You are on page 1of 157

Contents

CHAPTER 1 ............................................................................................................................ 4
DYNAMICS OF SYSTEM OF PARTICLES.............................................................................. 4
1.1 External and Internal Forces .......................................................................................... 4
1.1.1 Translational Motion.............................................................................................. 4
1.1.2 Rotational Motion .................................................................................................. 6
1.2 The Method of Lagrange ............................................................................................... 8
1.3 Conserved Quantities .................................................................................................. 14
1.3.1 Conservation of Linear Momentum...................................................................... 16
1.3.2 Conservation of Angular Momentum ................................................................... 16
1.3.3 Conservation of Energy ....................................................................................... 17
1.4 Two-Body Problem in Central Force ........................................................................... 18
1.4.1 Center-of-Mass (CM) and Relative Coordinates ................................................... 18
1.5 The Kepler Problem .................................................................................................... 22
1.5.1 Calculation of r ( ) .............................................................................................. 22
1.5.2 The Orbits ............................................................................................................ 23
1.5.3 Proof of Conic Orbits ........................................................................................... 24
1.5.4 Kepler's Laws ...................................................................................................... 26
1.6 Elastic Collisions of Two Particles .............................................................................. 28
1.6.1 Kinematics of Elastic Collisions........................................................................... 35
1.6.2 Inelastic Collisions ............................................................................................... 39
1.7 Motion of Systems with Variable Mass ....................................................................... 42
1.7.1 Rocket Motion in Free Space ............................................................................... 42
1.7.2 Vertical Ascent Under Gravity ............................................................................. 45
1.8 Check List................................................................................................................... 49
1.9 Chapter 1 Exercises..................................................................................................... 49
CHAPTER 2 .......................................................................................................................... 52
RIGID BODY DYNAMICS ..................................................................................................... 52
Introduction ........................................................................................................................... 52
2.1 Rotating Coordinate Systems ...................................................................................... 52

1
2.2 The Inertia Tensor ....................................................................................................... 56
2.3 Rotation About an Axis Through the Origin ................................................................ 58
2.4 Principal Axes of Inertia ............................................................................................. 62
2.5 Steiner’s Theorem (Parallel Axis Theorem) ................................................................ 67
2.6 The Dynamics of Rigid Body (Euler’s Equations of Motion) ...................................... 68
2.6.1 Symmetrical Top Free of Torques ........................................................................ 69
2.7 Euler Angles ............................................................................................................... 71
2.8 Check List................................................................................................................... 75
2.9 Chapter 2 Exercises .................................................................................................. 76
CHAPTER 3 .......................................................................................................................... 80
THEORY OF SMALL OSCILLATIONS ................................................................................. 80
Introduction ........................................................................................................................... 80
3.1 Before Theory of Small Oscillations ........................................................................... 80
3.2 General Theory of Small Oscillations .......................................................................... 86
3.3 Free Oscillations ......................................................................................................... 91
3.4 Orthogonality of the Eigenvectors ............................................................................... 95
3.5 Normal Coordinates .................................................................................................... 96
3.6 Weak Coupling ........................................................................................................... 99
3.7 Sympathetic Vibrations and Beats ............................................................................. 104
3.8 Molecular Vibrations ................................................................................................ 106
3.9 The Loaded String .................................................................................................... 109
3.10 Dissipative Systems And Forced Oscillations ............................................................ 116
3.11 Check List................................................................................................................. 120
3.12 Chapter 3 Exercises................................................................................................... 121
WAVE PROPAGATION ........................................................................................................ 124
Introduction ......................................................................................................................... 124
4.1 Vibrating String ........................................................................................................ 124
4.1.1 Equation of Motion ............................................................................................ 124
4.1.2 General Solution: Normal Modes of Vibration ................................................... 126
4.2 Wave Propagation in General .................................................................................... 131
4.3 Lagrange Formulation of a Vibrating String .............................................................. 136

2
4.3.1 Energy And Power ............................................................................................. 136
4.3.2 Power ................................................................................................................ 138
4.4 Behavior of a Wave at Discontinuity (Reflection & Transmission): Energy Flow ..... 139
4.5 Wave Interference ..................................................................................................... 143
4.6 Wave Polarization ..................................................................................................... 148
4.6.1 Linear Polarization............................................................................................. 150
4.6.2 Circular Polarization .......................................................................................... 151
4.6.3 Elliptical Polarization......................................................................................... 153
4.7 Check List................................................................................................................. 155
4.8 Chapter 4 Exercises................................................................................................... 155
REFERENCES ....................................................................................................................... 157

3
CHAPTER 1

DYNAMICS OF SYSTEM OF PARTICLES

General Properties of Many-Particle Systems


Classical mechanics provides us with general principles for modeling any material body, be it a
solid, liquid or gas, as a system of interacting particles. To analyze the behavior of a system, we
must separate it from its environment. This is done by distinguishing external and internal
variables. The external variables describe the system as whole and its interaction with other
(external) systems. The internal variables describe the (internal) structure of the system and the
interactions among its parts. Here we analyze the general case of an n-particle system.

Consider now a system of n particles with positions ri , i = 1,….n, in flat space. The configuration

of the system then has 3n coordinates (configuration space is R 3n ), and the phase space has 6n
 
coordinates ri , pi  .

1.1 External and Internal Forces



Let Fi be the total force acting on particle i. It is the sum of the forces produced by each of the

other particles and that due to any external force. Let F ji be the force particle j exerts on particle i

and let Fi E be the external force on particle i. Using Newton's second law on particle i, we have
    
Fi  Fi E   F ji  p i  mi vi ,
j

Define the center of mass and total mass



  mi ri
R , M   mi .
 mi

1.1.1 Translational Motion

Then if we define the total momentum



   d  dR
P   p i   mi vi   mi ri  M ,
dt dt

4
we have

dP     
 P   p i   Fi   Fi E   F ji .
dt i ij

Let us define F E to be the total external force. If the weak form of Newton’s 3rd law holds,
mutual forces of any two particles on one another are equal and opposite so the inter-particle
forces are related by
  
F ji   Fij , so  ij  0 ,
F
ij

   
Thus, P  FE . Or P  MR (1.1)
That means, the internal forces cancel in pairs in their effect on the total momentum, which
changes only in response to the total external force. Eqn. (1.1) along with its interpretation is the
Center of Mass (CM) Theorem. As an obvious but very important consequence the total
momentum of an isolated system is conserved.
The CM theorem describes the average motion of a system as equivalent to that of a single

particle of mass M located at the CM, R . Thus, it separates external and internal motions,
allowing us to study them independently. And when we are not interested in internal structure,
the CM theorem justifies treating the entire system as a single particle.
Internal and external properties of the system can be separated by introducing the internal

variables. Let ri  be the position of the ith particle w.r.t. the CM co-ordinate system & it’s related

with ri by
  
ri   ri  R (1.2)
describing the position of each particle relative to the CM. The internal velocities are therefore
  
ri   ri  R (1.3)
This leads us to the following decomposition of the total kinetic energy T for the system:
1 1 1    
T  mi vi2   mi ri 2   mi (ri   R )  (ri   R )
2 2 2

1 1 d 
  mi ri 2  MR 2  R   mi ri 
2 2 dt i
1 1
  mi ri 2  MR 2
2 2

5
Thus the kinetic energy of the system can be viewed as the sum of the kinetic energies of the
constituents about the center of mass, plus the kinetic energy the system would have if it were
collapsed to a particle at the center of mass.
The total angular momentum is also just a sum over the individual particles, in this case of the
individual angular momenta:
   
L   Li   ri  pi . (1.4a)

The total angular momentum depends on the point of evaluation, that is, the origin of the
coordinate system used. We now show that it consists of two contributions, the angular
momentum about the center of mass and the angular momentum of a fictitious point object
located at the center of mass.
      
L   mi ri  vi   mi (ri   R )  (ri   R )
i i

       
  mi ri  ri    mi ri  R  R   mi ri   MR  R
i i

   
  ri  pi  R  P . (1.4b)
i

The first term of the final form is the sum of the angular momenta of the particles about their
center of mass (and is sometimes called the internal angular momentum), while the second term
is the angular momentum the system would have if it were collapsed to a point at the center of
mass (and is sometimes known as the orbital angular momentum of the system). Thus the total
angular momentum is simply the sum of internal and external (orbital) angular momenta.

1.1.2 Rotational Motion

To describe rotational motion of the system as a whole, we derive an equation of motion for the total
angular momentum.
Its rate of change with time is

dL         
 L   vi  pi   ri  Fi  0   ri  Fi E   ri  F ji . (1.5)
dt i i ij

The total external torque is naturally defined as


  
   ri  Fi E , (1.6)
i

6
so we might ask if the last term vanishes due the Third Law, which permits us to rewrite
 1  
 
F ji  F ji  Fij . Then the last term becomes
2
  1   1  
ij i ji 2 ij i ji 2 ij ri  Fij
r  F  r  F 

1   1  
 
2 ij
ri  F ji   r j  F ji
2 ij
1   
 
2 ij
(ri  r j )  F ji . (1.7)

This is not automatically zero, but vanishes if one assumes a stronger form of the Third Law,
namely that the action and reaction forces between two particles acts along the line of separation
of the particles. If the force law is independent of velocity and rotationally and translationally
symmetric, there is no other direction for it to point *.
Making the assumption that the strong form of Newton's Third Law holds, we see that the
rotational equation of motion is

dL    
 L   ri  Fi E   o . (1.8)
dt i

where  o is known as the torque (about the origin). It is readily verified that a displacement of
 
the origin changes the value of L and  o without altering the form of Eqn. (1.8).
The most notable property of the rotational equation of motion, Eqn. (1.8), is the fact that
internal forces do not contribute directly to the torque. Note, however, that the torque in Eqn.

(1.8) depends on the values of ri whose time variations depend on the internal forces. Therefore,

the torque is indirectly dependent on the internal forces. Of course, for an isolated system the
torque vanishes for arbitrary internal forces. Hence the angular momentum of an isolated system
is conserved.

_____________________________________________________________________________
*For spinning particles and magnetic forces the argument is not so simple in fact electromagnetic forces between
moving charged particles are really only correctly viewed in a context in which the system includes not only the
particles but also the fields themselves.

7
It is usually desirable to separate the total angular momentum into its external and internal parts,
for the parts satisfy independent equations of motion. Time variation of the external (orbital)
angular momentum is determined by the CM Eqn. (1.4); thus,
   
d
dt
 
MR  R  MR  R  R  F E

On the other hand


 d      
L
dt
 
MR  R    R  F E   .
    
Where,    mi ri  ri    ri  p i
i i

Substituting this, we get the equation of motion for the internal angular momentum:
      
 
   ri  R  Fi E   ri  Fi E   . (1.9)
i i

In contrast to Eqn. (1.8), this equation is independent of the origin, or rather, it differs from Eqn.
(1.8) by a shift of the arbitrary fixed origin to an origin intrinsic to the system, the center of mass.
The angular momentum equation of motion Eqn. (1.9) is the second major result of general
many-particle systems theory. Let us call it the angular momentum theorem. To make use of this
theorem, however, we need further results relating the angular momentum to kinematical
variables of the system.
The conservation laws are very useful because they permit algebraic solution for part of the
1 2 
velocity. Taking a single particle as an example, if E  mv  U (r ) is conserved, the speed
2
  
v (t ) is determined at all times (as a function of r ) by one arbitrary constant E. Similarly if L is
 
conserved, the components of v which are perpendicular to r are determined in terms of the
 
fixed constant L . With both conserved, v is completely determined except for the sign of the
radial component.

1.2 The Method of Lagrange

Lagrange developed a systematic method for describing a system in terms of an arbitrary set of
variables.

8
In the Newtonian approach, an N-particle system is described by specifying the position
 
ri  ri (t ) of each particle as a function of time. To describe the system, instead, in terms of some

set of scalar variables q  q (t );   1,2,...n , we must express the positions as functions of the

new variables,
 
ri  ri (q1 , q 2 ,...q n ; t ) (1.10)

where i  1,2,...N . The scalar variables q are called generalized coordinates.


  
The set of particle positions r1 , r2 ,...rN  is called a configuration of the system. Since each of the

N position vectors ri is a vector in a 3-dimensional space, it takes n  3N generalized coordinates

q to specify all possible configurations of an N-particle system. Therefore, the q are coordinates of a
point q=q(t) in a 3N-dimensional space. This space is called configuration space. The motion of
an entire system is thus described by a single trajectory q=q(t) in the 3N-dimensional
 
configuration space instead of a set of N trajectories ri  ri (t ) in the 3-dimensional position

space. This helps us apply our intuition and knowledge of the dynamics of a single particle to the
dynamics of a many-particle system. It is an important conceptual advantage of Lagrange’s
method.

It may happen that the position variables ri are related by holonomic constraints specified by K
scalar equations
  
 J (r1 , r2 ,...rN ; t )  0 (1.11)

where J  1,2,..., K . The set of K constraints confines the system to a (3 N  K ) -dimensional


surface. Consequently, we can use the equations of constraint to eliminate K variables and
specify the system by n  3 N  K independent generalized coordinates q1 , q 2 ,..., q n . Such

independent coordinates are sometimes called degrees of freedom, so n  3 N  K is the number


of degrees of freedom of the system.
Our problem now is to convert the Newtonian equations of motion in position space to an
equation of motion for the system in configuration space. Newton’s equation for the ith particle
can be put in the form
   
mi ri   riU  f i  N i , (1.12)

9
  
where U  U (r1 , r2 ,...rN , t ) is the potential for conservative forces,
        
f i  f i (r1 , r2 ,...rN ; r1 , r2 ,...rN , t )  0 is the force function for non-conservative forces, and N i is
the resultant force of constraint.

The J-th equation of constraint (1.11) determines a constraining force  J  ri J , which can be
interpreted as the force required to keep the i-th particle on the J-th surface of constraint. The
resultant constraining force is therefore
    
N i    J  ri J   ri    J  J . (1.13)
J  J 
  
The equations of constraint  J (r1 , r2 ,...rN ; t )  0 are among the givens of the problem, however

the scalars  J are among the unknowns and must be obtained from the solution if the

constraining force is to be found.


Since the q are taken to be variables independent of the constraint equations, we must have

 q  J  0 , which by application of the chain rule, gives us K equations

 J  
 q  J 
q
 
   q ri   ri J  0. (1.14)
i

These equations can be used to eliminate the constraining forces from the equations of motion;
for
   
N   r 
  r 
 i q i  J  q i   ri J  0,
  
i J i

hence, from Eqn. (1.12) we get ;


 
 r   
 
i
m i i xiU  f i   q ri  0 , (1.15)

for   1,2,....n  3N  K .
The n equations, Eqn. (1.15), can be re-expressed as equations of motion for the by employing
the q chain rule for differentiation. Thus, from Eqn. (1.10) we obtain
  
ri   q  q ri   t ri . (1.16)

Differentiating this, we establish


 
 q ri   q ri ,

10
 d 
 q ri 
dt

 q ri  .

Hence, the first term in Eqn. (1.15) can be brought into the form
  d    d  
 m r   q r     dt m r   q r   m r   dt  q r 
i i i i i i i i i
i i  
d 1  1 
    q  mi ri 2    q  mi ri 2  
i  dt 2  2 
d
 
  T   q T ,
dt q
 (1.17)

1
where T   mi ri 2 is the total kinetic energy of the system.
i 2

By substitution, the potential can be expressed as a function of the generalized coordinates


 
U r1 q1 , q 2 ,..., q n ; t ,....rN q1 , q 2 ,..., q n ; t   U q1 , q 2 ,..., q n ; t  . (1.18)

Hence, the second term in (1.15) becomes


 
 r 
 q i   xiU   q U .  (1.19)
i

This can be combined with the first term (1.17), giving


   d
 m r   U   q r   dt  q L   q L ,
i
i i xi i (1.20)

where we have introduced a new function


L  T  U  Lq1 ,...., q n ; q1 ,..., q n ; t  (1.21)

is the Lagrangian of the system.


For the third term in (1.15) we introduce the notation
 
F  F q1 ,...., q n ; q1 ,..., q n ; t    f i   q ri .   (1.22)
i

The quantity F is called the q -component of the generalized force. The right side of Eqn.

(1.22) shows that F can be interpreted as the component of force on the system in the

“direction” of a change in q this “direction” is a direction in configuration space rather than

position space. Using Eqns. (1.20) and (1.22), we get Eqn. (1.15) finally in the form
d

  L   q L  F
dt q
 (1.23)

11
where   1,2,..., n . These are called Lagrange’s equations for the system.
These are the desired equations of motion for the system in terms of generalized coordinates.
Before Lagrange’s equations can be used, the Lagrangian and the generalized force must be
expressed in terms of the generalized coordinates. For the potential this is done by simple
substitution, as in Eqn. (1.18). To do it for the kinetic energy, we must use the chain rule; thus,
from
2
1   
T     q  q ri   t ri  ,
2 i  
we get the kinetic energy T  T q1 , q 2 ,..., q n ; q1 , q 2 ,..., q n ; t  in the form

1
T  m q q   
2  , 
B q  C , (1.24)

where,
 
  
m   mi  q ri   q ri  m q1 ,..., q n ; t  , (1.25a)
i

 
 
B   mi  q ri   t ri   B q1 ,..., q n ; t  , (1.25b)
i

1  2
 mi  t ri   C q1 ,..., q n ; t  ,
C (1.25c)
2 i
 
The explicit time-dependence of ri  ri q1 ,..., q n ; t  results only from time dependent constraints.
Consequently, for time independent constraints (or no constraints at all) the kinetic energy
assumes the form
1
T  m q1 ,.., qn q q 
2  ,
(1.26)

where the m are to be obtained from (1.25a).

By deriving Lagrange’s equations, we have carried out once and for all the steps required to
introduce any set of coordinates into Newton’s equations and eliminate the holonomic forces of
constraint. From now on we can avoid those steps by constructing Lagrange’s equations
straightaway. This is Lagrange’s method.
Lagrange’s Method can be summarized as a series of steps for attacking a given dynamical
problem:

12
Step I

Express any holonomic constraints in parametric form by determining the particle positions ri as

explicit functions ri q1 ,..., q n ; t  of an appropriate set of independent generalized coordinates.

Diagrams are often a valuable guide to the selection of coordinates. Sometimes it is best to begin
with dependent coordinates and then eliminates some of them by applying constraints in the
nonparametric form Eqn. (1.12). The best set of coordinates is usually determined by symmetry
properties of the potential energy function, but it may be difficult to find.
Step II
Express the Lagrangian and, if needed, the generalized force as explicit functions of the
generalized coordinates q and their velocities q .

Step III
Solve Lagrange’s equations. The equations may be quite complicated even for fairly simple
problems. No single mathematical method suffices to handle all problems.
____________________________________________________________________________
Example 1.1
A particle of mass m is suspended from a flexible string of length  in a uniform gravitational
field. While hanging motionless in equilibrium, it is struck a horizontal blow resulting in an
initial angular velocity  o . Treating the system as one with two degrees of freedom and a

constraint, answer the following:


(a) Compute the Lagrangian, the equation of constraint, and the equation of motion.
(b) Compute the tension in the string as a function of angle 
Solution:
(a) The Lagrangian is
1
L
2
 
m r 2  r 2 2  mgr cos  .

The constraint is r   . The equations of motion are


mr  mr 2  mg cos   
mr 2  2mrr  mg sin   0
(b) Energy is conserved, hence

13
1 2 2 1
m   mg cos   m 2o2  mg cos  o .
2 2
We take  o  0 and o   o . Thus

 2   o2  2 2 (1  cos  ) ,

where   g  . Substituting this into the equation for  , we obtain

 2 
  mg 2  3 cos   o2  .
  

_____________________________________________________________________________

1.3 Conserved Quantities

Consider a Lagrangian Lq a , q a  that depends on a number of generalized coordinates q a and a

number of generalized velocities q a . The quantities

L
pa  (1.27)
q a
feature prominently in the EL equations, which can be written in the form
L
p a  (1.28)
q a

The quantities p a are the generalized momenta of the mechanical system. There is one

generalized momentum p a for each generalized coordinate q a .

The generalized momenta can represent either a component of the linear-momentum vector or a
component of the angular-momentum vector. Generally speaking, whenever q a represents a

linear variable the corresponding p a will be a linear momentum; and whenever q a represents an

angular variable its corresponding p a will be an angular momentum. Consider, for example, the

Lagrangian of a free particle in cylindrical coordinates


1
L
2
 
m  2   2 2  z 2 .

The generalized momenta are


L
p   m ,


14
L
p   m 2 ,


L
pz   mz .
z
In the case of p  and p z we clearly have quantities that represent components of a linear

momentum vector. But the case of p is different. Here we have p  m ( ) , and this clearly

represents the component of an angular-momentum vector.


Suppose now that a Lagrangian Lq1 , q1 , q 2 , q 2 ,.... happens not to depend explicitly on one of its
generalized coordinates, say q k . Then

L
0 (1.29)
q k

and it follows from the EL equation for q k that

dp k
 0, (1.30)
dt
where p k  L q k is the generalized momentum associated with the coordinate q k .This

equation states that p k is a constant of the motion, and we have established the following
theorem:
Whenever the Lagrangian of a mechanical system does not depend explicitly on a
generalized coordinate q k , the corresponding generalized momentum p k  L q k is a

constant of the motion.


A coordinate q k that does not appear in L is sometimes called a cyclic coordinate. A Lagrangian

may contain any number of cyclic coordinates.


As an example consider the following Lagrangian, again in cylindrical coordinates,
1
L
2
 
m  2   2 2  z 2  U (  ) .

Here it is assumed that the potential energy U depends only on ρ; the mechanical system is
cylindrically symmetric. This implies that  and z are cyclic coordinates and that p  m 2

and p z  mz are constants of the motion.

15
This theorem on cyclic coordinates and conserved quantities is extremely important and very
useful. To find all the constants of the motion is usually a key step during the integration of the
equations of motion, and the theorem provides a very efficient algorithm to identify at least some
of them.

1.3.1 Conservation of Linear Momentum

As a very elementary example, consider a particle under a force given by a potential which
depends only on y and z, but not x. Then
1
L
2
 
m x 2  y 2  z 2  U ( y, z )

is independent of x, x is an ignorable coordinate and


L
px   mx
x
is conserved. This is no surprise, of course, because the force is F   U and Fx   U x  0 .
Note that, using the definition of the generalized momenta
L
pk  , (1.31)
q k
Lagrange's equation can be written as
d L T U
pk    .
dt q k q k q k
Only the last term enters the definition of the generalized force, so if the kinetic energy depends
on the coordinates, as will often be the case, it is not true that dp k dt  Qk . In that sense we

might say that the generalized momentum and the generalized force have not been defined
consistently.

1.3.2 Conservation of Angular Momentum

As a second example of a system with an ignorable coordinate, consider an axially symmetric


system described with inertial polar coordinates (r , , z ) , with z along the symmetry axis. The
1 2 1 2 2 1 2
kinetic energy of the system is T  mr  mr   mz . The potential is independent of  ,
2 2 2
because otherwise the system would not be symmetric about the z-axis, so the Lagrangian

16
1 2 1 2 2 1 2
L mr  mr   mz  U (r , z ) (1.32)
2 2 2
does not depend on  , which is therefore an ignorable coordinate, and
L
p   mr 2  Constant. (1.33)

We see that the conserved momentum p is in fact the z-component of the angular momentum,

and is conserved because the axially symmetric potential can exert no torque in the z-direction:
   U

 z   r  U 
z   r U  
  r 2

 0.

Finally, consider a particle in a spherically symmetric potential in spherical coordinates. We


know in classical mechanics I that the kinetic energy in spherical coordinates is
1 2 1 2 2 1 2
T mr  mr   mr sin 2  2 , so the Lagrangian with a spherically symmetric potential
2 2 2
is
1 2 1 2 2 1 2
L mr  mr   mr sin 2  2  U (r ) .
2 2 2
Again,  is an ignorable coordinate and the conjugate momentum p is conserved. Note,

however, that even though the potential is independent of  as well as  ,  does appear
undifferentiated in the Lagrangian, and it is not an ignorable coordinate, nor is p conserved.

1.3.3 Conservation of Energy

We may ask what happens to the Lagrangian along the path of the motion.
dL L dqi L dq i L
  
dt i q i dt i q  i dt t
L L L
 qi   qi  (1.34)
i qi i qi t
In the first term the first factor is
d L
dt q i
by the equations of motion, so

dL d  L  L
   q i   .
dt dt  i q i  t

17
Or, rearranging this expression gives

d  L  L
  q i  L    (1.35)
dt  i q i  t

We expect energy conservation when the potential is time invariant and there is not time
dependence in the constraints, i.e. when L dt  0 , so we rewrite this in terms of
L
H (q, q, t )   q i  L   q i p i  L . (1.36)
i q i i

Then for the actual motion of the system,


dH L
 . (1.37)
dt t
If L dt  0 , H is conserved.
H is essentially the Hamiltonian, although strictly speaking that name is reserved for the
function H (q, p, t ) on extended phase space rather than the function with arguments (q, q , t ) .

1.4 Two-Body Problem in Central Force

Consider two particles interacting via a potential U (r1 , r2 )  U ( r1  r2 ) . Such a potential, which

depends only on the relative distance between the particles, is called a central potential. The
Lagrangian of this system is then
1 1
L  T U  m1 r12  m1r22  U ( r1  r2 ). (1.38)
2 2

1.4.1 Center-of-Mass (CM) and Relative Coordinates

The two-body central force problem may always be reduced to two independent one-body
 
problems, by transforming to center-of-mass ( R ) and relative ( r ) coordinates, (see Figure 2) viz
 m1r1  m 2 r2   m2 
R r1  R  r
m1  m2 m1  m2

     m1 
r  r1  r2 r2  R  r (1.39)
m1  m2
We then have

18
1 1 1  1 
L m1r12  m1r22  U ( r1  r2 )  MR 2  r 2  U (r )
2 2 2 2

 
Figure 1.1. Center-of-mass ( R ) and relative ( r ) coordinates.

where
M  m1  m2 (total mass)

m1m 2
 (reduced mass) (1.40)
m1  m2
Solution to the CM problem
 
We have L R  0 , which gives Rd  0 and hence
  
R(t )  R(0)  R(0)t .
Thus, the CM problem is trivial. The center-of-mass moves at constant velocity.
Solution to the relative coordinate problem
    
Angular momentum conservation: We have that   r  p  r  r is a constant of the motion.
 
