Sect. 6.

3: Free Vibration
Frequencies & Normal Coordinates
• We just saw, for a system displaced slightly from equilib:
Eqtns of motion: Tijηj + Vijηj = 0 (i = 1,2,3, … n) (1)
are satisfied by oscillatory solutions: ηj  Caje-iωt

• In general: n eigenfrequencies ωk (n allowed oscillation

frequencies). (1) is a linear differential equation.
 The complete solution of (1) is a linear combination
(superposition) of oscillations at all allowed frequencies.
 A system displaced slightly from equilib performs small
oscillations about equilib with frequencies:
ω1,ω2, ... ωn.  Frequencies of “free” vibration
 Resonant or natural frequencies of the system
• Equations of motion:
Tijηj + Vijηj = 0 (i = 1,2,3, … n) (1a)
Or: (Vijaj - ω2Tijaj ) = 0 (i = 1,2,3, … n) (1b)
• General solution (summation convention):
ηj(t) = Ckajkexp(-iωkt) (2)
ajk  eigenvector elements (make up columns & rows of “eigenvector
matrix” A). Ck  a scale factor (complex).
• (1b) & (2)
 For each eigenvalue λk = (ωk)2  Seems as if there may be 2
resonant frequencies:  ωk. But recall, the actual motion is the real
part of (2). Text (p. 250) goes into great detail about how this means that
there is only one physical resonant frequency.
• General solution (summation convention):
ηj(t) = Ckajkexp(-iωkt) (2)
ajk  eigenvector elements (make up columns & rows of
“eigenvector matrix” A). Ck  a scale factor (complex).
• Ck: Determined by the initial conditions. (As the solution to
2nd order differential equations, need 2 initial conditions to
specify the solution). Text (p. 250-251) derives relations between
the complex factors Ck & the initial generalized displacements ηj(0)
& velocities ηj(0):
ηj(0) = ∑kRe(Ckajk) ηj(0) = ∑kIm(Ckajkωk)
and
Re(Ci) = ∑(j,k)ajiTjkηk(0) Im(Ci) = (ωi)-1∑(j,k)ajiTjkηk(0)
• General solution (summation convention):
ηj(t) = Ckajkexp(-iωkt) (2)
= sum of simple harmonic oscillations at all frequencies ωk
satisfying the eqtn of motion (secular eqtn). Eqtns of motion for the
ηj = coupled simple harmonic oscillator eqtns
• It is possible, & quite useful, to transform from the
generalized coords ηj(t) to a new set of generalized coords
 ζj(t) called the Normal Coordinates, in which each ζj
oscillates separately at only one of the resonant
frequencies ωj. Equations of motion for the ζj = uncoupled
simple harmonic oscillator eqtns. That is, each ζj satisfies a simple
harmonic oscillator equation of motion at a different ωj & there is
no coupling between them!
• Normal Coordinates, ζj(t): Text (p. 251) shows that
these coords are obtained from the ηj by:
η = Aζ or ζ = A-1η = Ãη
where A = matrix of the eigenvectors & η,ζ = column
vectors of the generalized displacements.
• Text (p. 251-252) shows: Both the KE T & the PE V are
diagonal in the Normal Coordinate representation.
That is, a similarity transformation with A diagonalizes both
T & V (summation convention):
T = (½)ζiζi V = (½)(ωi)2ζiζi
Note: with this definition, factors of the masses are included in the
definition of ζi, so the mass doesn’t appear explicitly in either T or V!
• In terms of Normal Coordinates:
T = (½)ζkζk V = (½)(ωk)2ζkζk
 The Lagrangian is:
L = T - V = (½)[ζkζk - (ωk)2 ζkζk]
 Lagrange’s eqtns of motion:
(d/dt)[(L/ζk)] - (L/ζk) = 0 (k = 1,2,3, … n)

Give: ζk + (ωk)2ζk = 0 (k = 1,2,3, … n)

