a means of explaining the re-
sult of the Michelson-Morley experiment. The transformation became the
main mathematical means of verifying whether a proposed equation can
describe effects at speeds comparable with the speed of light: all one needs
to check is whether the equation is invariant under the Lorentz transfor-
mation, or, in other words, whether the equation treats the space and time
variables in a manner consistent with special relativity.
The Galilean transformation goes back to Galilei's 1638 book Discorsi
e Dimostrazioni Matematiche, intorno a due nuoue scienze, known in the
English translation as Discourse on Two New Sciences.
The Lorentz transformation implies a number of important, and some-
times counter-intuitive, results of special relativity. Let us summarize some
of these results.
(i) RELATIVITY OF SPACE AND TIME INTERVALS: The length I of a rod
placed along the z-axis and moving with speed v, as measured by an ob-
server in the fixed frame O, is smaller than the length £Q of the same rod
measured in the frame 0\ that moves with the rod; see Figure 2.4.1. Sim-
ilarly, the time interval At between two events in a frame moving with
speed v, as measured in the fixed frame, is longer than the time interval
Aio between the same events, measured in the moving frame:
* = W l ~ ("Vc 2 ), A t = ° =• (2.4.12)
y/1 - {V2/C2)
(ii) RELATIVITY OF MASS: the r e s t mass mo of an object is either its

inertial or gravitational mass measured in an inertial frame when the object
is not moving relative to that frame. The r e l a t i v i s t i c mass m and the
t o t a l energy £ of the same object moving with velocity v are
m0 £ = mc 2 . (2.4.13)
m
Vl-(« 2 /c 2 )
EXERCISE 2.4.1? Verify (24-12). Hint: Use the Lorentz transformation. For
£, use (2.4.11), since £0 = Axi = (AT - vAt)(l - {v/c)2))~1/2 with At = 0,
because the difference of the coordinates is recorded at the same time in the fixed
frame. Similarly, for At, use (2.4-8) with Ax\ = 0, as the events happen at the
same point of the moving frame.
The relativity of mass (2.4.13) is a consequence of the relativity of time

and is a necessary consequence of the required invariance of the laws of
mechanics under the Lorentz transformation. This is one of the foremost
examples of how a mathematical model, namely, the Lorentz transforma-
tion, leads to discoveries of new physical phenomena. Relations (2.4.13)
have been verified experimentally on various particles. In 2005, the journal
Nature reported an experimental verification of the relation £ = roc2 to
within 0.00004%.
We will now derive the main equation of relativistic kinematics. Recall

that Newton's Second Law for the point mass m can be written as F =
dp/dt, where p = mr is the momentum of m; see page 40. The extra
requirement of invariance of the law under the Lorentz transformation leads
to the definition of the r e l a t i v i s t i c momentum
p = mr = . , (2.4.14)
x/i-(IHI/c) 2 '
where m is the relativistic mass from (2.4.13). The relativistic form of
Newton's Second Law is therefore
F = ^ , or F = * ( * ) . (2.4.15)
dt> ^Wi-dHIA)2/
Equation (2.4.15) is the main equation of relativistic kinematics. It is invari-
ant under the Lorentz transformation, just as the non-relativistic Second
Law of Newton is invariant under the Galilean transformation; see Problem
3.2 on page 423. Note that, in the limits c ^ o o o r |jdr||/c —> 0, meaning
speed of motion much smaller than the speed of light, equation (2.4.14)
becomes the familiar Second Law of Newton (2.1.1): F = mo r.
EXERCISE 2.4.2. (a)c Verify that
||p|| 2 + mlc2 = m2c2. (2.4.16)
Hint: direct computation using (2.4-14).

(b)B Show that the quantity c2(m — mo) is natural to interpret as the rel-
ativistic kinetic energy, that is, £ = mc2 = £K + m^c2. Hint: if C is the
trajectory, v = ||r|| is the speed, and the initial speed is zero, then the work done
by a force F is
m V
£K= J F-dr= f rd{mr)= f' ydl ° , ., ) =mc2-m0c2,
Jc Jo Jo V C1 - (l/7c 2 )) ' /
where the last equality follows after integration by parts.
One consequence of the second equality in (2.4.12) is the Twin Paradox,
which is often described as follows. Consider two identical twin brothers,
Peter and Paul, who live in an inertial frame O. Peter stays in that frame
O, and Paul makes a round trip in a spaceship, travelling out and back in
a fixed direction with fixed speed v = (24/25)c. Then the total travel time
A*o = 2*i, as measured by Paul in the moving spaceship's frame, will be
A* = 25A*o/7 in the frame O as measured by Peter; the key relation here,
beside (2.4.12), is 25 2 = 242 + 7 2 . In particular, if A*0 = 14 years, then
A* = 50 years, that is, Peter will be waiting for 50 years for Paul to return
from what is 14-year long trip for Paul. If Paul's biological clock also obeys
(2.4.12), then, by the end of the trip, Paul will be 26 year younger than
Peter.
For further discussions of (2.4.12) and (2.4.13), including the Twin Para-
dox, see, for example, the books Concepts of Modern Physics by A. Beiser,
2002, and Spacetime Physics by E. F. Taylor and J. A. Wheeler, 1992.
We conclude this section with a brief discussion of the geometry of
relativistic space-time. The reader will see that this geometry is not exactly
Euclidean, thus getting a gentle introduction to general relativity.
R e l a t i v i s t i c space-time is a combination of time with the three spa-
cial dimensions of our physical space, and is modelled mathematically as a
four-dimensional vector space. Fixing a coordinate system in this space
represents every point by four coordinates (x,y,z,t). A curve r(s) =
(x(s),y(s), z(s),t(s)), so < s < si, in this space is called a world l i n e . If
this were a truly Euclidean space with cartesian coordinates, then, by anal-
ogy with formula (1.3.10) for the arc length on page 29, the length of this
world line would be £,,,,„„ = f£ ^\x'(s)\2 + |y'(s)| 2 + | z ' ( s ) | 2 + \t'(s)\2 ds.

