Oscillations

M. Siddikov

December 4, 2019
Announcement - please don’t forget

Homework 1
Homework 1 is available in Aula Virtual
• Part 1A:
# Quiz (multiple options), relatively simple questions. Do
not require a lot of evaluations.
# You will have 1 hour from the start of the quiz to finish.
• Part 1B: 3 problems to analyze and solve.
# You may return your solutions in paper form
(during class) or by email, my mailbox is
[email protected].
#
•If You use Mathematica (or similar programs), please send by email
commented .nb-file
• Deadline for both parts: Friday, October 11 (Today 23:59).
Outline
Today’s topics:
Short recap
Oscillations: linear, forced, non-linear
Motion of pointlike mass in external field
The first law of Newton Case of motion in potential field U(x):
If the net force (the vector sum of all forces
Minimal action principle
acting on an object) is zero, then the velocity
of the object is constant. For real trajectories the action S
Z t2
The second law of Newton S[q] = L (qi , q̇k , t) dt = min
t1

d~p d~v ~ (~r , ~v , t)

=m = m ~a = F (1)
dt dt
L=T −U = (Lagrangian)
2
The third law of Newton m q̇ d f (q, t)
= − U(q)+
2 dt
All forces between two objects exist in equal
magnitude and opposite direction: if one Extremum of the action:
object A exerts a force FA on a second object δS ∂L d

∂L

B, then B simultaneously exerts a force FB 0= = − , (2)
~ A = −F
~B . δqa ∂qa dt ∂ q̇a
on A, and F
in Cartesian system coincides with
The three laws of Newton describe all the dy- (1) since pa = ∂L/∂ q̇a
namics of the mechanical system, given initial - is covariant w.r.t. change of coor-
condition, can predict evolution (Laplace’s de- dinates qa → qa (r1 , r2 , r3 )
terminism)
Oscillations-Introduction
Oscillations are quite common for many We will start with simplest 1-dimensional
systems near equilibrium (minimum of po- case
tential energy): (we can integrate equations exactly using
energy conservation)
Later we’ll generalize to multidimensional
system, when general solution is not known
or is too complicated for analysis

Systems and degrees of freedom might

differ considerably, yet for small oscillations
they share certain common characteristics.
 p
Oscillations-case of one particle x(t) = A cos (ωt + φ) , ω= k/m. (1)
The simplest system which might have A is amplitude, φ is the phase (both should
oscillations: be fixed from initial conditions).
The period T = 2π ω
does not depend on
amplitude A (energy).
This is valid only for harmonic oscillator

~ = −k ~r
m ~a = F
m ẍ = −k x = F
For students who are familiar
with complex numbers:
-elastic force F = −k x is proportional to
Note that we can rewrite (1) as
deformation and opposite in direction.
Description in Lagrangian formalism:
 
x(t) = Re A e i ωt , A = A e i φ
m ẋ 2 m ẋ 2 k
L= − U (x) = − x2
2 2 2
 
ẋ(t) = Re iωA e i ωt ,
k ≡ U 00 (0)  
ẍ(t) = Re (iω)2 A e i ωt ,
In this notations derivative d /dt → iω
The real part symbol Re is frequently
omitted (but is always implied, physical co-
ordinate is real)
Oscillations-case of one particle
In 1D case, any nonsingular potential may The potential energy might have more
be approximated near its equilibrium than one minimum, and we can approxi-
mate it near each of the minimums:
U(r ) ≈ U (r0 ) +U 0 (r0 ) (r − r0 )
U 00 (r0 ) E=const
+ (r − r0 )2 +O (r − r0 )3
2
E=const
U 0 (r0 ) = 0 in the equilibrium. Use nota-

U(r)
k2 (r - r02 )2
tion x = r − r0 . U(r)≈
k1 (r - r01 )2 2
U(r)≈
2
Uosc (r)~const(r-r0 )2
r
U(r)

