Seminar 1: LAGRANGES EQUATIONS

Problem 1. Atwoods Machine

Atwoods machine consists of two weights of mass m1 and m2 connecting by a string

of length l that passes over a pulley of a radius a and moment of inertia I (see Figure
in the Set of Problems). The string is assumed massless and inextensible, and the
pulley is frictionless. The number of degrees of freedom is obviously one, while the
total number of the coordinates describing the positions of the two masses are six.
However there exist five holonomic constraints: four of them prevents motion of the
masses in y and z directions, whereas the fifth has the form

x1 + x2 + a = l, (1)

where x1 and x2 are the vertical positions of each mass relative to the center of the
pulley. So the number of the generalized coordinates must be 6 5 = 1. The natural
choice for this single generalized coordinate is x1 x. The Lagrangian is expressed
through this coordinate as follows:
2
T = 21 m1 x 2 + 12 m2 x 2 + 12 I xa 2
V = m1 gx m2 g(l a x)
!
1 I
L= 2
m1 + m2 + a2
x 2 + (m1 m2 )gx + m2 g(l a). (2)

Lagranges equations:

d L L
dt x
= x
!
I
m1 + m2 + a2
x = (m1 m2 )g. (3)

The latter of these equations gives the final acceleration of the system in the form
(m1 m2 )g
x = . (4)
m1 + m2 + aI2
As we see, if m1 > m2 , then m1 falls with constant acceleration while m2 rises with the
same acceleration. If m1 < m2 , the converse is true. At last, if m1 = m2 , each mass
remains at rest (or moves at constant velocity). Of course, you are well familiar with
these conclusions from elementary physics.
 Problem 2. Double Atwoods Machine

Let replace one of the weights in the simple Atwood machine by a second simple
Atwood machine. Then we obtain the system which is known as a compound, or a
double Atwood machine that is shown in Figure from the Set of Problems. For the
motion of this system there are two degrees of freedom: one is the freedom of mass 1
(and the attached movable pulley) to move up and down about the fixed pulley, and the
second one is the freedom of mass 2 (and the attached mass 3) to move up and down
about the movable pulley. In general, to describe the configuration of the system we
need to have 12 coordinates (3 for each mass m1 , m2 , m3 plus 3 for a movable pulley).
Thus there must be 12-2=10 constraints. 8 of those constraints limits the motion of all
the components of the machine to only a single direction. To formulate the remaining
two constraints we assume for simplicity that the pulleys are massless, and their radii
are small compared with the lengths of the constraining strings l and l0 . Then the
constraints can be written in a simplified form as

(xp + x1 ) l = 0 and (2x1 + x2 + x3 ) (2l + l0 ) = 0, (5)

where xi and xp are the vertical positions of the masses and movable pulley relative
to the center of the fixed pulley (Note that the second constraint is coming from the
formula (x2 xp )+(x3 xp ) = l0 and by using the first constraint in the form xp = lx1 ).
Let choose the generalized coordinates x and x0 , as shown in Figure. Then the
kinetic and potential energies as well as the resultant Lagrangian can be written down
as follows:
0 2 0 2

1 2 1 1
T = 2 m1 x + 2 m2 (x + x ) + 2 m3 (x x )

V = m gx m g(l x + x0 ) m g(l x + l0 x0 )
1 2 3
L = 1 m x 2 + 1 m (x + x 0 )2 + 1 m (x + x 0 )2 + (m m m )gx + (m m )gx0 + const

2 1 2 2 2 3 1 2 3 2 3
(6)
Lagranges equations (
d L
dt x
= L
x
,
d L L (7)
dt x 0
= x0
yield
x x0 ) + m3 (
m1 x + m2 ( x + x0 ) = (m1 m2 m3 )g

0 (8)
m2 (x + x ) + m3 (x + x0 ) = (m2 m3 )g
The accelerations can be found from an algebraic solution of this system of equations
as
m1 m3 4m2 m3 + m1 m2
x = g (9)
m1 m3 + 4m2 m3 + m1 m2
and
2m1 (m2 m3 )
x0 = g (10)
m1 m3 + 4m2 m3 + m1 m2
We can see if m1 = m and m2 = m3 = m2 than x = 0 and x0 = 0.
Problem 3. Particle Sliding on a Movable Inclined Plane

Consider a particle of mass m which is allowed to slide free along an inclined plane
of mass M . As shown in Figure from the Set of Problems, the inclined plane itself
is not fixed but is also free to slide along a horizontal surface. Since both objects
are constrained to move along a single dimension, in this case there are two degrees
of freedom. So to describe the configuration of the system we need two generalized
coordinates: one to specify the position of the inclined plane relative to some reference
point on the motionless horizontal surface and the other to specify the position of the
mass m on the inclined plane relative, say, to the top of the plane. We denote those
coordinates x and x0 , respectively. To calculate the total kinetic energy of the system
we introduce the velocity of the plane,

V = ex x,
(11)

and the velocity of the mass m in the laboratory reference,

v = V + v0 = ex x + e x 0 , (12)

where ex and e are the unit vectors in the surface and in the plane, the latter being
directed down the plane at an angle relative to the horizontal surface.
The total kinetic energy is
T = TM + Tm , (13)

where

TM = 21 M V V = 12 M x 2
Tm = 21 mv v = 21 m(ex x + e x 0 ) (ex x + e x 0 ) = 21 m(x 2 + x 02 + 2x x 0 cos ) (14)

The expression for the potential energy depend on the choice of zero point. If we choose
it to be the top of the plane, then we may write

V = mgx0 sin . (15)

Therefore, the Lagrangian of the system is

1 1
L = M x 2 + m(x 2 + x 02 + 2x x 0 cos ) + mgx0 sin , (16)
2 2
and the equations of motion are
(
d L
dt x
= L
x
d L L (17)
dt x 0
= x 0,

or (
d
dt
[m(x + x 0 cos ) + M x]
=0
d 0 (18)
dt
(x + x cos ) = g sin .
We notice that the time derivative of the quantity

[m(x + x 0 cos ) + M x]
(19)

iz zero. This quantity is therefore a constant of motion. Close examination of this

quantity reveals that it is the total linear momentum of the system in the x direction.
From the Newtonian viewpoint, it means that there is no net force on the system in
the x direction - and this result seemingly falls out of the Lagrangian formalism! Note
also that this result follows from the fact that the Lagrangian is independent of the
L
coordinate x, i.e., x
= 0.
Carring out the time derivatives in (18), we obtain the equations
x + x0 cos ) + M x = 0

m(
(20)
x0 + x cos = g sin .
Finally, solving these equations for x and x0 , we find the accelerations
g sin cos
x = (21)
(m + M )/m cos2
and
g sin
x0 = (22)
1 m cos2 /(m + M )
This particular example illustrates clearly the ease with which quite complicated
problems of mechanics fall apart when attacked within the Lagrangian approach. You
could certainly try to solve such the problems using the conventional Newtonian meth-
ods, but such an attempt would require a great deal more thought and physical insight
than demanded if Lagranges equations are used.

Problem 4. Simple Model of Coupled Harmonic Oscillators

Consider two identical particles of mass m attached to the three springs of stiffness
k, as shown in Figure from Set of Problems. We assume that the masses are restricted
to move in a straight line, so that the number of degrees of freedom is 2. The natural
choice for the generalized coordinates are the positions of the masses, which we denote
x1 and x2 . The kinetic and potential energies in terms of these variables are
(
T = 12 m(x 21 + x 22 )
(23)
V = 21 k(x21 + x22 + (x2 x1 )2 )
This yields the Lagrangian
1
L = T V = m(x 21 + x 22 ) k(x21 + x22 x1 x2 ) (24)
2
and Lagranges equations
mx1 + k(2x1 x2 ) = 0

(25)
mx2 + k(2x2 x1 ) = 0
By adding and subtracting these equations we derive the equations

m(
x1 + x2 ) + k(x1 + x2 ) = 0 (adding) (26)

and
x1 x2 ) + 3k(x1 x2 ) = 0
m( (subtracting) (27)

These are the harmonic equations with respect to x1 x2 , and hence their solutions
can be written as
x1 + x2 = a1 cos(1 t 1 ) (28)

and
x1 x2 = a2 cos(2 t 2 ), (29)

where s s
k 3k
1 = , 2 = (30)
m m
aj and j (j = 1, 2) are the arbitrary constants.
Adding and subtracting once again, we finally obtain the general solution of the
equations of motion in the following form:
h i
x1 = 1 a1 cos(1 t 1 ) + a2 cos(2 t 2 )
2h i (31)
x = 1 a cos( t ) a cos( t )
2 2 1 1 1 2 2 2

This solution reveals the existence of the two distinct modes. The mode with the eigen-
frequency 1 corresponds to the movement of the particles with the same amplitudes
and the same phases. In process of such oscillations the length of the spring between
the particles doesnt change. The mode with the eigenfrequency 2 corresponds to the
movement of the particles with the same amplitudes but opposite phases. In this case
the center of mass is motionless. The general solution is the superposition of these two
modes. We will discuss this problem in the frame of theory of small oscillation during
the Seminar 6 (Problem 23 from the Set of Problems).
 SEMINAR 2. PENDULUMS

Problem 7. Simple Pendulum

A simple pendulum means a mass m suspended by a string or weightless rigid rod of
length l so that it can swing in a plane. The y-axis is directed down, x-axis is directed
horizontally,i.e. x = l sin , y = l cos . The kinetic energy is then
1 1 1 2.
T = mv 2 = m(
x2 + y2 ) = m(l) (1)
2 2 2
If we put the potential energy to be zero when the string is horizontal, then at angle
it is
V = mgl cos . (2)

So the Lagrangian is
1
L = T V = ml2 2 + mgl cos , (3)
2
which yields Lagranges equation of motion
d 2 2
ml + mgl sin = 0, (4)
dt
or
g
+ sin = 0. (5)
l
This equation looks simple but, in general, it is not easy to solve. However, if we
assume that the oscillations are small (say, << /2), then sin can be approximated
by , and Eq. (5) takes the form of the usual linear equation for a simple harmonic
motion, namely
g
+ = 0, (6)
l
or
+ 2 = 0, (7)

where
g
2 = . (8)
l
The solution of this equation is well-known:

= C cos(t + ), (9)

where and are the angular frequency and the phase of the oscillations, while C is
arbitrary constant which determine the amplitude of the oscillations. The period of
the oscillations is then s
2 l
T = = 2 . (10)
g
Note that the period of oscillations is independent of the amplitude, provided the
amplitude is small enough so that Eq. (7) is a good approximation.
Now we return back to the consideration of the pendulum equation in its general
form (5). Multiplying both sides of it by and integrating, we obtain
g
= sin , (11)
l
or
= g sin d,
d (12)
l
and
1 2 g
= cos + const. (13)
2 l
For a moment, let use this equation for finding the period in a particular case of large-
amplitude swinging when the pendulum is going back and force between the turning
points 90 and +90 . By definition, in these points

= 0, (14)

so the constant in (13) must be zero, and we have

1 2 g
2
= l
cos ,
d
q
2g

dt
= l
cos ,
q
d 2g
cos
= l
dt. (15)

By observation that in our particular case the change in from = 0 to = 90

corresponds just one-quarter of a period, it follows
/2 s T /4 s
Z
d 2g Z 2g T
= dt = . (16)
cos l l 4
0 0

that is the period of the 180 swings is

/2
s
l Z d
T =4 . (17)
2g cos
0

You might be familiar with the function involved in the r.h.s. of this expression: it is
nothing but a particular case of the Beta or B-function.
Finally, let consider swings of any amplitude, say . Then we may use the turning-
point-condition (14) with = which leads to the relation
2g
2 = (cos cos ). (18)
l
Hence Eq. (17) must be changed to
s
l Z d
T = 4 . (19)
2g cos cos
0

Here T is the period for swings from to + and back. The integral involved in
this expression can be transformed to a table integral as follows:

cos = 1 2 sin2
2

cos = 1 2 sin2
2

cos cos = 2(sin2

2
sin2 2 )
R R R
I= d
= d
= 1 r d
(20)
cos cos 2(sin2
sin2 2 ) 2 sin2
0 0 2 0 1 2 2 sin
sin 2 2

Introduce new variable:

sin 2 1 cos 2
x= dx = d (21)
sin 2 2 sin 2
In terms of this variable
Z1 dx  
I= 2 q 2K sin , (22)
(1 x2 )(1 x2 sin2 2 ) 2
0

where we used the notation K for an elliptic integral. Thus the period (19) takes the
form
l  
s s
l  
T = 4 2K sin =4 K sin . (23)
2g 2 g 2
This expression for the period can be used to exactify the value of the period as com-
pared with its small-oscillation approximation given by Eq. (10) due to the existence
of the following expansion for an elliptic integral:
!
   1 2  1 3 2 4
K sin = 1+ sin2 + sin + ... . (24)
2 2 2 2 24 2
For small enough so that sin /2 can be approximated by /2, it follows
2
!
 
K sin = 1+ + ... , (25)
2 2 16
and hence the period can be approximated as
s
2
!
l
T = 2 1+ + ... . (26)
g 16
We see that this formula differs from our previous one for simple harmonic motion,
q
T = 2 l/g, by the presence of the second- (and higher-)order terms on . Naturally,
for very small this difference is negligible. However, for somewhat large , say, =
0.5 radian (about 30 ), we get
s !
l 1
T = 2 1+ + ... . (27)
g 64
It means, for example, that a pendulum started at 30 would get exactly out of phase
with a pendulum started at very small angle in about 32 periods.
Physically, the motion of a pendulum at different amplitudes can be easily under-
stood if we consider the sum of the kinetic and potential energy,
1
T + V = ml2 2 mgl cos = E, (28)
2
where E is the initial energy level of the system. The potential energy V () is
mgl cos . We see that for
mgl < E < mgl, (29)

the motion is oscillating one because of the existence of the turning point where the
total energy is equal the potential energy. On the other hand, for

E > mgl, (30)

there is no turning point, and the motion is nonoscillatory: is steadily increases or

steadily decreases, while oscillates between a maximum and minimum value, as cabn
be shown in the phase diagram = f () . In this case a pendulum has enough energy
to swing around in a complete circle. Note that this motion is not oscillatory but
still periodic, a pendulum making one complete revolution each time increases or
decreases by 2. Finally, for
E = mgl, (31)

there exist the positions

= (2n + 1), (n = 0, 1, ...) (32)

which are called the bifurcation points of the solution of the equation of motion for a
simple pendulum.