This means that the motion r (t ) is confined to a plane perpendicular to  . It is convenient to
 
adopt two-dimensional polar coordinates (r ,  ) . The magnitude of  is

  r 2  2A (1.41)


1 2
where dA  r d is the differential element of area subtended relative to the force center.
2
The relative coordinate vector for a central force problem subtends equal areas in equal times.
This is known as Kepler’s Second Law.
Energy conservation: The equation of motion for the relative coordinate is

19
d  L  L U
   r   .
dt  r  r r

Taking the dot product with r , we have
U
0  r  r   r
r
d 1 2  dE
  r  U (r )  (1.42)
dt  2  dt
Thus, the relative coordinate contribution to the total energy is itself conserved. The total
1  2
energy is of course E tot  E  MR .
2
  
Since  is conserved, and since r    0 , all motion is confined to a plane perpendicular to

 . Choosing coordinates such that zˆ  ̂ , we have
1 2 1 2
E r  U (r )  r 2   U (r )
2 2 2r 2
1 2
 r  U eff (r ) (1.43)
2
2
U eff (r )   U (r ) (1.44)
2 r 2
Integration of the Equations of Motion, Step I: The second order equation for r (t ) is

dE 2 dU (r ) dU eff (r )
0  r  3
  . (1.45)
dt r dr dr
However, conservation of energy reduces this to a first order equation, via

dr
2 2
r   E  U eff (r )   dt   . (1.46)
 2
E  U (r )
2 r 2
This gives t(r), which must be inverted to obtain r(t). Note that a constant of integration also
appears at this stage – call it r0  r (t  0) .

Integration of the Equations of Motion, Step II: After finding r(t) one can integrate
to find  (t ) using the conservation of  :

20
 2 
 2  d  dt (1.47)
r r 2 (t )
This gives  (t ) , and introduces another constant of integration – call it 0   (t  0) .

Confined to the plane perpendicular to  , the relative coordinate vector has two degrees of
freedom. The equations of motion are second order in time, leading to four constants of
integration. Our four constants are E , , ro , and o .

The original problem involves two particles, hence six positions and six velocities, making for 12
 
initial conditions. Six constants are associated with the CM system: R (0) and R(0) . The six
remaining constants associated with the relative coordinate system are  (three components),
E , ro , and o .

Geometric Equation of the Orbit: From   r 2 , we have


d  d
 2 , (1.48)
dt r d
leading to
2
d 2 r 2  dr  r 4
    F (r )  r (1.49)
d 2 r  d  2

where F (r )   dU dr is the magnitude of the central force. This second order equation may be
reduced to a first order one using energy conservation:
1 2
E r  U eff (r )
2
2
2  dr 
    U eff (r ) . (1.50)
2 r 4  d 
Thus,
 dr
d    , (1.51)
2 r 2
E  U eff (r )

which can be integrated to yield  (r ) , and then inverted to yield r ( ) . Note that only one
integration need be performed to obtain the geometric shape of the orbit, while two integrations –
one for r(t) and one for  (t ) – must be performed to obtain the full motion of the system.

21
The sign ambiguity from the square root is only because r may be increasing or decreasing, but
time, and usually   , are always increasing.

1.5 The Kepler Problem

1.5.1 Calculation of r ( )

Our goal in this subsection will be to obtain r as a function of  , for a gravitational potential.
Let’s assume that we’re dealing with the earth and the sun, with masses m and Ms, respectively.
The gravitational potential energy of the earth-sun system is

U (r )   , where   GM S m . (1.52)
r
for simplicity, let us consider the sun to be at the origin of our coordinate system (since Ms >> m,
this is approximately true for the earth-sun system). The equation of motion that relates r and 
(Eqn. 1.49) becomes
2
 1 dr  2mE 1 2m
 2   2  2  (1.53)
 r d   r r 2
The reader may find by himself/herself to show that the solution to Eqn. (1.53) is given by

1 m  2 E 2 
 2 1  1  cos(   0 )  (1.54)
r   m 2 
where  0 is a constant of integration. It is customary to pick the axes so that  0 = 0, so we’ll

drop the  0 from here on. Therefore, Eqn. (1.54) takes a simpler form given by

1 m
 2 (1   cos  ) , (1.55)
r 
where

2 E 2
  1 (1.56)
m 2
is the eccentricity of the particle’s motion. We will see shortly exactly what  signifies.
At any rate, we just discovered the basic motion of objects under the influence of gravity. This
completes the derivation of r ( ) r(µ) for the gravitational potential, U (r )  1 r .

22
What are the limits on r in Eqn. (1.55). The minimum value of r is obtained when the right
 
hand-side reaches its maximum value, which is m  2 (1   ) . Therefore,

2
rmin  . (1.57)
m (1   )
What is the maximum value of r ? The answer depends on whether  is greater than or less than
1. If   1 (which correspond to the circular or elliptical orbits) then the minimum values of the
 
right-hand of Eqn. (1.55) is m  2 (1   ) . Therefore,

2
rmax  (if  < 1). (1.58)
m (1   )
If   1 (which corresponds to parabolic or hyperbolic orbits), then the right-hand side of Eqn.
(1.55) can become zero (when cos    1  ). Therefore,
rmax   (if   1 ). (1.59)

1.5.2 The Orbits

Let's examine in detail the various cases for  .


1. Circle (  = 0)
If   0 , then Eqn. (1.56) says that E   m 2 2 2 . The negative E simply means that the
potential energy is more negative than the kinetic energy is positive. The particle is trapped in
the potential well. Eqns. (1.57) and (1.58) give rmin  rmax   2 m . Therefore, the particle

moves in a circular orbit with radius  2 m . Equivalently, Eqn. (1.55) says that r is independent

of  . For a given  , the energy  m 2 2 2 is the minimum value that the E can take. (To
achieve the minimum, we certainly want r  0 . And you can show that minimizing the effective
potential,  2 2mr 2   r , yields this value for E.) If we plot Ueff (r), the particle is trapped at the
bottom of the potential well, so it has no motion in the r direction.
2. Ellipse (0 <  < 1)
If 0    1 , then Eqn. (1.56) says that  m 2 2 2  E  0 . Eqns. (1.57) and (1.58) give rmin and

rmax . If we plot Ueff (r), we see that the particle oscillates between rmin and rmax . The energy is

negative, so the particle is trapped in the potential well.

23
3. Parabola (  = 1)
If   1 , then Eqn. (1.56) says that E = 0. This value of E implies that the particle barely makes it
out to infinity (its speed approaches zero as r   ). Eqn. (1.57) gives rmin   2 2m , and Eqn.

(1.59) gives rmax   . If we plot Ueff (r), we see that the particle does not oscillate back and forth
in the r-direction. It moves inward (or possibly not, if it was initially moving outward), turns
around at rmin   2 2m , and then heads out to infinity forever.
4. Hyperbola (  > 1)
If   1 , then Eqn. (1.56) says that E > 0. This value of E implies that the particle makes it out to
infinity with energy to spare. (The potential goes to zero as r   , so the particle’s speed
approaches the nonzero value 2 E / m as r   .) Eqn. (1.57) gives rmin , and Eqn. (1.59)

gives rmax   .

If we plot Ueff (r), we see that, as in the parabola case, the particle does not oscillate back and
forth in the r-direction. It moves inward (or possibly not, if it was initially moving outward),
turns around at rmin , and then heads out to infinity forever.

1.5.3 Proof of Conic Orbits

Let’s now prove that Eqn. (1.55) does indeed describe the conic sections stated above. We will
also show that the origin (the source of the potential) is a focus of the conic section.
In what follows, we will find it easier to work with Cartesian coordinates. For convenience, let
2
k . (1.60)
m
Multiplying Eqn. (1.55) through by kr , and using cos   x r gives
k  r  x . (1.61)
Solving for r and squaring yields
x 2  y 2  k 2  2kx   2 x 2 . (1.62)
Let’s look at the various cases for  . We will invoke without proof various facts about conic
sections (focal lengths, etc.).

24
1. Circle (  = 0)
In this case, Eqn. (1.62) becomes x 2  y 2  k 2 . So we have a circle of radius k   2 m , with
its center at the origin (see Figure 1.2).
2. Ellipse (0 <  < 1)
In this case, Eqn. (1.62) may be written as (after completing the square for the x terms, and
expending some effort)
2
 k 
x 
 1 2  y2
 2  1,
a2 b
k k
where a , and b (1.63)
1 2 1 2
This is the equation for an ellipse with its center located at (  k (1   2 ) , 0). The semi-major

and semi-minor axes a and b, respectively, and the focal length is c  a 2  b 2  k (1   2 ) .


Therefore, one focus is located at the origin (see Figure 1.3). Note that c a equals the
eccentricity  .

Figure 1.2. Figure 1.3.

3. Parabola (  = 1)
In this case, Eqn. (1.62) becomes y 2  k 2  2kx . This may be written as y 2  2k  x  k 2 . This

is the equation for a parabola with vertex at ( k 2 ,0 ) and focal length k 2 . So we have a parabola
with its focus located at the origin (see Figure. 1.4).
4. Hyperbola (  > 1)
In this case, Eqn. (1.62) may be written (after completing the square for the x terms)

25
2
 k 
x 2 
  1 y2
  1,
a2 b2
k k
where a 2
, and b . (1.64)
 1  2 1
This is the equation for a hyperbola with its center (defined to be the intersection of the

asymptotes) located at ( k ( 2  1) , 0). The focal length is c  a 2  b 2  k ( 2  1) .


Therefore, the focus is located at the origin (see Figure 1.5). Note that c a equals the
eccentricity  .

Figure 1.4. Figure 1.5.

1.5.4 Kepler's Laws

Kepler’s laws assume that the sun is massive enough so that its position is essentially fixed in
space. This is a very good approximation, but the following section on reduced mass will show
how to modify them and solve things exactly.
First Law: The planets move in elliptical orbits with the sun at one focus.
The proof of this first law is left as an exercise. Of course, there are undoubtedly objects flying
past the sun in hyperbolic orbits. But we don’t call these things planets, because we never see the
same one twice.
Second Law: The radius vector to a planet sweeps out area at a rate that is independent
of its position in the orbit.

26
This law is nothing other than the statement of conservation of angular momentum. The area
swept out by the radius vector during a short period of time is dA  r (rd ) / 2 , because rd is

the base of the thin triangle in Figure 1.6. Therefore, we have (using   mr 2 )
dA 1 2  
 r  , (1.65)
dt 2 2m
which is constant, because  is constant for a central force.

Figure 1.6.

Third Law: The square of the period of an orbit, T, is proportional to the cube of the
semimajor-axis length, a. More precisely,
4 2 ma 3 4 2 a 3
T2   , (1.66)
 GM S
where Ms is the mass of the sun. Note that the planet’s mass, m, does not appear in this equation.

Figure 1.7.

27
1.6 Elastic Collisions of Two Particles

For the next few sections, we apply the conservation laws to the interaction of two particles.
When two particles interact, the motion of one particle relative to the other is governed by the
force law that describes the interaction. This interaction may result from actual contact, as in the
collision of two billiard balls, or the interaction may take place through the intermediary of a
force field. For example, or an   particle may be scattered by the electric field of an atomic
nucleus. We demonstrated in the previous section that once the force law is known, the two-body
problem can be completely solved. But even if the force of interaction between two particles is
not known, a great deal can still be learned about the relative motion by using only the results of
the conservation of momentum and energy. Thus, if the initial state of the system is known (i.e.,
if the velocity vector of each of the particles is specified), the conservation laws allow us to
obtain information regarding the velocity vectors in the final state.
On the basis of the conservation theorems alone, it is not possible to predict, for example, the
angle between the initial and final velocity vectors of one of the particles; knowledge of the force
law is required for such details. In this section and the next, we derive those relationships that
require only the conservation of momentum and energy. Then, we examine the features of the
collision process, which demand that the force law be specified. We limit our discussion
primarily to elastic collisions, because the essential features of two-particle kinematics are
adequately demonstrated by elastic collisions. The results obtained under the assumption only of
momentum and energy conservation are valid (in the non-relativistic velocity region) even for
quantum mechanical systems, because these conservation theorems are applicable to quantum as
well as to classical systems.
The description of many physical processes is considerably simplified if one chooses coordinate
systems at rest with respect to the system's center of mass. In the problem we now discuss—the
elastic collision of two particles—the usual situation is one in which the collision is between a
moving particle and a particle at rest. On the other hand, although it is indeed simpler to describe
the effects of the collision in a coordinate system in which the center of mass is at rest, the actual
measurements are made in the laboratory coordinate system in which the observer is at rest. In
this system, one of the particles is normally moving, and the struck particle is normally at rest.
We here refer to these two coordinate systems simply as the CM and the LAB systems.

28
We wish to take advantage of the simplifications that result by describing an elastic collision in
the CM system. It is therefore necessary to derive the equations connecting the CM and LAB
systems.
We use the following notation:
m1  moving 
Mass of the   particle
m2  struck 
In general, primed quantities refer to the CM system:

u1  Initial 
  Velocity of m1 in the LAB system
v1  Final 

u1  Initial 
  Velocity of m1 in the CM system
v1  Final 
    
and similarly for u 2 , v 2 , u 2 , and v 2 (but u 2  0 ):

To  LAB 
Total initial kinetic energy in   system
To  CM 
T1  LAB 
Final kinetic energy of m1 in   system
T1  CM 
and similarly for T2 and T2 ,

V  velocity of the center of mass in the LAB system
  angle through which m1 is deflected in the LAB system
  angle through which m2 is deflected in the LAB system
  angle through which m1 and m2 are deflected in the CM system

Figure 1.8 below illustrates the geometry of an elastic collision in both the LAB and CM
systems. The final state in the LAB and CM systems for the scattered particle m1 may be
conveniently summarized by the diagrams in Figure 1.9. We can interpret these diagrams in the
 
following manner. To the velocity V of the CM, we can add the final CM velocity v1 of the
scattered particle. Depending on the angle  at which the scattering takes place, the possible
 
vectors v1 lie on the circle of radius v1 whose center is at the terminus of the vector V . The LAB

29

velocity v1 and LAB scattering angle  are then obtained by connecting the point of origin of
 
V with the terminus of v1 .
  
For V  v1 , only one possible relationship exists between V , v1 , v1 , and  (see Figure 1.9a). But
 
if V  v1 , then for every set V , v1 , there exists two possible scattering angles and laboratory
 
velocities: v1,b ,  b , and v1, f ,  f (see Figure 1.9b), where the designations b and f stand for

backward and forward. This situation results from the fact that if the final CM velocity v1 is

insufficient to overcome the velocity V of the center of mass, then, even if m1 is scattered into

the backward direction in the CM system (    2 ), the particle will appear at a forward angle

in the LAB system (    2 ). Thus, for V  v1 , the velocity v1 in the LAB system is a double-
 
valued function of v1 . In an experiment, we usually measure , not the velocity vector v1 , so that
a single value of  can correspond to two different values of  . Note, however, that a
  
specification of the vectors V and v1 always leads to a unique combination v1 ,  ; but a
 
specification of V and only the direction of v1 (i.e.,  ) allows the possibility of two final
 
vectors, v1,b and v1, f , if V  v1 .

Having given a qualitative description of the scattering process, we now obtain some of the
equations relating the various quantities.
According to the definition of the center of mass, we have
  
m1r1  m 2 r2  MR (1.67)
Differentiating with respect to the time, we find
  
m1u1  m2 u 2  MV (1.68)

But u 2  0 and M  m1  m2 , the center of mass must therefore be moving (in the LAB system)

toward m2 with a velocity


 m u
V  1 1 (1.69)
M
By the same reasoning, because m2 is initially at rest, the initial CM speed of m2 must just equal
V:

m1u1
u 2  V  (1.70)
M
30
Note, however, that the motion and the velocities are opposite in direction and that
 
vectorially u 2  V .
The great advantage of using the CM coordinate system is because the total linear momentum in
such a system is zero, so that before the collision the particles move directly toward each other
and after the collision they move in exactly opposite directions. If the collision is elastic, as we
have specified, then the masses do not change, and the conservation of linear momentum and
kinetic energy is sufficient to provide that the CM speeds before and after collision are equal:
u1  v1 , u 2  v 2 (1.71)

Term u1 is the relative speed of the two particles in either the CM or the LAB system,

u1  u1  u 2 . We therefore have, for the final CM speeds.


m1u1
v 2  (1.72a)
m1  m2

m 2 u1
v1  u1  u 2  (1.72b)
m1  m2

(a) Initial condition (b) Initial condition

(c) Final condition (d) Final condition


Figure 1.8. Geometry and notations of an elastic collision in the LAB and CM systems,

(a) Initial condition with u 2  0 in the LAB system, (b) initial condition in the CM
system, (c) final condition in the LAB system, and (d) final condition in the CM system.
Note carefully the scattering angles.

31
V  v1 V  v1
(a) (b)
Figure 1.9. The final state of mass »ij for the elastic collision of two particles for the case
(a) V  v1 for which there is one trajectory and (b) V  v1 for which there are two
possible trajectories (b stands for backward and f for forward).

we have (see Figure 1.9a)


v1 sin   v1 sin  (1.73a)
and
v1 cos  V  v1 cos (1.73b)
Dividing Eqn. (1.73a) by (1.73b)
v1 sin  sin 
tan   (1.74)
v1 cos   V cos   V v1 

According to Eqns. (1. 69) and (1.72b), V v1 is given by

V m1
 (1.75)
v1 m2

Thus, the ratio m1 m 2 governs whether Figure 1.9a or Figure 1.9b describes the scattering
process:
Figure 1.9a: V  v1 , m1  m2
Figure 1.9b: V  v1 , m1  m 2
If we combine Eqns. (1.74) and (1.75) and write

32
sin 
tan  (1.76)
cos   m1 m2 

we see that if m1  m2 , the LAB and CM scattering angles are approximately equal; that is, the

particle m2 is but little affected by the collision with m1 and acts essentially as a fixed scattering
center. Thus,

  , m1  m2 (1.77)
However, if m1  m 2 , then
sin  
tan   tan
cos   1 2
and

  , m1  m2 (1.78)
2

and the LAB scattering angle is one half the CM scattering angle. Because the maximum value
of  is 180°, Eqn. (1.78) indicates that for m1  m 2 , there can be no scattering in the LAB
system at angles greater than 90°.
Let us now refer to Figure 1.8c and construct a diagram for the recoil particle m2 similar to
Figure 1.9a. The situation is illustrated in Figure , from which we find
v 2 sin   v 2 sin  (1.79a)

v 2 cos   V  v 2 cos  (1.79b)


Dividing Eqn. (1.79a) by Eqn. (1.79b), we have
v 2 sin  sin 
tan   
V  v 2 cos  V v 2   cos 

But, according to Eqns. (1.70) and (1.72a), V and v 2 are equal. Therefore,
sin  
tan    cot (1.80)
1  cos 2
which we may write as
  
tan   tan  
 2 2
Thus

33
2       (1.81)

Figure 1.10 The final state of recoil mass m2 in the elastic collision of two particles.

For particles with equal mass, m1  m 2 , we have   2 . Combining this result in Eqn. (1.81),
we have

   , m1  m2 (1.82)
2

Hence, the scattering of particles of equal mass always produces a final state in which the
velocity vectors of the particles are at right angles if one of the particles is initially at rest (see
Figure 1.11).

Figure 1.11. For the elastic scattering of two particles of equal mass ( m1  m 2 ) with
one of them initially at rest in the LAB system, the final velocities (trajectories)
of the two masses are at right angles to each other. Two such possibilities are shown.
____________________________________________________________________________
Example 1.2. What is the maximum angle that  can attain for the case V  v1 ? What is  max

for m1  m2 and m1  m 2 ?

34
Solution. For the case of max , Figure 1.9b becomes as shown in Figure 1.11. The angle between
 
v1 and v1 is 90° for  to be a maximum.
v1
sin  max  (1.83)
V
According to Eqn. (1.75), this is just
m2
sin  max 
m1
from which
m 
 max  sin 1  2  (1.84)
 m1 
For m1  m2 ,  max = 0 (no scattering), and for m1  m 2 ,  max = 90°. Generally, for m1  m 2 , no

scattering of m1 backward of 90° can occur.


______________________________________________________________________________

1.6.1 Kinematics of Elastic Collisions

Relationships involving the energies of the particles may be obtained as follows. First, we have
simply
1
To  m1u12 (1.85)
2
and, in the CM system.
1
To 
2

m1u1 2  m2 u 22 
which, on using Eqns. (1.72a) and (1.72b), becomes
1 m1m 2 2 m2
To  u1  To (1.86)
2 m1  m2 m1  m2

This result shows that the initial kinetic energy in the CM system To is always a fraction

m2 m1  m2   1 of the initial LAB energy. For the final CM energies, we find
2 2
1 1  m2  2  m2 
T1  m1v1 2  m1   u1    To (1.87)
2 2  m1  m2  m
 1  m 2 

and

35
2
1 1  m1  2 m1 m2
T2  m 2 v 2 2  m 2   u1  To (1.88)
2 2  m1  m2  m1  m 2

To obtain T1 in terms of To , we write

1
m1v12
T1 2 v12
  2 (1.89)
To 1
m1u1 u1
2

2
Referring to Figure 1.9a and using the cosine law, we can write
v1 2  v12  V 2  2v1V cos
or
T1 v12 v1 2 V 2 vV
 2  2  2  2 1 2 cos (1.90)
To u1 u1 u1 u1
From the previous definitions, we have
v1 m2 V m1
 and  (1.91)
u1 m1  m2 u1 m1  m2
The squares of these quantities give the desired expressions for the first two terms on the right-
hand side of Eqn. (1.90). To evaluate the third term, we write, using Eqn. (1.73a).
v1V  sin   V
2 2
cos  2 v1   2 cos (1.92)
u1  sin   u1
The quantity of v1 V u12 can be obtained from the product of the equations in
Eqn. (1.91), and using Eqn. (1.76), we have
sin  cos sin  m
  cos  1
sin  tan m2
so that

v1V 2m1m2  m 
2 2
cos   cos  1  (1.93)
u1 (m1  m2 ) 2  m2 

Substituting Eqns. (1.91) and (1.93) into Eqn. (1.90), we obtain


2 2
T1  m 2   m1  2m1m 2  m 
       cos  1 
To  m1  m2   m1  m2  (m1  m2 ) 2  m2 

which simplifies to

36
T1 2m1m 2
 1 (1  cos  ) (1.94a)
To (m1  m 2 ) 2

Similarly, we can also obtain the ratio T1 To in terms of the LAB scattering angle :
2
2   m2 
2 
T1 m 
 1
cos     sin  
2
(1.94b)
To (m1  m2 ) 2  m
 1 
 
where the plus (+) sign for the radical is to be taken unless m1  m 2 —in which case the result is
double-valued, and Eqn. (1.84) specifies the maximum value allowed for  .

The LAB energy of the recoil particle m2 can be calculated from


T2 T 4m1 m2
1 1  2
cos 2  ,   2 (1.95)
To To (m1  m 2 )

If m1  m 2 , we have the simple relation

T1
 cos 2  , m1  m2
To (1.96a)

with the restriction noted in the discussion following Eqn. (1.78) that   90 o . Also,

T2
 sin 2  , m1  m2 (1.96b)
To

Several further relationships are

m1T1
sin   sin  (1.97)
m 2T2

sin 2
tan  (1.98)
m1 m2   cos 2
sin 
tan  (1.99)
m1 m2   cos
_____________________________________________________________________________
Example 1.3. In a head-on elastic collision of two particles with masses m1 and m2 , the initial
  
velocities are u1 and u 2  u1 (  0) . If the initial kinetic energies of the two particles are equal

37
in the LAB system, find the conditions on u1 u 2 and m1 m 2 so that m1 will be at rest in the LAB
system after the collision. See Figure 1.12.
Solution. Because the initial kinetic energies are equal, we have
1 1 1
m1u12  m 2 u 22   2 m2 u12
2 2 2
or
m1
2 (1.100)
m2

If m1 is at rest after the collision, the conservation of energy requires


1 1 1
m1u12  m 2 u 22  m2 v 22
2 2 2
or
1
m1u12  m2 v 22 (1.101)
2
The conservation of linear momentum states that
   
m1u1  m 2 u 2  m1  m2 u1  m 2 v 2 (1.102)

Substituting v 2 from Eqn. (1.102) into Eqn. (1.101) gives
2
21  m  m 2  2
m u  m2  1
1 1
 u1
2  m2 
or
2
1 m 
m1  m 2  1    (1.103)
2  m2 

Figure 1.12. Example--. Velocities are indicated for two particles of different masses
in a head-on elastic collision before and after the collision.

Substituting m1 m2   2 from Eqn. (1.100) gives

38
2 2  ( 2   ) 2
with the result
  2  1  0.414
 2  0.172
so that
m1
  2  0.172
m2
and
u2
   0.414
u1
Because   0 , both particles are traveling in the same direction; the collision is shown in Figure
1.12.
_____________________________________________________________________________

1.6.2 Inelastic Collisions

When two particles interact, many results are possible, depending on the forces involved. In the
previous two sections, we were restricted to elastic collisions. But, in general, multi-particles
may be produced if large changes of energy are involved. For example, when a proton collides
with some nuclei, energy may be released. In addition, the proton may be absorbed, and the
collision may produce a neutron or alpha particle instead. All these possibilities are handled with
the same methods: conservation of energy and linear momentum. We continue to restrict our
considerations to the same particles in the final system as were considered in the initial system.
In general, the conservation of energy is
1 1 1 1
Q m1u12  m2 u 22  m1v12  m 2 v 22 (1.104)
2 2 2 2
where Q is called the Q-value and represents the energy loss or gain in the collision.
Q = 0: Elastic collision, kinetic energy is conserved
Q > 0: Exoergic collision, kinetic energy is gained
Q < 0: Endoergic collision, kinetic energy is lost

39
Figure 1.13. Direct head-on collision between two bodies indicating the initial
conditions, the collision, and the resulting situation.