We expect that for small oscillations near

each of the minimums should have similar
dynamics and solutions, e.g.
r x1 ≡ r − r01 = A cos (ω1 t + φ) ,
We expect that for small oscillations
ω = k1 /m, k1 ≡ U 00 (r01 )
p
near the equilibrium all such systems should
have similar dynamics. Oscillations exist only near minima:
Small oscillations approximation: can ap- U 0 (r0 ) = 0, U 00 (r0 ) > 0.
ply if O (r − r0 )3 terms are negligible
Oscillations-case of one particle
The simplest system which might have
oscillations: µẍ = −k x = F
-elastic force F = −k x is proportional to
deformation and opposite in direction.
p
x(t) = A cos (ωt + φ) , ω = k/m.
~ = −k ~r
m ~a = F A is amplitude, φ is the phase (should be
fixed from initial conditions)
m ẍ = −k x = F
Find the dependence on time t of the
-elastic force F = −k x is proportional to kinetic and potential energy, and
deformation and opposite in direction. demonstrate that their sum does not
p depend on time
x(t) = A cos (ωt + φ) , ω = k/m.

A is amplitude, φ is the phase (both should

be fixed from initial conditions).
Description in Lagrangian formalism:

m ẋ 2 k
L= − x2
2 2
k ≡ U 00 (0)
Oscillations-case of one particle
The simplest system which might have
oscillations: p
x(t) = A cos (ωt + φ) , ω= k/m.

A is amplitude, φis the phase (should be

fixed from initial conditions)

~ = −k ~r
m ~a = F Find the dependence on time t of the
kinetic and potential energy and
demonstrate that the sum does not
m ẍ = −k x = F depend on time
-elastic force F = −k x is proportional to
deformation and opposite in direction.
A2
ω }2 sin (ωt + φ)2
p
x(t) = A cos (ωt + φ) , ω = k/m. T = |m{z
2
≡k
A is amplitude, φ is the phase (both should
k A2
be fixed from initial conditions). U= cos (ωt + φ)2
Description in Lagrangian formalism: 2
Energy of the system:
m ẋ 2 k
L= − x2
2 2 k A2
E =T +U = = const
2
k ≡ U 00 (0)
Oscillations-notion of phase space
p2 x2
x(t) = A cos (ωt + φ), ω=
p
k/m. + =1
2mE 2E /k
√
p(t) = −m A ω sin (ωt + φ) , -ellipse with
p semiaxes pmax = 2mE and
xmax = 2E /k

A2
T =m ω }2 sin (ωt + φ)2
| {z 2
≡k

k A2
U= cos (ωt + φ)2
2
Energy of the system:
The space with axes (x, p) is called phase
k A2 space
E =T +U = = const
2 The trajectory of particle moving in os-
cillatory potential in phase space is ellipse
Phase space trajectories in any poten-
tial normally should not cross itself due to
Laplace’s determinism
Oscillations-notion of phase space
.
p E=const
x(t) = A cos (ωt + φ), ω= k/m.

p(t) = −m A ω sin (ωt + φ) , E=const

U(r)
k2 (r - r02 )2
p2 x2 U(r)≈
+ =1 k1 (r - r01 )2 2
2mE 2E /k U(r)≈
√ 2
-ellipse,
p pmax = 2mE and xmax =
2E /k r
p
p(r ) = ± 2m(E − U(r ))

p(r)
r

If potential has different minima, trajec-

tories are ellipses near each of he minimum
The space with axes (x, p) is called phase Red trajectory which crosses itself at one
space point is called separatrix, separates differ-
Phase space trajectories in any potential ent regimes (oscillations near a minimum
normally should not cross due to Laplace’s vs motion in the whole space)
determinism We’ll see later the properties of the phase
space in more details
Oscillations-one particle in 2D
Lagrangian: We can diagonalize kij , choose some ref-
erence frame where
ẋi2
P
m X kij
L= − xi xj 
k1

2 2 kij = ;
ij
k2
Equations of motion (EOM):
in this system
X p
µẍi = − kij xj = Fi xi (t) = Ai cos (ωi t + δi ) , ωi = ki /m.
j