Problem 8. Double Pendulum

Consider the motion of a double pendulum that consist of two simple pendula, each
of mass m and lenght l, as shown in Figure from the Set of Problems. The first one is
attached to a fixed support, and the second one is attached to the mass of the first. (a)
Assuming that the pendulum executes small oscillations confined to a single plane, find
the modes of oscillations. (b) Find numerically the general solution of the equations of
motion.
The configuration of the system is specified by the two angles and , as shown
in the Figure. The Cartesian coordinates of the two masses relate to these generalized
coordinates according to:


x1 = l sin
y1 = l cos


(33)



x2 = x1 + l sin = l(sin + sin )

y2 = y1 + l cos = l(cos + cos ).
The corresponding velocities are
x 1 = l cos



y 1 = l sin


(34)

x 2 = l(cos + cos )



y 2 = l(sin + sin ),

so that the kinetic and potential energies are calculated as

1 1
T = m(x 21 + y 12 + x 22 + y 22 ) = ml2 [22 + 2 + 2 cos( )]
(35)
2 2
and
V = mgy1 mgy2 = mgl(2 cos + cos ). (36)

The Lagrangian
1
L = T V = ml2 [22 + 2 + 2 cos( )]
+ mgl(2 cos + cos ) (37)
2
creates Lagranges equations of motion
2 + cos( ) + 2 sin( ) + 2gl sin = 0;
(
(38)
+ cos( ) 2 sin( ) + gl sin = 0.
In general, this system of equations is rather complicated and it must be solved
numerically. The analytical solution can be obtained, as usual, in the case of small
oscillations when we can use the small angle approximation for all sine and cosine
functions involved, that is

cos( ) 1, sin( ) ( ), sin sin . (39)

Substituting these expressions into the equations of motion and neglecting the higher-
order terms, we obtain the simplified version of these equations,
(
2 + + 202 = 0
(40)
+ + 02 = 0,
where
g
r
0 = . (41)
l
By substituting
= Aeit

, (42)
= Beit
we transform Eqs. (40) to the set of the homogeneous algebraic equations for the
amplitudes A and B,
2( 2 + 02 )A 2 B = 0

(43)
2 A + ( 2 + 02 )B = 0
This set of equations has the nontrivial (i.e. nonzero) solution only if its determinant
is equal zero,
2( 2 + 02 ) 2
 
det = 0, (44)
2 ( 2 + 02 )
or
2( 2 + 02 )2 4 = 0, (45)

which yields the equation

4 402 2 + 204 = 0, (46)

whose roots are q

1,2 = 2 2 0 . (47)

These roots determines the two possible modes of small oscillations of a double pen-
dulum.
To understand the physical sense of these modes we substitute the frequencies 1
and 2 back into the set of equations (43). For = 1 , we obtain
(
(2 2 2)A + (2 2)B = 0
(48)
(2 2)A + (1 2)B = 0,
which yields

B= 2A. (49)

For = 2 , we have (
(2 + 2 2)A + (2 + 2)B = 0
(50)
(2 + 2)A + (1 + 2)B = 0,
which yields

B = 2A. (51)

Hence the modes 1 and 2 can be naturally called the symmetric and antisymmet-
ric modes, respectively. It is interesting to notice that the ratio of these two mode
frequencies is independent of all the parameters m, l and g and is equal to
" #1/2
2 (2 + 2)
= = 2.414, (52)
1 (2 2)
that is the oscillation in the faster, antisymmetric, mode has a frequency about two
and one-half times that of the slower, symmetric, mode.
 Seminar 3: LAGRANGES EQUATIONS. II

Problem 10.

Two masses, 2m and m, are suspended from a fixed frame by two elastic springs
of elastic constant k, as shown in Figure from the Set of Problems. Consider only
vertical motion.
(a) Find the eigenfrequencies and normal modes of oscillations of this system.
(b) The upper mass 2m is slowly displaced downwards from the equilibrium po-
sition by a distance l and then let go. Consider the subsequent motion of the lower
mass m.

Solution:

The system is specified by the two generalized coordinates y1 and y2 , and its kinetic
energy is
1 1 1
T = 2my 12 + my 22 = m(2y 12 + y 22 ). (1)
2 2 2
The potential energy is the sum

V = V1 + V2 , (2)

where V1 and V2 are the contributions from the two masses and two springs, respec-
tively. The first contribution is

V1 = 2mgy1 mgy2 = mg(2y1 + y2 ). (3)

To calculate the second contribution we denote the natural length of the upper and
lower springs l1 and l2 . Then the changes of lengths of the springs due to the presence
of the two masses are as follows:[see Figure]

y1 l1 (4)

and
y2 y1 l2 . (5)

Thus the contribution into the total potential energy from the springs, V2 , is
1 1
V2 = k(y1 l1 )2 + k(y2 y1 l2 )2 . (6)
2 2
The resultant Lagrangian is
1 1
L = T V = T V1 V2 = m(2y 12 + y 22 )+2mgy1 +mgy2 k[(y1 l1 )2 +(y2 y1 l2 )2 ].
2 2
(7)
Lagranges equations
! !
d L L
= 0 (i = 1, 2) (8)
dt y i yi
yield
y1 + 2ky1 ky2 = 2mg + kl1 kl2 ,
2m

(9)
y2 + ky2 ky1 = mg + kl2 .
m
To simplify these equations we introduce new variables

1 = y1 y1 0 and 2 = y2 y2 0 , (10)

where y1,2 0 are the equilibrium positions of the masses 2m and m. These positions
may be found from the force equations
3mg = k(y1 0 l1 )

(11)
mg = k(y2 0 y1 0 l2 ),
which give
ky1 0 = 3mg + kl1

. (12)
ky2 0 = 4mg + kl1 + kl2
In terms of new variables Eqs. (9) are rewritten as
1 + 2k1 k2 = 2mg + kl1 kl2 2ky1 0 + ky2 0 ,
2m

(13)
2 + k2 k1 = mg + kl2 ky2 0 + ky1 0 .
m
Using the equilibrium values y1,2 0 , given by Eq. (12), you can easily check that the
right-hand-sides of Eqs. (13) are identically zero, so these equations are in fact the
homogeneous equations, namely
1 + 2k1 k2 = 0,
2m

(14)
2 + k2 k1 = 0.
m
With the trial solution of the type

1 = Aeit , 2 = Beit , (15)

Eqs. (14) transform to the set of the linear algebraic equations which is written in
matrix form as
2k 2m 2 k A
  
= 0. (16)
k k m 2 B
This set has the nontrivial (i.e. nonzero) solution only if the secular equation
 
 2k

2m 2 k 
=0 (17)
k k m 2 


holds. This equation has two positive roots
s
k 1 
1,2 = 1 , (18)
m 2
which define the eigenfrequencies of oscillations in the system. The corresponding
modes of oscillations follows from the relation
B 2k 2m 2
= = 2. (19)
A k
We see that there are two modes,
1 1 .
   
and (20)
2 2
These modes are usually called the normal modes of oscillations.

(b) To solve this problem we need to use the general solution of the problem which
we take in the form
1 = A1cos(1 t + 1 ) + A2 cos(
2 t + 2 )

. (21)
2 = 2A1 cos(1 t + 1 ) + 2A2 cos(2 t + 2 )
To determine the constants involved, we apply the initial conditions at t = 0:

1 (0) = 2 (0) = l, 1 (0) = 2 (0) = 0, (22)

which yield
A1 cos 1
+ A2 cos
2 = l



2A1 cos 1 + 2A2 cos 2 = l

(23)

1 A1 sin 1
2 A2 sin 2 = 0


21 A1 sin 1 22 A2 sin 2 = 0.
The last two equations in this set gives

1 = 2 = 0. (24)

Then the rest of equations in (23) is simplified to the form

A1+ A2 =l

(25)
2A1 + 2A2 = l,
and we thus obtain
! !
l 1 l 1
A1 = 1 A2 = 1+ . (26)
2 2 2 2
With these amplitudes and phases, we can easyly calculate the coordinate y2 which
describe the motion of the mass 2m. Namely, we have

y = y 2 0 + 2
! "r 2 # ! "r #
   
4mg 1 1 k 1 1 1 k 1
= l1 + l2 + k
+ 2
2
l cos m
1+ 2
t + 2
+ 2
l cos m
1 2
(27)
t
 Problem 11.

The block B attached to a string of stiffness k with the mass m at the end oscillates
in the vertical direction,
s = A sin t. (28)

Show that the motion of the mass m is described by the formula

A02
q(t) = C sin(0 t + ) + sin t, (29)
02 2
where s
k
0 = . (30)
m

Solution:

From Figure in Set of Problems it follows

q(t) + l = l1 (t) + s, (31)

where l is the equilibrium value of the string length. Denoting l0 the natural length
of the spring, we may express l as

l = l0 + l, (32)

where l is determined from an elementary balance of forces,

k l = mg, (33)

that is
mg
l = . (34)
k
Now we are able to calculate the kinetic, T , and the potential, V , energies as
1
T = mq2 , (35)
2
and
1 1 1  mg 2
V = mgq+ k(l1 l0 )2 = mgq+ k(l+qs)2 = mgq+ k +qs . (36)
2 2 2 k
Note that in these expressions the contributions from the block are absent because
the motion of the block is assumed to be known from the very beginning (it is given
by Eq. (28)).
Therefore, the Lagrangian of the system which we are interested in is
1 1  mg 2
L = mq2 + mgq k +qs . (37)
2 2 k
Lagranges equation !
d L L
=0 (38)
dt q q
yield the equation of motion
 mg 
q mg + k
m + q s = 0, (39)
k
or
q + 02 q = 02 s, (40)

where the use of the notation (30) has been made.

The solution of the eqution (40) consist of the sum of the general solution of the
homogeneous equation
q + 02 q = 0 (41)

and the particular solution of the nonhomogeneous equation

q + 02 q = 02 A sin t. (42)

The former is
q0 = C sin(0 t + ), (C and are constants), (43)

while the latter can be found by the substitution

q1 = D sin t, (44)

which gives
D 2 sin t + 02 D sin t = 02 A sin t. (45)

It follows
A02
D= , (46)
02 2
and thus the general solution may be written as
A02
q = q0 + q1 = C sin(0 t + ) + 2 sin t, (47)
0 2
as required.
 Problem 12.

A particle of mass m can slide freely on the inside surface of a rolling cylindrical
tube of mass M and radius a. The system starts at t = 0 in the position (see Figure in
Set of Problems) when the mass m is placed on the same height under the horizontal
plane where the cylinder is rolling. Derive the equation of motion in terms of the
coordinate which determines the position of the particle on the cylinder. Suppose
that all the mass of the tube is concentrated in its periphery.

Solution:

Let x be the coordinate of the center of the cross-section of the cylinder at time
t. Then the kinetic energy of the cylinder, Tc , can be written as the sum
1 1
Tc = M x 2 + Ic 2 , (48)
2 2
where Ic is the momentum of inertia of the cylinder,

Ic = M a2 , (49)

and is the angular frequency of its roll, that is

x
= . (50)
a
Substituting these values for Ic and into Eq. (48) we obtain
1 1 x 2
Tc = M x 2 + M a2 2 = M x 2 . (51)
2 2 a
The position of the mass can be specified by the coordinates
xm = x + a cos

(52)
ym = a a sin ,
which gives for the velocities
(
x m = x a sin (53)

y m = a cos .
Therefore the kinetic and potential energies of the particle are
1
Tm = m(x 2 + a2 2 2a sin x)
(54)
2
and
Vm = mgym = mg(a a sin ) = mga(1 sin ). (55)
Combining (51), (54) and (55), we obtain the total Lagrangian of the system in the
form
1
L = M x 2 + m(x 2 + a2 2 2a sin x)
mga(1 sin ). (56)
2
We see that the Lagrangian doesnt depend on the generalized coordinate x, and
hence there is the first integral

(2M + m)x ma sin = A. (57)

The constant A is determined from the initial condition at t = 0,

x = 0, = 0, x = 0, (58)

giving A = 0. Hence the first integral can be rewritten as

d
[(2M + m)x + ma cos ] = 0, (59)
dt
which yields
(2M + m)x + ma cos = B, (60)

where B is the constant whose value is immediately obtained from the same initial
condition (58) as
B = ma. (61)

With this value we obtain

ma(1 cos )
x= . (62)
2M + m
Next, we consider the remaining Lagranges equation, that is we derive the equa-
tion of motion for the generalized coordinate . Namely, by substituting the La-
grangian into Lagranges equation
d L L
= 0, (63)
dt
we have
h i
d 1
dt 2
ma2 2 + ma cos x mga cos
ma sin x)
= ma2 ma sin x ma cos x + ma cos x mga cos = 0, (64)

or
a = sin x + g cos . (65)

Then we use Eq. (62) to express x through the derivatives of as

x = ma
m+2M
sin
x = ma
m+2M
[cos 2
+ sin ]. (66)
Substituting this expression into Eq. (65), we finally obtain after a little algebra
 M  1  M g
2 + cos2 sin 2 2 1 + 2 cos = 0. (67)
m 2 m a
We see that this is the nonlinear differential equation of the second order whose
analytical equation is unknown. In general, its solution can be obtained numerically.
However, if the quantity of interest is the velocity of a particle at the bottom point
= 0, it can be found from the energy consideration as follows.
Let vf be the velocity of the particle at the final ( = /2) point, that is, in
accordance with Eq. (53),

vf = x a sin = x a. (68)
2
Seminar 4: CHARGED PARTICLE IN ELECTROMAGNETIC FIELD

Introduction

Let take Lagranges equations in the form that follows from DAlemberts principle,
!
d T T
= Qj , (1)
dt qj qj
and suppose that the generalized force is derivable not from the potential V (qj ) but
from a more general function U (qj , qj ), by the prescription
!
U d U
Qj = + . (2)
qj dt qj
Substituting this expression for the force into Eqs. (1) yields
! !
d T T U d U
= + , (3)
dt qj qj qj dt qj
or !
d (T U ) (T U )
= 0. (4)
dt qj qj
It follows that these equations can be written in the form of Lagranges equations,
!
d L L
= 0, (5)
dt qj qj
if we use as the Lagrangian the function

L = T U. (6)

The function U is called usually the velocity-dependent potential (sometimes the

term generalized potential is also used). It can be thought that the possibility of using
such a strange potential is purely academic but this is not the case! On the contrary,
it appears that all the fundamental forces in physics can be expressed in the form
(2), for a suitably chosen potential function U . Its near practical importance relates
to the theory of an electric charge in an electromagnetic field. As you know, a charge
q moving with the velocity v in an electromagnetic field, containing both an electric,
E, and magnetic, B, fields, experiences a force (Note: 1/c appears in Gauss system
of units, in SI system it will be absent)
h i
F = q E + (v B) , (7)

which is called the Lorentz force. Both vectors E(r, t) and B(r, t) are continuous
functions of time t and position r = (x, y, z) derivable from the scalar and vector
potentials (r, t) and B(r, t) by
A
E = (8)
t
and
B = A. (9)

Here, is the differential operator defined in Cartesian coordinates by

!

, , , (10)
x y z

so that

= +
x +
y z (11)
x y z
and
 z 
 x y
 
A=  x y z 
, (12)

Ax Ay Az
, y
where x and z are the unit vectors along x-, y- and z axes, respectively.
Notice that the electromagnetic field defined by Eqs. (8-9) dont change when the
potentials are transformed according to:

0 = , A0 = A + , (13)
t
where is an arbitrary function of the coordinates and time. These transformations
are known as the gauge transformations.
Problem 14. Lagrangian of Charged Particle in Electromagnetic Field

Show that the Lagrangian of a particle with the charge q moving with the velocity
v in an electromagnetic field given by the scalar and vector potentials and A is
1
L = mv 2 q + qA v. (14)
2

Solution:

Taking the vectors of the electric and magnetic fields E and B represented in terms
of the scalar and vector potentials in accordance with Eqs. (8) and (9), we have for
the Lorentz force " #
A  
F = q + v ( A) . (15)
t
Let calculate, say, the x-component of this vector force,
" #
Ax  
Fx = q ()x + v ( A) . (16)
t x

Using the definitions (11) and (12), we obtain

()x = (17)
x
and
     
Ay Ax Ax Az
v ( A) = vy x
y
vz z
x
x

= vy A
x
y
+ vz A
x
z
+ vx A
x
x

vy A
y
x
vz A
z
x
vx A
x
x

dAx Ax
= x
(v A) dt
+ t
, (18)

since
Ax Ay Az
(v A) = (vx Ax + vy Ay + vz Az ) = vx + vy + vz (19)
x x x x x
and
dAx Ax Ax Ax Ax
= + vx + vy + vz . (20)
dt t x y z
Substituting Eqs. (17) and (18) into Eq. (16), we finally obtain
h i h   i
Fx = q
x
Ax
t
+
x
(v A) dAx
dt
+ Ax
t

= q x vA dAx
dt
=
h    i
d
q x vA + dt vx
vA , (21)

which can be written as

U d U
Fx = + , (22)
x dt vx
where we introduce the function

U = q qv A = q qA v. (23)

Note that the term in the right-hand side of this equation coincides with the potential
V defined by Eq. (2.119) from our Lecture notes in a particular case of a single particle
of charge q.
Comparing the expression for the Lorentz force in the form (22) with the definition
(2) of the generalized force in terms of the velocity-dependent potential, we see that
in our case this potential is defined just by the equation (23). This observation
immediately yields the Lagrangian in the desired form given by Eq. (6),
1
L = T U = mv 2 q + qA v. (24)
2
 Problem 15. Calculate the conjugate momentum p and the energy function h
for a particle of the mass m and charge q in an electromagnetic field given by the
scalar and vector potentials and A.