An inelastic collision is an example of an endoergic collision. The kinetic energy may be


converted to mass-energy, as, for example, in a nuclear collision. Or it may be lost as heat
energy, as, for example, by frictional forces in a collision. The collisions of all macroscopic
bodies are endoergic (inelastic) to some degree. Two silly putty balls with equal masses and
speeds striking head-on may come to a complete stop, a totally inelastic collision. Even two
billiard balls colliding do not completely conserve kinetic energy; some small fraction of the
initial kinetic energy is converted to heat.
A measure of the inelasticity of two bodies colliding may be considered by referring to a direct
head-on collision (see Figure 1.12) in which no rotations are involved (translational kinetic
energy only). Newton found experimentally that the ratio of the relative initial velocities to the
relative final velocities was nearly constant for any two bodies. This ratio, called the coefficient
of restitution (  ), is defined by
v 2  v1
 (1.105)
u 2  u1

This is sometimes called Newton's rule. For a perfectly elastic collision,  = 1; and for a totally
inelastic collision,  = 0. Values of  have the limits 0 and 1.

40
Figure 1.14. An oblique collision between two bodies. For smooth surfaces, the velocity
components along the plane of contact bb' are hardly changed by the collision.

We must be careful when applying Eqn. (1.105) to oblique collisions, because Newton's rule
applies only to the velocity components along the normal (aa') to the plane of contact (bb')
between the two bodies, as shown in Figure 1.14. For smooth surfaces, the velocity components
along the plane of contact are hardly changed by the collision.
_____________________________________________________________________________
Exercise 1.1. For an elastic head-on collision described in the previous Sections, show that
 = 1. The mass m2 is initially at rest.
______________________________________________________________________
During a collision (elastic or inelastic), the forces involved may act over a very short period of
time and are called impulsive forces. A hammer striking a nail and two billiard balls colliding
are examples of impulsive forces. Newton's Second Law is still valid throughout the time period
t of the collision:
 d 
F  ( mv ) (1.106)
dt
After multiplying by dt and integrating, we have
t2  
 F dt  P (1.107)
t1

41

where t  t 2  t1 . Eqn. (1.107) defines the term impulse P . The impulse may be measured
experimentally by the change of momentum. An ideal impulse represented by no displacement
during the collision would be caused by infinite force acting during an infinitesimal time.

1.7 Motion of Systems with Variable Mass

We will now apply the conservation laws discussed in the middle of this chapter to some
particular situations. The conservation laws are applicable to any definite system of particles,
which may be chosen arbitrarily by including and excluding certain parts so long as it does not
exclude the forces acting on the chosen part of the system. Another restriction concerns the law
of conservation of kinetic and potential energy. It holds good so long as no mechanical energy is
converted into other forms of energy, such as heat produced by frictional forces, unless such
converted amounts are taken into account.
Rocket technology is based on the simplest principle of conservation of linear momentum. A
rocket is propelled in a forward direction by ejecting mass in a backward direction in the form of
gases resulting from the combustion of fuel. Thus the forward force on the rocket is the reaction
to the backward force of the ejected gases (burned-out fuel).
The two cases we examine are rocket motion in free space and the vertical ascent of rockets
under gravity. The first case requires an application of the conservation of linear momentum. The
second case requires a more complicated application of Newton's Second Law.

1.7.1 Rocket Motion in Free Space

We assume here that the rocket (space ship) moves under the influence of no external forces. We
choose a closed system in which Newton's Second Law can be applied. In outer space, the
motion of the space ship must depend entirely on its own energy. It moves by the reaction of
ejecting mass at high velocities. To conserve linear momentum, the space ship will have to move
in the opposite direction. The diagram of the space ship motion is shown in Figure 1.15. At some
time t, the instantaneous total mass of the space ship is m, and the instantaneous speed of the
space ship is v with respect to an inertial reference system. We assume that all motion is in the x
direction and eliminate the vector notation.

42
During a time interval dt, a positive mass dm is ejected from the rocket engine with a speed  u
with respect to the space ship. Immediately after the mass dm' is ejected, the space ship's mass
and speed are m  dm  and v  dv , respectively.
Initial momentum = mv (at time t) (1.108)
Final momentum = (m  dm )(v  dv)  dm(v  u ) (at time t  dt t) (1.109)
space ship less dm rocket exhaust dm


Figure 1.15. A rocket moves in free space at velocity v . In the time interval dt, a mass

dm is ejected from the rocket engine with velocity u with respect to the rocket ship.

Notice that the speed of the ejected mass dm with respect to the reference system is v  u . The
conservation of linear momentum requires that Eqns. (1.108) and (1.109) be equal. There are no
external forces ( Fext  0 ).
pinitial  p final
p(t )  p(t  dt ) (1.110)
mv  (m  dm)(v  dv)  dm (v  u )
mv  mv  mdv  vdm  dmdv  vdm   udm 
mdv  udm (1.111)
dm 
dv  u
m
where we have neglected the product of two differentials dmdv . We have considered dm being
a positive mass ejected from the space ship. The change in mass of the space ship itself is dm,
where

43
m   dm  (1.112)
and
dm
dv  u (1.113)
m
because dm must be negative. Let mo and vo be the initial mass and speed of the space ship,

respectively, and integrate Eqn. (1.113) to its final values m and v.


v m dm
 dv  u  m
v m
o o

m 
v  v o  u ln  o  (1.114)
 m

m 
v  vo  u ln  o  (1.115)
 m
The exhaust velocity u is assumed constant. Thus, to maximize the space ship's speed, we need
to maximize the exhaust velocity u and the ratio mo m .

Because the terminal speed is limited by the ratio mo m , engineers have constructed multistage

rockets. The minimum mass (less fuel) of the space ship is limited by structural material.
However, if the fuel container itself is jettisoned after its fuel has been burned, the mass of the
remaining space ship is even less. The space ship can contain two or more fuel containers, each
of which can be jettisoned.
For example, let
mo = Initial total mass of space ship
m1  m a  mb

ma = Mass of first-stage payload


mb = Mass of first-stage fuel containers, etc.
v1  terminal speed of first-stage of "burnout"
after all fuel is burned
m 
v1  vo  u ln  o  (1.116)
 m1 

44
At burnout, the terminal speed v1 of the first stage is reached, and the mass mb is released into

space. Next, the second-stage rocket ignites with the same exhaust velocity, and we have
ma = Initial total mass of space ship second stage
m2  mc  md

mc = Mass of second-stage payload


md = Mass of first-stage fuel containers, etc.
v1  Initial speed of second-stage
v 2  Terminal speed of second-stage at burnout

m 
v 2  v1  u ln  o  (1.117)
 m2 

m m 
v 2  v o  u ln  o a  (1.118)
 m1m2 
The product mo ma m1 m2 can be made much larger than just mo m . Multistage rockets are more

commonly used in ascent under gravity than in free space.


We have seen that the space ship is propelled as a result of the conservation of linear momentum.
But engineers and scientists like to refer to the force term as rocket "thrust." If we multiply Eqn.
(1.113) by m and divide by dt, we have
dv dm
m  u (1.119)
dt dt
Since the left side of this equation "appears" as Ota (force), the right side is called thrust:
dm
Thrust  u (1.120)
dt
Because dm dt is negative, the thrust is actually positive.

1.7.2 Vertical Ascent Under Gravity

The actual motion of a rocket attempting to leave Earth's gravitational field is quite complicated.
For analytical purposes, we begin by making several assumptions. The rocket will have only
vertical motion, with no horizontal component. We neglect air resistance and assume that the
acceleration of gravity is constant with height. We also assume that the bum rate of the fuel is
constant.

45
We can use the results of the previous case of rocket motion in free space, but we no longer
have Fext  0 . The geometry is shown in Figure 1.16. We again have dm' as positive,

with dm   dm  . The external force Fext is

Figure 1.16. A rocket in vertical ascent under Earth's gravity. Mass dm' is ejected from

the rocket engine with velocity u with respect to the rocket ship.

d
Fext  (mv)
dt
or
Fext dt  d (mv)  dp  p(t  dt )  p (t ) (1.121)

over a small differential time.


For the space ship system, we found the initial and final momenta in Eqns. (1.108) - (1.113). We
now use the results leading up to Eqn. (1.113) to obtain
p(t  dt )  p (t )  mdv  udm (1.122)

In free space, Fext  0 but in ascent, Fext   mg . Combining Eqns. (1.121) and (1.122) gives

Fext dt   mgdt  mdv  udm

 mg  mv  um (1.123)

46
Because the fuel burn rate is constant, let
dm
m    ,  0 (1.124)
dt
and Eqn. (1.123) becomes
  
dv    g  u dt
 m 
This equation, however, has three unknowns (v, m, t), so we use Eqn. (1.124) to eliminate time,
giving
g u
dv    dm (1.125)
 m 
Assume the initial and final values of the velocity to be 0 and v respectively and of the mass mo
and m respectively, so that
v m
g u
0 dv  m    m dm
o

g m 
v (mo  m)  u ln  o  (1.126)
  m
We can integrate Eqn. (1.124) to find the time:
m t

 dm    dt
mo 0 (1.127)
mo  m  t
Eqn. (1.126) becomes
m 
v   gt  u ln  o  (1.128)
 m
We could continue with Eqn. (1.126) and integrate once more to determine the height of the
rocket. Such integrations are tedious, and the problem is more easily handled by computer
methods. Even at burnout, the rocket will continue rising because it still has an upward velocity.
Eventually, with the preceding assumptions, the gravitational force will stop the rocket (because
we assumed a constant g not decreasing with height).
An interesting situation occurs if the exhaust velocity u is not sufficiently great to make v in Eqn.
(1.128) positive. In this case, the rocket would remain on the ground. This situation occurs
because of the limits of integration we assumed leading to Eqn. (1.126). We would need to burn

47
off sufficient fuel before the rocket thrust would lift it off the ground. Of course, rockets are not
designed this way; they are made to lift off as the rockets reach a full bum rate.
___________________________________________________________________________
Example 1.4. Consider the first stage of a Saturn V rocket used for the Apollo moon program.
The initial mass is 2.8  10 6 kg, and the mass of the first-stage fuel is 2.1  10 6 kg. Assume a
mean thrust of 37  10 6 N. The exhaust velocity is 2600 m/s.
Calculate the final speed of the first stage at burnout. Also calculate the vertical height at
burnout.
Solution. From the thrust (Eqn. 1.120), we can determine the fuel burn rate:

The final rocket mass is ( 2.8  10 6  2.1  10 6 kg) or 0.7  10 6 kg. We can determine the rocket
speed at burnout ( vb ) using Eqn. (1.126)

The time to burn out t b , from Eqn. (1.127), is

or about 2.5 min.


We may obtain the height at burnout y b as:

(1.129) (63)

Note that the actual height is only about two-thirds of this value.
_____________________________________________________________________________

48
In analyzing the motion of rigid bodies in the following sections, we also find it convenient to
use non-inertial reference frames and therefore make use of much of the development presented
here.

1.8 Check List

Check whether you can answer these questions or not.


 Can you differentiate between internal and external variables?
 Can you express the total angular momentum as a summation of internal and external
momenta?
 Can you discuss the method of Lagrange for describing a system in terms of an arbitrary
set of variables?
 Can you show the conservation of linear momentum, angular momentum and energy?
 Can you explain two body problem in central force?
 Can you state and prove Kepler’s laws?
 Can you show the kinematics of elastic collisions?
 Can you discuss the motion of a rocket?

1.9 Chapter 1 Exercises

1.1. A particle slides on the inside surface of a frictionless cone. The cone is fixed with its tip
on the ground and its axis vertical. The half-angle at the tip is  (see Figure. 1.17
below). Let r (t ) be the distance from the particle to the axis, and let  (t ) be the angle
around the cone. Find the equations of motion.
If the particle moves in a circle of radius ro , what is the frequency,  , of this motion? If

the particle is then perturbed slightly from this circular motion, what is the frequency,  ,
of the oscillations about the radius ro ? Under what conditions does    ?

Figure 1.17. For problem 1.1

49
1.2. A particle moves in a circular orbit in a force field given by
F (r )   k r 2 .
Show that, if k suddenly decreases to half its original value, the particle's orbit becomes
parabolic.

1.3. Investigate the motion of a particle repelled by a force center according to the law
F (r )  kr . Show that the orbit can only be hyperbolic.

1.4. An Earth satellite moves in an elliptical orbit with a period  , eccentricity  , and semi-


major axis a. Show that the maximum radial velocity of the satellite is 2a  1   2 . 
 
1.5. Two particles of mass m1 and m2 interact via an inverse-square force F21  k r r 3 ,
   
where r is the relative vector r  r2  r1 . There are no external forces and the initial
conditions in the CM frame are
 
r1 (0)  m2 d M ,0,0  , v1 (0)  0, vo m2 m1 ,0 
 
r2 (0)   m1d M ,0,0  , v2 (0)  0,vo ,0  ,

where d and vo are constants and M = m1 + m2.


(a) Show that the total energy E and angular momentum  are given by

 v2 k
E   o2  1 and   m2 vo d ,
 ve d
where

2m1 k
ve  .
m2 M d

(b) Show that the eccentricity e and semi-major axis a of an elliptical relative orbit
r ( ) are
d
  1  4 2 (1   2 ) and a ,
2(1   2 )
where   vo ve (< 1).

50
1.6. A rocket has an initial mass of 60,000 kg, and the speed of the burned exhaust gases is
6000 m/s. What should be the minimum mass flow rate of the gases to ensure life-off
from the surface of Earth?

1.7. A rocket of 60,000-kg mass is burning gases at a rate of 150 kg/s, and the speed of the
exhaust gases is 6000 m/s. If the rocket is fired vertically upward from the surface of
Earth, what will be its height and speed after 45,000 kg of fuel is expended? Graph the
velocity and height as a function of time.

51
CHAPTER 2
RIGID BODY DYNAMICS
Introduction

In previous chapter we have dealt with the motion of a particle or a system of particles under the
influence of external forces. In actual, everyday motions, we have to deal with rigid objects of
different shapes and sizes which may or may not reduce to equivalent point masses. We will
show now that to describe the motion of rigid bodies and apply conservation laws, we must
understand the full meanings of center of mass, moment of inertia, and radius of gyration.
Discussion of angular motion is complex in such cases, so simple cases of rotation about a fixed
axis and rotation about an axis passing through a fixed point will be discussed. Furthermore, in
this chapter we will assume that the bodies are rigid and do not deform, which is true in ideal
cases only.

2.1 Rotating Coordinate Systems

Let us consider two sets of coordinate axes. Let one set be the "fixed" or inertial axes, and let the
other be an arbitrary set that may be in motion with respect to the inertial system. We designate
these axes as the "fixed" and "rotating" axes, respectively. We use xi as coordinates in the fixed

system and xi as coordinates in the rotating system. If we choose some point P, as in Figure 2.1,

we have
  
r  R  r (2.1)
 
where r  is the radius vector of P in the fixed system and r is the radius vector of P in the

rotating system. The vector R locates the origin of the rotating system in the fixed system.
We may always represent an arbitrary infinitesimal displacement by a pure rotation about some
axis called the instantaneous axis of rotation. For example, the instantaneous motion of a disk
rolling down an inclined plane can be described as a rotation about the point of contact between

the disk and the plane. Therefore, if the xi , system undergoes an infinitesimal rotation  ,
corresponding to some arbitrary infinitesimal displacement, the motion of P (which, for the
moment, we consider to be at rest in the xi system) can be described as

52
 
dr  fixed    r (2.2)

where the designation "fixed" is explicitly included to indicate that the quantity dr is measured
in the xi , or fixed, coordinate system. Dividing this equation by dt, the time interval during

which the infinitesimal rotation takes place, we obtain the time rate of change of r as measured
in the fixed coordinate system:
 
 dr  d 
   r (2.3)
 dt  fixed dt

or, because the angular velocity of the rotation is



 d
 (2.4)
dt
we have

 dr   
  r (for P fixed in xi system) (2.5)
 dt  fixed

Figure 2.1. The xi ' are coordinates in


the fixed system, and xi are
coordinates in the rotating system. The
vector R locates the origin of the
rotating system in the fixed system.


If we allow the point P to have a velocity dr dt rotating with respect to the xi system, this
  
velocity must be added to   r to obtain the time rate of change of r in the fixed system:
 
 dr   dr   
    r (2.6)
 dt  fixed  dt  rotating
_____________________________________________________________________________

Example 2.1. Consider a vector r  x1eˆ1  x 2 eˆ2  x 3 eˆ3 in the rotating system. Let the fixed and

rotating systems have the same origin. Find r  in the fixed system by direct differentiation if the

angular velocity of the rotating system is  in the fixed system.

53
Solution. We begin by taking the time derivative directly

 dr  d  
     xi eˆi 
 dt  fixed dt  i 


  xi eˆi  xi e̂i  (2.7)
i


The first term is simply rr in the rotating system, but what are the e̂i ?

  dr 
rr   
 dt  rotating

 dr  
   rr   xi eˆi (2.8)
 dt  fixed i

Look at Figure 2.2 and examine which components of  i tend to rotate ê1 .

We see that  2 tends to rotate ê1 toward the  ê3 direction and that  3 tends to

rotate ê1 toward the + ê2 direction. We therefore have


deˆ1
  3 eˆ2   2 eˆ3 (2.9a)
dt

Figure 2.1. The angular velocity


components  i , rotate the system
around the êi axis, so that, for
example,  3 tends to rotate ê1 toward
the + ê2 direction.

Similarly, we have
deˆ2
  3 eˆ1  1eˆ3 (2.9b)
dt
deˆ3
  2 eˆ1  1eˆ2 (2.9c)
dt
_____________________________________________________________________________

54
In each case, the direction of the time derivative of the unit vector must be perpendicular to the
unit vector in order not to change its magnitude.
Eqns. (2.9a) - (2.9c) can be written

eˆi    eˆi (2.10)
and Eqn. (2.8) becomes

 dr   
   rr     x i eˆi
 dt  fixed i

  
 rr    r (2.11)
which is the same result as Eqn. (2.6).

Although we choose the displacement vector r for the derivation of Eqn. (2.6), the validity of
 
this expression is not limited to the vector r . In fact, for an arbitrary vector Q , we have
 
 dQ   dQ   
       Q (2.12)
 dt   dt 
  fixed   rotating
Eqn. (2.12) is an important result, and that we are going to apply it in the kinematics and
dynamics of rigid bodies.

We note, for example, that the angular acceleration  is the same in both the fixed and rotating
systems:
 
 d   d    
           (2.13)
 dt  fixed  dt  rotating
  
because    vanishes and  designates the common value in the two systems.
Description of a Rigid Body
A rigid body is defined as a system consisting of a large number of point masses, called particles,
such that the distances between the pairs of point masses remain constant even when the
body is in motion or under the action of external forces.
Forces that maintain constant distances between different pairs of point masses are internal
forces and are called forces of constraint. Such forces come in pairs and obey Newton's third law
in the strong form; that is, they are equal and opposite and act along the same line of action.
Hence we can apply the laws of conservation of linear momentum and angular momentum to the
description of the motion of rigid bodies. Furthermore, in any displacement, the relative
distances and the orientations of different particles remain the same with respect to each other;
55
hence no net work is done by the internal forces or the forces of constraint. This implies that for
a perfectly rigid body the law of conservation of mechanical energy holds as well.
Suppose a rigid body consists of N particles. Since the position of each particle is specified by
three coordinates, we may be led to conclude that we need 3N coordinates to describe the
position of the rigid body. This would be true only if the positions and the motions of all
particles were independent. But this is not so. The distance rkl between any pair of particles is

constant, and there are many such pairs. The rigid body requires only six coordinates, three to
specify the position of the center of mass and three to specify the body's orientation. Further, we
shall see that the motion of a rigid body can be divided into two separate simpler problems, the
translational motion of the center of mass and the rotation of the body around the CM.
Let us consider the motion of a rigid body constrained to rotate about a fixed point.

2.2 The Inertia Tensor

Since the dynamics of each point of a rigid body is strongly constrained by numerous conditions
rij  const, it is one of the most important fields of application of the Lagrangian formalism.

Consider the general motion of a rigid body as shown in Figure 2.2.

Figure 2.2.

Let the rigid body is rotating about an axis passing through an arbitrary fixed point O. This fixed

point (point O) is itself moving with velocity vo relative to a fixed lab frame. Consider any
 
infinitesimal mass-point at position r with respect to point O. The velocity v of the infinitesimal
mass-point relative to the lab frame is then
   
v  vo    r

56
The first thing we need to know for this approach is the kinetic energy of the body, which may
be readily calculated as a sum of kinetic energies of all its infinitesimal mass-points. Assuming
that the lab frame is inertial:

T 
dm 2
v  vo    r 2   dm vo2   dmvo    r    dm   r 2
dm 
V
2 V
2 V
2 V V
2
. (2.14)
1     dm   2
T  Mvo2   dmvo    r      r 
2 V V
2
Let us apply to this expression two general formulas of the vector analysis:
        
 
a  b  c  b  c  a   c  a  b   (2.15)
to the second term, and
a  b  c  d   a  c   b  d   a  d  b  c  (2.16)
to the third term. The result is
1 2    dm 2 2   2
T
2
Mvo   dmr  vo     
2
 r    r  ,   (2.17)
V V

where M   dm is the total mass of the body. This expression may be further simplified at a
V

specific choice of point O, namely if we use for this point the center of mass of the body whose
position is defined by equation
 1 
R  dmr . (2.18)
MV

In the lab frame bound to the center its own radius-vector is of course zero, so that in that frame
the integral over the volume V of the rigid body in the second term of Eqn. (2.17) vanishes, and
we may present the kinetic energy as a sum of two separate terms:
M 2 dm 2 2   2
T  Ttran  Trot , Ttran 
2
v cm , Trot  
2

 r    r  ,  (2.19)
V

  
where vcm  R is the center-of-mass velocity in the inertial lab frame, while all r have to be

measured in the center-of-mass frame.



Since the angular velocity vector  is common for all points of a rigid body, it is more
convenient to re-write the rotational energy in a form in which the components of this vector are
separated from the integral over all the points:

57
dm  3 2 3 2 3 3 
Trot    i  i  i i   j r j 
 r   r
V
2  i i i j 
3
dm
Trot   2

 i  j  ij ri 2   i  j ri r j 
i . j 1V

1 3
Trot  
2 i. j

 i  j  dm  ij ri 2  ri r j  (2.20)
V

where the integral term in Eqn. (2.20) is a 3×3 matrix with components

I ij   dm  ij ri 2  ri r j  (2.21)
V

Thus, the rotational kinetic energy, (Eqn. (2.20)) then becomes


1 3
Trot   i j I ij
2 i, j
(2.22)

The Kronecker-delta notation used in Eqn. (2.19) may disguise the fact that actually the matrix
has a very simple, symmetric structure:
 y2  z2  xy  xz 
 
I    dm  yx x z2 2
 yz  (2.23)
  zx  zy x 2  y 2 

I  is called the inertia tensor of the body.
Note that the inertia tensor I  has the following properties:

1. I  is a symmetric matrix. There are therefore only six independent entries, instead of nine.
2. In the case where the rigid body is made up of a collection of point masses, mi , the entries in

the matrix are just sums. For example, the upper left entry is  mi  y i2  z i2  .

3. I  depends only on the geometry of the object, and not on  .

4. To construct an I  , you not only need to specify the origin, you also need to specify the x, y,
z axes of your coordinate system. (These basis vectors must be orthogonal, because the cross-
product calculation above is valid only for an orthonormal basis.)

2.3 Rotation About an Axis Through the Origin



Consider a rigid body that rotates with angular velocity  about a fixed point – consider being
the origin of our coordinate system. Figure 2.2. Consider a little piece of the body, with mass dm

58
   
and position r . The velocity of this piece is v    r . So the angular momentum (relative to the
    
origin) of this piece is equal to r  p  (dm)r  (  r ) .
The angular momentum of the entire body is therefore
   
L   r  (  r )dm , (2.24)

where the integration runs over the volume of the body. In the case where the rigid body is made
up of a collection of point masses, mi , the angular momentum is simply
   
L   mi ri  (  ri ) (2.25)
i

Figure. 2.3.

This double cross-product is tackled as follows. First, we have


xˆ yˆ zˆ
 
  r  1  2 3
x y z

= ( 2 z   3 y ) xˆ  ( 3 x  1 z ) yˆ  (1 y   2 x ) zˆ . (2.26)
Therefore,
xˆ yˆ zˆ
  
r  (  r )  x y z
( 2 z   3 y ) ( 3 x  1 z ) (1 y   2 x )

  
 1 ( y 2  z 2 )   2 xy   3 zx xˆ   2 ( z 2  x 2 )   3 yz  1 xx yˆ  (2.27)
  ( x
3
2 2
 y )  1 zx   2 yz zˆ
The angular momentum in Eqn. (2.24) may therefore be written in the nice, concise, matrix form,

59
 L1    ( y  z )   
2 2
  xy   xz
 1 
 
 L2      xy  ( z  x )   yz   2 
2 2

 L    
 ( x  y )  3 
2 2
 3     zx   zy

 I xx I xy I xz  1 
  
  I yx I yy I yz   2 
I I zy I zz   3 
 zx

= I  (2.28)
For sake of clarity, we have not bothered to write the dm part of each integral.
The same technique we used to write Trot in tensor form can now be applied here. But the angular
momentum is a vector, so for the ith component, we write

Li   I ij  j
j (2.29)
The 3 x 3 matrix, I, is the inertia tensor. It acts on one vector (the angular velocity) to yield another
vector (the angular momentum). Thus, the inertia tensor relates a sum over the components of the angular
velocity vector to the ith component of the angular momentum vector.
Thus one important thing we should note is that using Eqn. (2.28) for the angular momentum, we
can write the rotational kinetic energy of the rigid body (Eqn. 2.23) as
1  
Trot    L (2.30)
2
_____________________________________________________________________________
Example 2.2. (Point-mass in x-y plane): Consider a point-mass m traveling in a circle (centered
at the origin) in the x-y plane, with frequency  3 . Let the radius of the circle be r (see Figure 2.4).

Using  = (0,0,  3 ) , x 2  y 2  r 2 , and z = 0 in Eqn. (2.28) (with a discrete sum of only one

object, instead of the integrals), the angular momentum with respect to the origin is

L  0,0, mr 2  3  . (2.31)
The z-component is mrv, as it should be. And the x- and y-components are 0, as they should be.

60
Figure 2.4. Figure 2.5. Figure 2.6.

Example 2.3. (Point-mass in space): Consider a point-mass m traveling in a circle of radius r,


with frequency  3 . But now let the circle be centered at the point (0,0, z o ) , with the plane of the
circle parallel to the x-y plane (see Figure 2.5).