-force is “linearly proportional”, but might Ai is amplitude, δi is the phase (should be

be not collinear to ~x fixed from initial conditions)

Particle trajectory: case ω1 = ω2

Ellipse with semiaxes Ax , Ay ;
orientation depends on δxy = δx − δy
y(t)

δxy =0

δxy =π/2

x(t)

Different colors=different phase shift δxy

Oscillations-one particle in 2D
We can diagonalize kij , choose some ref-
erence frame where
  Particle trajectory: case ωx = 2 ωy
k1 Ellipse with semiaxes Ax , Ay ;
kij = ;
k2 orientation depends on δxy = δx − δy
in this system
p
xi (t) = Ai cos (ωi t + δi ) , ωi = ki /m.

y(t)
δxy =π/2

Ai is amplitude, δi is the phase (should be δxy =0

fixed from initial conditions)
x(t)

Particle trajectory: case ω1 = ω2 Different colors=different phase shift δxy

Full cycle ymin → ymax → ymin ⇒ parti-
Ellipse with semiaxes Ax , Ay ;
cle passes twice xmax → xmin → xmax
orientation depends on δxy = δx − δy
⇒ ωx = 2 ωy
y(t)

δxy =0

δxy =π/2

x(t)

Different colors=different phase shift δxy

Lissajous curves in 2D &p3D
xi (t) = Ai cos (ωi t + δi ) , ωi = ki /m. Particle trajectory: case ω1 /ω2 6= m/n
This curves are not closed,
Ai is amplitude, δi is the phase (should be
Motion as a whole is not periodic
fixed from initial conditions)
(though x(t) and y (y ) are periodic with
Particle trajectory: case ω1 /ω2 = m/n their own periods)
Trajectories fill the whole square region
between (xmin , ymin ) and (xmax , ymax )

Particle trajectory in 3D: 3D Lissajous

curves possible if ratios of frequencies are
rational numbers (ω1 /ω2 /ω3 = m/n/k)

To plot such trajectories,

You may use in Mathematica
ParametricPlot[{x[t],y[t]},{t,0,T}]
Oscillations-general case
Lagrangian: Small oscillations: introduce displace-
ment from the minimum
1X
L= aij (q)q̇i q̇j − U(q) (min)
2 xi := qi − qi
ij
  k
in general aij (q) might depend on gener- U(q) ≈ U q (min) +
ij
xi xj + O x 3


alized coordinates q., e.g. if we work in 2
∂ 2 U 

spherical coordinates
kij ≡
m 2  ∂qi ∂qj q(min)
T = ṙ + r 2 θ̇2 + r 2 sin2 θφ̇2
2 Matrix kij is symmetric and positively de-
fined near the minimum
 
1 1X 1X
kij xi xj + O x 3


aij = m  r2 L≈ mij ẋi ẋj −
 2 2
ij ij
r 2 sin2 θ
≡ m diag 1, r 2 , r 2 sin2 θ

 
mij ≈ aij q (min)
n o
(min)
Minimum of the potential: qi
Write out the Euler-Lagrange equations
(in terms of variables xi ) and integrate

∂U 
∀i, =0 them
∂qi q(min)
Oscillations-general case
Lagrangian: 1X 1X
kij xi xj + O x 3


L≈ mij ẋi ẋj −
2 2
1X ij ij
L= aij (q)q̇i q̇j − U(q)
2  
ij mij ≈ aij q (min)
n o
(min) X X
Minimum of the potential: qi mij ẍj = − kij xj
 j ij
∂U  -system of N linear equations with constant
∀i, =0
∂qi q(min) coefficients, look for solutions in the form
Small oscillations: introduce displace- xj = Aj e λt = Aj e i ωt
ment from the minimum X
−ω 2 mij + kij Aj = 0