Solution:

The x-component of the conjugate momentum is

( )
L 1
px = = m[vx2 + vy2 + vz2 ] q + q[Ax vx + Ay vy + Az vz ] = mvx + qAx . (25)
x vx 2
The same for y- and z-components

py = mvy + qAy (26)

and
pz = mvz + qAz . (27)

The second terms in these expressions play the role of a potential momentum.
The energy function is
 
L 2
j qj qj L = j qj pj L = vx px + vy py + vz pz mv q + qA v
P P
h=
       
1
= vx mvx + qAx + vy mvy + qAy + vz mvz + qAz 2
mv 2 q + qA v
 
1
= mv 2 + qA v 2
mv 2 q + qA v = 21 mv 2 + q = T + q. (28)

If A and are independent of t, then L does not depend on t explicitly and the
energy function is constant, that is

T + q = const. (29)

Since the second term in this equation is nothing but the potential energy of a charge
particle, we see that in this case the total energy of the system is conserved as it
might be.

Problem 16. A particle of the mass m and charge q moves in a constant magnetic
field
B = (0, 0, B). (30)

Show that the orbit of a particle is a helix.

 Solution:

Let specify the scalar and vector potentials for the case when the electric field is
absent,
E = 0, (31)

and magnetic field has only non-zero component Bz = B. Keeping in mind that this
component is expressed in terms of the vector potential A as
Ay Ax
Bz = , (32)
x y
we can choose the potentials in the form

A = (0, Bx, 0), = 0. (33)

With this choice, the Lagrangian is

m 2
L= (x + y 2 + z 2 ) + qBxy.
(34)
2
The corresponding Lagranges equation are written as
d
qB y
dt (mx) =0
d
dt
(my + qBx) =0 (35)
d

(mz)
= 0.

dt

From the second and third equations it follows

y = C x (36)

and
z = z0 + Dt, (37)

where C, D, z0 are constants and the substitution

qB
= (38)
m
has been made. With the value of the y given by (36), the first equation from (35)
takes the form
x (C x) = 0, (39)

or
x + 2 (x x0 ) = 0, (40)

where
C
x0 = . (41)

The general solution of Eq. (40) can be written as

x x0 = a cos(t + ). (42)

Then
y = C x = (x0 x) = a cos(t + ). (43)

Integrating this equation yields

y = a sin(t + ) + y0 . (44)

Finally, combining Eqs. (42) and (44) we obtain

(x x0 )2 + (y y0 )2 = a2 . (45)

Together with Eq. (37) for z-component, this defines a helix as a trajectory of a
particle.

Problem 17. Find the eigenfrequencies for an isotropic three-dimensional har-

monic oscillator realized as a particle of charge q placed in a uniform magnetic, B,
and electric, E, fields which are mutually perpendicular and take their directions
along z- and x-axes, respectively.

Solution:

As the particle is an isotropic harmonic oscillator and has the charge q, its potential
energy may be written as
1 1
V = kr2 + q qA v m02 r2 + q qA v, (46)
2 2
where we introduced the natural angular frequency of an isotropic oscillator
s
k
0 = , (47)
m
and r = (x, y, z) is the displacement of the particle from the origin. We can easily
check also that the configuration of electric and magnetic field given in the problem
can be realized by the choice of the scalar and vector potentials in the form
 1 1 
= Ex, A = By, Bx, 0 . (48)
2 2
Indeed, with this choice we have

E = = x
E (49)
and
Ay Ax h1 i h 1 i 1 1
Bz = = Bx By = B + B = B, (50)
x y x 2 y 2 2 2
as it should be. It is instructive to notice that in previous problem to obtain the
same result for B we used another choice of A given by Eq. (33).
Now we are able to write the total Lagrangian as
1 1 1
L = m(x 2 + y 2 + z 2 ) m02 (x2 + y 2 + z 2 ) + qEx + qB(xy
+ xy).
(51)
2 2 2
This Lagrangian create Lagranges equations
+ 02 x qB qE

x

m
y m
=0
2 qB (52)
y + 0 y + m x = 0
z + 02 z = 0.

The general solution of the last equation is

z = z0 cos(0 t + ), (53)

that is the oscillation in z-direction takes place with the natural angular frequency
0 . By the change of variables
qE
x = x0 + , (54)
m02
the first two equations can be represented in a more symmetric form
 0
x + 02 x0 L y = 0
(55)
y + 02 y + L x 0 = 0,
where
qB
L =
. (56)
m
This is the system of the coupled linear differential equations of the second order,
and hence we can try a solution of type

x0 = Aeit , y = Beit . (57)

Then the system (55) changes to the system of the algebraic equations which is
written in matrix form as
 2
2 iL A
 
0
= 0. (58)
iL 02 2 B
So the secular equation is
 
 2 2
 0 iL 
 = (02 2 )2 (L )2 = 0. (59)
 iL 02 2 

This equation is equivalent two the pair of equations,
 2
+ L 02 = 0;
(60)
L 02 = 0,
2

which has two positive roots

q
+ = 21 [L + L2 + 402 ],
q
= 12 [L + L2 + 402 ]. (61)

Hence the oscillator in a combined electric and magnetic field exhibits the three
eigenfrequencies 0 , + and . Note that first mode does not depend on the applied
fields at all, and the last two modes are caused by the magnetic field alone, whereas
qE
the electric field only causes a displacement m02
along its direction. For a weak
magnetic field,
L << 0 , (62)

the frequencies are approximated to the form

L
+ = 0 + 2
,
L
= 0 2
, (63)

while in an opposite case of a strong magnetic field,

L >> 0 , (64)

we have
202 02
h  i
1
+ 2
L + 1 + L2 = L + L
,
202 02
h  i
1
2
L 1 + L2 = L
. (65)
 Seminar 5: LAGRANGE MULTIPLIERS

Problem 19.

A sphere of radius a and mass m rolls over the surface of a sphere of radius b > a.
Where the first sphere will leave the second one if initially it was slightly knocked in
the top position?

Solution

Choosing the generalized coordinates r, , as shown in Figure, we may write the

Lagrangian as
1 1 1
L = mr 2 + mr2 2 + Im 2 mgr cos . (1)
2 2 2
Here is the angular frequency of the rotation of a small sphere which can be
expressed in terms of the generalized coordinates and according to

= + , (2)

since the total angle of rotation of this sphere is equal + .

Next we notice that to describe the configuration of the system we introduced three
generalized coordinates though the system has one degree of freedom. Therefore,
there are two conditions of constraint. They follows from the equalities
r =a+b

(3)
a = b.
By inroducing the functions of the generalized coordinates
f1 = a + b r

(4)
f2 = a b,
we can rewrite Eqs. (4) in the standard form of the equations of constraint
f1 = 0

. (5)
f2 = 0
The generalized forces of constraint are thus

f1 f2
Qr = 1 r + 2 r = 1


f1 f2
Q =
1 +
2 2 = b (6)
Q = f1 + f2 = a,


1 2 2

where 1 and 2 are the Lagrange multipliers.

 Substituting these forces into Lagranges equations we obtain the set of three
equations of motion

r mr2 + mg cos = 1

m

mr2 + Im ( + )
mgr sin = 2 b (7)


Im ( + )
= 2 a

On the other hand, from the equations of constraint we have

b
r = a + b r = 0 a = b = . (8)
a
Then the last equation in (7) gives
1 b 1 2
2 = Im
1 + = Im 2 (a + b) = m(a + b). (9)
a a a 5
Here the use of the expression for the moment of inertia for a sphere,
2
Im = ma2 , (10)
5
has been made. Now the second equation can be transformed as

m(a + b)2 + 25 ma2 (a + b) a1 mg(a + b) sin = 52 m(a + b)b

(a + b) + 52 a g sin = 25 b
7
5
(a + b) g sin = 0. (11)

Multiply this equation by to obtain

7
(a + b) g sin = 0 (12)
5
or
7 1 d 2 d
(a + b) ( ) = g cos . (13)
5 2 dt dt
It follows
10g
2 = (cos 1), (14)
7(a + b)
and the first of Eqs. (7) gives
10 10 17
1 = mg(1 cos ) mg cos = mg(1 cos ). (15)
7 7 10
Therefore the generalized force Qr is equal
10 17
Qr = 1 = mg( cos 1). (16)
7 10
Note that at = 0, i.e. at the top of the larger sphere, this force is positive. As
increase, the force decreases, and at some angle = 0 it becomes zero. What might
happen later? The answer is clear, for by definition the normal force which is exerted
by the fixed sphere on the moving one can only be directed outward, never inward.
In means nothing but the stop of rolling of a smaller sphere and the beginning of
its flying off the fixed one. So the answer to the question in the statement of the
problem follows from the condition

1 = 0, (17)

which in view of Eq. (15) yields

10
= arccos 54 . (18)
17
Note that this angle is independent of neither the radius of moving sphere radius a,
nor the radius of the fixed one b.

Problem 20.

Let stand a homogeneous plain bar of the mass m and length 2l in an unstable
position near a vertical wall as shown in Figure from the Set of Problems. Under
some disturbance the bar starts to move in the vertical xy-plane. When the bar will
loose the contact with the vertical wall ? Assume that the system is free of friction.

Solution

Let (xB , yB ) be the Cartesian coordinates of the center of mass of the bar, B, and
is the angle of rotation of the bar about its center of mass. Then the potential
energy of the bar is
V = mgyB , (19)

while the kinetic energy consists of the contributions from the motion of the center
of mass and rotation about the center of mass, that is
1 1
T = m(x 2B + y B2 ) + I 2 , (20)
2 2
where I is the momentum of inertia of the bar,
1
I = ml2 . (21)
3
In principle, the coordinates (xB , yB ) can be expressed in terms of the angle as
xB = l sin

, (22)
yB = l cos
and the problem could be solved by using a single generalized coordinate in ac-
cordance with the existence of only one degree freedom. However, we can obtain
the answer to the problem in an essentialy simpler way by choosing two generalized
coordinates, x = xB and . With this choice, the first of Eqs. (22) will serve as an
equation of constraint,
x = l sin . (23)

In standard notation,
f (x, ) = x l sin = 0, (24)

so that
f f
= 1, = l cos , (25)
x
and the corresponding generalized forces are
f f
Qx = = Q = = l cos . (26)
x
To obtain Lagranges equations with these generalized forces we rewrite the po-
tential and kinetic energies, given by Eqs (19) and (20) , in terms of (x, ) as

V = mgl cos (27)

and
1 1
T = m(x 2 + l2 sin2 2 ) + I 2 . (28)
2 2
Hence the Lagrangian is
1 1
L = T V = m(x 2 + l2 sin2 2 ) + I 2 mgl cos , (29)
2 2
and Lagranges equations,
 
d L L

dt  x 
x
=
d L L
(30)

dt

= l cos ,

take the form

m x=

(31)
d
dt
(ml2 sin2 + I )
ml2 sin cos 2 mgl sin = l cos

Note that the first of this equations immediately gives the criterium for loosing
the contact between the bar and the wall: it happens if

= 0. (32)

Hence to find the answer to our problem we must only express as a function of .
A great advantage of using the Lagranges multiplier method is that it can be done
without direct solving the equations of motion. The procedure is as follows. Use the
equation of constraint (23) to calculate

x = l cos
x = l sin 2 + l cos
ml(cos sin 2 ) =
ml2 cos (cos sin 2 ) = l cos . (33)

Substituting the final result of this calculation into the second equation of motion
(31) leads after some algebra to the equation
3g
= sin . (34)
4l
Starting from this equation we can easily calculate also the quantity 2 :
 
2 = 2 3g
4l
sin
 
d 2
dt
( ) = 2 dtd 3g
4l
cos
 
2 = 2 3g
4l
cos + C. (35)

From initial condition ( = 0 at = 0) it follows

3g
C= , (36)
2l
and we thus have
3g
2 = (1 cos ) (37)
2l
Finally, substituting Eqs. (35) and (37) into the expression for from Eq. (33),
we obtain

= ml(cos sin 2 )
 
= ml cos 3g
4l
sin sin 3g
2l
(1 cos )
 
3
= mg sin 4
cos 32 + 23 cos
= 34 mg sin (3 cos 2). (38)

It follows that takes zero values when

= 0 or 3 cos = 2. (39)

The first value is out of our interest since it coincides with the starting value of ,
while the second equality determines the angle
2
= arccos , (40)
3
which we search for.
 Seminar 6: COUPLED HARMONIC OSCILLATORS

1. Lagrangian and Equations of Motion

Let consider a system consisting of two harmonic oscillators that are coupled
together. As a model, we will use two particles attached to elastic strings, as shown
in Figure from the Set of Problems. For simplicity, we assume that the oscillators
are identical and are restricted to move in a straight line. The coupling between
oscillators is represented by a spring of stiffness K 0 . The system has two degrees
of freedom, and we thus need to use two generalized coordinates to represent the
configuration of the system. The natural choice is x1 and x2 , the displacement of the
particles from their equilibrium positions.
The Lagrangian of the system is
1 1 1 1 1
L = T V = mx 21 + mx 22 Kx21 K 0 (x2 x1 )2 Kx22 . (1)
2 2 2 2 2
To derive Lagranges equations of motion we calculate

L


x 1
= mx 1 ,
L

= mx 2 ,


x 2
(2)



L
x1
= Kx1 + K 0 (x2 x1 ) = (K + K 0 )x1 + K 0 x2 ,
L
= Kx2 K 0 (x2 x1 ) = (K + K 0 )x2 + K 0 x1 ,



x2

which yields
d L L
x1 + (K + K 0 )x1 K 0 x2 ;
(
dt x 1
x1
= m
(3)
d L
dt x 2
L
x2
x2 + (K + K 0 )x2 K 0 x1 .
= m
Therefore, Lagranges equations of motion are as follows:
x1 + (K + K 0 )x1 K 0 x2 = 0;
m

(4)
x2 K 0 x1 + (K + K 0 )x2 = 0,
m
or
T11 x1 + T12 x2 + V11 x1 + V12 x2 = 0;

(5)
T21 x1 + T22 x2 + V21 x1 + V22 x2 = 0.
where we introduce the matrices
T11 T12 m 0
   
T = (6)
T21 T22 0 m
and
V11 V12 K + K0 K 0
   
V = . (7)
V21 V22 K 0 K + K0
Note that Eqs. (5) can be rewritten in a form of a single matrix equation

T ~ + V ~ = 0, (8)
where ~ is the column matrix (or vector) whose components represents the configu-
ration or the state of the system as a whole,
x1
 
~ . (9)
x2
The matrix equation (8) in an explicit component form is
m 0 x1 K + K0 K 0 x1
     
+ = 0. (10)
0 m x2 K 0 K + K0 x2
Finally, using Einstein convention we can write this equation also in the form

Tij j + Vij j = 0 (i, j = 1, 2). (11)

Note that in any form these equations are coupled: say, in the initial form (5) the
cross terms are nonzero, or in matrix form (8) the V-matrix has nonzero off-diagonal
elements.