Using  = (0,0,  3 ) , x 2  y 2  r 2 , and z  z o in Eqn. (2.28), the angular momentum with

respect to the origin is



 
L  m 3  xz o , yz o , r 2 . (2.32)
The z-component is mrv, as it should be. But, surprisingly, we have nonzero L1 and L2, even
though our mass is simply rotating around the z-axis.
Example 2.4. (Two point-masses): Add another point-mass m to the previous example. Let it
travel in the same circle, at the diametrically opposite point (see Figure 2.6).

Using  = (0,0,  3 ) , x 2  y 2  r 2 , and z  z o in Eqn. (2.28), you can show that the angular

momentum with respect to the origin is



L  2m 3 0,0, r 2  . (2.33)

The z-component is 2mrv, as it should be. And L1 and L2 are zero, unlike in the previous

example, because these components of the L ’s of the two particles cancel. This occurs because
of the symmetry of the masses around the z-axis, which causes the I zx and I zy entries in the inertia

tensor to vanish (because they are each the sum of two terms, with opposite x values, or opposite
y values).
Example 2.5. Find the elements Iij of the tensor of inertia I  for a cube of uniform density of
side b, mass M, with one corner placed at the origin.

61
Solution.
b b b

  x 
2
I 11    2  x 32 dx1 dx 2 dx3 .
0 0 0

2 5 2
The result of the three-dimensional integral is I 11  b  Mb 2 .
3 3
b b b
1 1
I 11     x1 x 2 dV       x x dx dx dx
1 2 1 2 3   b 5   Mb 2 .
V 0 0 0
4 4

We see that all the other integrals are equal, so that


2
I 11  I 22  I 33  Mb 2
3

I ij   14 Mb 2

i j

leading to the following form of matrix

 2 2 1 1 
 Mb  Mb 2  Mb 2 
 3 4 4 
I ij     1 Mb 2 2
Mb 2
1
 Mb 2

(2.34)
4 3 4
 1 2 1 2 
  Mb  Mb 2 Mb 2 
 4 4 3 
_____________________________________________________________________________

2.4 Principal Axes of Inertia



It should be clear that a considerable simplification in the expressions for T and L would result if
the inertia tensor consisted only of diagonal elements. If we could write
I ij  I i  ij (2.35)

then the inertia tensor would be


I1 0 0
I    0 I2 0

(2.36)
0 0 I 3 

We would then have
Li   I ij  ij  j  I i  i (2.37)
ij

62
and
1 1
Trot   I ij  ij  i  j   I i  i2 (2.38)
2 ij 2 i

Thus, the condition that I  have only diagonal elements provides quite simple expressions for
the angular momentum and the rotational kinetic energy. We now determine the conditions under
which Eqn. (2.35) becomes the description of the inertia tensor. This involves finding a set of
body axes for which the products of inertia (i.e., the off-diagonal elements of I  vanish. We call
such axes the principal axes of inertia.
If a body rotates around a principal axis, both the angular velocity and the angular momentum
are, according to Eqn. (2.37), directed along this axis.
Then, if I is the rotational inertia (moment of inertia) about this axis, we can write
 
L  I (2.39)

Equating the components of L in Eqns. (2.29) and (2.39), we have
L1  I1  I111  I 12 2  I13 3 

L2  I 2  I 211  I 22 2  I 23 3  (2.40)
L3  I 3  I 311  I 32  2  I 33 3 

Or, collecting terms, we obtain


( I 11  I )1  I12 2  I 13 3  0 

I 211  ( I 22  I ) 2  I 23 3  0 (2.41)
I 311  I 32 2  ( I 33  I ) 3  0 

The condition that these equations have a nontrivial solution is that the determinant of the
coefficients vanishes:
( I 11  I ) I 12 I 13
I 21 ( I 22  I ) I 23 0 (2.42)
I 31 I 32 ( I 33  I )

The expansion of this determinant leads to the secular or characteristic equation for I  , which
is a cubic. Each of the three roots corresponds to a moment of inertia about one of the principal
axes. These values, I 1 , I 2 , and I 3 , are called the principal moments of inertia. If the body

rotates about the axis corresponding to the principal moment I 1 , then Eqn. (2.39) becomes
    
L  I1 - that is, both to  and L are directed along this axis. The direction of  with respect

63
to the body coordinate system is then the same as the direction of the principal axis
corresponding to I 1 . Therefore, we can determine the direction of this principal axis by

substituting I 1 for I in Eqn. (2.41) and determining the ratios of the components of the angular-
velocity vector: 1 :  2 :  3 . We thereby determine the direction cosines of the axis about which

the moment of inertia is I 1 . The directions corresponding to I 2 , and I 3 can be found in a similar

fashion. The principal axes determined in this manner are indeed real and orthogonal.
The fact that the diagonalization procedure just described yields only the ratios of the

components of  is no handicap, because the ratios completely determine the direction of each
of the principal axes, and it is only the directions of these axes that are required. Indeed, we
would not expect the magnitudes of the  i to be determined, because the actual rate of the
body's angular motion cannot be specified by the geometry alone. We are free to impress on the
body any magnitude of the angular velocity we wish.
For most of the problems encountered in rigid-body dynamics, the bodies are of some regular
shape, so we can determine the principal axes merely by examining the symmetry of the body.
For example, anybody that is a solid of revolution (e.g., a cylindrical rod) has one principal axis
that lies along the symmetry axis (e.g., the center line of the cylindrical rod), and the other two
axes are in a plane perpendicular to the symmetry axis. It should be obvious that because the
body is symmetrical, the choice of the angular placement of these other two axes is arbitrary. If
the moment of inertia along the symmetry axis is I 1 , then I 2 , and I 3 for a solid of revolution—

that is, the secular equation has a double root.


If a body has I 1  I 2  I 3 , it is termed a spherical top; if I 1  I 2  I 3 , it is termed a symmetric

top; if the principal moments of inertia are all distinct, it is termed an asymmetric top. If a body
has I 1  0, I 2  I 3 as, for example, two point masses connected by a weightless shaft, or a

diatomic molecule, it is called a rotor.


_____________________________________________________________________________
Example 2.6. Find the principal moments of inertia and the principal axes for the cube in the
example given in the class room (for a cube with origin at one corner).
Solution. In the Example, we found that the moment-of-inertia tensor for a cube (with origin at
one corner) had nonzero off-diagonal elements. Evidently, the coordinate axes chosen for that

64
calculation were not principal axes. If, for example, the cube rotates about the x3 axis, then
 
   3 ê3 and the angular momentum vector L (see Eqn. (2.40)) has the components
1 1 2
L1    3 , L2    3 , L3   3
4 4 3
Thus,
  1 1 2 
L  Mb 2 3   eˆ1  eˆ2  eˆ3 
 4 4 3 

which is not in the same direction as  .
To find the principal moments of inertia, we must solve the secular equation
2 1 1
 I    
3 4 4
1 2 1
   I   0 (2.43)
4 3 4
1 1 2
     I
4 4 3
The value of a determinant is not affected if we add (or subtracting) any row (or column) from
any other row (or column). Eqn. (2.43) cab be solved more easily if we subtract the first row
 11 
from the second (and factoring out    I  from the second row):
 12 
2 1 1
 I    
 11 3 4 4
   I  1 1 0 0
 12  1 1 2
     I
4 4 3

Expanding, we have
2
 11   2  1 2 1 2 
   I     I         I   0
 12   3  8 4 3 

which can be factored to obtain


1  11  11 
   I    I    I   0
6  12  12 
Thus, we have the following roots, which give the principal moments of inertia:
1 11 11
I1  , I2  , I3  
6 12 12

65
The diagonalized moment-of-inertia tensor becomes
1 
6  0 0 
 11 
I    0  0  (2.44)
 12 
 0 11 
0 
 12 
Because two of the roots are identical, I 2  I 3 , the principal axis associated with I 1 must be an

axis of symmetry.
To find the direction of the principal axis associated with I 1 , we substitute for I in Eqn. (2.41)
1
the value I  I 1  :
6

2 1  1 1 
    11   21   31  0 
3 6  4 4 
1 2 1  1 
 11       21   31  0
4 3 6  4 
1 1 2 1  
 11   21       31  0
4 4 3 6  
where the second subscript 1 on the  i signifies that we are considering the principal axis

associated with I 1 . Dividing the first two of these equations by  4 , we have

211   21   31  0 
 (2.45)
 11  2 21   31  0

Subtracting the second of these equations from the first, we find 11   21 . Using this result in
either of the Eqn. (2.45), we obtain 11   21   31 and the desired ratios are

11 :  21 :  31  1 : 1 : 1
Therefore, when the cube rotates about an axis that has associated with it the moment of inertia
1 1 
I1    Mb 2 , the projections of  on the three coordinate axes are all equal. Hence, this
6 6
principal axis corresponds to the diagonal of the cube.
Because the moments I2 and I3 are equal, the orientation of the principal axes associated with
these moments is arbitrary; they need only lie in a plane normal to the diagonal of the cube.
_____________________________________________________________________________

66
2.5 Steiner’s Theorem (Parallel Axis Theorem)

For certain geometrical shapes, it may not always be convenient to compute the elements of the
inertia tensor at the center of mass of the body, because the center of mass point itself may be
accelerating so that it may not be convenient to discuss the dynamics of the body using such a
coordinate system. In such cases, we therefore consider some other set of coordinate axes, also
fixed with respect to the body and having the same orientation as the fixed (laboratory)-axes but
with an origin Q that does not correspond with the origin O located at the center of mass of the
body coordinate system. Origin Q may be located either within or outside the body under
consideration. Our aim here is to relate the moment of inertia (or inertial tensor) measured at the
new coordinate system with that of the value measured at the center of mass of the rigid body.
We suppose that the system x1 , x 2 , x3 has the origin in the center of mass of the RB. A second

system X 1 , X 2 , X 3 , has the origin in another position w.r.t. the first system. The only imposed

condition on them is to be parallel. The position of any arbitrary point P on the RB w.r.t the xi

system (CM coordinate system) is r  ( x1 , x 2 , x3 ) , and with that of X i system
   
is R  ( X 1 , X 2 , X 3 ) . Then we have a relation R  r  a , or in component form

X i  xi  ai (2.46)

where a  (a1 , a 2 , a3 ) is a vector that connects the origins of the two systems.

Let J ij be the components of the tensor of inertia w.r.t. the X i system,

 
J ij   dm  ij  X k2  X i X j  (2.47)
 k 
We substitute Eqn. (2.46) into Eqn. (2.47),
 
J ij   dm  ij   x k  a k    x i  a i x j  a j 
2

 k 

   
  dm  ij  x k2  xi x j    dm  ij  a k2  ai a j 
 k   k 
(2.48)
  
   2a k  ij  dmx k   a j  dmx j  ai  dmxi 
 k 
 

67
 
Since r is the position of a mass point w.r.t the CM, the integral over all points,  dmr , is zero;
i.e.,

  dmr 
r
M
0   dmr  0
or in component forms

 x dm  0 ,
i  x dm  0 ,
j  x dm  0
k

Thus the last term in Eqn. (2.48) vanishes, and the quantity J ij becomes

   
J ij   dm  ij  x k2  xi x j    dm  ij  a k2  ai a j 
 k   k 
or
  
J ij   dm  ij x 2  x i x j   dm  ij a 2  ai a j  (2.49)

The first term on the right-hand side of Eqn. (2.49) is the inertial tensor of the RB w.r.t the CM,
Iij therefore, Eqn. (2.49) becomes
J ij  I ij  M (a 2  ij  ai a j ) (2.50)

or if you know the elements of the inertial tensor Jij w.r.t. any arbitrary origin, you can calculate
the elements of the inertial tensor Iij w.r.t the center of mass system by:
I ij  J ij  M (a 2 ij  ai a j ) (2.51)

This is known as the Steiner’s Theorem or the parallel axes theorem.

2.6 The Dynamics of Rigid Body (Euler’s Equations of Motion)


 
The rate of change in time of the angular momentum L is equal to the torque. If L is defined

about an inertial (fixed) origin, the torque  given by:

 dL  
   (2.52)
 dt 
  inertial
At the beginning of this chapter, we have discussed a relation for the description of dynamical
variable w.r.t. a system fixed to the body with that of an inertial frame by the operator identity
d  d  
     . (2.53)
 dt  inertial  dt  body

68
Applying this relation to the term on the left side of Eqn. (2.52), we have
 
 dL   dL   
      L (2.54)
 dt 
  inertial  dt  body
Therefore, instead of Eqn. (2.52) we shall have

 dL    
     L  (2.55)
 dt 
  body
Now we project the Eqn. (2.55) onto the principal axes of inertia, that we call ( x1 , x 2 , x3 , ); then

Trot and L take by far simpler forms,
1 1
Li  I i  i , and Trot   Li  i   I i  i2 (2.56)
2 i 2 i
The ith component of the torque, Eqn. (2.55), is
dLi
  ijk  j Lk   i (2.57)
dt
Now projecting onto the principal axes of inertia and using the Eqn. (2.56), one can put Eqn.
(2.57) in the form:
d i
Ii   ijk  j  k I k   i (2.58)
dt
since the principal elements of inertia are independent of time. Thus, we obtain the following
system of equations known as Euler's equations
 1  I 11   I 3  I 2  3 2 ,
 2  I 2 2  I 1  I 3 1 3 , (2.59)
 3  I 3 3  I 2  I 1  21 .
These are known as Euler’s equations. You need only remember one of them, because the other
two can be obtained by cyclic permutation of the indices.

2.6.1 Symmetrical Top Free of Torques

Let us consider the symmetric top I 1  I 2  I 3 . In this case the axis x3 is the axis of symmetry.
The Euler equations projected onto the principal axes of inertia read

69
0  I 1  I 3  I  3 2 ,
0  I 2   I  I 3 1 3 , (2.60)
0  I 3 3 .

The last equation says that  3 is a constant. If we then define

I I
 3  3 , (2.61)
 I 
the first two equations become
1   2  0 , and  2  1  0 . (2.62)

Taking the derivative of the first of these, and then using the second one to eliminate  2 , gives

1   21  0 , (2.63)


and likewise for  2 . This is a nice simple-harmonic equation. The solutions for 1 (t ) and (by

using Eqn. (2.62))  2 (t ) are

1 (t )  A cos(t   ) ,  2 (t )  A sin( t   ) . (2.64)



Therefore, 1 (t ) and  2 (t ) are the components of a circle in the body frame. Hence, the 

vector traces out a cone around x̂3 (see Figure 2.7), with frequency  , as viewed by someone

standing on the body. The angular momentum is



L  ( I11 , I 2  2 , I 3 3 )   IA cos(t   ), IA sin( t   ), I 3 3  , (2.65)

so L also traces out a cone around x̂3 (see Figure 2.7), with frequency  , as viewed by
someone standing on the body.
The frequency,  , in Eqn. (2.61) depends on the value of  3 and on the geometry of the object.

But the amplitude, A, of the  cone is determined by the initial values of 1 and  2 .

Note that  may be negative (if I  I 3 ). In this case,  traces out its cone in the opposite

direction compared to the  > 0 case.

70
Figure 2.7.

2.7 Euler Angles

The rotation about the z-axis of the global coordinate frame is called the precession angle, the
rotation about the x-axis of the local coordinate frame is called the nutation angle, and the
rotation about the z-axis of the local coordinate frame is called the spin angle. The precession–
nutation–spin rotation angles are also called Euler angles. The kinematics and dynamics of rigid
bodies have simpler expression based on Euler angles.
The Euler angle rotation matrix B RG to transform a position vector from G(OXYZ) to B (Oxyz),
B  
r  B RG G r (2.66)

is
B
RG  R z , R x , R z ,

 CC  CSS CS  CCS SS 


  CS  CC S  SS  CCC SC  (2.67)
 S S   CS C 

So the Euler angle rotation matrix G R B to transform a position vector from B(Oxyz) to G(OXYZ),
G G 
r  RB B r (2.68)
is
RB  B RG1  B RGT  Rz , R x , R z , 
G T

CC  CSS  CS  CCS S S  


 CS  CCS  SS  CCC  CS  (2.69)
 S S  SC C 

71
B
When an Euler rotation matrix RG is given, we may calculate the equivalent precession,

nutation, and spin angles from

(a) (b)

(c)
Figure 2.8. (a) first Euler angle, (b) second Euler angle, (c) third Euler angle.

  cos 1 r33  (2.70)

r 
   tan 1  31  (2.71)
 r32 

r 
  tan 1  13  (2.72)
 r23 
provided that sin   0 . In these equations, rij indicates the element of row i and column j of the

precession–nutation–spin rotation matrix (Eqn. 2.67).

72
Euler Angle Application in Motion of Rigid Bodies The zxz Euler angles seem to be natural
parameters to express the configuration of rotating rigid bodies with a fixed point. Euler angles
that show the expression of a top are shown in Figure 2.9 as an example. The rotation of the top
about its axis of symmetry is the spin ψ. The angle between the axis of symmetry and the z-axis
is the nutation θ, and the rotation of the axis of symmetry about the z-axis is the precession ϕ.

Figure 2.9. Application of Euler angles in describing the configuration of a top.

Angular Velocity Vector in Terms of Euler Frequencies We may define an Eulerian local frame
E O, uˆ , uˆ , uˆ  by introducing the unit vectors û , û , and û as are shown in Figure 2.9. The

Eulerian frame is not necessarily orthogonal; however, it is very useful in rigid-body kinematic
analysis.
The angular velocity vector G  B of the body frame B(Oxyz) with respect to the global frame
G(OXYZ) can be written in the Euler frame E as the sum of three Euler angle rate vectors:
E
G  B  uˆ  uˆ uˆ (2.73)

The rates of the Euler angles,  ,  , and  are called Euler frequencies. These angular speeds
are also called precession, nutation, and spin, respectively.

73
To find G  B , we define the unit vectors û , û , and û along the axes of the Euler angles, as

shown in Figure 2.9, and express them in the body frame B. The precession unit vector û points

in z-direction of the global frame, so


0
uˆ  0  Kˆ (2.74)
1

is a vector in the global frame and can be transformed to the body frame after three rotations:
 sin  sin 
 
B
uˆ  B RG Kˆ  Rz , Rx, Rz , Kˆ  sin  cos  (2.75)
 
 cos 
The nutation unit vector, û , is in the intermediate frame Oxyz

1
 
uˆ  0  iˆ (2.76)
 
0
and it needs to get two rotations R x, and R z , to be transformed to the body frame:

 cos 
 
B
uˆ  B Roxyz iˆ  Rz , Rx, iˆ   sin  (2.77)
 
 0 
The spin unit vector û is already in the body frame:

0
 
uˆ  0  kˆ (2.78)
 
1
Therefore, G  B can be expressed in the body coordinate frame as

 sin  sin   cos  0


     
B
G
B   sin  cos     sin   0
     
 cos   0  1

74
B
G B  ( sin  sin   cos )iˆ  ( sin  cos   sin ) ˆj
(2.79)
 ( cos   )kˆ
The components of G  B in the body frame B (Oxyz) are related to the Euler frame E (Oϕθψ) by
the relationship
B
G  B  B RE GE B (2.80)

   sin  sin cos 0  


 x   
   sin  cos  sin 0   (2.81)
 y   
   cos 0 1  
 z
Then, G  B can be expressed in the global frame using an inverse transformation of the Euler
rotation matrix (Eqn. 2.67)
 sin  sin   cos 
 
  B 1 B
R   B 1 
R   sin  cos   sin 
G B G G B G  
  cos  

G
 B  ( cos   sin  sin  ) Iˆ  ( sin   cos  sin  ) Jˆ
 (2.82)
 (  cos ) Kˆ
and hence components of G  B in the global coordinate frame G(OXYZ) are related to the Euler

angle coordinate frame E(Oϕθψ) by the relationship

G  B  G RE GE B (2.83)

  0 cos  sin  sin    


 X   
    0 sin   cos  sin     (2.84)
 Y   
  1
 Z
0 cos   

2.8 Check List

Check whether you can answer

 Can you express the velocity in a fixed system in terms of a rotating coordinate system?
 Can you manipulate the inertia tensor?

75
 Can you explain the properties of inertia tensor for a rigid body that rotates with angular

velocity  about a fixed point
 Can you differentiate between spherical top, symmetric top and asymmetric top?
 Can you discuss Steiner’s theorem?
 Can you obtain the Euler’s equations of motion?
 Can you mention the different applications of Euler angles?

2.9 Chapter 2 Exercises


2.1. Consider a thin rod of length l and mass m pivoted about one end. Calculate the moment
of inertia. Find the point at which, if all the mass were concentrated, the moment of
inertia about the pivot axis would be the same as the real moment of inertia. The distance
from this point to the pivot is called the radius of gyration.

2.2. Consider a physical or compound pendulum as shown in Figure 2.10. For small
oscillations, find the frequency and period of oscillation if the mass of the body is M and
the radius of gyration is k.

Figure 2.10: Problem 2.2. The physical


or compound pendulum. The body
rotates about an axis passing through O.
The body rotates due to the gravitational
force acting at the center of mass.

2.3. Consider the pendulum shown in Figure 2.11 composed of a rigid rod of length b with a
mass m1 at its end. Another mass (m2) is placed halfway down the rod. Find the frequency
of small oscillations if the pendulum swings in a plane.

2.4. Calculate the moments of inertia I1, I2 and I3 for a homogeneous sphere of radius R and
mass M. (Choose the origin at the center of the sphere.)

76
Figure 2.11. Problem 2.4: A rigid rod
rotating as a pendulum has a mass m1 at
its end and another mass m2 halfway.

2.5. Show that none of the principal moments of inertia can exceed the sum of the other two.

2.6. A three-particle system consists of masses mi and coordinates ( x1 , x2 , x3 ) as follows:

m1  3m, (b,0, b)
m2  4m, (b, b,b)
m3  2m, (b, b,0)

Find the Inertia tensor, principal axes, and principal moments of inertia.

2.7. Determine the principal axes and principal moments of Inertia of a uniformly solid
hemisphere of radius b and mass m about Its center of mass.

2.8. Consider a thin homogeneous plate with principal momenta of inertia


I1 along the principal axis x1
I2 > I1 along the principal axis x2
I3 = I1 + I2 along the principal axis x3
Let the origins of the xi and xi systems coincide and be located at the center of mass O

of the plate. At time t = 0, the plate is set rotating in a force-free manner with an angular
velocity  about an axis inclined at an angle  from the plane of the plate and
perpendicular to the x2 -axis. If I1 I 2  cos 2 , show that at time t the angular velocity
about the x2 -axis is

 2 (t )   cos  tanh(t sin  )

77
2.9. Consider a heavy symmetrical top; that is, one that spins on a table, under the
influence of a uniform gravity (see Figure 2.12). Assume that the tip of the top is fixed on
the table by a free pivot. Find the Lagrange's equations of motion for the varying  ,  ,
and  , where ( ,  , and  ) are Euler angles.

Figure 2.12: Problem 2.9. A symmetric


heavy top with its tip of the top fixed on a
table by a free pivot.

2.10. The Lagrangian for a heavy symmetrical top spinning with one point fixed in a uniform
gravitational field (problem 3.9 above) is given by
1 2 1
L  T U 
2
  2
I 1    2 sin 2   I 3  cos      mgl cos  ,
2
where ( ,  , and  ) are Euler angles and I3, I1 = I2 are constant momentums of inertia.
By the symmetries of the problem, there are three constants of the motion. Identify them
and represent them as functionals of the angular coordinates and velocities.

2.11. Consider a system depicted in Figure 2.13 below in which a mass M moves horizontally
while attached to a spring of spring constant k. Hanging from this mass is a pendulum of
arm length l and bob mass m. Assume a convenient set of generalized coordinates is
( x, ), where x is the displacement of mass M relative to the equilibrium extension a of
the spring, and  is the angle the pendulum arm makes relative to the vertical
(a). Find the Lagrangian of this system.
(b). Find the canonical (generalized) momenta and the canonical (generalized) forces.
(c). Assuming small oscillations (i.e., both x and  are small), find the equations
(Lagrange's equations) of motion.
(d). Find the solution(s) for the equations of motion you obtained in (c).

78
Figure 2.13. Problem 2.11.

79
CHAPTER 3
THEORY OF SMALL OSCILLATIONS
Introduction

As we extended the motion of single particles to the motion of rigid bodies, we now investigate
the oscillatory motion of system of particles. One of the most deeply investigated concepts in
modem physics is that of oscillatory motion of atoms in the field of molecular physics and solids
in the field of solid state physics.
We will describe the oscillatory motion of many coupled oscillators in terms of normal
coordinates and normal frequencies. Theory of small oscillations in analyzing coupled oscillatory
motion uses methods of Lagrange's equations together with matrix tensor formulation. Also, at
the end of this chapter, we will briefly touch the topic of dissipative systems under forced
oscillators.