Even if we disregard t h a t t has the dimension of time and not distance, this
is still not the correct way to measure the length: as the reader will soon
discover, this formula is not invariant under the Lorentz transformation.
Accordingly, let us try to find an invariant distance formula in the form
/ 2 \fa\\x'(s)\2 + a,2\y'{s)\2 + 03|z'(s)| 2 + 6|t'(s)| 2 ds for some real numbers
ai,ci2,a3,6; the requirement of invariance takes priority over the require-
ment of positivity. Following formula (1.3.12) for the arc length on page
29, it is convenient to write the distance formula in the differential form:
(ds)2 = ax(dx)2 + a2(dy)2 + a3(dz)2 + b (dt)2. (2.4.17)
E X E R C I S E 2.4.3? Verify that (2-4-17) is invariant under the Lorentz trans-

formation if and only if a\ = a2 = a^ = —b/c2, where c is the speed of
light. Hint: direct computations. Start with the frame moving along the x-axis
so that x — ax\ + avt\, y = yi, z — z\, t = at\ + 7x1, with 0 , 7 from (2-4-9)
on page 100. Compute dx = adx\ + avdt\, etc., and plug in (2-4-17). Then the
invariance requirement (ds) 2 = a$dx{)2 + ai{dy$2 + a3(dz\)2 + b(dt\)2 implies
c2ai = —b. Letting the frame move along y axis and then along z axis, conclude
that c2a2 = —b, c2a^ = —b.
As a result, the general form of ds, invariant under the Lorentz trans-
formation, is (ds)2 = a((dx)2 + (dy)2 + (dz)2 -c2(dt)2), and the two n a t u r a l
choices of a are 1 and — 1; other choices would introduce uniform stretching
or compressing. Both a = 1 and a = — 1 are used in different areas of
physics. In what follows, our selection will be a = 1, corresponding to
(ds)2 = (dx)2 + (dy)2 + (dz)2 - c2(dt)2. (2.4.18)
To avoid negative distances, we have t o modify the arc length formula

(1.3.10) and compute the length of the world line as follows:
L(s1)S2) = f" ê (\x'(s)\2 + \y'(s)\2 + \z'(s)\2 - c2\t'(s)\2) ds, (2.4.19)
where
_ f 1, if |a:'(s)| 2 + \y'(s)\2 + \z'(s)\2 - c2\t'(s)\2 > 0 (space-like world line),

\-l, if \x'(s)\2 + \y'(s)\2 + \z'(s)\2 - c2\t'(s)\2 < 0 (time-like world line).
This special way to measure distance is the main difference between

relativistic space-time and ordinary Euclidean space. As a result, the
vector space E 4 in which the distance is measured by (2.4.18) is often
Einstein's Field Equations 105
called the Minkowski space, after the German mathematician HERMANN

MINKOWSKI (1864-1909), who introduced relativistic space-time in 1907 as
a part of a new theory of electrodynamics.
Mathematics takes (2.4.17) to a whole new level of abstraction by writ-
ing
(ds)2 = {dx, dy, dz, dt) g (dx, dy, dz, dt)T', (2.4.20)
where (dx,dy,dz,dt) is the row vector, g is the m e t r i c t e n s o r , that is,

a 4 x 4 symmetric matrix with non-zero determinant, and (dx, dy, dz, dt)T
is the column vector, Formula (2.4.20) defines a metric in M4, and ds is
called the l i n e element corresponding to the metric g. As a result, the
same vector space K4 can lead to different metric spaces, corresponding to
different choices of the matrix g. While this construction is possible in any
number of dimensions, and is the subject of pseudo-Riemannian geometry
(the standard Riemannian geometry deals with positive-definite matrices
g), we will only consider the four-dimensional case. The matrix g can, in
principle, be different at different points (x, y, z, t). A metric corresponding
to a constant matrix g is called f l a t ; in particular, the metric (2.4.18)
used in special relativity is often called the f l a t Minkowski metric. In
general r e l a t i v i t y , the metric is not flat, and the matrix g is different
at different points, that is, each element fljj of g is a function of (x,y, z,t).
We will discuss general relativity in Section 2.4.3.
If the matrix g is the same at all points, then, after a suitable change
of basis, we can assume that g is diagonal, see Exercise 8.1.5 on page 454.
For a non-constant matrix, a simultaneous reduction to a diagonal form at
all points at once might not be possible. A non-constant matrix g that is
diagonal everywhere still gives rise to a non-flat metric.
EXERCISE 2.4.4. c (a) What matrix corresponds to the usual Euclidean
metric? Hint: the identity matrix, (b) Write the components of the matrix g
corresponding to the metric (2.4-18). Hint: the matrix is diagonal.
2.4.3 Einstein's Field Equations and General Relativity