2. Solution of Lagranges equations

Let search for the solution of Eq. (8) in the form

~ = a cos(t ), (12)

whose components are therefore

x1 = a1 cos(t ), x2 = a2 cos(t ). (13)

Hence we search for the solution whose components has both the same frequency
and phase but a different amplitude. Equation (8) with the assumed solution (13)
becomes
V a = 2 T a, (14)

or
K + K0 K 0 a1 m 0 a1
     
= 2 . (15)
K 0 K + K0 a2 0 m a2
We are well familiar with equations of such type [see: Eq. (3.19) in Lecture Notes].
This is nothing but the eigenvalue equation with respect to the eigenvector a, and
2 is its eigenvalue. This equation is equivalent to a system of linear, homogeneous
equations given by
K + K 0 2m K 0 a1
  
= 0. (16)
K 0 K + K 0 2m a2
These equations have a nontrivial solution (that is, solutions other than a1 = a2 = 0)
if and only if
 K + K 0 2mK 0
 

  = 0. (17)
 K 0 K + K 0 2m 

On expanding this determinant, we obtain

(K + K 0 2 m)2 K 02 = 0. (18)

Rearranging this as follows:

(K + K 0 2 m)2 K 02 = ( 2 m K K 0 )2 K 02
= ( 2 m K K 0 + K 0 )( 2 m K K 0 K 0 ) = ( 2 m K)[ 2 m (K + 2K 0 )] = (19)
0,

we see that the roots of Eq. (17) are given by

K K + 2K 0
12 = , 22 = . (20)
m m
Next we may substitute the eigenfrequencies 1 and 2 back into equation (16)
to find the relations between the components of the eigenvectors]. For convenience,
we denote a specific eigenvector ai (i = 1, 2), so that the ith component of the jth
eigenvector will be aij (i, j = 1, 2) [Dont confuse a specific eigenvector ai with the
scalar component ai of some generalized eigenvector]. For the first eigenvector, the
matrix equation (16) becomes
K + K 0 12 m K 0 a11
  
0 = 0. (21)
K K + K 0 12 m a21
The first component of this matrix equation

[(K + K 0 ) 12 m]a11 K 0 a21 = 0 (22)

yields the solution

a11 = a21 . (23)

For the second eigenvector, Eq. (22) changes to

[(K + K 0 ) 22 m]a12 K 0 a22 = 0, (24)

which yields components

a12 = a22 . (25)

Substituting these eigenvectors into Eq. (12), we obtain two particular solutions
1
 
~
1 = a cos(1 t 1 ),
1 11

(26)
1

~2 = a22 cos(2 t 2 ),


1
whose sum determines the general solution
x1 (t)
 
~ = ~1 + ~2 = , (27)
x2 (t)
where
x1 (t) = A1 cos(1 t 1 ) A2 cos(2 t 2 )

(28)
x2 (t) = A1 cos(1 t 1 ) + A2 cos(2 t 2 ).
Here we introduced two new constants A1 and A2 , such that A1 /A2 = a11 /a12 . The
constants A1 , A2 , 1 and 2 must be considered as the four unknows which might be
determined from the initial conditions for the positions and velocities of each mass.
To apply these conditions we may differentiate the functions x1 (t) and x2 (t) to find
the velocities
x 1 (t) = 1 A1 sin(1 t 1 ) + A2 2 sin(2 t 2 )

(29)
x 2 (t) = 1 A1 sin(1 t 1 ) 2 A2 sin(2 t 2 ).

3. Initial Conditions and Results

Consider a specific initial configuration when the first particle is held at x1 = 0,

while the second particle is displaced one unit to the right, and then they both are
released from rest. To describe this specific configuration we first notice that at time
t = 0 Eqs. (28) and (29) become
x1 (0) = A1 cos 1 A2 cos 2

(30)
x2 (0) = A1 cos 1 + A2 cos 2 .
and
x 1 (0) = 1 A1 sin 1 A2 2 sin 2

(31)
x 2 (0) = 1 A1 sin 1 + 2 A2 sin 2 .
Solving these equation with respect to the amplitudes and phase shifts, we obtain

A21 = 1 [x1 (0) + x2 (0)]2 + 1
[x (0) + x 2 (0)]2
4 412 1
A22 = 1 [x2 (0) x1 (0)]2 + 1 (32)
4
[x (0)
4 2 2
x 1 (0)]2
2

and
tan 1 = x 1 (0)+x 2 (0)
1 [x1 (0)+x2 (0)]
x 2 (0)x 1 (0) . (33)
tan 2 =
2 [x2 (0)x1 (0)]
In a specific conditions of our problems

x1 (0) = 0, x2 (0) = 1, x 1 (0) = x 2 (0) = 0, (34)

and the general relations (32) and (33) are reduced to

1
A1 = A2 = , 1 = 2 = 0. (35)
2
Inserting these into Eqs. (28) yields
(
x1 (t) = 21 (cos 1 t cos 2 t)
. (36)
x2 (t) = 21 (cos 1 t + cos 2 t)

Note that the signs of A1 and A2 are taken in such a way that Eqs. (34) are satisfied.
The expressions (36) describe the well-known phenomenon of beats because x1 and
x2 are represented by the sum and difference of two simple, equal-amplitude, harmonic
motions whose frequencies are different. Moreover, these sum and difference can be
easily transformed to the products of sine and cosine of a sum and a difference of
frequencies, namely
" # " #
1 1
x1 (t) = sin (1 + 2 )t sin (1 2 )t (37)
2 2
and " # " #
1 1
x2 (t) = cos (1 + 2 )t cos (1 2 )t . (38)
2 2
Figure 11.3.2 (see hard copy which will be given during the seminar) shows the
functions x1 and x2 for the spring constants K = 4 and K 0 = 1 (in arbitrary units.
The motion has been plotted over one complete period. We see that the amplitude
of oscillations of the first mass slowly builds up and then dies away in step with the
dying away and buildup of the amplitude of oscillations of the second mass. That is
exactly which we call the beats.
Next it is instructive to define new time-dependent variables according to:
1
(
1 = 2
cos 1 t
1
(39)
2 = 2
cos 2 t.

In terms of these variables

1 (1 2 )
(
x1 = 2
1 (1
(40)
x2 = 2
+ 2 ),
or, in matrix notation,
x1 1 1 1 1
    
~ = = ~
A. (41)
x2 2 1 1 2
By this, we introduce the matrix A,
1 1 1
 
A= . (42)
2 1 1
This relation can be easily inverted to obtain 1 and 2 as functions of x1 and x2 ,
1 1 1 1 x1
    
~ = = A1 ~ . (43)
2 2 1 1 x2
By this, we define the inverse matrix A1 ,
1 1 1
 
A1 = . (44)
2 1 1
Indeed you can easily check that
1 0
 
A1 A = = I, (45)
0 1
as it might be.
Note that the matrix A and A1 describe /4 rotations of a two dimensional
coordinate system. This suggest that we can interpret x1 and x2 or 1 and 2 as com-
ponents of one and the same vector q whose eindpoint represents the time-dependent
configuration of the coupled oscillator in either of two different ccordinate systems,
as shown in Figure 11.3.3 (in hard copy). As times goes on, the eindpoint of q traces
out a path in either of two configuration spaces, (x1 , x2 ) or (1 , 2 ). These paths are
shown in Figure 11.3.4 (in hard copy), (a) and (b). We see that in both cases the
trajectory of a system point is confined to a box. But in (x1 , x2 )-space the bound-
aries of the box make 45 lines with the coordinate axes, while in 1 , 2 -space they
are parallel to the axes. This suggests that 1 , 2 -space is better choice to express
the equations of motion of the system. But even more fascinating thing here is that,
although the coordinates x1 and x2 exchange by their energies, 1 and 2 not, because
they are functions only single frequences 1 and 2 . These peculiarities is inherent in
the normal coordinates, and we can thus call (1 , 2 ) the normal coordinates for the
system at hand.

4. Normal Modes

Up to now, we dealt with the specific problem when the boundary conditions were
given by Eq. (34). In process, we introduced the normal coordinates 1 , 2 according
to (39). As a result, we obtained the relations between initial coordinates x1 , x2 and
normal coordinates given by Eqs. (40). Now we suppose that these equations are
valid for arbitrary initial conditions, and consider two particular cases

x1 (0) = x2 (0) = 1, x 1 (0) = x 2 (0) = 0, (46)

and
x2 (0) = x1 (0) = 1, x 1 (0) = x 2 (0) = 0. (47)
In both cases the two masses are initially displaced from their equilibrium positions
by equal amounts but in the same direction in the case of Eq. (1.46) and in opposite
directions in the case of Eq. (47).
Let consider both cases in more detail. Substituting Eq. (46) into Eq. (40), we
obtain (
1 (0) = 2
(48)
2 (0) = 0,
that implies the time-dependent solution
(
1 (t) = 2 cos 1 t
(49)
2 (t) = 0.

Then Eq. (40) yields

x1 (t) = x2 (t) = cos 1 t. (50)

That is the two masses vibrate back and forth as though they were completely in-
q
dependent simple harmonic oscillators with identical frequencies, 1 = K/M . The
system is said to execute a normal mode of oscillations called the symmetric mode
which is pictured in Figure 11.3.5 (in hard copy).
On the other hand, in the case of the boundary conditions given by Eq. (47) we
obtain
1 (0) = 0

(51)
2 (0) = 2,
that implies the time-dependent solution
1 (t) =
0

(52)
2 (t) = 2 cos 2 t
Then Eq. (40) yields
x2 (t) = x1 (t) = cos 2 t. (53)

In this case each mass vibrates 180 out of phase with the other at the single frequency
q
2 = (K + 2K 0 )/m, as shown in Figure 11.3.6 (in hard copy). Naturally, this
normal mode of oscillations is called the antisymmetric, or, for obvious reason, the
breathing mode. Notice that in the vibration of this kind no energy can be transferred
from one mass to the other across the central point which is called a nodal point in
the vibration.
So we see that the system of the two coupled harmonic oscillators can be started
off such that the two masses vibrate at a single fixed frequency and never exchange
energy. There are two ways for this but in both cases one of the normal coordinate
is chosen to be zero while the other oscillates with one of the eigenfrequencies, 1
or 2 . Such a situation implies the existence of a two separate differential equations
for the normal coordinates 1 and 2 whose solutions represent two decoupled simple
harmonic oscillators. We can guess that these equations might be obtained directly
from Lagranges equations by transforming the Lagrangian to a function of the nor-
mal coordinates.

5. Transformation to Normal Coordinates

To find such a transformation, we first rewrite the kinetic and potential energies
given by Eq. (1) in a slightly different form introducing the row vectors

~ = (x1 , x2 ), ~ = (x 1 , x 2 ). (54)

Then the kinetic and potential energies become

T = 12 mx 21 + 12 mx 22 =
m 0 x 1
  

1
= 2 (x 1 , x 2 ) = 12 ~ T ~ (55)
0 m x 2
and

V = 21 (K + K 0 )x21 + 12 (K + K 0 )x22 K 0 x1 x2
K + K0 K 0 x1
  
= 21 (x1 , x2 ) = 12 ~V ~ . (56)
K 0 K + K0 x2

Applying the transformation between the vectors ~ and ~ given by Eq. (41), we
obtain
g
~ A~ = 1 ~ AT
T = 21 AT A~ =
2
!
1 1 m 0 1 1 1
   
= 41 (1 , 2 )
1 1 0 m 1 1 2
 !
m 0

1
= 21 (1 , 2 )
0 m 2
= 12 m12 + 12 m22 (57)

and

V = 12 AV
g ~ A~ = 1 ~AV
A~ =
2
1 1 K + K0 K 0 1 1 1
    
= 41 (1 , 2 )
1 1 K 0 K + K0 1 1 2
K 0 1
  
= 12 (1 , 2 ) 0
0 K + 2K 2
= 21 K12 + 12 (K + 2K 0 )22 . (58)

Here the use of the matrix identity A~ = ~A has been made.
g

Hence the Lagrangian is

1 1 1 1
L = T V = m12 + m22 K12 (K + 2K 0 )22 . (59)
2 2 2 2
As anticipated, it contains no cross terms, and the resulting equations of motion are
(
m1 + K1 = 0, (60)
m2 + (K + 2K 0 )2 = 0,
or in matrix notation
!
m 0 1 K 0 1
    
+ = 0. (61)
0 m 2 0 K + 2K 0 2
These equations describe the motion of two uncoupled, simple harmonic oscillators,
and their solutions are
1 = B1 cos(1 t 1 ),

(62)
2 = B2 cos(2 t 2 ),
where
K K + 2K 0
12 =, 22 = . (63)
m m
Substituting the solutions (62) into the transformation formula (41), we obtain
1 (1 2 ) = A1 cos(1 t 1 ) A2 cos(2 t 2 )
(
x 1 = 1 = 2
1 (1
(64)
x 2 = 2 = 2
+ 2 ) = A1 cos(1 t 1 ) + A2 cos(2 t 2 ).

These expressions are identical with those of Eq. (28), obtained by direct solving the

starting coupled equations written in coordinates 1 , 2 (the factor 1/ 2 factor has
been absorbed).