3.1 Before Theory of Small Oscillations

Before we start about theory of small oscillation, we spend some time on this section about what
coupled systems and coupled differential equations means.
Consider a pair of “coupled” differential equations
2 x   2 (5 x  3 y )  0,
(3.1)
2 y   2 (5 y  3 x)  0

We’ll assume  2  0 here, but this isn’t necessary. We call these equations “coupled” because
there are x’s and y’s in both of them, and it is not immediately obvious how to separate them to
solve for x and y. There are two methods (at least) of solving these equations.
First method: Sometimes it is easy, as in this case, to find certain linear combinations of the
given equations for which nice things happen. If we take the sum, we find
( x  y)   2 ( x  y )  0 . (3.2)
This equation involves x and y only in the combination of their sum, x + y. With z = x + y, Eqn.
(3.2) is just friendly, z   2 z  0 . The solution is
x  y  A1 cos(t  1 ) , (3.3)

80
where A1 and 1 are to be determined from initial conditions. We may also take the difference of
Eqns. (3.1). The result is
( x  y)  4 2 ( x  y )  0 . (3.4)
This equation involves x and y only in the combination of their difference, x  y . The solution is
x  y  A2 cos(2t   2 ) , (3.5)
Taking the sum and difference of Eqns. (3.3) and (3.5), we find that x(t) and y(t) are given by
x(t )  B1 cos(t  1 )  B2 cos(2t   2 )
, (3.6)
y (t )  B1 cos(t  1 )  B2 cos(2t   2 )
where the Bi’s are half of the Ai’s.
We’ve managed to solve our equations for x and y. However, the more interesting thing that
we’ve done is produce the equations (3.3) and (3.5). The combinations (x + y) and ( x  y ) are
called the normal coordinates of the system. These are the combinations that oscillate with one
pure frequency. The two simple modes of oscillation represented by these special solutions
(Eqns. 3.3 & 3.5) are called normal modes of oscillation. Their frequencies are called normal
frequencies, short for ‘normal-mode frequencies’.
The motion of x and y will, in general, look rather complicated, and it may be difficult to tell that
the motion is really made up of just the two frequencies in Eqn. (3.6). But if you plot the values
of (x + y) and ( x  y ) as time goes by, for any motion of the system, then you will find nice
sinusoidal graphs, even if x and y are each behaving in a rather unpleasant manner.
Second method: In the above method, it was fairly easy to guess which combinations of Eqn.
(3.1) produced equations involving just one combination of x and y, Eqns. (3.2) and (3.4). But
surely there are problems in physics where the guessing isn’t so easy. What do we do then?
Fortunately, there is a fail-proof method for solving for x and y. It proceeds as follows.
Let us try a solution of the form x  Ae it and y  Be it , which we will write, for convenience,
as
 x   A  it
    e . (3.7)
 y  B
It is not obvious that there should exist solutions for x and y that have the same t dependence, but
let’s try it and see what happens.
Plugging our guess into Eqn. (3.1), and dividing through by e it , we find

81
2 A( 2 )  5 A 2  3B 2  0,
(3.8)
2 B ( 2 )  5 B 2  3 A 2  0,
or equivalently, in matrix form,
  2 2  5 2  3 2  A   0 
      . (3.9)
  3 2
  2 2  5 2  B   0 
This homogeneous equation for A and B has a nontrivial solution only if the determinant equals
zero. Because we seek a nontrivial solution, we must therefore have

 2 2  5 2  3 2
0
 3 2  2 2  5 2 (3.10)
 4 4  20 2 2  16 4 .
The roots of this equation are    and   2 . We have therefore found four types of
solutions. If    , then we can plug this back into Eqn. (3.9) to obtain A = B. (Both equations
give this same result. This was essentially the point of setting the determinant equal to zero.) And
if   2 , then Eqn. (3.9) gives A   B . (Again, the equations are redundant.) Note that we
cannot solve specifically for A and B, but only for their ratios. Adding up our four solutions
according to the principle of superposition, we see that x and y take the general form (written in
vector form for the sake of simplicity and bookkeeping),
 x 1 1 1 1
   A1  e it  A2  e it  A3  e 2it  A4  e  2it (3.11)
 y 1 1   1   1
The four Ai are determined from the initial conditions.
We can rewrite Eqn. (3.11) in a somewhat cleaner form. If the coordinates x and y describe the
positions of particles, they must be real. Therefore, A1 and A2 must be complex conjugates, and
likewise for A3 and A4. If we then define some  ’s and B’s via A2*  A1  B1 2e i1 and

A4*  A3   B2 2e i2 , we may rewrite our solution in the form, as you can verify,

 x 1 1
   B1   cos(t  1 )  B2   cos(2t   2 ), (3.12)
 y 1   1
where the Bi and i are real (and are determined from the initial conditions). We have therefore
reproduced the result in Eqn. (3.6).

82
It is clear from Eqn. (3.12) that the combinations x + y and x  y (the normal coordinates)
oscillate with the pure frequencies,  and 2  , respectively. The combination x + y makes the B2
terms disappear, and the combination x  y makes the B1 terms disappear.
It is also clear that if B2 = 0, then x = y at all times, and they both oscillate with frequency  .
And if B1 = 0, then x   y at all times, and they both oscillate with frequency 2  . These two
pure-frequency motions are called the normal modes. They are labeled by the vectors (1, 1) and
( 1,1 ), respectively. In describing a normal mode, both the vector and the frequency should be
stated.
_____________________________________________________________________________
Example 3.1 (Two masses, three springs): Consider two masses, m, connected to each other and
to two walls by three springs, as shown in Figure 3.1. The three springs have the same spring
constant k. Find the positions of the masses as functions of time. What are the normal
coordinates? What are the normal modes?

Figure 3.1.
Solution: Let x1 (t ) and x 2 (t ) be the positions of the left and right masses, respectively, relative

to their equilibrium positions. Then the middle spring is stretched a distance x 2  x1 . Therefore,
the force on the left mass is  kx1  k ( x1  x 2 ) , and the force on the right mass is

 kx 2  k ( x 2  x1 ) . With these forces, F = ma on each mass gives, with  2  k / m ,

x1  2 2 x1   2 x 2  0,
(3.13)
x2  2 2 x 2   2 x1  0.
These are coupled equations, with the sum and difference are the useful combinations to take.
The sum gives
( x1  x2 )   2 ( x1  x 2 )  0, (3.14)
and the difference gives
( x1  x2 )  3 2 ( x1  x 2 )  0. (3.15)
The solutions to these equations are the normal coordinates,

83
x1  x 2  A cos(t    ),
(3.16)
x1  x 2  A cos( 3t    ).
Taking the sum and difference of these normal coordinates, we have
x1 (t )  B cos(t    )  B cos( 3t    ),
(3.17)
x 2 (t )  B cos(t    )  B cos( 3t    ),
where the B’s are half of the A’s. They are determined from the initial conditions, along with the
 ’s.
__________________________________________________________________________
Remark: (Alternative Method). We can also derive Eqn. (3.17) by using the determinant method.
Letting x1  Ae it and x 2  Be it Eqn. (3.13), we see that for there to be a nontrivial solution for
A and B, we must have
  2  2 2 2
0
2   2  2 2 (3.18)
  4  4 2 2  3 4 .
The roots of this equation are    and    3 . If    , then Eqn. (3.13) gives

A = B. If    3 , then Eqn. (3.13) gives A   B . The solutions for x1 and x 2 therefore take
the general form

 x1   1  1 1 1
   A1  e it  A2  e it  A3  e 3it
 A4  e  3it

 x2   1  1   1   1
1 1
 B   cos(t    )  B   cos( 3t    ) . (3.19)
1   1
This is equivalent to Eqn. (3.17).

The normal modes are obtained by setting either B_ or B+ equal to zero in Eqn. (3.17) or (3.19).
Therefore, the normal modes are (1, 1) and (1,1 ). How do we visualize these?
The mode (1, 1) oscillates with frequency  . In this case (where B_ = 0), we have x1(t) = x2(t) =
B cos(t    ) at all times. So the masses simply oscillate back and forth in the same manner,
as shown in Figure 3.2a. It is clear that such motion has frequency  , because as far as the
masses are concerned, the middle spring is effectively not there, so each mass moves under the
influence of only one spring, and therefore has frequency  .

84
The mode ( 1,1 ) oscillates with frequency 3 . In this case (where B+ = 0), we have

x1 (t )   x 2 (t )  B cos( 3t    ) at all times. So the masses oscillate back and forth with
opposite displacements, as shown in Figure 3.2b. It is clear that this mode should have a
frequency larger than that for the other mode, because the middle spring is stretched (or
compressed), so the masses feel a larger force. But it takes a little thought to show that the
frequency is 3 .

Figure 3.2a Figure 3.2b

Remark: The normal mode (1, 1) above is associated with the normal coordinate x1 + x 2 .
They both involve the frequency  . However, this association is not due to the fact that the
coefficients of both x1 and x 2 in this normal coordinate are equal to 1. Rather, it is due to the fact
that the other normal mode, namely ( x1 , x 2 )  (1, -1), gives no contribution to the sum x1 + x 2 .
There are a few too many 1's floating around in the above example, so it's hard to see
which results are meaningful and which results are coincidence. But the following example
should clear things up. Let's say we solved a problem using the determinant method, and we
found the solution to be
 x  3  1 
   B1   cos(1t  1 )  B2   cos( 2 t   2 ). (3.20)
 y  2   5
Then 5x + y is the normal coordinate associated with the normal mode (3, 2), which has
frequency 1 . (This is true because there is no cos( 2 t   2 ) dependence in the combination
5x + y.) And similarly, 2 x  3 y is the normal coordinate associated with the normal mode
( 1,5 ), which has frequency  2 (because there is no cos(1t  1 ) dependence in the
combination 2 x  3 y ).

85
3.2 General Theory of Small Oscillations

We will employ the Lagrange formulation of mechanics developed in previous lessons. We


consider a conservative N-particle system described in terms of a set of generalized coordinates
q k and the time t. If the system has n degrees of freedom, then k = 1, 2,..., n. The interactions can

be described by a potential energy function U  U q1 , q 2 ,..., q n  . This includes the possibility that

the system is subject to time independent external constraints. We also specify that a
configuration of stable equilibrium exists for the system and that at equilibrium the generalized
coordinates have values q k 0 .

A system of particles is said to be in static equilibrium (in a given reference system) if all the
particles remain at rest. This is possible only if the net force on each particle vanishes. In the
Lagrange formulation, this equilibrium condition is expressed as a vanishing of the generalized
force:
Fk   qk U q1o , q 2o ,..., q no   0 ,

Or

U
 0, k = 1, 2, ..., n (3.21)
q k 0

where the subscript 0 designates that the quantity is evaluated at equilibrium. If the function
U  U q1 , q 2 ,..., q n  is known, then this is a system of n equations which can be solved for the

equilibrium values q k 0 . However, we shall see that in some physical problems the equilibrium

values are known while the function U is not. In either case the equilibrium condition, Eqn.
(3.21), is satisfied. Then the variables directly describe departures from the static equilibrium
state.
To simplify the discussion, we may choose the values the generalized coordinates when the
system is at the stable equilibrium is q1o  q 2o  ...  q no  0 . Thus, the generalized coordinates

q k be measured from the equilibrium positions.


Now, if the system is displaced slightly from the equilibrium position, the effects of small
departures from equilibrium can be described by approximating the potential with a Taylor series
expansion about the equilibrium. Writing U (q)  U q1 , q 2 ,..., q n  we have

86
U 1  2U
U q1 , q 2 ,..., q n   U 0   qk   q j qk     (3.22)
k q k 0 2 j , k q j q k 0

The second term in the expansion vanishes in view of Eqn. (3.21), and since only difference in
potential is physically meaningful—without loss of generality—we may choose to measure U in
such a way that U0 = 0. Then, if we restrict the motion of the generalized coordinates to be small,
we may neglect all terms in the expansion containing products of the q k of degree higher than

second. This is equivalent to restricting our attention to simple harmonic oscillations, in which
case only terms quadratic in the coordinates appear. Thus,
1 1
U ( q)  
2 j ,k
U jk q j q k   k jk q j q k
2 j ,k
(3.23)

where we define

 2U
k jk   U jk (3.24)
q j q k 0

Note that, because the order of differentiation is immaterial (if U has continuous second partial
derivatives), the quantity k jk is symmetrical; that is, k jk  k kj .

The potential Eqn. (3.23) is called the harmonic approximation and higher order terms in the
Taylor expansion are said to be anharmonic. A departure from equilibrium is said to be “small”
if the anharmonic contribution to the potential energy is negligible to the accuracy desired.
In the harmonic approximation, the potential (3.23) corresponds to a generalized force
Fk   qk U k (q )    k kj q j . (2.25)
j

The coefficients k jk are called force constants; they are measures of the coupling strength

between different degrees of freedom.


We have specified that the motion of the system is to take place in the vicinity of the equilibrium
configuration, and we have shown (Eqn. 3.21) that U must have a minimum value when the
system is in this configuration. Because we have chosen U = 0 at equilibrium, we must have, in
general, U  0 . In other words, we can express this statement as follows: The equilibrium point
q k 0  0 is said to be stable if the quadratic potential energy (Eqn. 3.23) is positive definite, that

is, if

87
1
U ( q)   k jk qk q j  0
2 j ,k
(3.26)

for all values of the q k .

By setting all q k but one to zero, we see that Eqn. (3.26) implies k kk  0 so the corresponding

generalized force  k kk q k draws the system back to equilibrium. Actually, Eqn. (3.26) is merely a

sufficient condition for stability.


What about the kinetic energy of the system? To answer this, we first note that in the
configuration of stable equilibrium (i.e., when the system is at the stable equilibrium), Lagrange's
equations are satisfied by
qk  q k 0 , q k  0 , qk  0 , k = 1, 2, ..., n (3.27)

Every nonzero term of the form d / dt L q k  must contain at least either q k or qk , so all such

terms vanish at equilibrium. From Lagrange's equation, we therefore have

L T U
  0 k = 1, 2, ..., n (3.28)
q k 0
q k q k 0
0

where again the subscript 0 designates that the quantity is evaluated at equilibrium.
We assume that the equations connecting the generalized coordinates and the rectangular
coordinates do not explicitly contain the time; that is, we have
x ,i  x ,i (q j ) or q j  q j ( x ,i ) (3.29)

Thus, as we know from chapter one, in terms of the generalized coordinates the system kinetic
energy T is a positive definite quadratic function of the q k :

1
T  m jk q j qk  0 .
2 j,k
(3.30)

Therefore, in general,

T
0, k = 1, 2, ..., n (3.31)
q k 0

In general, the mass coefficients m jk are functions of the coordinates as well as the masses of the

particles. But consistent with our approximation to the potential energy, we regard the mass
coefficients as constants
m jk (q )  m jk (0)  m jk  mkj (3.32)

88
From the fact that every particle has a mass it follows that mkk  0 for all k .

Eqns. (3.23) and (3.30) are of a similar form:


1
T  m jk q j q k
2 j ,k
(3.33)
1
U   k jk q j q k
2 j ,k
with T  0 , and U  0 .
In Eqn. (3.33), the m jk and the k jk are n  n arrays of numbers (or matrix) specifying the way

the motions of the various coordinates are coupled. For example, if mrs  0 for r  s , then the
th th
kinetic energy contains a term proportional q r q s ,, and a coupling exists between the r and s

coordinate. If, however, m jk is diagonal, so that m jk  0 for j = k but vanishes otherwise, then

the kinetic energy is of the form


1
T  m r q r2
2 r
where mrr has been abbreviated to mr . Thus, the kinetic energy is a simple sum of the kinetic

energies associated with the various coordinates. As we see below, if, in addition, k jk is diagonal

so that U is also a simple sum of individual potential energies, then each coordinate behaves in
an uncomplicated manner, undergoing oscillations with a single, well-defined frequency. The
problem is therefore to find a coordinate transformation that simultaneously diagonalizes both
m jk and k jk and thereby renders the system describable in the simplest possible terms. Such

coordinates are the normal coordinates.


The equations of motion of the system with kinetic and potential energies given by Eqn. (3.33)
are obtained from Lagrange's equation
L d L
 0
q k dt q k
But because T is a function only of the generalized velocities and U is a function only of the
generalized coordinates, Lagrange's equation for the k th coordinate becomes
U d T
 0 (3.34)
q k dt q k
From Eqns. (3.33), we evaluate the derivatives:

89
U 
  k jk q j 
q k j 
 (3.35)
T
  m jk q j 
q k j 

The equations of motion then become

 k
j
kj q j  mkj q j   Fk (3.36)

where the Fk are components of the generalized force due to external agents.
The system of Eqns. (3.36) is the sum of the products of matrices, and can be written as a single
matrix equation
m  q  k  q  F . (3.37)

where the mass matrix [m] is defined by


 m11 m12 . . . m1n 
m m22 . . . m2 n 
 21 
 . . . . . . 
m   , (3.38)
 . . . . . . 
 . . . . . . 
 
 mn1 . . . . mnn 

and the matrix [k] is defined similarly. The notation q indicates the column of generalized

coordinates defined by

 q1 
q 
 2
.
q   (3.39)
.
.
 
q n 

90
 q1   F1 
q  F 
 2  2
.  . 
Of course, q    and F  
.  . 
.  . 
   
qn   Fn 
So now the expressions, Eqns. (3.33), for kinetic and potential energies can be rewritten as
2T  q m  q , (3.40)

2V  q k  q . (3.41)

Any physical system modeled by a matrix equation of motion of the form (3.37), where the
matrices [m] and [k] are positive definite and [m] is nonsingular is called a harmonic system,
because it generalizes the single particle harmonic oscillator model.
Anyway, the Lagrange’s equations of motion for the system pertaining small oscillations are
given by either Eqns. (3.36) or (3.37) , and give n equations of motion for the system:

 k
j
kj q j  mkj q j   Fk

3.3 Free Oscillations

In the absence of external forces, the matrix equation of motion (3.37) reduces to
m  q  k  q 0 or  k j
kj q j  mkj q j   0 (3.42)

This is a set of n second-order linear homogeneous differential equations with constant


coefficients. Our first task is to find the general solution of this linear matrix equation. The most
straightforward approach is to consider a trial solution. Because we are dealing with an
oscillatory system, we expect a solution of the form
q j (t )  a j e i (t  ) (3.43)

where the a j are real amplitudes and where the phase  has been included to give the initial

conditions which renders two arbitrary constants required by the second-order nature of each of
the differential equations. (Only the real part of the right-hand side is to be considered.) The
frequency  and the phase  are to be determined by the equations of motion.
With a solution of the form given by Eqn. (3.43, the equations of motion become

91
 k
j
kj 
  2 mkj a j  0 (3.44)

where the common factor expi t    has been canceled. This is a set of n linear,

homogeneous, algebraic equations that the a j must satisfy. For a nontrivial solution to exist, the

determinant of the coefficients must vanish:


k kj   2 mkj  0 (3.45)

To be more explicit, this is an n  n determinant of the form

k11   2 m11 k12   2 m12 ..... k1n   2 m1n


k12   2 m12 k 22   2 m22 ..... k 2n   2 m2 n
. . . . 0 (3.46)
. . . .
k1n   2 m1n k2n   2 m2 n .... k nn   2 mnn

where the symmetry of the k jk and m jk , has been explicitly included.

The equation represented by this determinant is called the characteristic equation or secular
equation of the system and is a polynomial of degree n in the variable  2 . There are, in general,
n roots we may label  r2 . Its n roots  r2 are real and positive, because the matrices [k] and [m] are

real, symmetric and positive definite, so a real, positive value for each  r is thereby determined.
The  r are called the characteristic frequencies (or Eigen frequencies) or normal (or natural

frequencies) of the harmonic system. (In some situations, two or more of the  r can be equal;
this is the phenomenon of degeneracy.) Each unique root is said to be non-degenerate. Just as in
the procedure for determining the directions of the principal axes for a rigid body, each of the
roots of the characteristic equation may be substituted into Eqn. (3.44) to determine the ratios
a1 : a 2 : a3 : ... : a n for each value of  r . Because there are n values of  r , we can construct n sets

of ratios of the a j . Each of the sets defines the components of the n-dimensional vector ar, called

an eigenvector of the system. Thus ar is the eigenvector associated with the Eigen frequency  r .
We designate by a jr the j th component of the r th eigenvector.

Because the principle of superposition applies for the differential equation (Eqn. 3.42), we must
write the general solution for q j as a linear combination of the solutions for each of the n values

of r:

92
q j (t )   a jr e i(r t  r ) (3.47)
r

Because it is only the real part of q j (t ) that is physically meaningful, we actually have

q j (t )  Re  a jr e i (rt  r )   a jr cos( r t   r ) (3.48)


r r

The motion of the coordinate q j is therefore compounded of motions with each of the n values

of the frequencies  r . The q j evidently are not the normal coordinates that simplify the problem.

Later, we will continue the search for normal coordinates.


_____________________________________________________________________________
Example 3.2. Consider a system of coupled motion in one dimension: two masses connected by
a spring to each other and by springs to fixed positions (Figure 3.3). Let each of the oscillator
springs have a force constant k: the force constant of the coupling spring is k12. We restrict the
motion to the line connecting the masses, so the system has only two degrees of freedom,
represented by the coordinates x1 and x 2 . Each coordinate is measured from the position of
equilibrium. Find the characteristic frequencies.

Figure 3.3.
Solution. If m1 and m2 are displaced from their equilibrium position by amounts x1 and x2,

respectively, the force on m1 is  kx1  k12 ( x1  x 2 ) , and the force on m2 is  kx 2  k12 ( x 2  x1 ) .


The potential energy of the system is
1 2 1 1
U kx1  k12 ( x 2  x1 ) 2  kx 22
2 2 2
1 1
 (k  k12 ) x12  (k  k12 ) x 22  k12 x1 x 2 . (3.49)
2 2
The term proportional to x1 x 2 is the factor that expresses the coupling in the system. Calculating

the k jk (or U jk ), we find

93
 2U 
U 11  k11   k  k12 
x12 0 
2 
U 
U 12  k12    k12  U 21  (3.50)
x1x 2 0 
2 
U 
U 22  k 22  2  k  k12
x 2 0 

The kinetic energy of the system is


1 1
T Mx12  Mx 22 (3.51)
2 2
According to Eqn. (3.33)
1
T  m jk x j x k
2 j ,k
(3.52)

Identifying terms between these two expressions for T, we find


m11  m 22  M 
 (3.53)
m12  m 21  0 

Thus, the secular determinant (Eqn. 3.46) becomes


k  k12  M 2  k12
0 (3.54)
 k12 k  k12  M 2
The expansion of this secular determinant yields
(k  k12  M 2 ) 2  k122  0 (3.55)
Hence,
k  k12  M 2   k12
Solving for  , we obtain

k  k12  k12
 (3.56a)
M
We therefore have two characteristic frequencies (or Eigen frequencies) for the system:
k  2k12 k
1  , 2  (3.56b)
M M
Thus, the general solution to the problem is

94
x1 (t )  B11 e i1t  B11 e i1t  B12 e i2t  B12 e  i2t 
 (3.57)
x 2 (t )  B21 e i1t  B21 e i1t  B22 e i 2t  B22 e i2t 
__________________________________________________________________________

3.4 Orthogonality of the Eigenvectors

We now wish to show that the eigenvectors ar form an orthonormal set.


Rewriting Eqn. 3.24 for the s th root of the secular equation, we have
 s2  m jk a ks   k jk a ks (3.58)
k k

Next, we write a comparable equation for the rth root by substituting r for s and interchanging j
and k:
 r2  m jk a jr   k jk a jr (3.59)
j j

where we have used the symmetry of the m jk and k jk . We now multiply Eqn. (3.58) by a jr and

sum over j and also multiply Eqn. (3.59) by a ks and sum over k:

 s2  m jk a jr a ks   k jk a jr a ks 
j ,k j ,k 
 (3.60)
 r2  m jk a jr a ks   k jk a jr a ks 
j ,k j ,k 
The right-hand sides of Eqns. (3.60) are now equal, so subtracting the first of these equations
from the second, we have
 2
r   s2  m jk a jr a ks  0 (3.61)
j ,k

We now examine the two possibilities r = s and r  s . For r  s , the term  r2   s2  is, in

general, different from zero. Therefore the sum must vanish identically:

m
j ,k
jk a jr a ks  0 , rs (3.62)

And, without loss of generalities, we can normalize the eigenvectors and set, for r = s:

m
j ,k
jk a jr a kr  1 (3.63)

The a jr are then said to be normalized. Combining Eqns. (3.62) and (3.63), we may write

95
m
j ,k
jk a jr a ks   rs (3.64)

Because a jr is the jth component of the rth eigenvector, we represent ar by

ar   a jr eˆ j (3.65)
j

The vectors ar defined in this way constitute an orthonormal set; that is, they are orthogonal
according to the result given by Eqn. (3.62), and they have been normalized by setting the sum in
Eqn. (3.63) equal to unity.

3.5 Normal Coordinates

As we have seen in Eqn. (3.47) the general solution for the motion of the coordinate q j must be

a sum over terms, each of which depends on an individual eigen frequency. In the previous
section, we showed that the vectors ar are orthogonal (Eqn. 3.62) and, as a matter of
convenience, we even normalized their components a jr (Eqn. 3.63) to arrive at Eqn. (3.64); that

is, we have removed all ambiguity in the solution for the q j , so it is no longer possible to specify

an arbitrary displacement for a particle. Because such a restriction is not physically meaningful,
we must introduce a constant scale factor  r (which depends on the initial conditions of the
problem) to account for the loss of generality introduced by the arbitrary normalization. Thus,

q j (t )    r a jr e i (r t  r ) (3.66)
r

To simplify the notation, we write


q j (t )    r a jr e ir t (3.67)
r

where the quantities  r are new scale factors (now complex) that incorporate the phases of  r .
We now define a quantity r ,
 r (t )   r e ir t (3.68)
so that

q j (t )   a jr (3.69)
r r

96
The  r , by definition, are quantities that undergo oscillation at only one frequency. They may be

considered as new coordinates, called normal coordinates, for the system. The  r satisfy
equations of the form
r   r2 r  0 (3.70)
There are n independent such equations, so the equations of motion expressed in normal
coordinates become completely separable.
____________________________________________________________________________
Example 3.3. Determine the Eigen frequencies (i.e., normal frequencies), eigenvectors, and
normal coordinates of the mass-spring system in Example 3.2 by using the procedure just
described. Assume k12  k .
Solution. The Eigen frequencies were determined in Example 3.2, where we found T and U (step
1). We can find the components for k jk , directly from Eqn. (3.50) or by inspection from Eqn.

(3.49), making sure k jk , is symmetrical.

k  k12  k12 
[k ]   (3.71)
  k12 k  k12 

The array m jk can easily be determined from Eqn. (3.51):

M 0
[m]   (3.72)
0 M 

We use Eqn. (3.46) to determine the Eigen frequencies  r .

k  k12  M 2  k12
0 (3.73)
 k12 k  k12  M 2
Which gives the result

k  2k12 k
1  , 2  (3.56)
M M
We use Eqn. (3.44) to determine the eigenvector components a jr . We have two equations for

each value of r, but because we can determine only the ratios a1r a 2 r , one equation for each r is
sufficient. For r = 1, k = 1, we have
k 11   
 12 m11 a11  k 21  12 m21 a 21  0 (3.74)

97
or, inserting the values for k11 , k 21 , 12 , and m11 , and using the simplification that k12  k ,

 3k 
 2k   M a11  ka 21  0
 M 
with the result
a11   a 21 (3.75)
For r = 2, k = 1, we have
 k 
 2k   M  a12  ka 22  0
 M 
with the result
a12  a 22 (3.76)
The general motion (Eqn. 3.69) becomes
x1  a111  a12 2 
 (3.77)
x 2  a 211  a 22 2 

Using Eqns. (3.75) and (3.76), this becomes


x1  a111  a 22 2 
 (3.78)
x 2   a111  a 22 2 

Adding x1 and x 2 gives


1
2  ( x1  x 2 ) (3.79)
2a 22

Subtracting x1 from x 2 gives


1
1  ( x1  x 2 ) (3.80)
2a11

The normal coordinate  2 can be determined by finding the conditions when the other normal

coordinate 1 remains equal to zero. From Eqn. (3.80), 1 = 0 when x1 = x 2 . Thus, for normal
mode 2 (  2 ), the two masses oscillate in phase (the symmetrical mode). The distance between
the particles is always the same, and they oscillate as if the spring connecting them were a rigid,
weightless rod. Similarly, we can find the conditions for the normal coordinate 1 by

determining when  2 = 0 ( x1   x 2 ). In normal mode 1 ( 1 ), the particles oscillate out of phase


(the anti-symmetrical mode).