As we discussed on page 95, Einstein adopted the hypothesis that physical
laws should be stated in a form that is invariant under coordinate trans-
formations between frames. In his theory of special relativity, the frames
move with constant relative velocity, and there is only one possible coordi-
nate transformation: the Lorentz transformation. In his theory of general
relativity, the frames can move with variable relative velocities, for exam-
ple, rotate, and there are many admissible coordinate transformations. To
achieve this general invariance, he used tensor analysis and non-Euclidean
geometry. Thus,the mathematical foundation of general relativity is ten-
sor analysis and pseudo-Riemannian geometry, and the main result is a
system of ten nonlinear partial differential equations. This is more than
enough to scare most people away from even trying to understand the sub-
ject. The good news is that, while deriving the equations indeed requires
all this advanced knowledge, solving these equations often requires little
beyond the basic theory of ordinary differential equations. Accordingly,
we start by introducing the equations. We will then describe the mean-
ing of all components of the equations, and finally derive and analyze one
particular solution, known as the Schwarzschild solution. For the sake of
completeness, we summarize the main facts about tensors in Section 8.3 in
Appendix.
At the end of Section 2.4.2, we discussed the metric geometry of rela-
tivistic space-time. This geometry is characterized by the matrix, or the
metric t e n s o r , Q in equation (2.4.20), page 105. In special relativity, the
metric is flat, that is, the matrix g is the same at every point, see (2.4.18)
on page 104.
In general relativity, the metric tensor can be different at different points
and is determined by the gravitational field. The corresponding metric in
space-time is no longer flat, but curved. This equivalence between gravita-
tion and curvature of space-time is the mathematical expression of the main
idea of general relativity. Let us e m p h a s i z e t h a t general relativity
deals only with gravitational forces.
Einstein's field equations, also known as E i n s t e i n ' s g r a v i t a t i o n
equations, or the field equations of general relativity, describe the rela-
tion between the metric tensor g and the gravitational field at every point
of relativistic space-time. To state these equations, we switch from the
(x,y,z,t) variables to (a:1 ) variables:
x^x1, y = x2, z = x3, t = x*, x = (x1,x2,x3,x4). (2.4.21)
Note the use of super-scripts rather than sub-scripts in this definition. The
position of indices is important in tensor calculus; for example, in tensor
calculus, g y is an object very different from Qij. The reader should pay
special attention to position of indices in all formulas in this section.
The equations, introduced by A. Einstein in 1915, are written as follows:
^•[fl]-^[fl]fl« = ^ T « [ f l ] , i,j = l,..A, (2.4.22)
where Rij, R, and T^- are non-linear partial differential operators acting
on the metric tensor Q.
Our most immediate goal now is to understand the meaning of every
component of (2.4.22).
• The metric tensor a = (gij(x), i,j = 1,...,4) is a symmetric 4 x 4
matrix defined at the points x = (a;1, a;2, a;3, a;4) of relativistic space-time.
This matrix has a nonzero determinant and continuous second-order partial
derivatives of all components %%j(x) for most x (say, all but finitely many).
This matrix defines a metric on space-time by defining the l i n e element
ar every point x according to the formula
4
(ds)2(x)= ] T gij(x)dxidxj; (2.4.23)
compare this with (2.4.20) on page 105. Summations of this kind will
appear frequently in our discussion. To minimize the amount of writing,
we will use E i n s t e i n ' s summation convention, assuming summation in
an expression over an index from 1 to 4 if the index appears in the expression
twice. With this convention, (2.4.23) becomes
(ds)2(x) = Qi:j{x) dxi dxj. (2.4.24)
EXERCISE 2.4.5. c The Kronecker symbol S is defined by
l=J
' (2.4.25)
Verify that, according to Einstein's summation convention, 5\ = 4.

• The operators R^ in (2.4.22) are denned as follows:
f)Vk (T) dT^(x)
J i
KM*) = -^ -—^F +rS(*)r^(a')-Cfc(x)r?i(x)> (2.4.26)
where
i 4
rS(-)-j.~(«)(^ + ^ - ^ ) P«7)
and gmn(x), m,n = 1,...,4, are the components of the inverse matrix
g~1(x), that is, gmn(x)g„j(x) = 6™ for every x, with 6™ defined in
(2.4.25).
• The operator R in (2.4.22) is denned as follows:
R[B](x)=aij(*)Rij[B\(.x)- (2-4.28)
c
EXERCISE 2.4.6. (a) With the summation convention now in force, write
all the missing sums in (24.26), (2.4.27), and (2.4.28). (b)B Verify that,
for all x and all i,j,k,m, TJ™- = V%, and so there are at most 40 distinct
values of T™y Hint: all you need is that flij(x) = Qji(x). (c)A Verify that
Rij[o\(x) — Rji[g](x). Hint: find a function f so that r ^ = df/dxj.
(d)B Verify that
8 y (x)^-[fl](x) - \gi\x)R[g\{x)gij{x) = -R[g](x). (2.4.29)
• The collection of operators T y , i,j = 1,...,4, in (2.4.22) is called

the s t r e s s - e n e r g y t e n s o r and is determined by a given distribution of
the gravitational matter in a region of space. This tensor is symmetric:
Ty = Tji (it has to be, since the left-hand side of (2.4.22) is symmetric). At
this level of generality, no specific representation for Ty can be provided:
the particular form of Ty depends on the particular problem. In what
follows, we will concentrate on the vacuum s o l u t i o n s of the field equations,
that is, the solutions of (2.4.22) corresponding to Ty = 0; accordingly, we
will not discuss the stress-energy tensor any further.
• The remaining letters in (2.4.22) have the following meaning: G «
6.67 • 1 0 - 1 1 m 3 /(kg- sec2) is the universal gravitational constant, c « 3 • 10 8
m/sec is the speed of light, n « 3.14.
EXERCISE 2 . 4 . 7 . C
Verify that (2.4-22) is a system of ten equations for the
ten components gy, 1 < i < j < 4 of the metric tensor.
The system of equations (2.4.22) does not depend on the particular
4
choice of the coordinate system (ar ) in space-time; in fact, the
functions g^, T™fc, Ri3 [g], and R[g] can be defined in a coordinate-free way,
and the formulas (2.4.26)-(2.4.28) turn out the same in every coordinate
system (of course, the particular form of fly will depend on the coordinate
system). The resulting independence of the field equations (2.4.22) of the
coordinate system is the very foundation of general relativity.
Equations (2.4.22) are too complicated to justify as a postulate or as
a generalization of experimental facts. Instead, the main postulate of gen-
eral relativity is that the physical laws are independent of a particular frame,
inertial or not, in space-time, that is, the equations are invariant under gen-
eral coordinate transformations from one frame to another; what makes spe-
cial relativity special is the restriction of this invariance to inertial frames.
When combined with differential geometry and tensor calculus, whose con-
structions do not depend on a particular coordinate system and are valid
in every space, Euclidean or not, this invariance assumption motivates the
formulation of equations (2.4.22). Before formulating the equations in 1915,
Einstein studied Riemannian geometry and also learned tensor calculus di-
rectly from one of the creators of the theory, the Italian mathematician
TULLIO LEVI-CIVITA (1873-1941). We give a brief and elementary sum-
mary of tensors on page 457 in Appendix.
In classical Newtonian theory, the gravitational field is described by the
partial differential equation <pxx + (pyy + tpzz = 4irGp, where <p is the poten-
tial of the field and p is the density of the gravitating mass; see Exercise 3.3.8
on page 168 below. Einstein started with a relation, -Ayfe] = 4irGTij[g],
with g instead of <p and the stress-energy tensor instead of the usual mass
density, and assumed that the tensor Aij depends linearly on the second-
order partial derivatives of 0. In tensor analysis, it is proved that such a
tensor A^ [g] must have the form
Ai:j [g\ = a Ri:j [g] + (3 R[g]gij + 7 0y
for some real numbers a, 0, 7, with R^ and R denned in (2.4.26) and

