Lecture 4 : THEORY OF SMALL

OSCILLATIONS
The lowest order Lagrangian
1
L T V = (Tij i j Vij ij )
(1)
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).

(2)

Eigenvalue Problem
These equation of motion are linear differential equations with constant coefficients,
and we thus try an oscillatory solution of
the form
i = Caieit
(3)
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 2Tij )aj = 0.

(4)

These are homogeneous equations, and hence

the nontrivial solution exists only if the determinant of the coefficients vanishes:




2
2

 V

T
.
.
.
V

T


11
11
1n
1n






.
.
.
.
.








.
.
.
.
.
 = 0. (5)









.
.
.
.
.




2
2

 Vn1 Tn1
. . . Vnn Tnn 

The simplified version of the general problem: the appropriate generalized coordinates are the Cartesian coordinates of the
system of point particles multiplied by the
square root of the particle mass (sometimes referred to as mass-weighted coordinates).
The kinetic energy takes the form:
1
1
T = i2 = (~ )2,
(6)
2
2
In this case Tij = ij and
Vij aj = ai,

(7)

where 2 = . These equations forms the

well-known eigenvalue problem of the linear
algebra. It is convenient to put this problem in matrix form as
V a = a,

(8)

where V is n n matrix with elements Vij

and a is n-dimentional vector with components ai. This equation is referred to as
the eigenvalue equations.

The values of for which the equation

is solvable are known as the characteristic
values, or eigenvalues of V , and the vector
solutions a are called the eigenvectors of V .
Two facts from the linear algebra:
(i) the eigenvectors of a given matrix are
orthogonal to each other;
(ii) the eigenvalues of V are real if V is
symmetrical and real.

For arbitrary symmetric and real matrix

Tij the eigenvalue equation is rewritten in
matrix form as,
V a = T a,

(9)

also stands the eigenvalue problem for a

given matrix V . Note, that here the effect
of V on the eigenvector a is not merely
the same vector times the factor , but
a multipe of the result of the other given
matrix T acting on a. Nevertheless, even
in this general case we are in a position
to show that the eigenvectors a are still
orthogonal, and the eigenvalues are all
real under some conditions imposed on V
and T .

We can form all the eigenvectors ak into

a square matrix A with components ajk .
Then we get one matrix equation:
A = I,
AT

(10)

the transposed matrix A is obtained from A

by interchanging rows and columns. This
condition expresses in the matrix form orthonormality property of the eigenvectors
ak, i.e. these vectors are orthogonal and
of unit magnitude - but in Riemanns space!
To show this, we must recall some facts
concerning to this space. Riemanns space,
or Riemann geometry is based on the quadratic
differential form in the variables x1, ..., xn
2

ds = (g11dx1 + g12dx2 + g13dx3 + ... + g1ndxn)dx1

+(g21dx1 + g22dx2 + g23dx3 + ... + g2ndxn)dx2
................................................................
+(gn1dx1 + gn2dx2 + gn3dx3 + ... + gnndxn)dxn(11)
whose coefficients, g11, ..., gnn, are generally
not constant but functions of the n variables x1, ..., xn. These coefficients form the
metric tensor G of the n-dimensional space.

Clearly for Cartesian coordinates the metric tensor is the unit matrix I. In more
general case of the curvilinear orthogonal
coordinates (such as polar coordinates) G will
be not unit but still diagonal matrix. The
dot (or scalar) product of two vectors a and
b in Riemanns space is given by
a b aigik bk
aGb.

(12)

ab=
aGb = 0,

(13)

If
the vectorsa and b are called orthogonal.
Furthermore, vector a is said to be normalized to unity if
aa=
aGa = 1.

(14)

The eigenvectors ak are orthogonal and

that they are normalized to have unit magnitude in a particular Riemanns space for
which T is the metric tensor. Note, that the
kinetic energy (8) can be represented in
this space as
1 1
1
1 2
1
T = Tij i j = ~ T ~ = ~ G~ ~ ~ (~ ) ,
2
2
2
2
2
(15)

. 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,

(16)

where Ck is a complex scale factor. Strictly

speaking this solution must be generalized
to
i = aik (Ck+ exp+ik t +Ck expik t),

(17)

since for each eigenvalue k (and thus for

each eigenvector ak ) there are two frequences
+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 )

(18)

For the determination of the constants fk

and k we put
i(0) = Re Ck aik

(19)

Similarly, the initial value of the velocities

is
i(0) = Im Ck aik k
(20)
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).