(min)
xi := qi − qi j
 kij

3
-homogeneous linear system, nontrivial so-
U(q) ≈ U q (min) + xi xj + O x lutions (Aj 6≡ 0) only if
2
∂ 2 U 

kij ≡ Characteristic equation
∂qi ∂qj  (min)
q
det −ω 2 mij + kij = 0


Matrix kij is symmetric and positively de-
fined near the minimum
-determines eigenfrequencies ω 2 of the sys-
tem
Oscillations-general case
(min)
xi := qi − qi
Prove that all roots of characteristic
equation are real and positive: ω 2 ∈ R,
1X 1X ω 2 > 0 and
kij xi xj + O x 3


L≈ mij ẋi ẋj −
2 2
ij ij kmin kmax
0< ≤ ω2 ≤
X X mmax mmin
mij ẍj = − kij xj
where kmin/max and mmin/max are the
j ij
minimal and maximal eigenvalues of
xj = Aj e λt = Aj e i ωt matrices kij -and mij
X
−ω 2 mij + kij Aj = 0


Physically, we expect that energy should
j
conserve, and system should oscillate with
-homogeneous linear system, nontrivial so- constant amplitude. Imaginary ω is unac-
lutions (Aj 6≡ 0) only if: ceptable if there is no dissipation of energy
Characteristic equation

det −ω 2 mij + kij = 0

Oscillations-general case
(min)
xi := qi − qi
Prove that all roots of characteristic
equation are real and positive: ω 2 ∈ R,
1X 1X ω2 > 0
kij xi xj + O x 3


L≈ mij ẋi ẋj −
2 2
ij ij
X
−ω 2 mij + kij A∗i Aj = 0


X X
mij ẍj = − kij xj ı̈j
j ij
kij A∗i Aj
P
λt i ωt 2 ij
xj = Aj e = Aj e ⇒ω = P
X ij mij A∗i Aj
−ω 2 mij + kij Aj = 0


Both matrices kij and mij are real and
j
symmetric

∗
Characteristic equation ⇒ ω2 = ω2

det −ω 2 mij + kij = 0

Matrix mij is always positively defined (ki-
netic energy)
Matrix kij is positively near the minimum
⇒ ω 2 always must be positive
Some comments from linear algebra
 
If the matrix K is hermitian, K † = K , ~ ∗m · (1) − (2) · φ
φ ~ ∗m · K − K † ·φ
~n = φ ~n
where (K † )ij = Kji∗ , then all its | {z }
:=0
eigenvalues are real, and all eigenvectors
are orthogonal and form a complete set of = (κn − κ∗m ) φ
~ ∗m · φn
vectors. (Obviously, if K is real and
⇒For m = n, φ ~ ∗n · φn ≡norm of vector φ 6=
symmetric, it is hermitian). ∗
0⇒ κn = κn
Let φn be one of the eigenvectors of K , ⇒For m 6= n, and κm 6= κn , φ ~ ∗m · φn = 0
(orthonormality proved!).
~ n = κn φ
Kφ ~n (1) If ∃m 6= n, κm = κn , then any linear
combination of φm and φn is also eigen-
or explicitly vector of the same eigenvalue κn , and we
X can use these linear transformations to
Kij (φn )j = κn (φn )i
orthonormalize the vectors.
j

where κn is eigenvalue. For conjugate we

have
(See e.g. Arfken, Weber, Chapter 3.5.
Kij∗ (φ∗m )j = κ∗m (φ∗m )i (2) Note that AW use bra- and ket- notations
|{z} ~ and hφ| ≡ φ
|φi for vector φ ~∗ )
(K † )ji
Some comments from linear algebra
~ as
Decompose any vector A
 
If the matrix K is hermitian, K † = K ,
X
~ =
A cn (φn )j
j
where (K † )ij = Kji∗ , then all its n
eigenvalues are real, and all eigenvectors
where cn are some coefficients. In view of
are orthogonal and form a complete set of
orthonormality of vectors φn ,
vectors. (Obviously, if K is real and
symmetric, it is hermitian). ~∗ · K · A
~
P 2
A n |cn | κn
= P 2
~∗ · A
A ~ n |cn |
If φn is a full orthonormal set of
Since ∀n, |cn |2 ≥ 0
eigenvectors, and κmin /κmax are minimal
and maximal eigenvalues, then for any |cn |2 κmin ≤ |cn |2 κn ≤ |cn |2 κmax (2)
vector A~
~∗ · K · A
~
P ∗ so after summation of (2) over n, we
A ij Ai Kij Aj
κmin ≤ ≡ P ∗ ≤ κmax prove (1).
~∗ · A
A ~ i Ai Ai
(1) X
⇒ κmin |A|2 ≤ kij A∗i Aj ≤ κmax |A|2
ij