(2.4.28), respectively. Physical considerations, such as dimension analysis
and conservation of energy and momentum, result in the values a = —2/3 =
c 4 /2, 7 = 0. For more details, see the book Theory of Relativity by W. Pauli,
1981.
To complete the general discussion of (2.4.22), let us mention some other
related terms and names:
• The collection of Rij[g], i,j = 1,...,4, is called the Ricci curvature
t e n s o r (also Ricci curvature or Ricci tensor), after the Italian mathemati-
cian GREGORIO RICCI-CURB ASTRO (1853-1925), whose most famous pub-
lication The Absolute Differential Calculus, written together with his former
student T. Levi-Civita, was published with his name truncated to Ricci.
• The number R[g] is called the s c a l a r curvature or Ricci s c a l a r .
• Each r L is called a C h r i s t o f f e l symbol, after the German mathe-
matician ELWIN BRUNO CHRISTOPFEL (1829-1900). Because of the way
the numbers n f c change when the coordinate system is changed, the col-
lection of Christoffel symbols is not a tensor.
• The tensor Rij[g] — {l/2)R[g)gij, which is the left-hand side of (2.4.22),

is called E i n s t e i n ' s tensor.
Even though equations (2.4.22) are complicated, let us assume that
we were able to solve them and found the metric tensor g. In classical
Newtonian mechanics, the ultimate goal is to compute the trajectory of a
moving object under the action of forces. The equation of the trajectory is
provided by the Second Law of Newton. In general relativity, the Second
Law of Newton is replaced by a postulate that a point mass in a curved
space moves along a geodesic. A geodesic in a curved space is what a
straight line is in a flat space. Once you think about it, a straight line in
a fiat space has two special properties: (a) it defines the path of shortest
distance between two points; (b) it is straight in the sense that the tangent
vectors at each point are parallel, all being parallel to the direction vector
of the line. Accordingly, each of these properties could be used to define
a geodesic in a curved space, and it turns out that the second property
results in a more convenient definition. By this definition, it is shown in
differential geometry that x% = xl{s), s > 0, i = 1,...,4, is the vector
representation of a geodesic if and only if
where the r*-fc are from (2.4.27) on page 107. Below, we outline the proof
that a curve defined by (2.4.30) is a path of shortest distance between two
points.
Therefore, a trajectory of a moving object in a gravitational field in
general relativity is computed as follows:
(a) Find the metric tensor g by solving the field equations (2.4.22) for the
specified stress-energy tensor T y ;
(b) Compute the functions Tljk according to (2.4.27);
(c) Solve the system of equations (2.4.30).
This summarizes the general relativity mechanics.

Let us now look more closely at equations (2.4.30). Without any gravi-
tation, the space is flat so that each fly is independent of x, which makes
each Tljk — 0. Then the geodesic equations (2.4.30) become d2 xl(s)/ds2 = 0
so that xl (s) — xl(0) + vxs, which is a straight line, corresponding to the
motion with constant velocity.
It turns out that a geodesic is always parameterized by its arc length

s, although not necessarily canonically. In other words, equations (2.4.30)
imply that ds = ads for some constant a > 0; canonical parametrization
corresponds to a = 1, see page 30.
EXERCISE 2.4.8. A (a) Verify that, if a vector function x = x(s) satisfies
(2.4.30), then
d ( . , ..dxUs) dxi(s)\
Ts{ ^ -dir^r)=0-
3 s))
(2A31>
Hint: differentiate and use (2.4-27). (b) Conclude that 5 = as + f3, where s
is the arc length of the curve and a, j3 are real numbers. Hint: the left-hand
side of (2.4.31) is d2s(s)/ds2.
By following the steps below, the reader can prove that the geodesic,
as defined by (2.4.30), can be interpreted as the shortest path between two
points in a curved space.
Step 1. It is easier to do the basic computations in an abstract setting in IRn.
Let A and B be two fixed points in Mn and F = F(q,p), a smooth M-valued
function of 2n variables q1,..., qn,px,... ,pn; we keep the convention of this
section and write the indices as superscripts. Let x — x(s), a < s < b, be
a smooth function with values in R n . Consider the f u n c t i o n a l
L(x)= I F(x(s),x'(s))ds,
Ja
where x'(s) = dx(s)/ds; being a rule assigning a number to a function, L

is indeed a functional. Later, we will take
F{q-,P) = y9ij((l)PiPj (remember the summation convention), (2.4.32)
so that L(x) is the distance from A to B along the curve x(s).