With solutions: ζk = Ckexp(-iωkt)

n uncoupled SHO equations!
 Each Normal Coordinate is a sinusoidal function of
only one of the resonant frequencies!
• In terms of Normal Coordinates:
L = T - V = (½)[ζkζk - (ωk)2 ζkζk]
 Eqtns of motion: ζk + (ωk)2ζk = 0 (k = 1,2,3, … n)
With solutions: ζk = Ckexp(-iωkt)
• Each Normal Coordinate corresponds to a vibration of the entire system with
only one frequency.
• It’s common to speak of Normal Modes of Vibration.
In one of these normal modes, each particle in the entire system is vibrating with
the same frequency.
Further, they are either vibrating in phase (if the corresponding
eigenvectors have the same sign) or π out of phase (if the eigenvectors
have opposite signs).
Relative amplitudes of particle displacements are given by appropriate ajk.
NORMAL MODES:
Collective properties of the system, not of any one particle!
 Sect. 6.4: Free Vibrations of a
Linear Triatomic Molecule
• Detailed example of this formalism: Normal modes of
vibration for a linear, triatomic molecule.
The figure shows the equilibrium configuration of a molecule:

Symmetric! Atomic masses: m1 = m3 = m, m2 = M.

Generalized coords x1, x2, x3. Equilib distances: x02 - x01 = b =
x03 - x02. A very simple model of the interatomic potential between
atoms: Assume ideal springs of spring constant k. Consider
only longitudinal vibrations (displacements in the x direction).
The PE is: V = (½)k(x2 - x1 -b)2 + (½)k(x3 - x2 -b)2
V = (½)k(x2 - x1 -b)2 + (½)k(x3 - x2 -b)2
• Displacement coordinates:
ηj  xj - x0j Note again: x02 - x01 = b = x03 - x02
 V = (½)k(η2 - η1)2 + (½)k(η3 - η2)2
Or: V = (½)k[(η1)2 + 2(η2)2 +(η3)2 - 2η1η2 - 2η2η3]
k -k 0
Tensor or matrix V = -k 2k -k
0 -k k
• The KE: T = (½) m[(x1)2 + (x3)2] + (½)M(x2)2
Or: T = (½)m[(η1)2 + (η3)2] + (½)M(η2)2
m 0 0
Tensor or matrix T = 0 M 0 Diagonal!
0 0 m
• We needn’t derive the eqtns of motion from the Lagrangian. We
could instead go directly to the secular matrix eqtn:
(Vijaj - ω2Tijaj ) = 0 (i = 1,2,3) (1)
• Its instructive, however, to briefly derive the differential equations
of motion from the Lagrangian: L = T - V =
(½)m[(η1)2 + (η3)2] + (½)M(η2)2- (½)k[(η1)2 + 2(η2)2 + (η3)2 - 2η1η2 - 2η2η3]
Lagrange’s Eqtns: (d/dt)[(L/ηi)] - (L/ηi) = 0

 mη1 + k(η1 - η2) = 0

Mη2 + k(2η2 - η1 - η3) = 0
mη3 + k(η3 - η2) = 0
• However, the power of our matrix (tensor) formalism is that we
needn’t derive these AND we needn’t try to solve them directly. We have
the general solution in terms of the matrices & the secular or eigenvalue
equation: So, go directly to the secular matrix equation:
(Vijaj - ω2Tijaj ) = 0 (i = 1,2,3) (1)
• Using k -k 0 m 0 0
V= -k 2k -k T= 0 M 0
0 -k k 0 0 m
(1)  |V - ω2T| = 0 or:
k - ω2m -k 0
-k 2k - ω2M -k =0
0 -k k - ω2 m
• Evaluating the determinant (algebra):
 ω2(k - ω2m)[k(M + 2m) - ω2Mm] = 0
• A cubic equation for the resonant or natural frequencies (3
solutions of course!):
ω1 = 0, ω2 = (k/m)½, ω3 = [(k/m)(1 + 2m/M)]½
• PHYSICS: ω1 = 0. At first, this may seem strange. No oscillation!
The corresponding equation of motion for the normal coordinate is: ζ1 +
(ω1)2ζ1 = 0 = ζ1  The normal coordinate ζ1 is actually a uniform
translation of the molecule along the x axis! We can translate the
molecule in this way without changing the PE. That is, this is a neutral
equilibrium! We’ve assumed the molecule has 3 vibrational degrees of freedom. In
reality, it has one translational degree of freedom & 2 vibrational degrees of freedom.
One translational normal mode & 2 vibrational normal modes.
• Some more points about ω1 = 0 & ζ1 uniform translation being one of
the solutions.