6. Diagonalization of Lagrangian

Note that both T and V matrices in the normal coordinate representation (see the
matrix equation (61)) are diagonal, and that each of those matrices was diagonalized
by the so-called congruent transformation

A Vdiag = AT
Tdiag = AT A, (65)

where the matrix A has been defined by Eq. (42). But notice: in view of the
expressions given by Eqs. (41) and (43)) this matrix is equivalent to the matrix
whose its columns are the (x1 , x2 components of the eigenvectors ai , that is,
a11 a12
 
A= . (66)
a21 a22
Each ai in this expression is a column vector which obeys the equation

V ai = i2 T ai (i = 1, 2), (67)

where aij refers to the jth component of the eigenvector ai , and i2 is its eigenfre-
quency.
Therefore, the matrix formed with each eigenvector as one of its column is the
desired matrix A that transforms the generalized coordinates into normal coordi-
nates and diagonalizes the V and T matrices that make up the Lagrangian of the
system. Moreover, the elements of the diagonal matrices Tdiag and Vdiag determine
the eigenfrequencies by means of the relations
i V ai
a
i2 = (i = 1, 2) (68)
i T ai
a
which follow directly from Eq. (67).
Everything sounds well, isnt it? However, pay your attention, please, in what
way we became so clever as were able to find the matrix which diagonalizes the
Lagrangian. First, we solved the coupled equations of motion in their original form,
and second we find the normal mode representation of these solution. In other words,
we were forced to solve the coupled equations of motion before finding the normal
modes that we need to decouple them. The question arises: Can we obtain the
normal modes first in a way, other than by educated guess-work? Unfortunately,
there is no general prescription how to do this but one method works quite well in
many cases of interest. The method is based on the use of some symmetry properties
of the Lagrangian for finding the normal coordinates.
To see how it works, let take the Lagrangian for any two-component coupled
oscillator in the following general form:
1 1 1 1
L = T11 x 21 + T22 x 22 + T12 x 1 x 2 V11 x21 V22 x22 V12 x1 x2 . (69)
2 2 2 2
Now suppose, in this Lagrangian we replace x2 with x1 and x1 with x2 /, and
after this operation the Lagrangian remains the same. Then it is said to be invariant
under an exchange operation. Carrying out the exchange

x1 x2 , x2 / x1 (70)

transforms the Lagrangian (69) to

01 x 22 1 1 x22 1
L = T11 2 + T22 x 1 + T12 x 2 x 1 V11 2 V22 2 x21 V12 x2 x1 .
2 2
(71)
2 2 2 2
We see that the two cross terms in L0 are identical to those in L, and hence the
invariance property will held if
T11 V11
2 = = . (72)
T22 V22
The second of these equalities,
T11 V11
= , (73)
T22 V22
must be a property of the Lagrangian for the system under consideration, while the
first condition,
T11
, 2 = (74)
T22
determines the ratio of the x-components that may be used for the Lagrangian will
be invariant under exchange operation. Namely, x2 -component must be times
that of the first.
This suggests the two eigenvectors a1 and a2 of the form
1 1
   
a1 = a2 = , (75)

and thus the A matrix that generates the transformation from generalized to normal
coordinates has to be taken as
1 1
 
A = (a1 a2 ) = . (76)

It is instructive to reexamine the problem of a two coupled oscillator in light of
this discussion. Looking at the diagonal elements of the matrices T and V in Eqs. (6)
and (7), we see that the condition (73) is satisfied automatically and the condition
(74) requires = 1. Hence
1 1
 
A = (a1 a2 ) = , (77)
1 1
and the matrices T and V are diagonalized according to the transformations
1 1 m 0 1 1
   
Tdiag =
1 1 0 m 1 1
m 0
 
=2 (78)
0 m
and
1 1 K + K0 K 0 1 1
   
Vdiag = 0
1 1 K K + K0 1 1
K 0
 
=2 . (79)
0 K + 2K 0
The ratio of the diagonal elements of Vdiag and Tdiag yield the eigenfrequences 12 and
22 obtained previously [see: Eqs. (62)]. Note that the multiplicative factor 2 that
occurs in Eqs. (78) and (79) cancels out in these ratios and is therefore irrelevant.
If we wish, it could be eliminated by normalizing the eigenvectors a1 and a2 by the

factor 1/ 2 (in fact, we did this early).
 Seminar 7: CENTRAL FORCE PROBLEM

Problem 26

A particle of mass m is attracted to a force center by a force which varies inversely

as the cube of its distance from the center. Solve the equations of motion for the or-
bits and discuss how the nature of the orbits depends on the parameters of the system.

Solution

From Lecture 2, we already know the equations of motion for a particle in the
presence of an arbitrary central force f (r). These equations are as follows:
r r2 ) = f (r)
(
m(
(1)
d
mr2 = 0.
dt
The second equation has the first integral,

mr2 = l, (2)

which is the angular momentum of the particle about the origin since it can be
represented as a product
l = r mr r mv , (3)
where v is the angular velocity of the particle.
For the force
kr
f = , (4)
r4
the first of Eqs. (1.1) is
k
m(r r2 ) = 3 , (5)
r

or, by substituting expressed in terms of l in accordance with Eq. (1.3),
l2 l2
!
k mk
r
m 3
+ 3
= m
r 3
1 2 = 0. (6)
mr r mr l
We know that r = r(), and that this equation can be essentially simplified if we
change this variable letting
1
r= . (7)
u
Indeed, then we have
lu2
= l
mr2
= m
2
r = d 1
dt u
= u12 du
d
= u12 du lu
d m
= ml du
d
2 2 2 2 2
r = ml d
d du
dt
= ml ddu2 = ml ddu2 lum = ml 2 u2 ddu2 . (8)
With these transformations, Eq. (6) becomes
l2 d2 u l2 u3
!
mk
m 2 u2 2 1 2 = 0, (9)
m d m l
or
d2 u
!
mk
2
+ 1 2 u = 0. (10)
d l
As we see, this is the linear homogeneous differential equation of the second order.
The form of its solution depends critically on the relation between the parameters
involved. Namely, if
l2 > mk, (11)

this solution is "s #

1 mk
u = cos 1 2 ( 0 ) , (12)
r0 l
where (r0 , 0 ) is a point on the orbit. In the opposite case, when

l2 < mk, (13)

the form of solution is changed to

"s #
1 mk
u = cosh 1 ( 0 ) . (14)
r0 l2
The corresponding equations for the orbits are
" s !#1
mk
r = r0 cos 1 2 ( 0 ) , (15)
l
and " s !#1
mk
r = r0 cosh 1 ( 0 ) , (16)
l2
respectively.

Problem 27

Consider a particle of mass m moving in a plane under a central force

k k0
f (r) = + . (17)
r2 r3
Find the equation for the orbit in the polar coordinates (r, ), assuming that k > 0
and l2 > mk 0 , where l is the orbital angular momentum.
 Solution

The force (17) differs from the one given by Eq. (4) in the preceeding problem by
the presence of the inverse-square term. Hence the equation of motion (6) is modified
to
l2 mk 0
!
k
r
m 1 + + = 0, (18)
mr3 l2 r2
or, with the use of the transformations (8),
d2 u mk 0
!
mk
2
+ 1 + 2
u= 2 . (19)
d l l
Opposite to our previous case [see: Eq. (10)], Eq. (19) contains in its r.h.s. the
mk
driving term l2
. This term is constant, and hence an obvious particular solution of
the inhomogeneous Eq.(19) is also constant,
mk 0
!
mk mk
u= 2 1+ 2 = . (20)
l l l2 + mk 0
This particular solution must be add to the general solution of the corresponding
homogeneous equation,
d2 u mk 0
!
+ 1 + u = 0. (21)
d2 l2
where l2 > mk 0 ,by assumption. Therefore we may write the solution of this equation
in the form given by Eq. (12), that is
s !
mk 0
u = A cos 1 + 2 ( 0 ) , (22)
l
The general solution of the inhomogeneous equation (19) is thus
s !
mk 0 mk
u = A cos 1 + 2 ( 0 ) + 2 , (23)
l l + mk 0
where A and 0 are constants. By a suitable choice of coordinates, 0 can be put to
zero, and hence the equation for the orbit can be written as
1
r= q  , (24)
mk0 mk
A cos 1+ l2
+ l2 +mk0

or
1
r= h q i , (25)
mk0
C 1 + cos 1+ l2

where C and e are new constants,
mk l2 + mk 0
C= =A . (26)
l2 + mk 0 mk
 To understand the character of this orbit, let first consider a particular case, when
k 0 = 0. Then Eq. (25) becomes
1
r= , (27)
C(1 + cos )
where
mk l2
C= , = A . (28)
l2 mk
We already met with this orbit in our Lectures [see: Eq. (4.125) in Lecture Notes],
and we know that this orbit is an ellipse with one of the foci at the origin of the plane
polar coordinates and with the eccentricity . Indeed, you know that an ellipse is
defined as the curve traced by a particle moving so that the sum of its distances from
two fixed points F, F 0 is constant. These points are the foci of the ellipse. Using the
notation indicated in Figure (it will be demonstrated during Seminar), we have

r0 + r = 2a, (29)

where a is a half the largest diameter (major axis) of the ellipse. In terms of the
polar coordinates with center at the focus F and with the negative x-axis through
the focus F 0 , the cosine law gives

r02 = r2 + 4a2 2 + 4ra cos , (30)

where a is the distance from the center of the ellipse to the focus. Combining Eqs.
(29) and (30), we find
a(1 2 )
r= . (31)
(1 + cos )
Comparing this equation with our orbit given by Eq. (27), we see that these expres-
sions are identical, if
1
C= . (32)
a(1 2 )
But this is really the case because for the perihelion point = 0 we have [see Figure
3.38]
r = a a = a(1 ), (33)

while Eq. (27) with = 0 reduces to

1
r= , (34)
C(1 + )
Therefore
1
a(1 ) = , (35)
C(1 + )
which yields
1
C= , (36)
a(1 2 )
that is the expression for C given by the standard equation (32). Note that the con-
stant A which determines the eccentricity according to Eq. (28) remains unknown:
it can be found from the formal solution of the equations of motion or from the energy
consideration but not from the differential equations for the orbit which we used in
our problem.
Now it is not difficult to guess that if the constant k 0 is not zero but small enough,
the orbit will be a modified ellipse whose perihelion will be shifted a little from its
conventional value given by Eq. (33). If necessary, this shift can be easily calculated
from the equation of orbit with k 0 6= 0, that is, Eq. (25). This problem is in no way
only of academic interest. Say, it can be shown that under the influence of the Earth
attraction Mercury moves in the force
GM m
F = 2
3, (37)
r r
where is a positive constant, that is, just in the force of the type, given by Eq. (17).
 Seminar 10. HAMILTON-JACOBI THEORY

Problem 40

Use the Hamilton-Jacobi theory for the solution of a simple problem about the
motion of a point particle of mass m in xy plane under the gravity force

F = (0, mg). (1)

Suppose that the particle starts its motion from the origin with the velocity

v0 = (v0 cos , v0 sin ), (2)

as shown in Figure.

Solution:

The Hamiltonian of the system is

1 2
H= (p + p2y ) + mgy, (3)
2m x
or " !2 !2 #
1 S S
H= + + mgy, (4)
2m x y
if we use the equations of transformation
S S
px = py = , (5)
x y
where S is the Hamilton principal function.
Hence the Hamilton-Jacobi equation has the form
" !2 !2 #
1 S S S
+ + mgy + = 0. (6)
2m x y t
Try the separable solution

S(x, y, t) = W1 (x) + W2 (y) 3 t. (7)

Then Eq. (6) becomes

!2 #
W1 2 
"
1 W2
+ + mgy = 3 , (8)
2m x y
which is equivalent to the equations
 2
1 W1


2m x
= 1
 2 (9)
1 W2


2m y
+ mgy = 2 ,
where 1 and 2 are the constants such that

1 + 2 = 3 . (10)

It follows

W1 = 2m1 W1 = 2m1 x
x q 3 3
W2 = 2m(2 mgy) W2 = 2 1
(2m2 2m2 gy) 2 = 3m12 g (2m2 2m2 gy) 2 ,
y 3 (2m2 g)
(11)
and
1 3
S(x, y, t) = 2
(2m2 2m2 gy) 2 (1 + 2 )t.
2m1 x (12)
3m g
To determine the coordinates x and y we may differentiate the function S(x, y, t)
with 1 and 2 , and then put the results of these differentiations to new constants
1 and 2 , respectively. In this way we obtain

s
S 1 1 m
1 = = 2mx t = x t (13)
1 2 1 21
and
!1
S 1 3 1 1 1 2 2y 2
2 = = 2 2m(2m2 2m2 gy) 2 t = (2m2 2m2 gy) 2 t = 2 2 t.
2 3m g 2 mg mg g
(14)
These equations yield s
21
x= (1 + t) (15)
m
and " #
g 2 2 g
y = 2 2 (2 + t)2 = (2 + t)2 . (16)
2 mg mg 2
To complete the solution we calculate the momenta
S
px = = 2m1 (17)
x
and
S 1
= (2m2 2m2 gy) 2 .
py = (18)
y
Finally, determine the four constants, 1 , 2 , 1 and 2 , using the initial conditions

x(0) = 0, y(0) = 0, px (0) = mv0 cos py (0) = mv0 sin . (19)

It follows
mv02



1 = 2
cos2
mv02

2 = sin2

2 (20)


1 = 0
2 = v0 sin


g
.
Substituting these constants into Eqs. (15-16), after a little algebra we obtain the
solution of the problem in its final form
x = v0 t cos


y = v0 t sin 12 gt2 . (21)

It is instructive to solve the same problem starting from the equation for the
Hamiltons characteristic function W (q, ),
!
W
H q, = 1 . (22)
q
In our case this equation becomes
" !2 !2 #
1 W W
+ + mgy = 1 , (23)
2m x y
where W = W (x, y, 1 , 2 ). To simplify this equation we notice that the x-component
of momenta is constant,
px = mv0 cos C. (24)

Then, by definition,
W
= px = C, (25)
x
and hence
W = Cx + Wy (y, 1 , 2 ), (26)

where Wy is a function of y. To find this function we substitute the representation

(26) into Eq. (23) which gives
" !2 #
1 Wy
C2 + + mgy = 1 . (27)
2m y
It follows
Wy q
= (2m1 C 2 ) 2m2 gy, (28)
y
and
s v
C2 22
u
q Z
1 q Z u 1
Wy = 2m2 g dy y 2m2 g dy t y. (29)
mg 2m2 g mg 2m2 g

Therefore the characteristic function (26) takes the form

v
22
u
q Z u 1
W (x, y, 1 , 2 ) = 2 x + 2m2 g dy t y. (30)
mg 2m2 g
Here the change of notation the constant C to 2 has been made.
Next we differentiate this characteristic function over 1 ,
s v
22
u
W q Z
1 1 1 2u
t 1
Q1 = = 2m2 g dy r = y (31)
1 2 mg 1 22 g mg 2m2 g
mg
2m2 g
y

and
! s v
22
u
W q Z
2 1 2 2 u
t 1
Q2 = = x+ 2m2 g dy r = x+ y
2 2m2 g 1 22 gm mg 2m2 g
mg
m2 g
y
(32)
Then, following the method of characteristic function, we put
Q1 = t + 1

(33)
Q2 = 2
and find the constants 1 and 2 letting t = 0 and applying the boundary conditions
given by Eq. (1.19). As a result, we obtain
s v
2 2 2 2
u
2u
t mv0 m v0 cos = v0 sin
1 = 2
(34)
g 2mg 2m g g

and s
2 mv0 cos v0 sin v 2 sin cos
2 = = 0 . (35)
g m 2g g
Now we are able to find x and y from the equations
(
t + 1 = F (y)
(36)
2 = x + m2 F (y),

where we put
s v
22
u
2u
t 1
F (y) = y (37)
g mg 2m2 g
Excluding F (y) from these equations, we obtain
v 2 sin cos mv0 cos
!
2 v0 sin
x = 2 + (t + 1 ) = 0 + + t = v0 t cos (38)
m g m g
Finally, using the first of Eqs. (36) we have
22
" #
2 2 1 2
(t + 1 ) = [F (y)] = y , (39)
g mg 2m2 g
which gives
22 gt2
" #
g 2 2 1
y = t + 21 t + 1 + = v0 sin t , (40)
2 mg 2m2 g 2
as it should be.

Problem 41

A charged particle moves in a plane in the presence of the force of the non-
electromagnetic nature whose potential is
1
V = kr2 , (41)
2
and the constant magnetic field B applied perpendicular to the plane, so that its
vector potential is given by
1
A = B r. (42)
2
Write down the Hamilton-Jacobi equation for this system in polar coordinates. Sep-
arate variables in this equation and integrate the resulting equations for different
coordinates. What will be the motion when p = 0 at t = 0?