(a) Let y = y(s), a < s < b, be a smooth function with values in ffi" so
that y(a) = y(b) = 0 and let x(s) be a function for which the value L(x)
is the smallest (assume that such a function exists.) For a real number A,
define /(A) = L(x + Xy). Argue that /'(0) = 0. (b) Show that
fb fdF(x(s),x'(s)) aF(£(3),x'(s))dy'(a)\
n V{S)+ d S
'~Ja V W d^ ds ) -
(c) Integrate by parts and use y%{a) = yl(b) = 0 to conclude that
m dF(x(s),x'(s)) d dF(x(s),x'(s))\
-I.( W Ts W
i(
)y{s)ds-
(d) Argue that, since /'(0) = 0 and y = y{s) is an arbitrary smooth
function, the expression in the big parentheses in the last equality must
vanish for all s and so, for all i = 1 , . . . , n,
d2F{x(s),x(s))dxi d2F{x(s),x(s))d2x^ dF{x(s),x{s))

dqidqi ds dfdpi ds2 dqi ~
(2.4.33)
System (2.4.33) is an example of the Euler-Lagrange equations of vari-
ational calculus.
Step 2. Verify that, with F as in (2.4.32) and s = s, equation (2.4.33)
becomes (2.4.30).
While in most situations of practical importance the geodesic does in-
deed define the shortest path between two points, and the path is unique up
to a re-parametrization, there are two main technical problems with defin-
ing the geodesic as the distance-minimizing path: (i) The a priori possibility
that several different paths have this property, and (ii) The possibility that
the sign of (2.4.32) is different in different parts of the space (even the flat
Minkowski metric (2.4.18) can change sign.) Definition (2.4.30) avoids these
problems by generalizing a different property of the straight line, namely,
that the tangent vectors at every two points of the curve are parallel. This
property formalizes the intuitive notion of going straight, and the functions
Tljk define the parallel transport of a vector along a curve by means of the
L e v i - C i v i t a connection. For more details, an interested reader should
consult a specialized reference on Riemannian geometry, such as Chapter 10
of the book Differentiable Manifolds: A First Course by L. Conlon, 1993.
The rest of the section will be an investigation of the important spe-

cial case of (2.4.22), corresponding to Tij = 0, namely, the equation
Rij [g] — (l/2)Rgij = 0. The solutions of this equation are called the vacuum
s o l u t i o n s of (2.4.22), because Ty = 0 in empty space.
EXERCISE 2.4.9.c Verify that Rij[g] - (1/2)11$^ = 0 if and only if
Rij[g]=0. (2.4.34)
Hint: use (2.4-28) and (2.4-29); the matrices g and j? - 1 are non-singular.
From now on, we will consider (2.4.34), which is still a system of 10

equations, and call this system the vacuum f i e l d equations. Every fiat
metric, that is, a constant non-singular matrix g, is a trivial solution of
(2.4.34). Indeed, for constant g, the numbers T\j are all equal to zero, and
then so are Rij. A relativistic vacuum solution must be invariant under the
Lorentz transformation. By Exercise 2.4.3 on page 104, such a solution g is
a constant multiple of the f l a t Minkowski metric (2.4.18) on page 104.
EXERCISE 2.4.10. c Verify the above assertion that every trivial solution of
(2.4-34), invariant under the Lorentz transformation, is a constant multiple
of the flat Minkowski metric (2.4-18).
In 1916, the German physicist K A R L SCHWARZSCHILD (1873-1916)
found the first non-trivial solution of the vacuum field equations. Vacuum
field equations do not necessarily describe a space without any gravitation
whatsoever; gravitating matter can be present, but in other parts of the
space, where we are not trying to solve the field equations. Accordingly,
in the derivation of the Schwarzschild solution, it is assumed that there is
only a stationary spherical object of mass M somewhere in space.
Introduce spherical coordinates (r, 0, <p) in the space part of space-time
(x,y,z,t): x = rcos9sintp, y = rsmOsimp, z = rcosip. Then we set
x1 = r, x2 = 6, x3 = ip, x4 = t. Schwarzschild's s o l u t i o n , or
Schwarzschild's metric, g is a diagonal matrix so that
d(s? = X_}R /r)(dr? + r 2 ( s i n M ^ ) 2 + W ' ) - c2(l - {R0/r)){dt)\

(Ro/r)
(2.4.35)
where
Ra = ™i (2.4.36)
is called the Schwarzschild r a d i u s of the mass M. Schwarzschild found

this metric by looking for a spherically symmetric solution of (2.4.34); see
below for an outline of the corresponding computations.
EXERCISE 2.4.11. c (a) Verify that R0 indeed has the dimension of the
length, (b) Compute the Schwarzschild radius of the following objects: (i)
The Sun (take M = 2 • 10 30 kg), (ii) The Earth, (Hi) yourself, (c) Find the
mass and the Schwarzschild radius of the ball whose density is p = 1000
kg/m3 and whose radius is R0. Hint: you have R0 = 2G(4n/3)Rlp/c2; find
R0 and then M = (4n/3)Rlp. (d) Repeat part (c) when p = 10 18 kg/m3, the
density of a typical atomic nucleus, (e) Verify that, as r —> oo, the metric
in (2.4-35) approaches the flat Minkowski metric in spherical coordinates