1. Corresponds to the case where all displacements ηi are equal:

η1 = η2 = η3. Corresponds to the case where V = 0.

2. Since ω1 = 0 does not affect the vibrational properties, we could have done
the problem from the beginning to exclude this root. This could be done
by imposing the constraint that the CM of the molecule be stationary at
the origin. Or:
m(x1 + x3) + Mx2 = 0 Constraint eqtn
Could use constraint eqtn to eliminate one of the coords xi (or one of the
displacements ηi) from T & V
 Reducing the system to 2 degrees of freedom.
• One more point about ω1 = 0 & ζ1 uniform translation being one of
the solutions.
3. Motion restriction along molecular axis results in 1 type of uniform
translational motion. For real molecules in 3d, vibrations in all 3
directions & must be considered. Will find 6 vanishing frequencies
corresponding to 3 translational degrees of freedom & 3
rotational degrees of freedom :
• A general point about zero frequencies: ω = 0.
Could also happen if both (V/qi)0 = 0 and Vij =
(2V/qiqj)0 = 0 at equilibrium. In this case, it may not be
possible to make the small oscillation approximation.
Could do so if the 4th derivatives vanish, but then motion
will not be simple harmonic!
• Back to resonant frequencies for the linear triatomic
molecule. Solutions to:
 ω2(k - ω2m)[k(M + 2m) - ω2Mm] = 0
• Solutions:
ω1 = 0, ω2 = (k/m)½, ω3 = [(k/m)(1 + 2m/M)]½ Consider
ω2: From elementary physics, this is the frequency of
oscillation for a single mass m suspended from a single spring
of constant k.
 Based on this, we speculate that only end the atoms, m, are involved in
the oscillations associated with this mode.
Expect the normal mode displacement ζ2 for eigenvalue ω2 has end atoms
moving & center atom M stationary. So, look at the eigenvectors for 3
modes.
 ω1 = 0, ω2 = (k/m)½, ω3 = [(k/m)(1 + 2m/M)]½
• To get the normal mode displacements (eigenvectors) put
these solutions back into secular equation. In this case, we
get 3 eqtns for each of the 3 eigenvalues ωj:

[k - (ωj)2m]a1j - ka2j +0 =0
- ka1j + [2k - (ωj)2M]a2j - ka3j =0
0 - ka2j + [k - (ωj)2m]a3j = 0

In addition, have normalization:

 m[(a1j)2 + (a3j)2] + M(a1j)2 = 1
• Algebra gives the eigenvector components:
For ω1 = 0: (The mode corresponding to uniform translation). a11 =
a21 = a31 = (2m +M)-½. All 3 equal is what we expect for uniform
translational motion. As in the figure:

For ω2 = (k/m)½ : (The mode expected to correspond to the end

atoms m vibrating & the center atom M not moving). Find
a12 = (2m)-½ = - a32. a22 = 0. Consistent with the expectation!
a22 = 0  the center atom is at rest. a12 = - a32  The 2 end atoms
vibrate exactly π out of phase. As in figure:
• Algebra gives the eigenvector components:

For ω3 = [(k/m)(1 + 2m/M)]½ :

a13 = a33 = [2m(1 + 2m/M)]-½ a23 = -2[2M(2 + M/m)]-½

Interpretation: The outer 2 atoms vibrate with the same

amplitude & phase. The inner atom vibrates out of phase with
them & with a different amplitude. As in the figure:
• Given the eigenvector components, we can compute the normal
mode displacements (algebra). Use ζ = A-1η = Ãη. Get:
For ω1 = 0: (Uniform translation).
ζ1 = (2m +M)-½[(m)½η1 + (M)½η2 + (m)½η3].
Clearly corresponds to:

For ω2 = (k/m)½ : (The end atoms vibrating, the center atom

not moving). ζ2 = (2)-½[η1 - η3]. Clearly corresponds to:
• Use ζ = A-1η = Ãη. Get:
For ω3 = [(k/m)(1 + 2m/M)]½ :
ζ3 = (2m +M)-½[(M/2)½(η1 + η3) - (2m)½η2].
Clearly corresponding to:

• Any general longitudinal vibration (excluding uniform translation) of the

molecule involves some linear combination of the normal mode ζ2 at ω2 =
(k/m)½ :
ζ2 = (2)-½[η1 - η3]exp(-iω2t) & the normal mode
ζ3 at ω3 = [(k/m)(1 + 2m/M)]½ :
ζ3 = (2m +M)-½[(M/2)½(η1 + η3) - (2m)½η2]exp(-iω3t)
• For example, D = C2 ζ2 + C3 ζ3 , C2, C3 complex. Relative sizes & phases
determined by initial conditions.
• This simple treatment: Considered only longitudinal (x directed)
displacements. A real molecule: also has normal vibrational modes
 molecular axis. In general, 9 degrees of freedom  9 normal
modes. Complicated, but doable algebra. The basic physics is
similar to what we just went through. Instead of detailed math, we
now do a qualitative physics discussion of expected results.
• In general, for a molecule with n atoms there are 3n degrees of
freedom. The vibrational frequency eigenvalue problem will have
3n frequencies as solutions. In general, there will be 6 frequencies
for which ω = 0. 3 of these correspond to the 3 translational degrees
of freedom, in which the normal modes are 3 uniform translations
along 3 mutually  axes. Three (usually) correspond to the 3
rotational degrees of freedom in which the normal modes are 3 rigid
rotations about 3 mutually  axes.
• For a linear molecule, things are a bit more complicated:
• There are 3 translational degrees of freedom, corresponding to
uniform translations along 3 mutually  axes. However, there are
only 2 rotational degrees of freedom, in which the normal modes
are 2 rigid rotations about the 2 mutually  axes  molecule itself.
Rotation with the rotation axis being the molecular (x) axis itself
makes no sense, since the molecule (in this approximation!) has no
extent in directions  x. (This is an artifact of our simple model!). So,
there will only be 2 normal modes corresponding to rigid rotations.
 # vibrational degrees of freedom = 9 - 3 (trans) - 2 (rot) = 4
 Expect 4 true vibrational degrees of freedom.
 4 vibrational frequencies & 4 (non-zero)
solutions to the eigenvalue problem.
• The linear triatomic molecule: 4 true vibrational degrees of
freedom.  4 vibrational frequencies & 4 (non-zero) solutions to
the eigenvalue problem.
• 2 of these 4 modes correspond to longitudinal vibrations. These are
the ones we’ve just computed. In fact, if coupling between
vibrations along the molecular (x) axis & vibrations  that direction
is neglected, the solutions for the frequencies and for the
eigenvectors are EXACTLY the ones we’ve already computed!
• 2 of these 4 modes correspond to transverse vibrations (vibrations 
x direction). Molecule is axially symmetric:
 No reason to expect a preference for one  direction (say y) over
another (say z).  The 2 frequencies corresponding to these 2
transverse vibrations must be equal. The 2 transverse modes are
degenerate.
• The 2 transverse modes are degenerate. They have the same
frequency. So, to determine the eigenvectors, we cannot directly
follow the procedure we used up to now, as degeneracy was excluded.
In computing the eigenvectors, any 2 mutually orthogonal axes  x
axis can be chosen. The motion of the molecule  x axis depends on
the amplitudes & relative phases of these 2 degenerate modes.
• If both modes are excited, & if they vibrate exactly in
phase, the atoms will move in a straight line passing
through the equilibrium configuration. But, if they are out
of phase, the total transverse motion is an elliptical
Lissajous figure (like the 2d isotropic SHO). In this case, the
motion in these modes is actually a rotation rather than a
vibration!
• From molecular symmetry: In all modes, the amplitudes of the displacements
of the end atoms must be equal. Detailed calculation of the degenerate, transverse
vibrational (rotational) modes shows that, in this mode, the end atoms travel in the
same direction along a Lissajous figure. In order to conserve angular momentum,
the center atom must move (revolve) in the opposite direction. See figure (which is
for the 2 degenerate modes which are 90° out of phase).

• See the book’s discussion of use of group theory & modern computational techniques to
find the vibrational normal modes of real, multi-atom molecules!