X
⇒ mmin |A|2 ≤ mij A∗i Aj ≤ mmax |A|2
ij
Oscillations-general case All roots of characteristic equation are real
(min) and positive: ω 2 ∈ R, ω 2 > 0 and are
xi := qi − qi bound by
kmin kmax
1X 1X 0< ≤ ω2 ≤ (1)
kij xi xj + O x 3 mmax mmin


L≈ mij ẋi ẋj −
2 2
ij ij
where kmin/max and mmin/max are the
X X minimal and maximal eigenvalues of
mij ẍj = − kij xj
matrices kij -and mij
j j

xj = Aj e λt = Aj e i ωt X
−ω 2 mij + kij A∗i Aj = 0


X
−ω 2 mij + kij Aj = 0


ı̈j
j
kij A∗i Aj
P
ij
⇒ ω2 = P
Characteristic equation ij mij A∗i Aj
Both matrices kij and mij are real and
det −ω 2 mij + kij = 0


symmetric

∗
⇒ ω2 = ω2

The condition (1) is a mere consequence of

the lemmas we have seen
Oscillations-general case
General solution is the sum of oscillations
(min)
xi := qi − qi with different eigenfrequencies ωn
X (n) i ω t
xj = Aj e n
1X 1X n
kij xi xj + O x 3


L≈ mij ẋi ẋj −
2 2 (n)
ij ij
Each individual solution Aj is called
X X “normal mode”, shows how the system
mij ẍj = − kij xj
oscillates “as a whole”. Should use (1) to
j j
fix them (up to normalization) for each
xj = Aj e λt = Aj e i ωt given eigenfrequency ωn :
X . ? Take ωn and fix some component
−ω 2 mij + kij Aj = 0


(1)
(e..g. A1 = const 6= 0)
j
? Solve (1) for variables A2 , ..., An
Characteristic equation (we’ll see an example later)

det −ω 2 mij + kij = 0

Show that if mij ∼ m δij , then
(n)
eigenvectors
P (m) (n) A are orthonormal, i.e.
j Aj Aj ∼ δmn .
Oscillations-general case
General solution is the sum of oscillations
(min)
xi := qi − qi with different frequencies ωn
X (n) i ω t
xj = Aj e n
1X 1X n
kij xi xj + O x 3


L≈ mij ẋi ẋj −
2 2
ij ij

X X Find the eigenfrequencies for the system

mij ẍj = − kij xj shown in the plot. Assume that masses m
j ij
may move only in horizontal direction
xj = Aj e λt = Aj e i ωt
X
−ω 2 mij + kij Aj = 0


j

Characteristic equation

det −ω 2 mij + kij = 0

Oscillations-general case
Assume that q1,2 are coordinates of the
1X 1X particles.
L≈ mij ẋi ẋj − kij xi xj
2 2 m 2
k1
ij ij L= q̇1 + q̇22 − (q1 − `1 )2
2 2
(min)
xi := qi − qi k3 k2
− (q2 − `1 − `2 )2 − (q1 − q2 − `2 )2
2 2
Find the eigenfrequencies for the system Introduce deviations from equilibrium:
shown in the plot. Assume that masses m x1 = q1 − `1 , x2 = q2 − `1 − `2
may move only in horizontal direction and
. m 2
k1 k3
springs are massless ⇒L= ẋ1 + ẋ22 − x12 − x22
2 2 2
k2
− (x1 − x2 )2 =
2
1X 1X
= mij ẋi ẋj − kij xi xj
2 2
ij ij