(dr)2+r2 (sin2 ip(d9)2 + (d<p)2)-c2(dt)2.
The reader can derive (2.4.35) by following the steps below. The com-
putations are not hard, but Steps 2 and 3 require a lot of care and patience.
Step 1. Argue that spherical symmetry and stationarity imply ds2 —
F(r)(dr)2 + r2(sin2(p{d6)2 + (dtp)2) + H(r)(dt)2 for some functions F,H
that are not equal to zero. With x1 = r, x2 = 0, x3 = tp, x4 = t, we have
the components of the metric tensor g n = F(r), 022 = r2 sin 2 <p, g 33 = r 2 ,
044 = H(r) and all other g^- are zero. Similarly, for the inverse tensor,
0 i i = l/F(r), g 22 = l / ( r 2 sin 2 <p), g 33 = 1/r2, g 44 = l/H(r) and all other
u
g are zero.
Step 2. Using formulas (2.4.27), verify that, of the 40 values of rj f c , only
nine are non-zero:
,i F'(r) pi rsin2<p ril r pi H'(r).
u 133 _ 44
~ 2 F ( r ) ' ™~ F(r) ' F(rY 2F(r)'
l2 1 p 2 COStp 3 1 3 4 ^'(r)
12 ~ ' l
23 —
• i i
13 "~ i L 22 — sin <p cos </>;F r 1 4 =
r sinw r
(2.4.37)
57ep 5. Using formulas (2.4.26), verify that, out of 10 values of R^, only
four are non-zero:
_ H"(r) (H'(r))2 F'(r) F'(r)H'(r)
11
~ 2H(r) 4H2(r) rF(r) AF(r)H(r)'
, , 1 rF'(r) rH'ir) \ . ,
(2.4.38)
_1 rF'(r) rH'(r) _ R22
F(r) 2F2(r) +
F(r)H(r) ~ sin2<p]
H"(r) F'(r)H'(r) H'(r) (H'(r))2
2F(r) AF2(r) rF(r) 4F(r)H(r)'
Step 4- (a) Multiply by 4rF(r)H2(r) the equation Rn = 0, multiply by
4rF2(r)H(r) the equation R44 — 0, and take the difference to conclude
that AH(r)(H(r)F(r))' = 0 or F(r)H(r) = A for some real number A. (b)
Multiply by 2F2(r)H(r) the equation R33 = 0 and use F(r)H(r) = A to
conclude that r F ' ( r ) = F(r)(l - F(r)) or F(r) = (1 + lf(Br))-1 for some
real number B.
Step 5. We have limôo F(r) = lim,.-^ H(r) = 1 and, by (2.4.18) on page
104, far away from the gravitating mass the metric should be flat, that is
(cte)2 = (dr)2 + r 2 sin 2 <p (d6)2 + r2{dip)2 - c2(dt)2. Conclude that A = - c 2 ,

where c is the speed of light. Keep in mind that (dx)2 + (dy)2 + (dz)2 =
{dr)2 + r2 sin2 <p (d6)2 + r2(d<p)2.
Step 6. Using (2.4.30) and (2.4.37), verify that the geodesic equations for
the trajectory (r(s),8(s),(p(s),t(s)) are as follows:
[S) +
2F(r(s))[V {S))
F(r(s)) {° [S
>> 2F(s) °'
9"(s) + -r'(s)9'(s) = 0;
r
j (2.4.39)
¥>"(*) + -ri:r'(s)ip'(s) - sin<p{s) cosip(s)(6'(s))2 = 0;
r(s)
t"(s) + ^r'(s)t'(s) = 0.
Keep in mind that, since rf- = T^, the terms in (2.4.30) often come in
pairs.
Step 7. Consider the circular trajectory of radius R in the equatorial
plane (<p = 7r/2) in the Newtonian approximation, so that r(s) = R,
ip(s) = 7r/2, t'(s) — 1; the absolute time t'(s) = 1 is the key feature
of Newtonian mechanics. Use the function H(r) = —c2(l + l/(Br)) we
found in Steps 4 and 5, to conclude from the first equation in (2.4.39)
(6'(s))2 = -H'(R)/(2R) = -c2/(2BR3). On the other hand, Newton's
laws imply that, for the circular trajectory in the gravitational field of
mass M, (6»'(s))2 = MG/R3. Conclude that B = -c2/(2MG) = -1/R0.
This completes the derivation of (2.4.35).
It is always a major accomplishment to find an explicit non-trivial so-

lution of a nonlinear partial differential equation. The importance of the
Schwarzschild solution is further enhanced by the following facts:
• This solution leads to a direct verification of general relativity in three
important problems, namely, perihelion precession of planets, gravitational
deflection of light, and gravitational red shift; we discuss these problems
below);
• The solution is a lot more general than it looks: in 1927, the American
mathematician G E O R G E DAVID B I R K H O F F (1884-1944) proved that, after
a suitable change of coordinates, every spherically symmetric solution of
the vacuum field equations becomes a Schwarzschild solution.
• The solution formula (2.4.35) leads to a natural, and mathematically pre-
cise, definition of a black hole as an object whose size is smaller than or
equal to the corresponding R0. As the preceding exercise shows, a black

hole can be produced either by compression or by vast accumulation of un-
compressed matter. Once you do Exercise 2.4.11, you realize that neither
you nor the Earth nor the Sun is likely to become a black hole.
The first experimental verification of general relativity was the calcula-
tion of the precession of the perihelion (the point closest to the Sun) for the
planet Mercury. Astronomical observations of the precession had been car-
ried out since the late 18th century, but the Newtonian theory of gravitation
was unable to calculate the observed precession correctly. In 1916, Einstein
used the Schwarzschild solution, with R0 equal to the Schwarzschild radius
of the Sun, to compute the correct value of the precession; see Problem 2.2
on page 414.
The reader can derive equation (7.2.7) on page 415, describing the tra-
jectory of Mercury (or any other planet, for that matter) by following the
steps below.
Step 1. Write three equations for the geodesic involving 0(s), <p(s), and
t(s) using (2.4.30) and (2.4.37); use g to denote dg(s)/ds, and Q', to de-
note dQ(r)/dr, for the appropriate functions g,Q:
§(s) + - ^ 6(s)f(s) = 0; (2.4.40)

r{s)
<p{s) + - ^ - ff>{s)r(s) - sin <p(s) cos ip(s) 62{s) = 0; (2.4.41)
r(s)
... . H'(r(s)) ., s , , ,
t{s)+
H(r(s))t{s)^s)=°- {2AA2)
Step 2. From (2.4.40) conclude that r2(s)6(s) — a, with the real number
a independent of s; from (2.4.41), that <p(s) = n/2 is a possible solution;
from (2.4.42), that i(s) = /?F(r(s)), with j3 independent of s.
Step 3. Keeping in mind that the geodesic is parameterized by the arc
length, take ds — cds to have s in time units, and then, with <p(s) = TT/2,
conclude that (2.4.35) implies
{dsf = c2H(r(s))(dt{s))2 - F(r(s))(dr(s))2 - r2(s)(d0(s))2,