(21)

A similar procedure with velocities gives

1
Im Cl = ajl Tjk k (0).
(22)
l

Normal Coordinates
The general solution for each coordinate
i being a sum of simple harmonic 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 potential energy can be written as
can be written as
1
1
1
1
1 2 2
V = AV A = = l lk k = l lk k k = k k .
2
2
2
2
2
(23)
In analogous way we obtain for the kinetic energy
1
1

= AT A,
(24)
T = T
2
2
which can be reduced to
1
1
T = = k2.
(25)
2
2

We notice that (i) both the potential and

kinetic are sums of squares only (any cross
terms vanish), and (ii) both have been diagonalized by A which defines the principal axes transformation. It follows then
that the principal axes transformation which
we employ here is the well-known algebraic process of simultaneous diagonalization
of two quadratic forms.
The Lagrangian in a new reference system
1
L = (k k k2k2),
(26)
2
which results in the Lagranges equations
for k
k + k2k = 0,
(27)
with an obvious solution
k = Ck exp (ik t).

(28)

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.

Example: Triatomic Molecule

We approximate the actual complicated
interatomic potential by two strings of force
constant k, and thus take the potential energy in the form
k
k
2
V = (x2 x1 b) + (x3 x2 b)2. (29)
2
2
Denoting the equilibrium position of the
atoms by x0i, i = 1, 2, 3, we have
x2 x1 b = x2 x02 (x1 x01) b + x02 x01 2
x3 x2 b = x3 x03 (x2 x02) b + x03 x02 3
where we introduce new coordinates
i = xi x0i

(31)

and took into account that

x02 x01 = b,

x03 x02 = b.

(32)

In terms of these new variables the potential is rewritten as

k
k
V = (2 1)2 + (3 2)2 =
2
2
k 2
(1 + 222 + 32 212 223)
2
1
1
Vij ij V .
(33)
2
2

where matrix V has the form

k
k
0

(34)
V = k 2k k .
0
k
k
The kinetic energy has an obvious form
m
M
T = ( 12 + 32) + 22,
(35)
2
2
or, in matrix form
1
,
T = T

(36)
2
where T is the diagonal matrix

0
0
m

T = 0 M 0
(37)
0
0 m
The characteristic equation , we obtain


2



 k m
k
0




2

2
2k M
k  = 0.
|V T | =  k


0
k
k 2m 
(38)
Direct evaluation of this determinant yields
2(k 2m)[ 2M m k(M + 2m)] = 0,

(39)

which immediately yields the following solutions:

v
v
u
u
u
u
u k
u k 
2m 
u
u
t
t
1 = 0, 2 =
, 3 =
1+
. (40)
m
m
M

The first solution represents an example of the neutral equilibrium mentioned

above: the molecule under consideration
is translated along the line without any
change in its potential energy - in full agreement with the corresponding equation for
the first normal coordinate
1 = 0,

(41)

which produces a uniform translational motion.

The second solution, 2, is recognized as
the frequency of oscillations for a mass m
without any reference to the central mass
M which remains stationary in this mode
of motion. And only the third solution, 3
depends on the value of the central mass
which points out the participation of this
mass in the motion. All these predictions
are completely verified by examining the
eigenvectors on the basis of the eigenvalue
equation.

For the components aij with the frequency

j , we have
(k 2m)a1j ka2j = 0,
2
ka
+
(2k

M )a2j ka3j = 0,
1j

ka
2j + (k m)a3j = 0.

(42)

The normalization condition

m(a21j + a23j ) + M a22j = 1

(43)

For the first mode 1 = 0, and the system

immediately gives
a11 = a21 = a31,

(44)

i.e. all three amplitudes are the same as

it should be in the case of a pure translational motion . The values of these amplitudes are fixed by the normalization conditions which gives
1
1
a11 =
, a21 =
(45)
2m + M
2m + M
1
a31 =
,
(46)
2m + M

For the second mode the amplitudes are

a22 = 0 and a12 = a32, i.e. the center atom
is at rest, whereas the two edge atoms vibrate exactly out of phase. The normalized values of the amplitudes are
1

a12 =
, a22 = 0,
(47)
2m
1

a32 =
,
(48)
2m
For the third mode when (k 2m) 6= 0,
a13 = a33. Combining this with the normalization condition, we obtain
1
2
s
,
a13 = s
,
a
=

23
2m
M
2m(1 + M )
2M (2 + m )
(49)
1
,
a33 = s
(50)
2m
2m(1 + M )
In this case the two edge atoms oscillates
with the same amplitudes, while the central atom vibrates out of phase with them
and has a different amplitude .