Solution

According to the geometry of the system, shown in Figure, we have

B = Bz


A = 2 Bz rr = 12 Br~
1

(43)
~ = 1 Br2 .
A v = 12 Br~ (r
r + r)


2

Hence the Lagrangian is

1 1 1 1 1 1
L = mv 2 kr2 + qA v = mr 2 + mr2 2 kr2 + qBr2 .
(44)
2 2 2 2 2 2
This Lagrangian creates the generalized momenta

pr = L
r
= mr r = pmr
2

qBr2
 (45)
p = L

= mr2 + qBr
2
= 1
mr2
p 2
,

and the Hamiltonian

qBr2
   
H = pr r + p L = mr 2 + mr2 + 2
1
2
mr 2 + 21 mr2 2 12 kr2 + 12 qBr2
p2r qBr 2 2
 
= 12 mr 2 + 12 mr2 2 + 21 kr2 = 2m
+ 1
2mr2
p 2
+ 12 kr2 . (46)

By the substitution
S S
pr = , p = (47)
r
into this Hamiltonian, we obtain the Hamilton-Jacobi equation
!2 #2
qBr2
" !
S 1 S 1 S 1 S
H(q, p) + = + + kr2 + = 0. (48)
t 2m r 2mr2 2 2 t
With the separable solution of the form

S(r, , t) = f1 (r) + f2 () 1 t. (49)

this equation becomes

!2 #2
qBr2
" !
1 f1 1 f2 1
+ + kr2 = 1 . (50)
2m r 2mr2 2 2
To simplify this equation we note that the Lagrangian L does not depend on , that
is is a cyclic variable, and thus
S f2
p = = = const 2 . (51)

It follows
f2 = 2 , (52)

and Eq. (50) is reduced to the equation containing only the function f1 (r),
!2 !2
1 f1 1 qBr2 1
+ 2 + kr2 = 1 . (53)
2m r 2mr2 2 2
This equation is easily integrated in the two steps,
f1 2 qBr 2 2
   
1
r
= 2m1 r2
2 2
kmr2
r
Rr 1

qBr 2 2

f1 (r) = dr 2m1 r2
2 2
kmr2 . (54)
0

Substituting this solution together with the expression of the function f2 () in the
form (49), we obtain
Zr
s
1 qBr2 2
S= dr 2m1 2 kmr2 + 2 1 t. (55)
r2 2
0

Next we differentiate this function over 1 and 2 and put these derivatives to the
constants 1 and 2 , respectively:
S Z
2m
1 = = dr r t (56)
1 1

qBr2 2

2 2m1 r2
2 2
kmr2

and  2

S Z r22 2 qBr
2
2 = = dr r + . (57)
2 1

qBr 2 2

2
2 2m1 r2 2 2 kmr
In the case of our problem the initial condition (p = 0) gives 2 = 0, and thus Eqs.
(56-57) are simplified to

m
= q 2 dr
R R

1 + t = dr q 2 2 2
=
2m1 q B4 r kmr2 1 ( 2 + 2 )r 2


m L
qB (58)
2 = dr q = q 2 L dr
R R


q 2 B 2 r 2 1 2 2 2
2 2m1 4
2 kmr m
(L + )r

where s
qB k
L = , = (59)
2m m
Combining these equations we have

L (1 + t) = 2 , (60)

which yields
= 2 L 1 L t L t. (61)

We notice also that the integral in the first of Eqs. (58) is elementary one, and we
can write q
L2 + 2
!
1
1 + t = arcsin r q . (62)
21 / m 2
L + 2
It follows
q !1
L2 + 2 q q q
r= sin(1 L2 + 2 + t L2 + 2 ) A| sin( L2 + 2 t + )|, (63)
21 / m
where s
21 q
A= , = 1 L2 + 2 (64)
m(L2 + 2 )
are the amplitude and the phase of the oscillations.
 Seminar 11. POISSON BRACKETS

Problem 43.

Calculate Poisson brackets [x, px ] and [Lx , Ly ].

Solution

According to the general definition of the Poisson brackets,

!
X u v u v
[u, v] = , (1)
i qi pi pi qi
we have !
X x px x px x px
[x, px ] = = = 1. (2)
i qi pi pi qi x px
In the same way we can easily prove the rest of the general formulas given by Eqs.
(6.55)-(6.56) from the Lecture Notes:

[qj , qk ] = [pj , pk ] = 0; [qj , pk ] = [pj , qk ] = jk . (3)

Using these formulas we can calculate also the Poisson brackets for the angular
momenta, since by definition,
i 
 j k 
L = (r p) = x y z . (4)
 

 
px py pz
It follows
Lx= ypz zpy

L = zpx xpz (5)

y
Lz = xpy ypx .
Hence

[Lx , Ly ] = [ypz zpy , zpx xpz ]

(ypz zpy ) (zpx xpz ) (ypz zpy ) (zpx xpz ) (ypz zpy ) (zpx xpz )
= x px
+ y py
+ z pz
z zpy ) (zpx xpz ) (ypz zpy ) (zpx xpz ) (ypz zpy ) (zpx xpz )
(ypp x x
py y
pz z

= py (x) ypx = xpy ypx = Lz . (6)

In the same way we can prove that

[Ly , Lz ] = Lx and [Lz , Lx ] = Ly . (7)

 Problem 44

Take the Hamiltonian for the three-dimensional isotropic harmonic oscillator,

3 3
1 X m 2 X
H= p2n + q2 , (8)
2m n=1 2 n=1 n
and consider the functions

Tkl = pk pl + (m)2 qk ql for k, l = 1, 2, 3. (9)

Show that these functions are the constants of motion for all k and l.

Solution

The equations of motion for the functions Tkl in Poisson bracket formulation of
mechanics are
dTkl Tkl
= [Tkl , H] + . (10)
dt t
Since Tkl does not depend on time t explicitly, this equation is reduced to
dTkl
= [Tkl , H]. (11)
dt
Hence the functions Tkl will be the constants of motion if the Poisson brackets for H
and Tkl are zero. So let calculate
h 3 3 i
1 m 2
p2n + qn2 ), pk pl + (m)2 qk ql
P P
[H, Tkl ] = ( 2m 2
n=1 n=1
1 P m 2
= 2m [pn pn , pk pl ] + 2 (m)2 [qn qn , qk ql ]
P
n n
(m)2 P m 2 P
+ 2m [pn pn , qk ql ] + 2 [qn qn , pk pl ]. (12)
n n

Taking into account the general relations

[pn , pk ] = [qn , qk ] = 0

(13)
[qn , pk ] = [pn , qk ] = nk ,
we immediately see that the first two terms in Eq. (12) are equal to zero, while the
rest two terms require some additional calculations. For the third term, we have
X pn pn qk ql X pn pn qk ql
[pn pn , qk ql ] = = [pn , qk ql ]pn + pn [pn , qk ql ], (14)
j qj pj j pj qj

since formally
pn pn pn pn
= pn + pn (15)
qj qj qj
and
pn pn pn pn
= pn + pn . (16)
pj pj pj
In the same way we can prove that

[pn , qk ql ] = [pn , qk ]ql + qk [pn , ql ], (17)

which by using Eq. (13) yields

[pn , qk ql ] = nk ql nl qk . (18)

Substituting these relations into Eq. (14) we finally obtain for the third term

[pn pn , qk ql ] = (nk ql nl qk )pn +pn (nk ql nl qk ) = ql pk qk pl pk ql pl qk . (19)

In an analogous way, we transform the fourth term in Eq. (12) as

[qn qn , pk pl ] = [qn , pk pl ]qn + qn [qn , pk pl ] = ([qn , pk ]pl + pk [qn , pl ])qn + qn ([qn , pk ]pl + pk [qn , pl ])
= (nk pl + nl pk )qn + qn (nk pl + nl pk ) = pl qk + pk ql + qk pl + ql pk . (20)

From the comparison between (19) and (20), it follows

[pn pn , qk ql ] = [qn qn , pk pl ], (21)

which yields
[H, Tkl ] = 0, (22)

and hence the functions Tkl are the constants of motion for all k and l, as required.
Next we obtain this result in a shorter way using some additional relations given
by Eqs. (6.54) in Lecture Notes,
u
[u, qk ] = p
(

u
k
(23)
[u, pk ] = q k
.

Let u = H, then ( H
[qk , H] = p k
H (24)
[pk , H] = q k
.
In our case the Hamiltonian H is given by Eq. (8), and thus

= pmk
( H
pk
H (25)
qk
= m 2 qk

and Eqs. (24) become

[qk , H] = pmk

(26)
[pk , H] = m 2 qk
 With the use of these relations, we have

[Tkl , H] = [pk pl + (m)2 qk ql , H] = [pk , H]pl + pk [pl , H] + (m)2 ([qk , H]ql + qk [ql , H])
pl
= m 2 (qk pl + pk ql ) + (m)2 ( pmk ql + qk m )
= m 2 (qk pl + pk ql pk ql qk pl ) = 0, (27)

as it should be.
 I. VARIATIONAL APPROACH TO MECHANICS

1. Hamiltons Variational Principle

Consider the integral
Zt2
I= L dt, (1)
t1

where the function L is the Lagrangian of the mechanical system defined as the
difference between the kinetic, T , and potential, V , energies of the system,

L T V. (2)

Hamiltons variational principle states that the integral I taken along a possible path
of motion of a physical system is an extremum when evaluated along an actual path
of motion. In other words, out of the innumerable ways in which a system could
change its configuration during a time interval between t1 and t2 , Nature chooses
the way that either maximizes or minimizes the integral I which is called the action
for the system at hand. Mathematically, this statement can be expressed as follows:
Zt2
I = L dt = 0, (3)
t1

where means a variation in the entire integral about its extremum value. We
assume that such a variation is obtained by varying the coordinates and velocities
of a system away from the values which are actually taken as the system evolves in
time from t1 to t2 , under the constraint that all system parameters are assumed to
be unvaried values at the endpoints of the motion at t1 and t2 .
2. Hamiltons Principle: General Mathematical Formulation
The generalized coordinates are any collection of independent coordinates qi (that
is, not connected by any equations of constraint) that just suffice to specify uniquely
the configuration of a mechanical system. It follows that the required number of
generalized coordinates is equal to the number of degrees of freedom for a system. If
fewer than this number is chosen, the state of the system will be indeterminate; if a
greater number would be chosen, then some of the coordinates should be determined
through the others by conditions of constraint.
In general, if N particles move in free space but their 3N Cartesian coordinates
are connected by M conditions of constraint,

fj (xi , yi , zi , t) = 0, i = 1, 2, ..., N ; j = 1, 2, ..., M, (4)

then there exists

n = 3N M (5)
independent generalized coordinates q1 , q2 , ..., qn sufficient to describe the position of
the N particles and n degrees of freedom is available for the motion this system. The
number n is a characteristic constant of the system which cannot be altered. Less
than n parameters are not enough to determine the position of the system. More
than n parameters are not required and should not be assigned without satisfing
certain additional conditions. For the analytical treatment, neither the number N

1
of particles, nor the coordinates of these particles are of importance but the gen-
eralized coordinates q1 , q2 , ..., qn are because we assume that Cartesian coordinates
are expressible in terms of the generalized coordinates as

x1 = x1 (q1 , q2 , ..., qn )
...................................
...................................
zN = zN (q1 , q2 , ..., qn ). (6)

With these expressions, the constraint conditions (4) are satisfied identically, and
hence there are no connections between the generalized coordinates.
However, sometimes it is preferable to use the generalized coordinates whose total
number n is larger than the number of degrees of freedom n0 ,

n = n0 + m, (7)

where 0 < m M . In such cases the generalized coordinates are not independent
but subject to m constraint conditions of the form

fi (q1 , q2 , ..., qn ) = 0 (i = 1, ..., m). (8)

In analytical mechanics, the system is called holonomic system if the constraints

are expressed in the form of algebraic equalities like Eqs. (4) or (8), and the non-
holonomic system otherwise. Say, in the case of a particle constrained to remain
outside the surface of a sphere of the radius R the condition of constraint is given
by inequality
R2 (x2 + y 2 + z 2 ) 0, (9)
and hence this system is nonholonomic. Another typical example of the nonholo-
nomic system is described by the constraints of the form
X
aji dqi + ajt dt = 0 (j = 1, 2, ..., m). (10)
i

when these equations are not integrable and thus cannot be expressed in the form
of Eqs. (8).
To formulate Hamiltons principle in a compact mathematical form it is conve-
nient also to introduce the concept of the configuration space which is one of the
most imaginative and useful concepts of the analytical mechanics. The idea is very
simple: if we associate with the three numbers x, y, z a point in a three-dimensional
space, why we should not do the same with the n numbers q1 , q2 , ..., qn , considering
these set as a point P in n-dimensional space? And if we adopt this, why we
should not to treat the equations

q1 = q1 (t)
.............
.............
qn = qn (t) (11)

as representing the motion of a point along some curve in this n-dimensional space
called configuration space?

2
 Note that the entire mechanical system is pictured as a single point in the con-
figuration space of no matter how numerous the particles constituting a system may
be, or how complicated are the relations existing between them. So, this concept
makes it possible to extend the mechanics of a single mass point to arbitrarily com-
plicated many-particle mechanical systems. It is necessary only to remind that the
space which carries this point is no longer the ordinary physical space. It is an
abstract space with as many dimensions as the nature of the problem requires. For
the sake of brevity, the points and curves in the configuration space which symbolize
the position and the motion of the mechanical system, are sometimes reffered to as
C-point and C-curve. C-point is frequently called also the system point. Any
C-curve, or the path of motion in configuration space, will not have any necessary
resemblance to the path in space of any actual particle, and each point on the path
in the configuration space, C-point, represents the entire system configuration at
some given instant of time t, so that the time can be considered as a parameter of
the curve.
SUMMARY: Now, we are able to formulate Hamiltons principle as follows:

The motion of an arbitrary conservative mechanical system occurs in such a way

that the variation of the definite line integral in the configuration space
Zt2
I= L(q1 , ..., qn ; q1 , ..., qn ; t) dt (12)
t1

is equal to zero, that is

Zt2
I = L(q1 , ..., qn ; q1 , ..., qn ; t) dt = 0, (13)
t1

where L T V is the Lagrangian of the system.

The integral I is often called the action integral. Then Hamiltons principle is
referred to as the principle of least action. Hamiltons principle in the form (13) is
the starting point for the derivation of Lagranges equations for conservative systems.