1 = c2H(r(s)) (i(s))2 - F(r(s))(f(s))2 - r2(s)(0(s))2,
(r(*))2 + 4^ (l ~ T T ) - ^TT = c 2 ^ " !)• (2-4-43)

J
r (s) \ r(s)J r(s)
Similar to (2.4.19) on page 104, we have to switch the sign in (2.4.35) to
ensure that the expression for (ds)2 is positive.
Step 4- Put u(9(s)) = l/r(s), using r{s) = -(u'(9)/u2(9))6(s) = -cm'(9),

to get
{u'(9))2 + u2{9){\ - RoU{9)) - c2R0u{9)/a2 = c2{(32 - I)/a2.
Step 5. Differentiate the last equality with respect to 6 to get
«"(9) + «(9) = § + Y« 2 («) I

which is exactly equation (7.2.7) on page 415, with A$ = c2R0/(2a2).
Another experimental verification of general relativity is the predicted
GRAVITATIONAL DEFLECTION OF LIGHT. For the flat Minkowski metric
(2.4.18), page 104, the trajectories of photons are lines satisfying x2 + y2 +
z2 = c2t2 or ds — 0. In a curved space, a photon geodesic is not necessarily
a line and can be curved (deflected) by the gravitational mass. Because of
the properties of g, this geodesic must still satisfy ds = 0.
EXERCISE 2.4.12.3 Verify that the trajectory of a photon satisfies ds = 0 in
every curved space. Hint: change the metric on the flat space; the zero on the
right will stay zero.
Using the Schwarzschild metric (2.4.35), Einstein calculated the deflec-

tion of light rays passing near the Sun. The experimental validation of the
calculation came in 1919, when, during a total solar eclipse, the British
astronomer Sir ARTHUR STANLEY EDDINGTON (1882-1944) showed that a
star whose light passed close to the Sun appeared displaced by the amount
that corresponded to the deflection value calculated by Einstein.
EXERCISE 2.4.13? Consider a photon moving radially, that is, d6 = dip — 0.
Use (2.4-35) to conclude that ds2 = 0 implies that, for r > R0, t — ± ( r +
ro + R0 In ((r/i? D ) — l)- Therefore, the equation of trajectory (world line)
of the photon is different from the equation of the straight line in four-
dimensional space-time.
Yet another experimental verification of general relativity is GRAVITA-
TIONAL RED SHIFT, that is, the increase of wavelength of photons that
are mowing away from the source of the gravitational field Red light has
the longest wavelength in the optical spectrum, whence the term red shift.
On a deep level, the gravitational red shift illustrates two major results of
general relativity, the slow-down of time and shortening of length in a grav-
itational field; both these results can be deduced from the Schwarzschild
solution. On an intuitive level, this shift can be explained without using

the Schwarzschild solution or any general relativity. Indeed, let a photon
of wavelength A be emitted from a far-away star of mass M and radius
R. At the moment of the emission, the photon has radiation energy hu,
mass m = hv/c2, and potential energy —mMG/R. Far away from the
star, with no other gravitating objects around, the potential energy of the
photon is approximately zero. This increase in potential energy of the
photon must be compensated by a decrease of its radiation energy. Thus,
mMG/R = -hAu or Av/v = -MG/{Rc2) = -Ra/{2R). Assuming that
the relative changes are small, and using v = c/A, we can write the result
as
EXERCISE 2.4.14. (a)c Estimate the gravitational red shift for the Sun.
(take M = 2 • 10 30 kg, R — 7 • 109 m). (b)c Consider the radiation emitted
at the surface of the Earth and received H meters above the surface, with H
much smaller than the radius of the Earth. Denote by g the gravitational
acceleration on the Earth surface. Show that the observed red shift should
be
£ -f • <*".>
(c)A Explain how the slow-down of the clock and shrinking of length in
the gravitational field can be deduced from the Schwarzschild solution, and
connect these effects with the gravitational red shift. Hint: for example, with
dO = d<p = dt = 0 in (24.35), we have ds = (1 - {Ra/r))~1/2dr.
Einstein predicted the gravitational red shift eight years before formu-
lating the theory of general relativity, but, because of the need for very high
accuracy, both in the generation of the radiation and in the measurements,
it was only in the 1960s that the gravitational red shift was observed in an
experiment (for H = 90m, the right-hand side of (2.4.45) is about 10~ 14 ).
The famous Pound-Rebka-Snider experiment was first conducted in 1960
at Harvard University by R. V. Pound and G. A. Rebka Jr.; in 1964, R.
V. Pound and J. L. Snider carried out a more accurate experiment; as of
2005, Robert V. Pound is Emeritus Professor of Physics at Harvard.
For an observer on the Earth, the clock on an orbiting satellite will
appear to be running slow because of the special relativity effects, see for-
mula (2.4.12) on page 101, and because of the gravitational red shift: the
gravitational field of the Earth is much weaker in space. With communica-

tion signals travelling at the speed of light, even a small discrepancy in the
clocks can cause problems. For example, a 10-meter error in space can be
caused by a 3.3 • 10~8-second error in time (the time it take a radio-signal
to travel 10 meters). The relativity effects can be 10~ 6 seconds per day or
more. The appropriate adjustment of the clock is necessary, and is indeed
implemented, on many navigation satellites.
EXERCISE 2.4.15. A+ Estimate the necessary time correction per day for a
satellite on a geostationary orbit (that is, a satellite moving in the equatorial
plane of the Earth in the direction of the rotation of the Earth and with
period of revolution around the Earth equal to the Earth period of rotation
about its axis). Take into account both special and general relativity effects.
The gravitational red shift is different from the cosmological red
s h i f t , which is attributed to the expansion of the Universe, and which
was first observed in 1929 by the American astronomer EDWIN POWELL
HUBBLE (1889-1953). The cosmological red shift is consistent with general
relativity, and is connected with non-vacuum solutions of the field equa-
tions; it is a popular topic of research in modern cosmology. The idea is to
assume a particular form of the stress-energy tensor Ty on the right-hand
side of the Einstein field equations (2.4.34) so that the solution has a pre-
scribed form. A popular form of the solution produces the following line
element:
(ds)2 = c2(dt)2 - R(t) ({dr)2 +r2(sm2p(de)2 + (dip)2)) - c2{dt)2,
where R(t) represents the radius of the Universe.