98
We may determine the components of the eigenvectors (Eqn. 3.65),
1 : a1 = a11eˆ1  a 21eˆ2
2 : a2 = a12 eˆ1  a 22 eˆ2 (3.81)

by using Eqns. (3.75) and (3.76).


a1 = a11 (eˆ1  eˆ2 )
a2 = a 22 (eˆ1  eˆ2 ) (3.82)

Although normally not required, we may determine the values of a11 and a 22 from the
orthonormality condition of Eqn. (3.64) with the result
1 
a11   a 21  
2M
 (3.83)
1 
a12  a 22 
2M 

_____________________________________________________________________________

3.6 Weak Coupling

Some of the more interesting cases of coupled oscillations occur when the coupling is weak—
that is, when the force constant of the coupling spring is small compared with that of the
oscillator springs: force constant k12  k . We know the frequencies 1 and  2 are (Eqn. 3.56)

k  2k12 k
1  , 2  (3.56)
M M
If the coupling is weak, we may expand the expression for 1

k  2k12 k
1  1  1  4 (3.84)
M k M
where
k12
  1 (3.85)
2k
The frequency 1 now reduces to

k
1  (1  2 ) (3.86)
M
The natural frequency of either oscillator, when the other is held fixed, is

99
k  k12 k
o   (1   ) (3.87)
M M
or

k
  o (1   ) (3.88)
M
Therefore, the two characteristic frequencies are given approximately by

k k 
1  (1  2 ), 2  
M M 
  o (1   )(1  2 )  (3.89)
  o (1   )  2   o (1   )


We can now examine the way a weakly coupled system behaves. If we displace Oscillator 1 a
distance D and release it from rest, the initial conditions for the system are
x1 (0)  D , x 2 (0)  0 , x1 (0)  0 , x 2 (0)  0 (3.90)
If we substitute these initial conditions into the solutions for x1 (t ) and x 2 (t ) , we find the
amplitudes to be
D
B1  B1  B 2  B2  (3.91)
4
Then, x1 (t ) becomes
D i1t
x1 (t ) 
4
  
e  e i1t  e i 2t  e i2t 
D
 cos 1t  cos  2 t 
2
   2     2 
 D cos 1 t   cos 1 t (3.92)
 2   2 
But, according to Eqn.(3.89),
1   2 1   2
 o ;   o (3.93)
2 2
therefore,
x1 (t )   D cos  o t  cos  o t (3.94a)

Similarly,

100
Figure 3.4. The solutions for x1 (t ) and x 2 (t ) have a high frequency component

(  o ) that oscillates inside a slowly varying component (  o ). Energy is

transferred back and forth between the two oscillators.

x 2 (t )  D sin  o t  sin  o t (3.94b)

Because  is small, the quantities D cos  o t and D sin  o t vary slowly with time. Therefore,

x1 (t ) and x 2 (t ) are essentially sinusoidal functions with slowly varying amplitudes. Although
only x1 is initially different from zero, as time increases the amplitude of x1 decreases slowly

with time, and the amplitude of x 2 increases slowly from zero. Hence, energy is transferred from
the first oscillator to the second. When t   2 o , then D cos  o t = 0, and all the energy has

been transferred. As time increases further, energy is transferred back to the first oscillator. This
is the familiar phenomenon of beats and is illustrated in Figure 3.4. (In the case illustrated,
  0.08 .)
____________________________________________________________________________
Example 3.4. (On Normal coordinates & Normal Modes). Determine the Eigen frequencies and
describe the normal mode motion for two pendula of equal lengths b and equal masses m

101
connected by a spring of force constant k as shown in Figure 3.5. The spring is unstretched in the
equilibrium position.
Solution. We choose 1 and  2 (Figure 3.5) as the generalized coordinates. The potential energy
is chosen to be zero in the equilibrium position. The kinetic and potential energies of the system
are, for small angles,
1 1
T m(b1 ) 2  m(b2 ) 2 (3.95)
2 2
1
U  mgb(1  cos 1 )  mgb(1  cos  2 )  k (b sin 1  b sin  2 ) 2 (3.96)
2
Using the small oscillation assumption sin    and cos   1   2 2 , we can write
mgb 2 1
U (1   22 )  kb 2 (1   2 ) 2 (3.97)
2 2
3 5 2 4
Note that sin          and cos  1    
3! 5! 2! 4!
The components of the matrices [k] and [m] are
mb 2 0 
[ m]    (3.98)
 0 mb 2 

mgb  kb 2  kb 2
[k ]   2  2
(3.99)
  kb mgb  kb 
The determinant needed to find the Eigen frequencies  is

mgb  kb 2   2 mb 2  kb 2
0 (3.100)
 kb 2 mgb  kb 2   2 mb 2
which gives the characteristic equation
b 2 (mg  kb   2 mb) 2  (kb 2 ) 2  0

(mg  kb   2 mb) 2  (kb) 2


or
mg  kb   2 mb   kb (3.101)

Taking the plus sign,   1 ,

mg  kb  12 mb  kb

102
Figure 3.5. Two pendula of equal lengths having equal masses are
connected by a spring.

g
12  (3.102)
b
Taking the minus sign in Eqn. (3.101),    2 ,

mg  kb   22 mb   kb
g 2k
 22   (3.103)
b m
Putting the values of 1 and  2 into Eqn. (3.44) gives, for k = 1,

(mgb  kb 2   r2 mb 2 )a1r  kb 2 a 2 r  0 (3.104)


If r = 1, then
 2 g 2 2
 mgb  kb  mb  a11  kb a 21  0
 b 
and
a11  a 21 (3.105)
If r = 2, then
 2 g 2 2k 
 mgb  kb  mb  mb 2 a12  kb 2 a 22  0
 b m 
and
a12   a 22 . (3.106)

We write the coordinates 1 and  2 in terms of the normal coordinates by

1  a111  a12 2 
 (3.107)
 2  a 211  a 22 2 

103
Using Eqns. (3.105) and (3.106), Eqns. (3.107) become
1  a111  a 22 2 
 (3.108)
 2  a111  a 22 2 

The normal modes are easily determined, by adding and subtracting 1 and  2 , to be

1 
1  (1   2 ) 
2a11 
 (3.109)
1
2  ( 2   1 )
2a 22 

Because normal coordinate 1 occurs when  2 = 0, then  2 = 1 for normal mode 1


(symmetrical). Similarly, normal coordinate  2 occurs when 1 = 0 ( 1   2 ), and normal
mode 2 is anti-symmetrical. The normal mode motions are shown in Figure 3.6. Notice that for
mode 1, the spring is neither compressed nor extended. The two pendula merely oscillate in
unison (harmony) with their natural frequencies ( 1   o  g b ). The higher frequency of

normal mode 2 is easily displayed for a stiff spring.

Figure 3.6. For the above Example: The two normal mode motions are shown.
____________________________________________________________________________

3.7 Sympathetic Vibrations and Beats

Sympathetic vibrations and beats are the results of weak coupling. To understand more about
beats and sympathetic vibrations, let us consider two simple oscillators (coupled pendula) each
of length b and mass m that are coupled by a spring constant k, as shown in Figure 3.5 (Example
3.4). If the spring offers a small resistance to the relative motion of the two pendulums, we say

104
that the system has weak coupling, whereas if the spring offers a greater resistance, the system is
said to have strong coupling. If the pendulums are not exactly equal in length or in mass, we say
that the two pendulums are out of tune or detuned.
For the present, let us assume that the two pendulums are exactly of equal length and mass, and
they are weakly coupled by a spring. We assume that they oscillate in the same plane. Let us
further assume that the one pendulum is excited by giving an initial displacement while the other
pendulum is at rest. As time passes, the resulting oscillations of the two pendulums are as shown
in Figure 3.5. As is clear, the oscillations are modulated, and the energy is continuously being
transferred from one pendulum to the other. When one pendulum is oscillating with maximum
amplitude, the other pendulum is at rest, and vice versa. This is the phenomenon of resonance or
sympathetic vibration between two systems. This is the theory of resonance. A slight detuning
leads to the phenomenon of beats.
If the two pendulums are slightly detuned (have slightly different frequencies), the energy
exchange will still take place, but this exchange is not complete. The initially excited first
pendulum reaches a certain minimum amplitude, but not zero amplitude. The second pendulum
initially at rest, does reach zero amplitude during its oscillations. This results in the phenomenon
of beats, as shown in Figure 3.7. Thus sympathetic vibration or resonance is upset by slight
detuning.
The preceding discussion for coupled mechanical oscillating systems can be extended to
electrical systems. Sympathetic oscillations are of great importance in electrical circuits. in
electrical systems, we have a primary and a secondary circuit that are usually inductively
coupled with each other. Thus, if the primary circuit is excited, the secondary circuit will also
oscillate strongly if there is a resonance. Unlike the coupled pendulums considered previously, in
electrical circuits damping must be included. As we shall discuss in 3.10, damping is equivalent
to ohmic resistance, mass corresponds to the self-inductance, and restoring force to the
capacitance effects. Furthermore, in electrical oscillations, we deal not only with "position
coupling" but also with "velocity and acceleration coupling".

105
x1  x2

x1  x2

t
Figure 3.7. The phenomenon of beats resulting from two slightly detuned,
weakly coupled oscillators (pendulums in this case).

3.8 Molecular Vibrations

We mentioned previously that molecular vibrations are good examples of the applications of the
small oscillations discussed in this chapter.
_____________________________________________________________________________
Example 3.5.
Consider a symmetrical linear triatomic molecule similar to (for example, CO2) Figure 3.8. The
central atom (carbon) has mass M, and the symmetrical atoms (oxygen) have masses m. Assume
that only longitudinal vibrations (that is, in-line or 1-dimensional vibration) is possible.
Determine the normal (or Eigen frequencies) and describe the normal mode motion of the
system.

106
Figure 3.8. A symmetrical linear triatomic molecule vibrating
in longitudinal direction.

Solution
Consider vibrations along the line of the molecule and let x1 , x 2 and x3 be the displacements of

the atoms from their equilibrium points. These are the generalized coordinates q1, q2 and q3 that
are most convenient here. Then
m 0 0
1 1 1  
T  mx12  Mx 22  mx 32  [ m]   0 M 0 (3.110)
2 2 2 0 0
 m 

and
 1 1 0 
1 1  
U  k ( x 2  x1 ) 2  k ( x 2  x 3 ) 2  [ k ]  k   1 2  1 (3.111)
2 2  0 1 1 
 
To find the normal frequencies, we solve the characteristic equation, which in this case is
k   2m k 0
2
k 2k   M k 0 (3.112)
0 k k   2m

which, after a little algebra, reduces to


 2 (k   2 m)(2km  kM   2 mM )  0 .
Thus the normal frequencies in ascending order are given by
12  0 ,  22  k m ,  32  k (2M 1  m 1 ) .
To find the normal modes for each normal frequency, we solve Eqn. (3.44).
For 1 , we have

k 11    
 12 m11 a11  k 21  12 m 21 a 21  k 31  12 m31 a 31  0 
k 12  12 m12 a  k
11 22  12 m22 a  k
21 32  12 m32 a 31 0
k 13  12 m13 a  k
11 23  12 m 23 a  k
21 33  12 m33 a31 0

107
Or substituting 1  0 , and using the values of k jk and m jk , the above set of equations becomes

ka11  ka 21  0  0  k k 0  a11 
  
 ka11  2ka 21  ka31  0 or  k 2k  k  a 21   0 (3.113)
0  ka 21  ka31  0  0 k k  a31 

The first equation of Eqn. (3.113) gives
a 21  a11 (3.114)
Using Eqn. (3.114) into the second & third equations of Eqn. (3.113) gives
ka11  ka31  0 
  a11  a31 (3.115)
 ka11  ka31  0

Therefore, the ratios of the components of the eigenvector a1 corresponding to the normal
frequency 1  0 is
a11 : a 21 : a31  1 : 1 : 1 (3.116)

Then, the eigenvector a1 is

 a11  1
   
a1 =  a 21   1 (3.117)
 a  1
 31   
The overall multiple can be fixed by requiring that the eigenvector is normalized with respect
to m jk ; i.e., using Eqn. (3.64) we have (for 1 , & r  s  1 )

m
j ,k
jk a j1 a k 1  1

m11 a11 a11  m 21 a 21 a11  m31 a31 a11  m12 a11 a 21  m22 a 21 a 21
(3.118)
 m32 a31 a 21  m13 a11 a 31  m 23 a 21 a31  m33 a31 a 31  1

Using m jk  0 for j  k , Eqn. (3.118) becomes

ma112  Ma 21
2 2
 ma 31 1

Or using Eqns. (3.114) & (3.115), we get


ma112  Ma112  ma112  1  2m  M a112  1
1 / 2
 a11   2m  M   a 21  a31 (3.119)

Then, the normalized eigenvector a1 for 12  0 is

108
 1
1 / 2  
a1 =  2m  M   1 (3.120)
 1
 
This normal mode (with the normal frequency   0 ), is known as a zero mode. It is just a rigid
translation of the system.
Similarly, for  22  k m , we have

 0 k 0  a12 
  
 k 2k  k M m  k  a 22   0 (3.121)
 0 k 0  a32 

so that (following the same steps) the normalized normal mode has generalized eigenvector
1
1 / 2  
a2 = 2m  0 (3.122)
  1
 

 2m  M 
And for  32  k   , we have
 mM 
  2km M k  a13  0
  
 k kM m  k  a 23   0 (3.123)
 0 k  2km M  a 33 

so the normalized normal mode has generalized eigenvector
 1 
 
a3 = 2m  4 m 2 M 
1 / 2
 2m M  . (3.124)
 1 
 
You should check that the three modes are pairwise orthogonal with respect to [m], as they must
be because 1 ,  2 and  3 are all distinct.

____________________________________________________________________________

3.9 The Loaded String

We now consider a more complex system consisting of an elastic string (or a spring) on which a
number of identical particles are placed at regular intervals. The ends of the string are
constrained to remain stationary. Let the mass of each of the n particles be m, and let the spacing

109
between particles at equilibrium be d. Thus, the length of the string is L  (n  1)d . The
equilibrium situation is shown in Figure 3.9.
We wish to treat the case of small transverse oscillations of the particles about their equilibrium
positions. First, we consider the vertical displacements of the masses numbered j  1, j , and j  1
(Figure 3.10). If the vertical displacements q j 1 , q j , and q j 1 are small, then the tension  in the

string is approximately constant and equal to its value at equilibrium. For small displacements,
the string section between any pair of particles makes only small angles with the equilibrium
line. Approximating the sines of these angles by the tangents, the expression for the force that
tends to restore the jth particle to its equilibrium position is

Fj   q j  q j 1    q j  q j 1  (3.125)
d d
The force F j is, according to Newton's law, equal to mq j , Eqn. (3.125) can therefore be written

as

q j  q j 1  2q j  q j 1  (3.126)
md
which is the equation of motion for the jth particle. The system is coupled, because the force on
the jth particle depends on the positions of the ( j  1 )th and ( j  1 )th particles; this is therefore an
example of nearest neighbor interaction, in which the coupling is only to the adjacent particles.

Figure 3.9. A schematic of the loaded string. In equilibrium, identical masses are
spaced equidistantly. The ends of the string are fixed.

110
Figure 3.10. Vertical displacements ( q j 1 , q j , and q j 1 ) of masses on the loaded string.

We have considered only the motion perpendicular to the line of the string: transverse
oscillations. It is easy to show that exactly the same form for the equations of motion results if
we consider longitudinal vibrations—that is, motions along the line of the string. In this case, the
factor / d is replaced by k, the force constant of the string.
Although we used Newton's equation to obtain the equations of motion (Eqn. 3.126), we may
equally well use the Lagrangian method. The potential energy arises from the work done to
stretch the n  1 string segments:
 n 1
U  (q j 1  q j ) 2
2d j 1
(3.127)

where qo and qn1 are identically zero, because these positions correspond to the fixed ends of

the string. Note that Eqn. (3.127) yields an expression for the force on the jth particle that is the
same as the previous result (Eqn. 3.125).
The kinetic energy is given by the sum of the kinetic energies of the n individual particles:
1 n 2
T m q j (3.128)
2 j 1

Because q n1  0 , we may extend the sum in Eqn. (3.128) to j = n + 1 so that the range of j is the

same as that in the expression for the potential energy. Then, the Lagrangian becomes
1 n 1  2  
L   mq j  (q j 1  q j ) 2  (3.129)
2 j 1  d 
It should be obvious that the equation of motion for the jth particle must arise from only those
terms in the Lagrangian containing q j or q j . If we expand the sum in L, we find

1 1 1
L       mq 2j  (q j 1  q j ) 2  (q j  q j 1 ) 2      (3.130)
2 2d 2d

111
where we have written only those terms that contain either q j or q j . Applying Lagrange's

equation for the coordinate q j we have


mq j  (q j 1  2q j  q j 1 )  0 (3.131)
d
Thus, the result is the same as that obtained by using the Newtonian method.
To solve the equations of motion, we substitute (a trial solution), as usual,
q j (t )  a j e it (3.132)

where a j can be complex. Substituting this expression for q j (t ) into Eqn. (3.131), we find

    
 a j 1   2  m 2  a j  a j 1 )  0 (3.133)
d  d  d
where j  1,2,3....n , but because the ends of the string are fixed, we must have ao  a n1  0 .
Eqn. (3.133) represents a linear difference equation that can be solved for the Eigen frequencies
 r by setting the determinant of the coefficients equal to zero. We therefore have the following
secular determinant:

  0 0 0 
d
 
   0 0 
d d
 
0    0 
d d 0 (3.134)
 
0 0    
d d
0 0 0   
    
    
where we have used

2  m 2 (3.135)
d
This secular determinant is a special case of the general determinant that results if the tensor m 

is diagonal and the tensor k  involves a coupling only between adjacent particles. Thus, Eqn.
(3.134) consists only of diagonal elements plus elements once-removed from the diagonal.
For the case n = 1 (i.e., a single mass suspended between two identical springs), we have   0 ,
or

112
2

md
We may adapt this result to the case of longitudinal motion by replacing  d by k; we then obtain
the familiar expression

2k

m
For the case n = 2, and with  d replaced by k; we have 2  k 2 , or

2k  k

m
which are the same frequencies as those found in the previous Section for two coupled masses.
The secular equation should be relatively easy to solve directly for small values of n, but the
solution becomes quite complicated for large n. In such cases, it is simpler to use the following
method. We try a solution of the form

a j  ae i( j  ) (3.136)

where a is real The use of this device is justified if we can find a quantity  and a phase  such
that the conditions of the problem are all satisfied. Substituting a j in this form into Eqn. (9) and

canceling the phase factor, we find


    
 e i   2  m 2   e i  0
d  d  d
Solving for  2 , we obtain
2   i  i 
2   e  e 
md md  
2
 (1  cos  ) (3.137)
md
4 2 
 sin
md 2
Because we know that the secular determinant is of order n and therefore yields exactly n values
for  2 , we can write

 
r  2 sin r , r = 1, 2, . . . , n (3.138)
md 2

113
We now evaluate the quantity  r and the phase  r by applying the boundary condition that the
ends of the string remain fixed. Thus, we have

a jr  ar e i ( j r  r ) (3.139)

or, because it is only the real part that is physically meaningful,


a jr  a r cos( j r   r ) (3.140)

The boundary condition is


a 0r  a(n 1)r  0 (3.141)

For Eqn. (3.140) to yield a jr  0 for j = 0, it should be clear that  r must be  2 (or some odd

integer multiple thereof). Hence,



a jr  ar cos( j r  )
2
 a r sin j r (3.142)
For j = n + 1, we have
a (n1)r  0  a r sin( n  1) r
Therefore,
(n  1) r  s , s = 1, 2, . . .

or
s
r  , s = 1, 2, . . .
n 1
But there are just n distinct values of  r because Eqn. (3.138) requires n distinct values of  r .
Therefore, the index s runs from 1 to n. Because there is a one-to-one correspondence between
the values of s and the values of r, we can simply replace s in this last expression by the index r:
r
r  , r = 1, 2, . . . , n (3.143)
n 1
The a jr (Eqn. 3.142) then becomes

 r 
a jr  a r sin  j  (3.144)
 n  1

The general solution for q j (see Eqn. 3.66) is

114
q j    r a jr e ir t
r

 r   i r t
   r a r sin  j e
r  n  1
 r  ir t
   r sin  j e (3.145)
r  n  1
where we have written  r   r a r . Furthermore, for the frequency we have

  r 
r  2 sin   (3.146)
md  2(n  1) 

Note that this expression yields the same results found for the case of two coupled oscillators
(Eqns. 3.56) when we insert n = 2, r = 1, 2 and replace  d by k ( = k12 ).
Notice also that if either r  0 or r  n  1 is substituted into Eqn. (3.144), then all the amplitude
factors a jr vanish identically. These values of r therefore refer to null modes. Moreover, if r

takes on the values n  2, n  3, ... 2n  1 , then the a jr are the same (except for a trivial sign

change and in reverse order) as for r = 1, 2,..., n; also, r = 2n + 2 yields the next null mode. We
conclude, therefore, that there are indeed only n distinct modes and that increasing r beyond n
merely duplicates the modes for smaller n. (A similar argument applies for r < 0.)
The normal coordinates of the system (Eqn. 3.68) are
 r (t )   r e ir t (3.147)
so that
 r 
q j (t )   r sin  j  (3.148)
r  n  1
This equation for q j is similar to the previous expression (Eqn. 3.69) except that the quantities

a jr are now replaced by sin  j r  n  1 .

Because  r may be complex, we write for the real part of q j

 r 
real: q j (t )   r sin  j (  r cos  r t   r sin  r t ) (3.149)
r  n  1
where
 r   r  i r (3.150)

115
The initial value of q j (t ) can be obtained from Eqn. (3.149):

 r 
q j (0)    r sin  j  (3.151)
r  n  1
 r 
q j (0)     r r sin  j  (3.152)
r  n  1
If we multiply Eqn. (3.151) by sin  j s  n  1 and sum over j, we find

 s   r   s 
q
j
j (0) sin  j     r sin  j
 n  1  j,r
 sin  j
 n 1  n  1
 (3.153)

Applying the following trigonometric identity for the sine term:


n
 r   s  n 1
 sin  j n  1  sin  j n  1  
j 1 2
 rs , r, s = 1, 2, . . ., n (3.154)

so that Eqn. (3.153) becomes

 s  n 1
q
j
j (0) sin  j    r
 n  1 r 2
 rs

n 1
 s
2
or
2  s 
s  
n 1 j
q j (0) sin  j 
 n  1
(3.155a)

A similar procedure for  s yields

2  s 
s    q j (0) sin  j  (3.155b)
 s (n  1) j  n  1
Thus, we have evaluated all the necessary quantities, and the description of the vibrations of a
loaded string is therefore complete.

3.10 Dissipative Systems And Forced Oscillations

So far in the discussion of small oscillations, we neglected the effects of viscous or frictional
forces. A common situation is one in which the viscous damping forces are proportional to the
first power of the velocity. In such situations, the motion of the ith particle may be described by
Newton's second law as

116
  
mi ri  Fi  ci ri (3.156)
which in component form may by written as
mi xi  Fix  ci x i (3.157a)

mi yi  Fiy  ci y i (3.157b)

mi zi  Fiz  ci z i (3.157c)



where ci are constants and Fix , Fiy , and Fiz are the components of a resultant force Fi that are

derivable from a potential, and the potential is a homogeneous quadratic function of the
coordinates.
Suppose the system has n degrees of freedom and is described by n independent coordinates:
q1 , q 2 ,....., q n . (3.158)
The relations between these and the x, y, and z coordinates are given by the following 3N
equations for N particles.
xi  xi q1 , q 2 ,....., q n 
y i  y i q1 , q 2 ,....., q n  (3.159)
z i  z i q1 , q 2 ,....., q n 
Note that there is no explicit dependence on time t because kinetic energy T is a homogeneous
quadratic function of time. Multiply each of Eqns. (3.157), respectively, by the
quantities xi q j , y i q j , and z i q j ; adding all three and summing over all the N particles

yields
N N

 m x x
i 1
i i i q j  yi y i q j  zi z i q j  =  Fix xi q j  Fiy y i q j  Fiz z i q j  
i 1

 c x
i 1
i i xi q j  y i y i q j  zi z i q j  (3.160)

where

d  T   T



First term on the left 
dt  q j   q
  j


U
First term on the right    Qi , the generalized force,
q j
excluding the dissipative forces

117
 1 N 2  Fr
Second term on the right  
q j
 2 2

 2  ci x i  y i  zi    q
 i 1  j

1 N
and  Fr  
2 i 1
 
ci xi2  y i2  zi2 half the rate at which the energy is being dissipated through the

action of frictional forces. Thus, Eqn. (3.160) may be written as

d  T   T


   U  Fr (3.161)
dt  q j   q
  j

 q j q j

Since L = T - U, Eqn. (3.157) or (3.161) takes form

d  L  L
  Qrj (3.162)
dt  q j  q
 j

where Qrj is the generalized damping force

Fr
Qrj   (3.163)
q j

For sufficiently small motions, the expressions for U, T, and Fr may be written as

U  a11 q12  2a12 q1 q 2  2a13 q1 q3       a nn q n2 (3.164a)

T  b11 q12  2b12 q1 q 2  2b13 q1 q 3      bnn q n2 (3.164b)

Fr  c11 q12  2c12 q1q 2  2c13 q1 q 3       c nn q n2 (3.164c)

where a nn ,     , bnn ,    , and c nn ,    , are constants.