II. LAGRANGIAN FORMULATION OF MECHANICS

1. Lagranges Equations for Conservative Systems
The Lagrangian equations of motion for a conservative system subject to holo-
nomic constraints of type (8):
!
d L L
= 0 (i = 1, 2, ..., n). (14)
dt qi qi
NOTE: The derivation of these equations you can find at page 13 in
Lecture Notes.
2. Forces of Constraint: Lagrange Multipliers
We consider first the simple case of a system, described by the two generalized
coordinates connecting by a single equation of constraint
f (q1 , q2 , t) = 0. (15)

3
These are Lagranges equations of motion for the system with constraint given by
Eq. (8).
d L L
= Qi , (i = 1, 2) (16)
dt qi qi
where
f
Qi = (t) (i = 1, 2). (17)
qi
NOTE: Derivation of these equations at pages 18-19 in Lecture Notes.
We can generalize these ideas to the system described by n generalized coordi-
nates with m equation of constraints of the form

fj (q1 , q2 , ..., qn ) = 0 (j = 1, 2, ..., m). (18)

Namely, if knowledge of the constraining forces is desired, then it is necessary to

solve the system of n Lagranges equations
d L L
= Qi , (i = 1, 2, ..., n) (19)
dt qi qi
where Qi are generalized forces of constraint given by
X fj
Qi = j (t) (i = 1, 2, ..., n). (20)
j qi

Equations (19) and (20) with (18) together constitute n + m independent equa-
tions for the set of n + m unknown functions of time: q1 (t), q2 (t), ..., qn (t), and
1 (t), 2 (t), ..., m (t). These equations when solved give one the possibility to deter-
mine not only the set of the generalized coordinates but also the forces of constraint
(20).
For the case of nonholonomic constraints:
d L L
= Qi , (i = 1, 2, ..., n), (21)
dt qi qi
where X
Qi = j (t)aji . (22)
j

These equations must be solved together with the set of constraint equations (10)
which may be considered as the first-order differential equations. Eqs. (22) define
the generalized forces of constraint for the nonholonomic systems.
3. Lagranges Equations: Derivation from DAlemberts Principle
DAlemberts principle, or the principle of virtual work for the dynamical systems
of N particles reads as follows:
N
X
(Fi p i ) ri = 0 (23)
i=1

Here, Fi represents all the forces acting on i-th particle, besides the inertial force
p i given by the second term in paranthesis in (23). Finally, ri are the so-called
virtual displacements. These displacements are akin variations: they are imagined
(the particle do not undergo any real displacements), they are assumed to be of

4
infinitesimal extent, and they assumed to take place instantaneously (i.e. t = 0) in
a way that is consistent with any imposed constraints.
For forces of any kind the Lagranges equations:
NOTE: derivation you can find at pages 24-25 in Lecture Notes:
!
d T T
= Qj , (24)
dt qj qj
In the the case of conservative forces, when the forces are derivable from a po-
tential energy function,
V
Fi = 5i V , (25)
xi
and hence
X xi X V xi V
Qj = Fi = = . (26)
i qj i xi qj qj
Substituting this into Eq. (24)
!
d T (T V )
= 0. (27)
dt qj qj
If the potential energy function V is independent of velocities qj , we can include
V into the first term of Eq. (27) as well and then to introduce the Lagrangian
T = L V reduce this equation to the form
! !
d (T V ) (T V ) d L L
= = 0. (28)
dt qj qj dt qj qj
We see that this equation is nothing but Lagranges equation of motion for conser-
vative systems that we derived previously from Hamiltons variational principle.
In general, for the Lagrangian of type
L(qj , qj , t) = T (qj , qj , t) V (qj ), (29)
Lagranges equations may be written as
!
d L L
= Q0j , (j = 1, 2, ..., n). (30)
dt qj qj
Here the Lagrangian includes the effect of all conservative forces acting on the sys-
tem, whereas all the nonconservative forces and forces of constraint comprise Q0j .
The forces of constraint may be obtained using the method of Lagrange multipliers.
The rest part of Q0j , that is, the generalized nonconservative forces may be calculated
immediately by means of expression:
X xi
Qj = Fi . (31)
i qj
if the nonconservative forces Fi are known, or they must be found from the process
of solving Lagranges equations itself.
4. First Integrals and Conservation Theorems
Under special circumstances a partial integration of the Lagranges equations can
be accomplished with the result
f (q1 , q2 , ...; q1 , q2 , ...; t) = const, (32)

5
which is the first-order differential equation. This relation is called the first integral
of the equations of motion. Such integrals can bring us much information about the
most general properties of a mechanical system. A particularly important example
is the case of the cyclic or ignorable variables for which the Lagrangian does not
contain a certain variable qj , although qj is present.
For a cyclic coordinate qj , the Lagrange equation of motion,
d L L
= 0, (33)
dt qj qj
reduces to
d L
= 0. (34)
dt qj
L
pj , (35)
qj
are the components of the generalized momentum. The terms canonical momentum
or conjugate momentum are often also used for this quantity.
The equation
pj = const (36)
is a first integral which expresses an important conservation theorem:
The generalized momentum conjugated to a cyclic coordinate is conserved.
If we introduce the quantity
X L X
t)
h = h(q, q, qj L pj qj L. (37)
j qj j

and assume that our system is scleronomic, i.e. L does not contain the time
explicitly, we obtain
dh
= 0, (38)
dt

h(q, q,
t) = const. (39)
This relation again constitues a first integral of Lagranges equation (sometimes it
is referred to as the Jacobis integral).
5. Laws of Momentum Conservation
Considering the case of the conservative systems by assuming that V is not a
function of velocities and that T does not depends on qi . The Lagrange equation of
motion
d L L d T V dpj V
= + + = 0, (40)
dt qj qj dt qj qj dt qj
which gives
dpj V
= Qj , (41)
dt qj
where the generalized force Qj is introduced.
Suppose now that the translation coordinate qj under discussion is cyclic. Then
qj cannot appear in V , and we obtain the momentum conservation law reformulated
in terms of the generalized forces via the relation
dpj
= Qj = 0. (42)
dt

6
For our special choice of generalized coordinate qj , this relation is equivalent to the
linear momentum conservation law in the Newton mechanics which stands:
SUMMARY: If a given component of the total applied force vanishes, the cor-
responding component of the linear momentum is conserved.

In the same way it can be shown that if a cyclic coordinate qj is such that dqj
corresponds to the rotation of the system of particles as a whole around some axis,
then the conservation of its conjugate momentum corresponds to the conservation
of an angular momentum.
Conservation law for angular momentum stands:
SUMMARY: If a given component of the total applied torque vanishes, the cor-
responding component of the angular momentum is conserved.

6. Law of Energy Conservation

For a very wide range of systems and sets of generalized coordinates, T is a
quadratic form in the velocities qi ,
1X
T = aik qi qk , (43)
2 i,k

aik being functions of qi , while V is independent of the velocities qi . Under these

circumstances n
T X
pj = = ajk qk , (44)
qj k=1
and thus
X X
h= pj qj L = ajk qj qk (T V ) = 2T T + V = T + V E, (45)
j j,k

where E is the total energy of the system. The conservation theorem with this
special choice of the Lagrangian takes the form

E = T + V = const, (46)

which expresses the well known law of the conservation of energy:

SUMMARY: The sum of kinetic and potential energies is constant during the
motion.
7. Elimination of cyclic variables
The process of eliminating an ignorable variable may be performed into three
steps:
1. Write down the equation for the kinosthenic momentum
dL
dqn
= cn (a)
2. Modify the given Lagrangian to
L = L cn qn (b)
3. Eliminate the ignorable velocity qn by solving the equation (a) for qn and
substituting this solution in (b).
As a result of this reduction process, the original variation problem of n degrees
of freedom is reduced to a new variation problem of (n 1) degrees of freedom.
The process remains the same if the given problem contains more than one cyclic

7
coordinates with only obvious difference: the modified Lagrangian has now to be
defined by X
L=L ck qk , (47)
k

the summation being extended over all the cyclic coordinates.

III. THEORY OF SMALL OSCILLATIONS

1. Lagrangian
The lowest order Lagrangian:
NOTE: derivation you can find at pages 37-38 in Lecture Notes
1
L T V = (Tij i j Vij i j ) (48)
2
which leads via the Lagranges equations to the following n equations of motion
of a mechanical system near the equilibrium.

Tkj j + Vkj j = 0 (k = 1, ..., n). (49)

2. Eigenvalue Problem
These equation of motion are linear differential equations with constant coeffi-
cients, and we thus try an oscillatory solution of the form

i = Cai eit (50)

where Cai is the complex amplitude of the oscillations for coordinate i , the scaling
factor C and the frequency being the same for all coordinates.
n linear algebraic equations for the amplitudes ai :

(Vij 2 Tij )aj = 0. (51)

These are homogeneous equations, and hence the nontrivial solution exists only if
the determinant of the coefficients vanishes:
V11 2 T11 . V1n 2 T1n 
 

 . .

 . . . . . 

. . . . .  = 0. (52)
 

 

 . . . . . 

 Vn1 2 Tn1 . . 2
. Vnn Tnn 
3. General Solution of Equations of Motion
The trial function of the oscillatory type satisfies the equations of motion not
for a single frequency but in general for a set of n frequencies k , k = 1, ..., n.
Mathematically, it means that the complete solution of the equations of motion
may be written as a sum over index k,

i = Ck aik expik t , (53)

where Ck is a complex scale factor. Strictly speaking this solution must be general-
ized to
i = aik (Ck+ exp+ik t +Ck expik t ), (54)

8
since for each eigenvalue k (and thus for each eigenvector ak ) there are two fre-
quences +k and k and accordingly two different scale factors Ck+ and Ck .
Nevertheless the real parts of these equations can be written in the same form

i = fk aik cos (k t + k ) (55)

For the determination of the constants fk and k we put

i (0) = Re Ck aik (56)

Similarly, the initial value of the velocities is

i (0) = Im Ck aik k (57)

These equations form a set of 2n equations from which the real and imaginary parts
of the scale factors Ck , ReCk and ImCk , may be evaluated.
In matrix relation,
Re Cl = ajl Tjk k (0). (58)
A similar procedure with velocities gives
1
Im Cl = ajl Tjk k (0). (59)
l
4. Normal Coordinates

The general solution for each coordinate i like (?) being a sum of simple har-
monic oscillations of the frequencies k is not itself a periodic function of time. It is
possible, however, to transform i to a new set of generalized coordinates that are
periodic function of time. Such set of variables is known as the normal coordinates.
The Lagrangian in a new reference system
1
L = (k k k2 k2 ), (60)
2
which results in the Lagranges equations for k

k + k2 k = 0, (61)

with an obvious solution

k = Ck exp (ik t). (62)
This equation demonstrates that each coordinate k is a periodic function of one
of the frequences k and amplitudes aik . Such states are said to define the normal
modes of vibration, and k are properly called the normal coordinates of the system.
5. Forced oscillations:
a) Sinusoidal driving force: In nature, the driving forces have frequently the
sinusoidal character varying in time as

Qi = Q0i cos (t + i ). (63)

The equation of motion becomes

i + i2 i = Q0i cos (t + i ). (64)

9
The particular solution
i = Bi cos (t + i ). (65)
where
Q0i
Bi = , (66)
i2 2
The complete motion is composed by the linear combinations of the normal modes
aij Q0j
i = aij j = cos (t + j ). (67)
j2 2

b) Dissipative forces:
Equations for the normal coordinates j :

i + Fi i + i2 i = 0, (68)

Solutions: " s #
F t
2i F2
i = Ci e exp i i2 i t. (69)
4
This
r expression represents nothing but the damping oscillations with the frequency
F2
i2 4i and damping coefficient F2i .
c) Combined Effect of Sinusoidal and Dissipative Forces
Equations:

Tij j + Fij j + Vij j = F0i eit , (70)

When searching for solution
j = Aj eit , (71)
it leads to the following set of the linear equations for the amplitudes Aj :

(Vij iFij 2 Tij )Aj = F0i . (72)

These equations are merely the inhomogeneous equations. As such, they can be
solved by the application of the Cramers rule which gives
Dj ()
Aj = , (73)
D()
where
D() = det(Vij iFij 2 Tij ) (74)
and Dj () is obtained from the determinant D() when the j-th column is replaced
by F01 , ..., F0n .
Equating the determinant D() to zero, we obtain the secular equation for free
oscillations. Its roots determine the complex frequencies of the free modes of oscil-
lations which can be treated as the resonance frequencies for the forced oscillations.

10
 IV. CENTRAL FORCE PROBLEM

1. Lagranges Equations and First Integrals

The Lagrangian in plane polar coordinates is expressed as
1
L = T V = m(r 2 + r2 2 ) V (r), (75)
2
Canonical momentum
L
p = mr2 , (76)

is conserved which is equivalent to the equation of motion
d = 0,
p = (mr2 ) (77)
dt
We notice that p is the angular momentum of a single particle

mr2 = l, (78)

determines the constant magnitude l of the angular momentum.

Differential area:
1 1
dA = r(rd) = r2 d, (79)
2 2
and the areal velocity is
1 2
dA r d 1 d 1
VA = 2 = r2 r2 . (80)
dt dt 2 dt 2

d = 2m d ( 1 r2 )
= 2m dVA = 0,
(mr2 ) (81)
dt dt 2 dt
dVA
=0 (82)
dt
i.e. the areal velocity is constant. Note that this statement sounds as a general
property of the central force motion, whereas the well-known Keplers second law of
planetary motion: the radius-vector sweeps out equal areas in equal times.
The equation for the remaining coordinate r.
h
L
r
= 1
r 2
m(r 2 + r2 2 ) V (r)] = mr;

L
r
= 1
[ m(r 2
r 2
+ r2 2 ) V (r)] = mr2 V (r)
r
, (83)

The Lagranges equation gives

d L L d V (r)
mr2 +
= (mr) . (84)
dt r r dt r
Using the designation
V (r)
f (r) = , (85)
r
f (r) being the value of the force along r

r mr2 = f (r).
m (86)

11
 l2
r
m = f (r). (87)
mr3
l2 d 1 l2 1 l2
! !
dV d
m
r = f (r) + = = V + (88)
mr3 dr dr 2 mr2 dr 2 mr2
1 l2 dr
!
d
rr =
m V + (89)
dr 2 mr2 dt
1 l2
! !
d 1 2 d
mr = V + (90)
dt 2 dt 2 mr2
d 1 2 1 l2
!
mr + +V =0 (91)
dt 2 2 mr2
1 2 1 l2
mr + + V = const, (92)
2 2 mr2
1 l2 1 2 4 2 mr2 2
= m r = , (93)
2 mr2 2mr2 2
1 2 1 2 2
mr + mr + V (r) = E, (94)
2 2
where E is the total energy of the system.
For r,
v !
2
u
u2 l
r = t EV (95)
m 2mr2
dr
dt = q . (96)
2 l2
m
(E V 2mr2
)
Zr
dr
t= q , (97)
2 l2
r0 m
(E V 2mr2
)
which gives t as a function of r. Inverting this function, we can find r as a function
of t. Then the solution
d l
= (98)
dt mr2 (t)
Zt
dt
(t) = l + 0 , (99)
mr2 (t)
0

where 0 is the initial value of (t).