EXERCISE 2A.16.C (a) Verify that the radial motion of the photon, corre-
sponding to above metric, with ds = 0 and 6 = <p = 0, satisfies
4l
Ht
(2A46)
/ Wr
(b) Assume that a pulse of electromagnetic radiation, lasting Aio seconds,
is emitted from a point in space at time moments to; the value of 1/Aio
can be interpreted as the frequency of the emitted radiation. The pulse is
received at a different point during the time interval [ti,ti + Aii], where
ti > h + Aio; the value 1/Aii can be interpreted as the frequency of the
received radiation. Assuming that both Ato and At\ are small, use (2.4-46)
to conclude that
At 0 A£i
(2.4.47)
R(h) RihY
Hint: r(ti) - r(t0) = r(h + Ati) - r(t0 + At0).
By observing the light from distant galaxies, Hubble determined that
Ato < Aî or 1/Ato > 1/Ati, that is, the frequency of the received radi-
ation shifts toward the lower-frequency, or red, part of the spectrum. By
(2.4.47), this observation implies R(t0) < R(h), that is, the Universe is
expanding.

Vectors

Uploaded by

Copyright:

Available Formats

Vectors

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vectors

Uploaded by

Copyright:

Available Formats

Chapter 2

Vector Analysis and Classical and

2.1 Kinematics and Dynamics of a Point Mass

Kinematics is the study of motion without reference to forces; the Greek

2.1.1 Newton's Laws of Motion and Gravitation

If the reference frame is inertial, then (2.1.1) becomes

Recall that a force F acting on the point mass m has a torque, or

Applying the rule (1.3.6), page 26, to formula (2.1.4) we find

Fig. 2.1.1 Simple Rigid Pendulum

The motion of m is a circular rotation in the (i, j) plane, and is best

L — mr x r = m(£r x 196) = m£29A, and

T=-(mg£ sin 9 + £2C9)K. (2.1.8)

m£e + d6 = -mg sin9. (2.1.9)

Equation (2.1.9) is a nonlinear ordinary differential equation and cannot

is obtained from (2.1.9) when c = 0 and 6 is small so that sin0 w 6; this

0(t) = 0O cos(cjt), where w = (£/g)1/2.

The basic ideas of modern astronomy go back to the Polish astronomer

recover Newton's argument by combining equation (1.3.29), page 36, with

In our derivation of (2.1.11), we implicitly used the equivalence principle

Following the historical developments, we derived relation (2.1.11) from

2.1.2 Parallel Translation of Frames

starting point is the analysis of the relative motion of frames.

Fig. 2.1.2 Translation of Frames

Clearly, ro(t) — roi(i) + Vi(t), and the absence of relative rotation

r o ( * ) = r 0 i ( t ) + fi(t), r(t)=roi(*) + ri(t). (2.1.12)

If the frame O is inertial, then, by Newton's Second Law, mr(t) = F(t),

mri(t) = F(t) - mroi(t) (2.1.13)

As an EXAMPLE, consider a golf ball in a moving elevator. We fix the

2.1.3 Uniform Rotation of Frames

Fig. 2.1.3 Rotation of Frames

By construction, roi = Rf. For a passenger riding in the car, the

ar = -(R + v0t)Jl r, ae = 2v0wo §• (2.1.14)

By (1.3.27) on page 36, r 0 i = -Rw$r. In frame 01: r^t) = 0. Then

r 0 ( 0 = roi(t) + r\(t) - v0ttj$r + 2v0uj0d. (2.1.15)

mf\ = F + (mRul + mvotwl) r — 2mv0ujo&- (2.1.16)

Thus, frame 0\ is not inertial. Similar to (2.1.13), inertial forces appear as

circle of radius R with uniform angular velocity u (Figure 2.1.4). Denote by

Fig. 2.1.4 Rotating Point

r'(t) = ( i x j ) x («f(<)) = iv x r{t) = w x r(t), (2.1.17)

EXERCISE 2.1.6.c Verify all equalities in (2.1.17).

D0R{t) = DiR(t) +UJX R(t). (2.1.22)

mfi = F - 2mw x n - m w x f w x r i ) (2.1.24)

in the rotating frame 0\. Similar to (2.1.16), inertial forces appear as

Fig. 2.1.5 Motion Along a Meridian

(1.3.27), page 36, we find ag = 0 and so

rl=ar = -j2Rr, (2.1.25)

where R is the radius of the Earth.

acor(t) = -27-Rsin(7i)b'(i). (2.1.26)

Finally, we use 9 and b to write the centripetal acceleration ac = w x ( u x r )

u>xr = Rsm(n/2-e)b' = RcosQb', u xb'=-uj2b. (2.1.27)

Combining (2.1.25), (2.1.26), and (2.1.28) in (2.1.23), we find the total

r(t) = --y2Rr(t) -2^Rsia{jt)b'(t) - cj2Rcos(-ft)b(t). (2.1.29)

EXERCISE 2.1.8. c Verify all the equalities in (2.1.27).

Newton's Second Law implies mr = F. By (2.1.24),

mf\ = F - 2mu> x f i - mu> x (u> x n). (2.1.30)

The force F is the sum of the Earth's gravitational force Fa =

FP(t) = mri + 2mu>u x r i + rnu x ( w x n) — Fa- (2.1.31)

Conversely, if the force Fp — Fp(t) is specified, then the resulting trajec-

mri = F — m acor — mac = FQ + Fp + Fcor + Fc.

0 = i • Ji = (i • )(ii • «) + (i • i)(ji • i) + (ii • K)C?I • «)• WA Veri/j/ i/iai