Assume that the length of each spring is

`1,2,3 , and there is no tension in
equilibrium
Oscillations-Technical remark
If the coordinates of the extremes of the
spring are (x1 , y1 ) and (x2 , y2 ), then the
m 2 k1 k3 elongation is
⇒L= 2

ẋ1 + ẋ2 − x12 − x22
2 2 2
k2
h
− (x1 − x2 )2 δ` = (` cos α + x2 − x1 )2
2 i1/2
The result which we got does not depend + (` sin α + y2 − y1 )2 −`
explicitly depend on `i , however we should ≈ cos α (x2 − x1 ) + sin α (y2 − y1 ) =
be very careful with this, in some cases `i ~
` · δ~r
appears even when there is no tension in =
`
equilibrium:
where ~ ` = (` cos α, ` sin α) is the vector in
the initial direction of the spring with absolute
value=length of the spring;
So the potential energy of the spring con-
necting points 1 − 2 is

k δ`2 k (δ~r )2
U= 6=
2 2
as could be naively assumed !!!
Only in 1 dimension δ` = |δ~r |
Oscillations-general case
Assume that q1,2 are coordinates of the
1X 1X particles.
L≈ mij ẋi ẋj − kij xi xj
2 2 m 2
k1
ij ij L= q̇1 + q̇22 − (q1 − `1 )2
2 2
(min)
xi := qi − qi k3 k2
− (q2 − `1 − `2 )2 − (q1 − q2 − `2 )2
2 2
Find the eigenfrequencies for the system Introduce deviations from equilibrium
shown in the plot. Assume that masses m (∂U/∂qi = 0):
may move only in horizontal direction and x1 = q1 − `1 , x2 = q2 − `1 − `2
springs are massless
. m 2
k1 k3
⇒L= ẋ1 + ẋ22 − x12 − x22
2 2 2
k2
− (x1 − x2 )2 =
2
1X 1X
= mij ẋi ẋj − kij xi xj
2 2
Assume that the length of each spring is ij ij

`1,2,3 , and there is no tension in

equilibrium
Write out explicitly the matrices mij , kij
and solve the characteristic equation.
Oscillations-general case
1X 1X 
k1 + k2 −k2

L≈ mij ẋi ẋj − kij xi xj mij = mδij , k =
2 2 −k2 k3 + k2
ij ij

X
−ω 2 mij + kij Aj = 0


(1) q
2
j Ω21 + 2Ω22 + Ω23 ± Ω21 − Ω23 + 4 Ω42
2
ω1, 2 =
det −ω 2 mij + kij = 0 2


r
ki
Ωi :=
m
Find the eigenfrequencies for the system
shown in the plot. Assume that masses m Fixing vector A:
may move only in horizontal direction and

(1,2) (1,2)
springs are massless 2
−m ω1,2 + k1 + k2 A1 − k2 A2 =0

~ (1,2) = const k2 , −m ω 2 + k1 + k2


⇒A 1,2

~ = 1.
where const should be fixed from |A|

Assume that the length of each spring is

`1,2,3 , and there is no tension in equilibrium
Oscillations-general case
1X 1X q
L≈ mij ẋi ẋj − kij xi xj Ω21 + 2Ω22 + Ω23 ± Ω21 − Ω23

2
+ 4 Ω42
2 2 2
ω1,
ij ij 2 =
2
r
ki
X 2

−ω mij + kij Aj = 0 (1) Ωi :=
j
m

det −ω 2 mij + kij = 0 ~ (1,2) = const k2 , −m ω 2 + k1 + k2

⇒A 1,2

Symmetric case k1 = k3 : Ω1 = Ω3
Find the eigenfrequencies for the system
2 k1 + 2 k2 k1 k2
shown in the plot. Assume that masses m ω1, = Ω21 + 2Ω22 = = +
may move only in horizontal direction and m m m/2
springs are massless k1
ω22 = Ω21 =
m

~ (1) = √1 (1, −1)

⇒A
2

~ (2) = √1 (1, 1)
⇒A
2
Assume that the length of each spring is -oscillations in the opposite and the same
`1,2,3 , and there is no tension in equilibrium directions!