The resulting differential equations of motion obtained from Eqn. (3.161) or (3. 162) are similar
to the undamped case, except that terms of the form q are present. To calculate normal modes,

we must find new coordinates that are linear combinations of q1 , q 2 ,....., q j so that U, T, and Fr ,

when expressed in terms of coordinates 1 , 2 ,....., n , do not contain cross terms; that is, they
contain the sum of the squares of the new coordinates and their time derivatives.
Because of the presence of Fr , it is not always possible to find such new coordinates. In some
situations it is possible to find a normal coordinate transformation, and the resulting differential
equations are of the form
m j j  c j j  k j j  0 (3.165)

which have solutions of the form

118
 j t
 j  Aje cos( j t   j ) (3.166)

Thus, unlike the case of undamped motion in which one observes oscillations, in the present case
the motion may be underdamped, critically damped, or overdamped, as the case may be; hence
the motion may be non-oscillatory. The normal coordinates and their phases are the same as in
the corresponding problem of undamped motion. The amplitude decreases exponentially with
time, while the frequencies are different from the ones in the undamped case.
First, we must assume that the driving forces are small enough so that the squares of the
displacements and velocities will be such that the equations of motion are still linear. If the
forces are constant, such as a system under gravitational force, the only change is in the
equilibrium position about which the oscillations take place. If the driving force is periodic, it is
possible to discuss motion in terms of normal coordinates. For convenience, let us assume that a
single harmonic force of the type Q jext cos t or Q jext e it is applied. The resulting equation of

motion in normal coordinates is of the form (in the presence of a linear restoring force,
dissipative force, and driving force)
m j j  c j j  k j j  Q jext e it (3.167)

If the driving frequency is equal to one of the normal frequencies of the system, the
corresponding normal mode will assume the largest amplitude in the steady state. Furthermore, if
the damping constants are small, not all normal modes are excited to any appreciable extent; only
one normal mode that has the same frequency as the driving force will be excited.
____________________________________________________________________________
Example 3.6. Let us consider once again the situation of two coupled pendula, as discussed
above in Example 3.4. Let us assume that the driving force is F cos t , and the frictional force
proportional to velocity is cx , where c is a constant. Discuss the solution of this problem.
Solution
The equations describing the system are
mg
mx1  x1  k ( x1  x 2 )  cx1  F cos t
l
mg
mx2  x 2  k ( x1  x 2 )  cx 2  F cos t
l

119
Equations involving normal coordinates X 1 and X 2 are (1  X 1  x1  x 2 and

 2  X 2  x1  x 2 )
c g 2F
X1  X 1  X 1  cos t
m l m
c  g 2k 
X 2  X 2    X 2  0
m l m
We should be able to recognize these differential equations, which have the following solutions:
2 F cost   

X 1  e c 2m t A1e i1t  A1e i1t  
m 2 2
o 2 2
  2c 2 
1/ 2

and

X 2  e  c 2m t A2 e i2t  A2 e  i2t 
where
1/ 2 1/ 2 1/ 2
g g c2   g 2k  c 2 
o    , 1    2
 ,  2      2 
l  l 4m   l m  4m 
c
tan   , for g l  c 2 4

m  o2   2  
___________________________________________________________________________

3.11 Check List

Check whether you can answer these questions or not.


 Can you mention the characteristics of small oscillations?
 Can you explain free oscillations?
 Can you elaborate the Orthogonality of Eigenvectors?
 Can you explain normal coordinates?
 Can you find the normal coordinates and modes for identical masses connected by a
spring which has a uniform spring constant?
 Can you explain molecular vibrations?
 Can you manipulate the loaded string?
 Can you describe all the characteristics of dissipative systems and forced oscillations?

120
3.12 Chapter 3 Exercises

3.1. Two blocks of identical masses mare attached by massless springs (with identical spring
constants) as shown in Figure 3.11.

Figure 3.11. Problem 3.1


x1 x2
k k

m m

where x1 and x2 denote departures from equilibrium.


(a). Find the Lagrangian for this system
(b). Derive the Euler-Lagrange equations for x1 and x2 .
(c). Show that the eigenfrequencies for small oscillations for this system are
 k2
 2  3  5,
2
where  k2  k m .

3.2. Assume that you are given a coupled equation of solution given by
 x  3  1 
   B1   cos1t  1   B 2   cos 2 t   2  .
 y  2   5
Determine the normal coordinates and the normal modes.

3.3. For a coupled system that undergo small oscillations, the  r , by definition, are quantities
that undergo oscillation at only one frequency. They may be considered as new
coordinates, called normal coordinates, for the system. Show that the normal
coordinates,  r satisfy equations of the form

r   r2 r  0 r = 1, 2, ..., n


so the equations of motion expressed in normal coordinates become completely
separable.

121
3.4. Two identical harmonic oscillators (with masses M and natural frequencies  o ) are

coupled such that by adding to the system a mass m common to both oscillators the
equations of motion become
x1  m M x2   o2 x1  0
x2  m M x1   o2 x 2  0
Solve this pair of coupled equations, and obtain the frequencies of the normal modes of
the system.

3.5. Consider the electrical circuit in Figure 3.12. Find the characteristic frequencies in terms
of the capacitance C, inductance L, and mutual inductance M. The Kirchhoff circuit
equations are
q1
LI1   MI2  0
C
q
LI2  2  MI1  0
C

Figure 3.12. Problem 3.5.

3.6. Consider a symmetrical linear tri-atomic molecule similar to, for example, CO2, Figure
3.8 The central atom has mass M, and the symmetrical atoms have masses m.
Assume that only longitudinal vibrations (that is, in-line or 1-dimensional vibration) is
possible.
(a). Determine the normal frequencies (or eigenfrequencies) of the system.
(b). Determine the eigenvectors for the zero mode oscillation.

3.7. Consider a loaded string consisting of three particles regularly spaced on the string.
At t = 0 the center particle (only) is displaced a distance a and released from rest.
Describe the subsequent motion.

122
3.8. Show that the equations of motion for longitudinal vibrations of a loaded string are of
exactly the same form as the equations for transverse motion, except that the factor  d
must be replaced by k, the force constant of the string.

123
CHAPTER 4

WAVE PROPAGATION
Introduction

This chapter is a continuation of the study of mechanics of continuous media such as strings.
Because a large number of particles are involved, it is cumbersome to apply the laws of
mechanics and investigate the resultant motion. Some simplifying assumptions must be made
and an overall picture of the motion obtained. First, we investigate transverse vibrations of
strings in one dimension. To start, we shall consider a simple case but then generalize it by the
methods of Lagrange formulation. The main problem involves setting up a wave equation
describing the given situation, followed by solving these differential equations by applying
appropriate boundary conditions.

4.1 Vibrating String

We investigate the propagation of waves along vibrating strings. Our discussion is divided in
two parts: the equation of motion and the general solution (normal modes of vibration).

4.1.1 Equation of Motion

Consider a homogeneous string of length L that is fixed at both ends: x = 0 and x = L. The string
has a linear density (mass per unit length)  and is under tension T throughout the string. The
string is in equilibrium along the X-axis, as shown in Fig. 4. l(a). We are interested in
investigating the motion of such a string following an initial lateral displacement from its
equilibrium position. Also, the displacements of the string are not large enough to change the
tension T appreciably. Furthermore, we assume that the force due to gravity (=  Lg) is small
compared to the tension T and may be neglected. Thus, we mean that there is no longitudinal
displacement of the string. That is, for small displacements of the string, we are concerned only
with the lateral or transverse motion (motion perpendicular to the length of the string), That is,
the string vibrates in the XY plane.
The motion of the string is described by a displacement function u(x, t) of each point x and at

124
an instant of time t. One can find (or drive by him/herself) from elementary wave mechanics that
a differential equation (called wave equation) that describes the motion of the string is given by

Figure 4.1 (a) A string of length L is horizontal when in equilibrium. (b) A small portion
ds of a string under a small displacement results in transverse vibrations.

 2u  2u
  T (4.1)
t 2 x 2
or
 2u   2u
 (4.2)
x 2 T t 2
The dimension of  T is the reciprocal of velocity squared. Hence the wave equation of the
vibrating string is
 2u 1  2u
 0 (4.3)
x 2 v 2 t 2
where

T
v (4.4)

v is not simply a velocity of propagation; it has a much deeper physical interpretation, which we
shall seek later. Here v may be identified as the wave velocity with which the wave propagates
along the string.

125
4.1.2 General Solution: Normal Modes of Vibration

Eqn. (4.3) is a partial differential equation for the function u(x, t) that describes the motion of a
vibrating string. To evaluate the function u(x, t), we make use of initial and boundary conditions.
Suppose at t = 0 the function u(x, t) satisfies the following initial conditions:
u ( x,0)  u 0 ( x ) (4.5)

 u 
 t   u 0 ( x ) (4.6)
t 0

where u0 ( x ) is the displacement and u0 ( x) is the velocity of the string at t = 0, and both are

functions of position x. Since the string is tied at the ends, it must satisfy the following boundary
conditions:
u (0, t )  u ( L, t )  0 (4.7)
That is, the displacement at the ends is zero at all times.
We now proceed to find the solution u(x, t) of the differential equation (Eqn. 4.3) by the method
of separation of variables. Let
u ( x, t )  X ( x )(t ) (4.8)
where X is a function of x alone and  is a function of t alone. Differentiating Eqn. (4.8) and
substituting these in Eqn. (4.3) (and after a little rearranging), we obtain
v 2 d 2 X 1 d 2
 (4.9)
X dx 2  dt 2
The left side of this equation is a function of x only, while the right side is a function of t only.
This is possible for all values of x and t only if each side is equal to a constant. Let this constant
be   2 . The minus sign indicates that the acceleration of the element of the string is always
directed toward the equilibrium position. Thus, from Eqn. (4.9),
v2 d 2 X d 2X 2
2
  2 or  2 X 0 (4.10)
X dx dx 2 v
and
1 d 2 d 2
2
  2 or 2
  2  0 (4.11)
 dt dt
where  may be interpreted as the angular frequency. The solution of Eqn. (4.10) is

126
 
X ( x)  C cos x  D sin x (4.12)
v v
and that of Eqn. (4.11) is
(t )  E cos t  F sin t (4.13)
where C, D, E, and F are the four constants of integration to be evaluated by using the initial and
boundary conditions given by Eqns. (4.5) to (4.7).
Thus, by substituting for X(x) and  (t) from Eqns. (4.12) and (4.13) into Eqn. (4.8), we get the
general solution:
   
u ( x, t )   C cos x  D sin x ( E cos t  F sin t ) (4.14)
 v v 
We may now apply the boundary conditions to evaluate the constants C and D. At x = 0,
u(0, t) = 0 for all values of t; that is X(0) = 0 in Eqn. (4.12):
   
0  C cos 0   D sin  0 
v  v 
which is possible only if C = 0; thus

X (t )  D sin x (4.15)
v
At x = L, u(L, t) = 0 for all values of t; that is, X(L) = 0 in Eqn. (4.15):

0  D sin L (4.16)
v
Since C = 0 and D cannot be zero because that would give a trivial solution, to satisfy Eqn.
(4.16), we must have
 
sin L0 or L  n (4.17)
v v
where n = 1, 2, 3, .... or, replacing  by  n , and v  T 

nv n T
n   (4.18)
L L 
Thus, with C = 0 and letting DE = An, and DF = Bn, we may write Eqn. (4.14) to be
n
u ( x, t )  ( An cos  n t  Bn sin  n t ) sin x (4.19)
L

127
where  n  2 n ,  n being the normal frequencies of vibrations. For a given value of n, we may

write
nx nv nx nv
u ( x, t )  An sin cos t  B n sin sin t (4.20)
L L L L
Eqn. (4.19) or (4.20) represents normal mode of vibration of the string in particular, the nth mode.
The velocity of the normal mode can be obtained by differentiating Eqn. (4.20):
d
u ( x, t )  u ( x, t )
dt
 nx  nv  nv  nx  nv  nv
 An sin     sin t  Bn sin    cos t (4.21)
 L  L  L  L  L  L
We can now evaluate the constants An, and Bn of the nth mode of vibration by using the initial
conditions that, at t = 0,
u ( x,0)  u 0 ( x ) and u ( x,0)  u 0 ( x ) (4.22)

Using these conditions in Eqns. (4.20) and (4.21), respectively, we obtain


nx
u 0 ( x)  An sin (4.23)
L
nv nx
u0 ( x )  Bn sin (4.24)
L L
We know from the theory of differential equations that if u1(x, t) and u2(x, t) are any two
solutions that satisfy the boundary conditions given by Eqn. (4.7), then u(x), which is a linear
combination of u1(x, t) and u2(x, t), that is
u ( x, t )  u1 ( x, t )  u 2 ( x, t ) (4.25)
is also a solution. A more general solution is obtained by adding together all the n particular
solutions using different constants An and Bn corresponding to different frequencies  n . Thus the
general solution of motion of a vibrating string is a linear combination of a large number of
normal modes [from. Eqn. (4.29)] and is given by

 nx nx 
u ( x, t )    An sin cos  n t  Bn sin sin  n t  (4.26)
n 1  L L 
where
nv
n  , n = 1, 2, 3, . . .
L

128
which is a solution containing an infinite number of arbitrary constants. In initial conditions
corresponding to different modes are given, that is, at t = 0
u ( x,0)  u 0 ( x ) and u ( x,0)  u 0 ( x ) (4.27)

then from Eqn. (4.26) we obtain



nx
u 0 ( x)   An sin (4.28)
n 1 L

nv nx
u 0 ( x)   Bn sin (4.29)
n 1 L L
The mode of vibration in which n = 1 is called the fundamental or first harmonic. The mode of
vibration for n = 2 is called the first overtone or second harmonic; similarly, n = 3 corresponds to
the second overtone or third harmonic. The frequency of the nth harmonic is n times that of the
fundamental frequency. In general, a string vibrates with several modes simultaneously.
The general solution given by Eqn. (4.26) consisting of sums of sines and/or cosines is called a
Fourier series. The general solution is completely known if the coefficients An and Bn are known.
These coefficients can be evaluated if the initial conditions, that is, the values of u0 ( x ) and

u0 ( x) , are known. We use the Fourier technique to evaluate these constants. Multiply both sides

of Eqn. (4.28) by sin mx L  , where m is an integer; and integrate from x = 0 to x = L.


L  L
mx nx mx
0 u 0 ( x ) sin
L
dx    An sin
L
sin
L
dx (4.30)
0 n 1

But all the terms on the right will vanish unless m = n. Thus integration yields
L L
nx nx L
 u 0 ( x) sin dx  An  sin 2 dx  An
0
L 0
L 2

or
L
2 nx
An   u 0 ( x) sin dx (4.31)
L0 L

Similarly, multiplying both sides of Eqn. (4.29) by sin mx L  and integrating from x = 0 to x =
L.
L  L
mx nx mx

0 0
u ( x ) sin dx    Bn sin sin dx
L 0 n 1
L L

yields, as before,

129
L
2 nx
Bn   u 0 ( x) sin dx (4.32)
nv 0 L

Thus Eqns. (4.31) and (4.32) state that, if displacements u0 ( x ) and velocity u0 ( x) are given for

all points of the string at one time, An and Bn can be evaluated. Once these constants are known,
the motion of the string at all subsequent times is known.
__________________________________________________________________________
Example 4.1 A string of length L of mass per unit length  is fixed at two ends and is under
tension T. The string is initially displaced a distance h (h << L) at the middle of the string and
then released. Evaluate the Fourier coefficients for the subsequent motion of the string.
Solution
Figure 4.2 shows the initial configuration of the string. Hence the initial conditions are
L u h 2h
For 0  x  ,  or u x (a)
2 x L2 L

L u h 2h
For  x L,  or u (L  x) (b)
2 Lx L 2 L

du 0 ( x )
At t = 0,  u 0 ( x )  0 (c)
dt
Substituting the value of u0 ( x) from Eqn. (c) in Eqn. (4.32) reveals that Bn,= 0 for all n. The

values of An, can be determined by using the initial conditions given by Eqns. (a) and (b).
Substituting these in Eqn. (4.31)

2  2h L / 2 nx 2h L nx 
An    x sin dx   ( L  x ) sin dx  (d)
L L 0 L L L/2 L 
Evaluating integrals for different values of n, we obtain
An  0 , if n is even (e)

8h n
An  2 2
sin , if n is odd (f)
n  L
Substituting these in Eqn. (4.26), we obtain the general solution of the form
8h x nvt 1 8h 3x 3vt 1 8h 5x 5vt
u ( x, t )  2
sin cos  2 2 sin cos  2 2 sin cos     (g)
 L L 3  L L 5  L L

130
Figure 4.2. Example 4.1
___________________________________________________________________________

4.2 Wave Propagation in General

Wave motion is not limited to the vibration of strings. It is a phenomenon that occurs in many
different branches of physics and involves such cases as sound waves, waves on a liquid surface,
and electromagnetic waves. One may be tempted to say that wave motion deals with those
phenomena that exhibit periodicity or oscillations. But this is not always true in general. For
example, a pulse that travels on a rope or a tidal wave does not exhibit periodicity.
A better definition of wave motion is discussed in terms of energy transport; when a wave
reaches a portion of a medium, it sets the particles of the medium into motion. After the wave
has passed the particles come to rest, while neighboring particles are set in motion. From this we
may conclude that one of the common characteristics of all wave motion is the following:
Wave motion provides a mechanism for transfer of energy from one point to another
without physical transfer of any material between the points.
Let us discuss the propagation of a single pulse in one dimension. Consider a stretched rope that
has been shaken at one end, resulting in a pulse traveling along its length and taking the form
shown in Fig. 4.3. This pulse, wave, or disturbance travels along the rope, say along the X-axis,
without distortion in form; that is, it has the same shape at tl as at any other later time, t2. We
have assumed an ideal case in which the form of the pulse does not change. In actual practice,
because of damping, there will be some change in form. The pulse form travels with a constant
velocity. The same remarks can be made about any wave disturbance or wave motion. Thus we
may define it as follows:
Wave motion is a disturbance that propagates itself with constant velocity without
changing its form or pattern.

131
Suppose a pulse or disturbance is traveling along the +X-axis with constant velocity v. Now we
view this pulse from the  -axis, which is moving with velocity v along and parallel to the X-
axis. Furthermore, if the origins of the X-axis and  -axis coincide at t = 0, we may write
  x  vt (4.33)
Thus, to an observer in the  system, the form and position of the disturbance remains
unchanged; that is, the disturbance has such a time dependence that it is a function of  alone.
Thus the wave propagating to the right is
u ( x, t )  f ( )  f ( x  vt ) (4.34)
where f ( ) is a completely arbitrary function. Eqn. (4.34) guarantees that it is a wave traveling
to the right. Thus, as t increases, x must increase so that  remains constant; hence
f ( ) represents a wave traveling to the right. Similarly, we define

Figure 4.3 Pulse in a rope traveling to the right and viewed by an observer moving with
velocity v along an axis parallel to the rope.

  x  vt (4.35)
and a wave propagating to the left is given by
u ( x, t )  g ( )  g ( x  vt ) (4.36)
where g ( ) is another arbitrary function. Once again, as t increases, x must decrease so that  is
constant, and then g ( ) represents a wave propagating to the left. f and g given by Eqns. (4.34)

132
and (4.36) are referred to as wave forms and represent the most general type of one-dimensional
motion.
By direct substitution of f and g from Eqns. (4.34) and (4.36) into Eq. (4.3), we can show that
these satisfy the wave equation.
The general expression for u is a combination of two functions, one of which depends only on 
and the other only on  ; that is, the sum of the two linear functions of  and  [individual
functions f ( ) and g ( ) are also solutions as long as they are linear] is
u ( x, t )  f ( )  g ( )  f ( x  vt )  g ( x  vt ) (4.37)
Thus the most general solution of the wave equation, Eqn. (4.3), is given by Eqn. (4.37) or any
other linear combination of f ( ) and g ( ) . That Eqn. (4.37) is a general solution is consistent
with the fact that the general solution of a second-order differential equation contains two
arbitrary functions.
Let us now proceed to evaluate these functions using initial conditions; that is, at t = 0,
u  u0 ( x) and u  u 0 ( x ) (4.38)

give
u ( x,0)  f ( x)  g ( x)  u 0 ( x ) (4.39)

 u   df df 
and  t    v d  v d   u0 ( x ) (4.40)
t 0   t 0
At t = 0,     x , and Eqn. (4.40) takes the form
d
v  f ( x)  g ( x)  u 0 ( x) (4.41)
dx
which on integration gives
x
1
u 0 ( x )dx  C (4.42)
v 0
 f ( x)  g ( x ) 

Adding and subtracting Eqns. (4.39) and (4.42) [we may replace x by  or  as the solutions
hold for any value of x, and dropping the constant C as it may be eliminated in linear
combinations of solutions], we obtain

1 1 
f ( )  u 0 ( )   u 0 ( )d  (4.43)
2  v0 

133

1 1 
g ( )  u
 0 ( )   u 0 ( )d  (4.44)
2  v0 
To see the connection between the general solution obtained in this section and those in the
previous section for vibrations of strings, we may write the solutions given by Eqns. (4.12) and
(4.13) (instead of in the form of sines and cosines) as in the following alternative form:
X ( x)  Ce i  v  x  De i  v  x (4.45)

(t )  Ee it  Fe it (4.46)


where C, D, E, and F are the constants to be determined from the boundary conditions. Thus the
general solution will be of the form
u ( x, t )  X ( x )(t )  Ae i  v x e it

 Ae i  v ( x vt ) (4.47)
where A is a constant. This states that the general solution u(x, t) is a linear combination of the
following terms:
 i  v ( x  vt ) (4.48)
Note that these solutions already contain the quantities that are functions of x + vt and x  vt . By
taking the real part or by adding the complex conjugate and dividing by 2, we obtain the
solutions

u ( x, t )  A cos ( x  vt ) (4.49)
v

u ( x, t )  A cos ( x  vt ) (4.50)
v
and by adding the imaginary parts or subtracting the complex conjugate and dividing by 2i, we
obtain

u ( x, t )  A sin ( x  vt ) (4.51)
v

u ( x, t )  A sin ( x  vt ) (4.52)
v
The solutions containing x  vt represent waves traveling to the right. While those containing
x  vt represent waves traveling to the left. These solutions do not satisfy the boundary
conditions because they represent traveling waves down the string or medium.

134
Furthermore, these equations are not satisfied by only one particular value of   2 ; many more
are possible. Thus the general solution is not only a linear combination of harmonic terms given
by Eqn. (4.47), but also must be summed over all possible frequencies. Thus the most general
solution is

u ( x, t )   An e i n v ( xvt ) (4.53)


n

Once the boundary conditions are known, the constants can be evaluated in a manner similar to
the case of evaluation of coefficients in infinite Fourier series. For our discussion, we shall write
the solution in the following form, it being understood that the complete solution is summed over
all frequencies:

u ( x, t )  Ae i  v ( x vt ) (4.54)
Using the quantity k, called the wave number, defined as

2 2  2
k  2 or k  or k (4.55)
v v 
Thus the wave equation for X, Eqn. (4.10) and its general solution Eqn. (4.54) take the forms
d2X
2
 k2X  0 (4.56)
dx
and

u ( x, t )  Ae ik ( x vt )  Ae i ( kx t ) (4.57)


Let us see what happens if we superimpose two waves, both of the same frequency and
amplitude, but one traveling to the right and the other to the left. Thus

u  u1  u 2  Ae i ( kxt )  Ae i ( kx t ) (4.58)

u  2 Ae it cos kx (4.59)


The real part of this equation yields
u ( x, t )  2 A cos kx cos t (4.60)
This wave has the property that it does not propagate forward with time. This superposition of
waves leads to the formation of standing waves. There are certain points where there is no
motion at all because of the cancellation of one wave by the other. Such points are called nodes.
Since at nodes no motion is possible, no energy is transmitted from one side to the other; hence

135
the pattern is named standing waves. From Eqn. (4.60), we can obtain the condition for the
position of the nodes to be
 
x  (2n  1)  (2n  1) (4.61)
4 2k

4.3 Lagrange Formulation of a Vibrating String

4.3.1 Energy And Power

If we calculate the kinetic energy and potential energy of a vibrating string, we can set up the
Lagrangian L and the Lagrange equations; hence we can calculate the normal modes of a
vibrating string. Furthermore, we know the total energy stored in the string and also the rate at
which the energy is being transferred from one portion of the string to the other.
Let us reconsider the vibrating string shown in Figure 4.1, which has length L and is fixed at both
ends. As shown in Figure 4.1 (b), the element of length dx when in equilibrium is stretched to
length ds when vibrating. The tension in the string is T when it is vibrating. Thus the amount of
potential energy stored in this vibrating element of the string is, assuming the potential energy to
be zero when the string is unstretched,
 ds 
dU  T (ds  dx)  T   1 dx (4.62)
 dx 
1/ 2
2
ds   u  
where  1     (4.63)
dx   x  

Substituting this in Eqn. (4.62), assuming u x << 1, and using the binomial theorem for
expansion, we obtain
2
T  u 
dU    dx (4.64)
2  x 
Thus the total potential energy stored in the string may be obtained by integrating Eqn. (4.64);
that is,
L 2
T  u 
U     dx (4.65)
2 0  x 

The mass of an element of length dx is dx ; hence its kinetic energy is (in order to avoid
confusion we will start using K for kinetic energy instead of T )

136
2
1  u 
dK  dx  (4.66)
2  t 
while the total kinetic energy of the string is obtained by integrating Eqn. (4.66):
L 2
  u 
K   dx
2 0  t 
(4.67)

To evaluate U and K, we make use of the solution given by Eqn. (4.19),


n x
u ( x, t )   n (t ) sin (4.68)
v
where
 n (t )  An cos  n t  Bn sin  n t (4.69)
and we have used the relation given in Eqn. (4.18),

nv n T
n   (4.18)
L L 
Thus, from Eqns. (4.68) and (4.18), we obtain (for all solutions)
u   nx
  n n cos (4.70)
x L n 1 L
u   nx
   n sin (4.71)
t n1 L
Substituting Eqn. (4.70) into Eqn. (4.65), we get
L
 2T     nx mx 
U    nm  m  cos cos dx  (4.72)
2 L2 
m 1 
n
L L
n 1 0 
On integrating, we find that only those terms are nonzero for which m = n, and each of these
terms on integration yields L/2. Thus
 2T 
2 2
U
4L
n 
n 1
n , n = 1, 2, 3, . . . (4.73)

Similarly, substituting Eqn. (4.71) into Eqn. (4.67), we obtain

     L nx mx 
K     n  m  sin sin dx  (4.74)
2 n 1 
m 1  L L
0 
Once again, on integrating we find that only those terms are nonzero for which m = n, and each
of those terms on integration yields

137
L   2
K  n
4 n 1
(4.75)

while the Lagrangian of the system may be written as

1    2  2T 2 2 
L  K U    L n  L n  n 
4 n 1 
(4.76)

Note that the potential energy is the sum of quantities of the form An  2n , and the kinetic energy
 2 . The Lagrangian equations
has terms of the form Bn  n

d  L  L
   0 (4.77)
dt  

n   n
take the form
2
   T n 2   0
 (4.78)
n n
L2
where  n is the dependent variable and t is the independent variable. The solutions of these

yield the normal coordinates  n . Since n varies from 1 to  , the number of normal coordinates

for a vibrating string is infinite.