2. Classification of Orbits-see Figures 3-3 to 3-11 in Lecture Notes
The particle velocity is determined in the form
s
2 
v= E V (r) . (100)
m

l2
m
r=f+ , (101)
mr3

12
Formally this equation corresponds to the particle of mass m subjected to a force
l2
f0 = f + . (102)
mr3
The physical sense of the extra term in this expression is established by the rela-
tionship
l2 2 mv ,
2
= mr (103)
mr3 r
which is the familiar centrifugal force.
The total force
V 0
f0 = , (104)
r
where V 0 is the equivalent one-dimensional potential, defined by
1 l2
V0 =V + . (105)
2 mr2
The energy conservation theorem
1
E = V 0 + mr 2 . (106)
2
Special case of an inverse square law of force,
k
f = (107)
r2
with k > 0, which ensures that the force acts toward the center of force. For this
force,
k
V = , (108)
r
and the equivalent one-dimensional potential is
k 1 l2
V0 = + . (109)
r 2 mr2
3. Equations for Orbits
The dependence of r upon :
ldt = mr2 d. (110)

ldr
d = q =
2 l2
mr2 m
(E V 2mr2
)
dr
q
2mE
r2 l2
2mV
l2
1
r2
du
q , (111)
2mE 2mV
l2
l2
u2
where the variable r is changed to u = 1/r.
Zu
du
= 0 q . (112)
2mE 2mV
u0 l2
l2
u2

13
 This integral solves the problem formally but it can be practically used only for
certain types of potential laws. The most important is the power-law potential,

V = arn+1 , (n 6= 1). (113)

Zu
du
= 0 q . (114)
2mE 2ma n1
u0 l2
l2
u u2
The total set of integral exponents n resulting in circular and elliptic functions is

n = 5, 3, 1, 0, 2, 3, 4, 5, 7. (115)

In these cases the solution can be investigated analytically because the properties
of the circular and elliptic functions.
The functions (u) (or u()) can be treated as the solutions of some differential
equation for the orbit. This equation can be obtained by converting the equation of
motion
l2
mr = f (r), (116)
mr3

d l d
dt
= mr2 d
,
 
d2 l d l d
dt2
= 2
mr d mr d 2 , (117)

and the Lagrange equation

l2
!
l d l d
2 2
= f (r). (118)
r d mr rd mr3
!
1 dr d 1 du
2
= . (119)
r d d r d
d2 u
!
m 1
+ u = f J(u), (120)
d2 l2 u2 u
In terms of potential as
d2 u
!
m d 1
2
+u= 2 V J(u), (121)
d l du u
since
d dr d 1 d
= = 2 . (122)
du du dr u dr
4. Kepler Problem
Introduce the inverse-square law,
k k
f (r) = , V (r) = , (123)
r2 r
!
1
f = ku2 (124)
u

14
into the general differential equation for the orbit
d2 u
!
m 1 mk
2
+ u = 2 2f = . (125)
d l u u l2
The change of the variable
mk
y =u (126)
l2
transforms this equation to
d2 y
+y =0 (127)
d2
with the solution
y = B cos( 0 ), (128)
B and being the two constant of integration. In terms of r this solution is
1 mk  
= 2 1 + e cos( 0 ) , (129)
r l
where
l2
e=B
. (130)
mk
To rederive this expression, starting from the formal solution, we represent the
latter in the form, containing indefinite integral:

0
Z
du
= q . (131)
2mE 2mku
l2
+ l2
u2

Setting
2mE 2mk
= , = , = 1, (132)
l2 l2
we obtain the standard integral
!
Z
du 1 + 2u
= arccos , (133)
+ u + u2 q
where the discriminant q is
!2
2El2
!
2 2mk
q = 4 = 1+ . (134)
l2 mk 2
l2 u
0 mk
1
= arccos q . (135)
2El2
1+ mk2

Converting this equation, we obtain the equation of the orbit in the form
s !
1 mk 2El2
= 2 1+ 1+ cos( 0 ) , (136)
r l mk 2
s
2El2
e= 1+ . (137)
mk 2

15
The general equation of a second order curve ( or conic section) with the focus at
the origin [any MATHEMATICAL HANDBOOK or Fowles, Cassiday. Analytical
mechanics, Appendix C],
1  
= C 1 + e cos( 0 ) , (138)
r
where C is constant and e the eccentricity of the conic section which defines the
type of the curve according to
e > 1, hyperbola,



e = 1, parabola,
(139)



e < 1, ellipse,
e = 0, circle.
From this comparison, it follows that the orbit is always a conic section, so that the
type of the conic depends on the total energy E according to


E > 0, hyperbola,
E

= 0, parabola,
(140)
E < 0, ellipse,
2

E = mk


2l2
, circle.
In the particular case of an elliptic orbit with the two apsidal distances r1 and r2
it is instructive to evaluate the so-called semimajor axis,
r1 + r2
a= . (141)
2
The most elegant method of this calculation is based on the use of the conservation
law which, for the apsidal points, where the radial velocity is zero, stands
l2 k
E 2
+ = 0, (142)
2mr r
or
k l2
r2 + r = 0. (143)
E 2mE
The roots of this equation are just the apsidal distances, so that
r1 + r2 k
a= = . (144)
2 2E
It is useful also to rewrite the eccentricity of the ellipse in terms of the semimajor
axis as s
l2
e= 1 . (145)
mka
From this expression we have
l2  
= a 1 e2 . (146)
mk
 
a 1 e2
r= (147)
1 + e cos( 0 )

16
For the apsidal distances which occur when 0 is 0 and , we have

r1 = a(1 e) (148)

and
r2 = a(1 + e), (149)
as is to be expected for an ellipse.

5. Keplers Laws
The third Keplers law which in its original formulation reads:
The squares of the times of circulation of the planets around sun are proportional
to the cubes of their middle distances to the sun.
r
mZ dr
r
t= q . (150)
2r k
l2
+ E
0 r 2mr2
r
mZ r dr
r
t= q . (151)
2k r r2
r 2a a(1e2 )
0 2
The integral in the r.h.s. of this equation is most conveniently evaluated through
the auxiliary variable , denoted in the terminology of medieval astronomy as the
eccentric anomaly and defined by the relation

r = a(1 e cos ). (152)

s
ma3 Z
t= (1 e cos ) d. (153)
k
0

m
r
3/2
= 2a . (154)
k
This equation states that, other things being equal, the square of the periods of the
various planets are proportional to the cube of their major axis. This statement can
be considered as more precise formulation of the third of Keplers laws.

V. HAMILTONIAN FORMULATION OF ANALYTICAL

MECHANICS

1. Hamiltons Equations
NOTE: derivation you can find at pages 73-74 in Lecture Notes
Hamilton discovered a remarkable transformation which renders the Lagrangian
function linear in the velocities which leads to the 2n first-order equations expressed
in terms of 2n independent variables , namely n generalised coordinates qi and n
generalised or conjugated momenta pi :
L
pi . (155)
qi
In the Hamilton formulations we can again speak of the system point and its path
but - in the space of 2n dimensions which has been called by the American physicist

17
Gibbs the phase space. The coordinates of the system point in the phase space
are the original coordinates of Lagrangian mechanics qi (t) along with the conjugate
momenta pi (t). To find the equation of motions in the variables qi and pi replacing
the Lagranges equation, we introduce the Hamiltonian
H = qi pi L, (156)
Applying the variational procedure we get the system of the 2n first order equations
qi = H
(
pi
,
H (157)
pi = qi ,

which are known as the canonical equations of Hamilton. For cyclic coordinate, the
conjugate momentum pi is constant.
2. Evaluating of Hamiltonian-see examples at pages 77-79 in Lecture
Notes
The Hamiltonian for each system can be constructed via the Lagrangian formu-
lation by performing the following sequence of steps:
1. With a chosen set of the generalized coordinates qi , the Lagrangian L(qi , qi , t)
is constructed.
2. The congugate momenta are defined as the functions of qi , qi and t by equation:
L
pi . (158)
qi
3. To form the Hamiltonian, or rather some mixed function of qi , qi , pi and t we use
the equation:
H = qi pi L, (159)
4. Equations from (2) are then inverted to obtain qi as functions of (q, p, t).
5. The results of the previous step are then applied to eliminate qi from H so as
to express it solely as a function of (q, p, t).
VI. CANONICAL TRANSFORMATIONS-see example: harmonic
oscillator in canonical transformation at pages 85-87 in Lecture Notes
Jacobi has discovered a special type of coordinate transformation which gives
rise to so highly simplification of the starting canonical equations that in the new
coordinates these equations become directly integrable. The essential feature of this
transformation is that it preserves the canonical equations themselves. In this sense,
the transformation is said to be the canonical transformation.
1. Generating functions
The transformed variables Qi , Pi , and the transformed Hamiltonian by K(Q, P, t),
we may write

Qi = Qi (q, p, t),

Pi = Pi (q, p, t),

Q i = P
K
, (160)
i


P = K .


i Qi

From the application of variational principle the equation

dF
pi qi H = Pi Q i K + (161)
dt

18
holds. four different possibilities of a choice of F as a function of mixed set of old
and new variables,

F1 (q, Q, t),
F (q, P, t) Q P ,

2 i i
F = (162)
q
i i

p + F 3 (p, Q, t),
qi pi Qi Pi + F4 (p, P, t).

The remarkable feature of any choice is that in each case the variables which are
complementary to the total set of old and new variables, (q, p, Q, P ), are determined
by the simple differentiation of the corresponding functions Fi according to:
pi = F
(
qi
1
,
F1 (q, Q, t) F1
Pi = Qi ;
pi = F
(
qi
2
,
F2 (q, P, t) F2
Qi = Pi ;
qi = F
(
pi
3
,
F3 (p, Q, t) F3
Pi = Qi ;
qi = F
(
pi
4
,
F4 (p, P, t) F4 (163)
Qi = Pi .

To complete the story, we notice that the connection between the new Hamiltonian
K and the old one H is always expressed by the same formula formula
Fi
K=H+ , (i = 1, 2, 3, 4). (164)
t
In particular case of the time-independent function F we have K = H, and in order
to obtain new Hamiltonian K it is enough to substitute into H the variables p, q,
expressed in terms of new variables P, Q.
2. Poisson Bracket Formulation of Mechanics Almost the entire framework
of Hamiltonian mechanics can be restated in terms of the so-called Poisson brackets.
For the two arbitrary functions of canonical variables u(p, q) and v(p, q), the Poisson
bracket is defined as
u v u v
[u, v] = . (165)
qi pi pi qi
Poisson bracket formulation of mechanics: let us suppose that any function of canon-
ical variables and time satisfies the generalized equation of motion
du u
= [u, H] + , (166)
dt t
where H is the Hamiltonian of the system, and [u, H] is the Poisson bracket.

VII. HAMILTON-JACOBI THEORY

1. Hamilton-Jacobi Equation
There is an alternative technique to use the generating function directly for seek-
ing a canonical transformation from the coordinates and momenta, (q, p), at time
t to a new set which is nothing but the 2n initial values of the coordinates and

19
momenta, q0 , p0 , at t = 0. With such a transformation, the equations of the trans-
formation relating the old and new canonical variables are then exactly the desired
solution of the mechanical problem:

q = q(q0 , p0 , t),
p = p(q0 , p0 , t). (167)

To obtain this equation, we notice that the new variables shall be surely constant in
time if we require that the transformed Hamiltonian K is identically zero, for then
the equations of motion are

Q i = P
K
i
= 0,
Pi = K = 0.
Qi
(168)

Further, we know that the new Hamiltonian K relates to the old Hamiltonian H by
means of the equation
F
K=H+ , (169)
t
where F is a generating function to be determined. Now, we see that K will be zero
if this function satisfies the equation
F
H(q, p, t) + = 0. (170)
t
The question arises: what variables the function F depends on? It is most
convenient to take F as a function of the old coordinates qi , the new (constant!)
momenta Pi , and the time . This generating function is F2 (q, P, t). The equation of
transformations
F2
pi = , (171)
qi
!
F2 F2 F2
H q1 , ..., qn ; , ..., ;t + = 0. (172)
q1 qn t
This equation is Hamilton-Jacobi equation. In the context of this equation, it is
customary to use the symbol S for the generating function F2 ,

F2 S. (173)

In this notation the Hamilton-Jacobi equation is rewritten in a more familiar form

!
S S S
H q1 , ..., qn ; , ..., ;t + = 0. (174)
q1 qn t
The function S is usually called the Hamiltons principal function.

2. Hamiltons Principal Function-see example of harmonic oscillator

in Hamilton-Jacobi theory, at pages 93-95 in Lecture Notes We can take
the solution in the form

S = S(q1 , ..., qn ; 1 , ..., n ; t), (175)

20
In this form the solution S is a function of n coordinates, the time t and some
n independent quantities i which we can freely identify with the new constant
momenta:
i P i i , (176)
These identities can be considered as the first part of canonical transformations.
The other part which provide the new coordinates appear as a result of calculation
of the right side of equations
S(q, , t)
Q i i = (177)
i
at the initial time t = t0 with the known initial values of qi . These equations can
then be turned inside out to furnish qi in terms of , and t:
qi = qi (, , t). (178)
By this, we solve the problem of finding the coordinates as functions of time and
the initial conditions. The congugate momenta pi ,
pi = pi (, , t) (179)
can be then found by substitution of qi into expressions
S(q, , t)
pi = (180)
qi
after the differentiations have been performed. Equations for qi and pi constitute
the complete solution of equations of motion in the Hamilton formulation.
3. Hamiltons Characteristic Function
What is the canonical transformation in which the new momenta are all constant
of the motion i , and where one of these constants (say, 1 ) is equal to H? The
unknown function of our interest by W (q, P ), where (q, P ) is the set of all coordinates
and momenta describing the system. The equation of transformation that we are
searching for, are
W W W
pi = , Qi = = . (181)
qi Pi i
By making agree the first of this transformations with the requirement
H(q, p) = 1 , (182)
we obtain the condition for determining the generating function W
!
W
H q, = 1 . (183)
q
We see that this condition has the form of a partial differential equation which can
be evidently referred to as the Hamilton-Jacobi equation for the function W . Since
this function does not depend on t, the new and old Hamiltonian are equal
H = K = 1 = P 1 . (184)
This leads to the canonical equations of motion
Pi = Q
K
= 0,
i
1, i = 1
Q i = P
K
= (185)
i 0, i 6= 1

21
with the immediate solutions

P i = i ,
t + 1 , i = 1

W
Qi = (186)
i i , i 6= 1
So, the function W which obeys the Hamilton-Jacobi equation (40) is the function
which we are searching for, since it generates n constant momenta i , with 1 = H.
As for the new coordinate, the only one is not simply a constant of the motion,
namely, Q1 , which is equal to the time plus a constant. The realtion between
Hamiltons principal and characteristic functions:

S(q, P, t) = W (q, P ) 1 t, (187)

4. Separable Variables
A coordinate qi is separable when it is possible to split Hamiltons principal
function into two additive parts, one of which depends only on the coordinate qi
and the other is entirely independent of it.

S(q1 , ..., qn , P, t) = S1 (q1 , P, t) + S 0 (q2 , ..., qn , P, t), (188)

The Hamilton-Jacobi equation

!!
S S S
H qi , , t, f q1 , + = 0, (189)
qi q1 t
where qi represents the set of all the generalized coordinate qi0 s except q1 . The
solution
S(q, P, t) = S1 (q1 , P ) + S 0 (qi , P, t), (190)

S 0 S 0
! !
dS1
H qi , , f q1 , ,t + = 0, (191)
qi dq1 t

 
f q1 , dS
dq1
1
= 1 ;
0 S 0
 
H qi , S
qi
, 1 , t + t
= 0, (192)

where 1 is an arbitrary constant.

For conservative systems:

S(q, P, t) = W (q, P ) + S1 (P, t), (193)

The Hamilton-Jacobi equation

!
W S1
H q, + = 0. (194)
q t

S1
= 1 ;
t 
H q, W
q
= 1 . (195)

22
The first of this equations shows that
S1 (P, t) = 1 t, (196)
We can apply the separation of variables to this equation and, if succeded, we can
obtain the complete integral of it in the form
n
X
W = Wk (qk ; 1 , ..., n ), (197)
k=1

where each function Wk is dependent of only one coordinate qk and i are the
constants of integration as always.
The further simplification is possible in the case of cyclic coordinates. The mo-
menta which corresponds to such coordinates are constant:
Wi
pi = = i . (198)
qi
Consider, for example, the case when all coordinates, besides q1 , are cyclic, so that
the Hamiltonian depends only on q1 . The Hamilton-Jacobi equation
!
W1
H q1 , , 2 , ..., n = 1 . (199)
q1

Wi = i qi . (200)
The total characteristic function
n
X
W = W1 + i qi . (201)
i=2

5. Action-Angle Variables -see Kepler problem in action-angle vari-

ables at pages 106-108 in Lecture Notes
For either type of periodic motion we can introduce the action variable J,
I
J= p dq, (202)

where the integration is to be carried over a complete period of libration or rotation.

Using this variable as the new generalized momentum, we can write Hamiltons
characteristic function as
W = W (q, J). (203)
Then the transformation equation gives the generalized coordinate w conjugate to
J:
W
w= . (204)
J
w is known as the angle variable.
The equation of motion for w is reduced to
H
w = = (J), (205)
J
where is a constant function of J only. This equation shows that w is always a
linear function of time,
w = t + . (206)

23
 1
= . (207)

Therefore the quantity can be interpreted as the frequency associated with the
periodic motion described by the function q(t).

24