It is simple to write the total energy E using Eqns. (4.73) and (4.75) as

1    2  2T 2 2 
E  K  U    L n  n  n  (4.79)
4 n1  L 
Since L  M (the mass of the string), and from Eqn. (4.18),

L2 n2 ML
T 2 2
 2 2  n2 (4.80)
n n
and, using Eqn. (4.69), we may write Eqn. (4.79) in the form

M
E
4
  A
n 1
2
n
2
n  Bn2  (4.81)

where An and Bn are constants.

4.3.2 Power

The rate of flow of energy, that is, power P, delivered from the left to right across any point x
along the string. To calculate power, we make use of the definition that P  Fu , where F is the

138

magnitude of the driving force F . F is equal in magnitude to tension T and must be applied in a

direction tangent to the string. Thus the component of F in the direction of transverse
displacement at point x is
u
F y  T (4.82)
x
while the component of velocity u at the point x is u t . Therefore,

 u  u 
P  Fy u    T   (4.83)
 x  t 
The value of P can be evaluated by using the values of u x and u t given by Eqns. (4.70)
and (4.71), respectively.
Example. Let us calculate P for a particular case. Consider a wave traveling to the right and
given by
u  f ( x  vt )  f ( ) (4.84)
Suppose f is a sinusoidal function of the form
u  f ( )  A cos(kx  t ) (4.85)
Evaluating u x and u t and substituting in Eqn. (4.83) yields

P  kTA 2 sin 2 (kx  t ) (4.86)


the average power P transmitted from left to right will be
1
P  kTA 2 (4.87)
2

4.4 Behavior of a Wave at Discontinuity (Reflection & Transmission):


Energy Flow

As an example of discontinuity, consider two semi-infinite strings of different linear mass


densities tied together at x = 0, as shown in Figure 4.4. The string that extends over    x  0
has a linear mass density 1 , and the wave traveling along this string has a velocity v1 , while the

string that extends over 0  x   has a linear mass density  2 , and the wave traveling along
this string has a velocity v 2 . Let the tension in the string be  . We want to investigate the effect
of a sudden change in density at x = 0 on a continuous harmonic wave.
Let an incident wave traveling from the left for x < 0 be represented by

139
u I  AI cosk1 x  t  (4.88)

where AI is the amplitude of the incident wave, k1   v1 , v1 being the wave velocity. When this
wave reaches x = 0, the point where the two strings join (the point of discontinuity), part of the
wave is reflected back along the first string, while the remaining wave is transmitted. The
reflected wave is represented by
u R  AR cosk1 x  t  (4.89)
where AR is the amplitude of the reflected wave. The transmitted wave is given by
uT  AT cosk 2 x  t  (4.90)
where AT is the amplitude of the transmitted wave and k 2   v 2 , v 2 being the velocity of the
wave on the second string to the right of x = 0. Note that we could have used solutions of the
following form:
i ( k1 x t )
u I  Re AI e (4.91)
where Re stands for the real part of the expression.
Our aim is to evaluate the reflected and transmitted amplitudes AR and AT in terms of the incident
amplitude AI. This can be done by imposing the boundary conditions that at the junction of the
two string (x = 0) the displacement u and its derivative u x must be continuous. These are the
continuity conditions and are valid for any other types of wave motion. The first condition
satisfies the requirement that there is no break in the string, while the second condition implies
that the restoring force resulting from a displacement y is the same on each side of the junction.
If this were not true, then a finite force acting on a vanishing small mass element would produce
an infinite acceleration. Thus the boundary conditions may be written as
uI  uR x 0
 uT x 0
(4.92)

and

 u I u R   u 
    T  (4.93)
 x x  x0  x  x 0

Using Eqns. (4.88), (4.89), and (4.90), the continuity of u, Eqn. (4.92), yields
AI  AR  AT (4.94)
while the continuity of u x , given by Eqn. (4.93), yields

k1  AI  AR   k 2 AT (4.95)

140
Figure 4.4. Two semi-infinite strings of different linear mass densities
tied together at x = 0.

Solving these two equations for AR AI and AT AI ,

AR k1  k 2
 (4.96)
AI k1  k 2

AT 2k1
 (4.97)
AI k1  k 2

Since k   v and v    , we may write these results as

AR v 2  v1 1   2
  (4.98)
AI v1  v 2 1   2

AT 2v 2 2 1
  (4.99)
AI v1  v 2 1   2

It is clear that the ratio AT AI is always positive; hence the transmitted wave is always in phase
with the incident wave. If the second medium is lighter, v 2  v1 or  2  1 , the ratio AR AI
will be positive; hence the reflected wave will be in phase with the incident wave. On the other
hand, if the second medium is denser than the first, v 2  v1 or  2  1 , AR AI will be negative.
This means that the reflected wave is out of phase by  with respect to the incident wave. This
type of behavior is typical of many kinds of wave motion.
The intensity, the rate of energy flow, for any type of wave motion is proportional to the square
of the amplitude. For this purpose, we define the reflection coefficient, R, to be the fraction of the
incident energy that is reflected back; that is,

141
2 2 2
A   k1  k 2  v v 
R   R      2 1  (4.100)
 AI   k1  k 2   v1  v 2 
While the transmission coefficient, T defined as the fraction of the incident energy that is
transmitted, must satisfy the condition
R T 1 (4.101)
or
4v1v 2
T  1 R  (4.102)
(v1  v 2 ) 2

[Note: From Eqn. (4.100), R becomes larger and larger as the difference between v1 and v 2
becomes larger, while correspondingly T becomes smaller.]
Finally, let us calculate the rate of energy flow dE dt across the junction at x = 0. This is equal
to the work done by the adjacent portion of the string on the particle at x = 0 and is equal to the
product of the restoring force   u x  and the velocity of the particle u t both evaluated at
x = 0. Then
dE  u   u 
     (4.103)
dt  x  x 0  t  x 0
If we want to calculate the energy transmitted to the left of the string at x = 0, we let
u  u I  uR
 AI cos(k1 x  t )  AR cos(k1 x  t ) (4.104)
Substituting this in Eqn. (4.103), we get
 dE 

2 2 2
  k1 AI  AR sin t   (4.105)
 dt  
Similarly, if we use
u  uT  AT cos(k 2 x  t ) (4.106)
in Eqn. (4.103), we get energy transmitted to the right as
 dE  2 2
   k 2AT sin t (4.107)
 dt  
1
Since the average value of sin 2 t over one complete cycle is , we may write Eqn. (4.105) as
2

142
 dE   1 2 1 2
    k 2AI  k1AR (4.108)
 dt    ave 2 2

where the first term on the right is the mean rate at which the energy is incident on the junction,
while the second term is the mean rate at which the energy is reflected back. Similarly, the mean
rate at which the energy is transmitted, from Eqn. (4.107), is
 dE   1 2
    k 2AT (4.109)
 dt    ave 2
This is the net rate at which the energy is supplied to the junction from left to right.

4.5 Wave Interference

Interference is what happens when two or more waves come together. Depending on how the
peaks and troughs of the waves are matched up, the waves might add together or they can
partially or even completely cancel each other. This is the result of superposition principle. The
principle of linear superposition says that- when two or more waves come together, the result is
the sum of the individual waves. Thus we can simply state that wave interference is the
phenomenon that occurs when two waves meet while traveling along the same medium.

Although the waves interfere with each other when they meet, they continue traveling as if they
had never encountered each other. When the waves move away from the point where they came
together, in other words, their form and motion is the same as it was before they came together.
This means, the waves pass through each other without being disturbed.

The principle of linear superposition applies to any number of waves, but to simplify matters, we
just consider what happens when two waves come together. For example, this could be sound
reaching you simultaneously from two different sources, or two pulses traveling towards each
other along a string. When the waves come together, what happens? The result is that the waves
are superimposed: they add together, with the amplitude at any point being the addition of the
amplitudes of the individual waves at that point. For the case of mechanical waves, the net
displacement of the medium at any point in space or time, is simply the sum of the individual
wave displacements. This is true of waves which are finite in length (wave pulses) or which are
continuous sine waves. We'll discuss interference as it applies to sinusoidal traveling waves, such

143
as sound waves, but it applies to other waves as well, including light waves, because most of the
mathematics and reasoning of mechanical waves applies to light waves.

Suppose that we have a situation where two real sinusoidal traveling waves
u1 ( z , t )  A1 cos(kz  t   1 ) (4.110a)
and
u 2 ( z , t )  A2 cos(kz  t   2 ) (4.110b)
are simultaneously present at the same point z, and have the same angular frequency  and thus
same wavelength λ, and wavenumber k but have different amplitudes A1, A2 and (absolute)
phases δ1, δ 2 {defined relative to a common chosen origin of time, t = 0}.
We can simply add the two waves together:
u 3 ( z , t )  u1 ( z , t )  u 2 ( z , t ) , (4.111)

however, this approach will involve some rather tedious algebra and use of trigonometric
identities to obtain A much easier method is to carry this out using complex notation:

u 3 ( z , t )  u1 ( z , t )  u 2 ( z , t )  Reu~1 ( z , t )  Reu~2 ( z , t )  Reu~3 ( z , t ) (4.112a)

with
~
u~3 ( z , t )  u~1 ( z , t )  u~2 ( z, t )  A3e i (kz t ) . (4.112b)

Thus
~ ~ ~
A3e i ( kzt )  A1e i ( kzt )  A2 e i ( kz t )
~ ~ ~ (4.113)
 A A A
3 1 2

or
A3e i 3  A1e i1  A2 e i 2 (4.114)

Writing the last relation out in its explicit complex form:


A3 cos  3  iA3 sin  3  A1 cos  1  iA1 sin  1  A2 cos  2  iA2 sin  2

Thus we see that


~
 
Re A3  A3 cos  3  A1 cos  1  A2 cos  2 (4.115a)
~
ImA   A sin 
3 3 3  A1 sin  1  A2 sin  2 (4.115b)

144
We can either use the so called phasor-diagram in the complex plane and or
wade through the tedious trigonometry and algebra.

Figure 4.5. Phasor diagram.

The use of phasor diagram doesn’t allow us to evade the use of algebra and trigonometry. What
we are essentially doing here is nothing more than adding two 2-dimensional vectors together;
i.e.,
  
C  A B
where

A  a x xˆ  a y yˆ  a cos 1 xˆ  a sin 1 yˆ ,

B  bx xˆ  b y yˆ  b cos  2 xˆ  b sin  2 yˆ , (4.116a)

C  c x xˆ  c y yˆ  c cos  3 xˆ  c sin  3 yˆ.

Then
c x  c cos  3  a x  bx  a cos 1  b cos  2

and c y  c sin  3  a y  b y  a sin  1  b sin  2 (4.116b)

  
The magnitude of A , B and C are
  
A  a  a x2  a 2y , B  b  bx2  b y2 , C  c  c x2  c 2y

The phase angles are  1  tan 1 a y a x  ,  2  tan 1 b y bx  and  3  tan 1 c y c x  . Thus, for the

addition of two complex amplitudes, we see that:

145
~ ~ ~*
A3  A3  A3  A~  A~  A~  A~ 
1 2 1 2
*
 A e
1
i 1

 A2 e i 2  A1e i1  A2 e i 2 
  
A12  A22  A1 A2 e i1 e i 2  e i1 e i 2  A12  A22  A1 A2 2 cos( 2   1 )

 A12  A22  2 A1 A2 cos( 12 ) (4.117)

where  12 is ( 2   1 )

The absolute phase angle  3 can be obtained from:

A3 sin  3 sin  3 A sin  1  A2 sin  2


tan  3    1
A3 cos  3 cos  3 A1 cos  1  A2 cos  2
i.e.,
 A sin  1  A2 sin  2 
 3  tan 1  1  (4.118)
 A1 cos  1  A2 cos  2 
Note also that the use of complex notation allows us to explicitly describe properly
mathematically the phase-shifts that can / do occur in the response of a system (a “black box”) to
an input stimulus / input signal:

Figure 4.6. A phasor diagram representing phase-shifts that can occur in


the response of a system

Consider two traveling waves interfering with each other in a non-dispersive medium with
different frequencies and amplitudes. For non-dispersive medium, this means that
v  f11  f 2  2  1 k1   2 k 2 with angular frequencies of 1  2f 1 and  2  2f 2 and
wavenumbers
Then

146
u~tot ( z , t )  u~1 ( z , t )  u~2 ( z , t )  A1e i ( k1 z 1t 1 )  A2 e i ( k2 z 2t  2 ) (4.119)

Easy cases of phase relations between the two waves:


1.  1   2 (in phase). Then  12   2   1  0 radians = 00.

Phasor diagram

Constructive Interference

2.  2   1   (1800 out of phase). Then  12   2   1   radians.


A2
Destructive Interference
AR A1
3. The General case

From the diagram we see that the magnitude of the resultant or net amplitude AR is:

AR u~tot ( z , t )  A12  A22  2 A1 A2 cos( 12 )

is simply the law of cosines.

AR  u~tot ( z , t )  u~tot
*
( z, t ) (4.120)

If two waves have equal amplitudes; i.e., A1 = A2 = A


1.  1   2 (in phase). Then  12   2   1  0 and hence cos( 2   1 )  1 , which implies
constructive interference. Resultant amplitude is

AR  A12  A22  2 A1 A2 cos( 12 )  4 A 2  2 A

147
2.  2   1   (1800 out of phase). Then  12   2   1   radians, cos( 2   1 )  1 ,
which implies destructive interference. Resultant amplitude is

AR  A12  A22  2 A1 A2 cos   0


Note that classical wave interference can/do occur at amplitude level even if e.g. sound
waves on strings
Note also that amplitude interference effects occur in the world of quantum mechanics – i.e.
matter waves – but is somewhat more complicated – e.g. by line width effects and/or uncertainty
principle effects. Only identical particles with the exact same quantum numbers (external and
internal) can interfere with each other…

4.6 Wave Polarization

Depending (largely) on the type of wave and the nature of the medium that the waves are
propagating through, the waves can have another degree of freedom known as polarization.
When there are two or more waves superimposed in a region the situation becomes more
 
complicated. For simplicity, we can restrict our attention to two waves, say u1 and u 2 . When the
waves are collinear, the resulting wave will simply combine to form another linearly polarized
wave. However, if the two waves are not collinear, then the second wave can be split into
collinear and perpendicular components. The collinear component will form a linearly polarized
wave again. The resultant wave formed by the perpendicular components can have various states
of polarization.

Waves propagating in  ẑ direction with small transverse displacement amplitude on a string are
known as transverse waves because the displacement of string (relative to its equilibrium

position) is transverse to the direction of propagation of the wave; i.e., ( v prop   zˆ direction).

Thus the transverse displacement amplitude u ( z, t )  u e i ( kz t ) is oriented, for example, in


o
 x̂ and/or  ŷ directions for a traveling transverse wave propagating along the  ẑ direction.
Thus transverse traveling waves have two polarization states, either the  x̂ or the  ŷ direction,
or equivalently 2 orthogonal {i.e., mutually-perpendicular} linear combinations of the  x̂ and
 ŷ basis states for waves propagating in the  ẑ direction:

148
Figure 4.7. Linear combination of two orthogonal waves.

Propagation of, for example, longitudinal sound waves in solid or non-solid media (example;
normal gases, liquids and solids) also have longitudinal polarization – because longitudinal
sound waves have longitudinal displacements of atoms / molecules – i.e., along / against (i.e.
parallel/anti-parallel to) the direction of propagation of the longitudinal wave, e.g. in the  ẑ
direction.

Here the longitudinal displacement amplitude is also of the form u ( z, t )  u e i ( kz t ) .


o
Longitudinal traveling waves have only one polarization state, e.g. the z direction.
Both longitudinal and transverse waves obey the same wave equation.
Longitudinal waves: e.g. sound waves/acoustic waves-liquids, gases and solids and
e.g. large amplitude’s in strings (compression waves)
Transverse waves: e.g. small and large amplitudes in strings, long solid rods and solid bars, etc.,
(shear waves). EM waves are transverse waves.
Two orthogonal polarization states for transverse waves thus we can represent the transverse
displacement amplitude as a vector quantity, indicating its polarization state.
Transverse displacement e.g.is in

Vertical plane → “vertical polarization” (up and down) u x ( z , t )  u o, x cos(kz  t ) xˆ .

Horizontal plane →” horizontal” polarization (sideways) u y ( z , t )  u o, y cos(kz  t   ) yˆ .

The polarization unit vector n̂ in the transverse plane defines the plane of polarization: (i.e.,
plane of transverse vibrations).
Note that nˆ  zˆ  0
Define the polarization angle  with respect to x̂ axis.

149
Figure 4.8. Transverse diaplacements.

nˆ  cos  x̂  sin  ŷ

Figure 4.9. Decomposition of a polarization unit vector n̂ into two orthogonal


unit vectors x̂ and ŷ .

4.6.1 Linear Polarization

Start with two orthogonal waves



u x ( z , t )  u o, x cos(kz  t ) xˆ (4.121)

and


u y ( z , t )  u o, y cos(kz  t   ) yˆ (4.122)

150
where  is the relative phase difference between the waves, both of which are traveling in the z
direction. The resulting superposition of these waves is

  
u P ( z, t )  u x ( z, t )  u y ( z, t ) . (4.123)

If   2m , m  0,  1,  2,  3,.... , the waves are said to be in phase. In this case,


u P  u o, x xˆ  u o, y yˆ cos(kz  t ) . (4.124)

This wave has a fixed amplitude equal to u o , x xˆ  u o , y yˆ  , which shows that this wave is also

linearly polarized. Similarly, if   (2m  1) , m  0,  1,  2,  3,.... , the resultant wave is


again linearly polarized, but now the original waves are said to be out of phase.

4.6.2 Circular Polarization

Another special case comes about when both of the original waves have equal amplitude and a
relative phase difference of   2m   2 , m  0,  1,  2,  3,.... . In this case

u x ( z , t )  u o cos(kz  t ) xˆ

and


u y ( z , t )  u o sin( kz  t ) yˆ ,

so that the resulting wave is given by


u R  u o cos(kz  t ) xˆ  sin( kz  t ) yˆ  (4.125)

Notice that the scalar amplitude, Ao, is a constant, but the direction of the amplitude varies with

time. At a fixed position in space, u ( z, t ) rotates counter clock wise CW (in the x-y plane) as
time increases for a wave propagating in the ±direction. This implies that the amplitude is not
restricted to a single plane as before, but instead rotates so that the axis of rotation (as given by
the right hand rule) is opposite the direction of motion. The angular frequency of the rotation
is  . This wave is said to be right circularly polarized and is depicted as in Figure 4.10.

151
Figure 4.10. Right-circular polarization.

In a similar way, if   2m   2 , m  0,  1,  2,  3,.... , then the resulting wave is


u L  u o cos(kz  t ) xˆ  sin( kz  t ) yˆ  (4.126)

In this case the wave rotates with the axis of rotation in the same direction as the motion, and the
wave is said to be left circularly polarized and is depicted as in Figure 4.11.

Figure 4.11. Left-circular polarization.

As an interesting aside, a linearly polarized wave can be constructed from two oppositely
polarized waves,

  
uP  uR  uL


u P  u o cos(kz  t ) xˆ  sin( kz  t ) yˆ   u o cos(kz  t ) xˆ  sin( kz  t ) yˆ 

 2u o cos(kz  t ) xˆ (4.127)

152
This wave has an amplitude of twice the original wave and is linearly polarized in the x plane.

4.6.3 Elliptical Polarization

The three polarization states that we have found so far are really special cases of elliptically
polarized wave. To find the magnitude of elliptical polarization, start with Eqn. (4.123)
u P  u o, x cos(kz  t )  u o, y cos(kz  t   )

 u o, x cos(kz  t )  u o, y cos(kz  t ) cos   sin( kz  t ) sin  


(4.128)
 u o, x  u o, y cos  cos(kz  t )  u o, y sin( kz  t ) sin 

From Eqn. (4.121), we have that

ux
cos(kz  t ) 
u o, x

so that Eqn. (4.128) becomes

1/ 2
2
u   u  
u E  u o, x  u o, y cos   x  u o, y 1   x 

 sin 
u o, x   u x ,o  
 
1/ 2
(4.129)
2
uo, y   u  
 ux  ux cos   u o, y 1   x 

 sin 
uo ,x   uo, x  
 

Since u E  u x  u y , this can be solved to yield

1/ 2
2
uy ux   u  
 cos   1   x 

 sin 
u o, y u o, x   u o, x  
 
1/ 2
2
uy u   u  
 x cos   1   x 

 sin 
u o, y u o, x   u o, x  
 

Squaring both sides and rearranging yields

2 2
 uy   ux   u  u x 
     2 y   cos   sin 2  . (4.130)
u  u  u  u 
 o, y   o, x   y ,o  x ,o 

153
This is the equation of an ellipse making an angle  with the ( u x , u y ) coordinate system such

that

2u x ,o u y ,o cos 
tan 2  . (4.131)
u x2,o  u 2y ,o

If we rotate the ( u x , u y ) coordinate system through an angle  , this takes on the more familiar

form,

2 2
 uy   ux 
     1. (4.132)
u  u 
 y ,o   x ,o 

To see that both circularly polarized and plane polarized waves are special cases of Eqn.
2m  1
(4.130), consider the following. If    , m  0,  1,  2,  3,.... , and u x ,o  u y ,o  u o ,
2
then Eqn. (4.130) becomes

u x2  u y2  u o2 (4.133)

which is the equation of a circle.

Similarly, if   m , m  0,  1,  2,  3,.... , Eqn. (4.130) is

u y ,o
uy   ux (4.134)
u x ,o

u y ,o
This is the equation of a straight line with a slope of  .
u x ,o

Using our results so far, we can now describe a particular wave in terms of its specific state of
polarization. If the wave is linearly, or plane, polarized, then we say it is in the P-state. Wave
that is in right (left) circularly polarized is in the R- (L-) state. Finally, elliptically polarized
wave is referred to as being in the E-state.

154
4.7 Check List

Check whether you can answer these questions or not.


 Can you explain the difference between dissipative and dispersive mediums?
 Can you drive the wave equation by considering a transversely displaced string from
equilibrium position?
 Can you verify Newton’s 2nd law using a mechanical wave on transversely displaced
string?
 Can you explain standing waves and superposition principle?
 Using complex notations can manipulate the linear superposition of two sinusoidal
waves?
 Can you show the difference between constructive and destructive interference using
wave intensity?
 Can you find the complex amplitudes for reflected and transmitted waves in terms of
complex amplitude of incident wave?
 Can you describe the difference between linearly, circularly and elliptically polarized
waves?

4.8 Chapter 4 Exercises

4.1. A string of length L and mass m is tied at both ends. The midpoint of the string is pulled
a distance h  L 10 in the vertical direction and released. Find an expression that
describes the motion of the String.

4.2. A uniform string of length L and linear mass density µ under tension T, is displaced
initially ( ), as shown in Figure 4.12. Find the general solution of the equation that
describes the motion of the vibrating string and evaluate the coefficients by using initial
conditions.

Figure 4.12. For problem 4.2.

155
4.3. A uniform string of length L and linear mass density  , under tension  , is initially in an
equilibrium position but has a velocity given by
L
v  ax , 0 x
2
L
v  a ( x  L) , xL
2
where a is a constant. Find the general solution of the equation that describes the motion
of the vibrating string and evaluate the coefficients by using initial conditions. Make
graphs to describe the nature of the vibrating string.

4.4. A string of length L and mass m is tied at x  0 and the end x  L is tied to a ring that
slides without friction on a vertical rod. Show that the boundary condition at end x  L is
u x  x L  0 and find the normal frequencies and normal modes of vibrations.

4.5. Calculate the characteristic frequencies and its amplitudes for different modes for a
vibrating string under the following initial conditions:
4 x( L  x)
u ( x,0)  and u ( x,0)  0 .
L2

4.6 A transversely vibrating string produces a wave that propagates along the string, and the
general wave equation for such wave is given by
 2u 1  2u
 0
x 2 v 2 t 2
where u ( x, t ) represents the transverse displacement of each point x on the string at an
instant of time t, and v is the wave velocity with which the wave propagates along the
string.
For a homogeneous string of length L that is fixed at both ends, find the general solution
for the nth normal mode of vibration of the string. [Hint: use the boundary conditions
u (0, t )  u ( L, t )  0 to evaluate the constants for all values of t.]

4.7. Explain polarization in case harmonic waves with circular and elliptical form.

156
REFERENCES

1. Walter Hauser; Introduction to principles of mechanics; Addison Wesley, 1966.


2. Stephen T. Thornton, Jerry B. Marion; Classical Dynamics Of Particles And Systems,
Fifth Edition; Brooks/Cole -- Thomson Learning, 2004.
3. David Hestenes; New Foundations for Classical Mechanics, Second Edition; Kluwer
Academic Publishers,
4. Gerd Baumann, Mathematica for Theoretical Physics. Classical Mechanics and
Nonlinear Dynamics, Second Edition; Springer.
5. David Morin; Introduction to Classical Mechanics, with Problems and Solutions, Second
Edition;
6. Tom W.B. Kibble, Frank H. Berkshire; Classical Mechanics, 5th edition; Imperial
College Press (London),
7. Murrey R. Speigle; Theory and problems of theatrical mechanics; Schaum’s Outline
series
8. R. Taylor; Classical Mechanics; Universal Science, 2005
9. H. Goldstein; Classical Mechanics, Third Edition; Addison Welsey, 2001.
10. K. R. Symon; Mechanics, Third Edition; Addison Welsey, 1971.

157

You might also like