Oscillations

M. Siddikov

December 4, 2019
Outline
Today’s topics:
Oscillations: linear, forced, non-linear
Free oscillations in N dimensions (short summary)
(min)
xi := qi − qi
1X 1X Theorem
kij xi xj + O x 3


L≈ mij ẋi ẋj −
2 2 All roots of characteristic equation are real
ij X ijX
mij ẍj = − kij xj and positive: ω 2 ∈ R, ω 2 > 0 and
j ij
kmin kmax
X 2
0< ≤ ω2 ≤
−ω mij + kij Aj = 0 mmax mmin
j where kmin/max and mmin/max are the
nontrivial solutions
Characteristic (Aj 6≡ 0) only if:
equation minimal and maximal eigenvalues of
matrices kij -and mij
det −ω 2 mij + kij = 0


If system is closed, then we will get some
Normal coordinates Qn (t) eigenfrequencies ωn = 0, correspond to
. collective degrees of
xj =
X (n)
Aj e i ωn t =
X (n)
Aj Qn (t) freedom:
n n
 Translations
 Rotations
1 Xh 2 i
L= Q̇n − ωn2 Qn2 (t)
2 n Corresponding normal coordinates
(ω=0)
Qn ⇔center-of-mass coordinate,
Q̈n + ωn2 Qn =0 orientation angle(s)
Damped oscillations (short summary)
~ damping = −α~v :
Case F Energy is not conserved, rate of ener-
gy loss depends on velocity
µ ẍ + αẋ + k x = 0.
s
α k
q
γ= , ω1 = − γ 2 = ω02 − γ 2
2µ µ
r
k −γt
p = mẋ(t) = −m A e sin (ω1 t + φ + δ) ,
m
 
γ
δ = arctan
ω1

Underdamping: γ 2 < mk
, ω1 ∈ R.
−γt
x(t) ∼ A e cos (ω1 t + φ)
Phase space: open curves
1.0 p

0.5 ±ⅇ-ℽ t
x(t)/A

0.0
x
-0.5

-1.0
0 5 10 15 20 25 30
t
Forced oscillations
~ damping = −α~v :
Case F Monochromatic case

µ ẍ + αẋ + k x = F (t). F (t) = F̃ (ω)e i ωt

F (t) is some external force xpart (t) = x̃(ω)e i ωt

Might depend on time
Should not depend on x(t),ẋ(t) ˜ ˜
|x /F |
µ ẍ + αẋ + k x = F (t). ˜ ˜
arg(x /F )

˜ ˜
x /F
Z
xpart (t) = d ω x̃(ω) e i ωt
Z
F (t) = d ω F̃ (ω) e i ωt
0.0 0.5 1.0 1.5 2.0
ω/ω0
F̃ (ω)/µ The amplitude depends on the difference
x̃(ω) = .
(−ω 2 + 2iωγ + ω02 ) ω − ω0 and damping γ
The particular solution does not have free
constants, but the full solution does:
x(t) = xhomog (t) + xpart (t)
xhomog (t) ∼ A e −γt cos (ω1 t + φ)
⇒fix (A, φ) after adding xpart (t) !!!
Forced oscillations (short summary)
Important mathematical tool: ~ damping = −α~v :
Case F
Fourier integral theorem
µ ẍ + αẋ + k x = F (t).
Any sufficiently smooth function f (t)
might be recovered from its Fourier image F (t) is some external force
Z +∞ Might depend on time
f˜(ω) = d τ e −i ωτ f (τ ) Should not depend on x(t),ẋ(t)
−∞

using inverse Fourier transform x(t) = xhomog (t) + xpart (t)

Z +∞ xhomog (t) -damped oscillations.
d ω i ωt ˜
⇒ f (t) = e f (ω)
−∞ 2π
Z t
xpart (t) = d τ G (t − τ ) F (τ )
−∞
Generalizes concept of Fourier series for
non-periodic fuctions in inifnte interval Green function
Closely related to Laplace’s transform
Very usable for diff. equations:
e i ωξ /µ
Z
dω
 n Z +∞ G (ξ) =
d d ω i ωt 2π (−ω + 2iωγ + ω02 )
2
f (t) = e (iω)n f˜(ω)
dt −∞ 2π Relation between Fourier images:
Very useful for diff equation with constant F̃ (ω)
coefficents. x̃(ω) =
(−ω 2 µ + iωα + k)
Green function evaluation*
Fromally, both integrals diverge, however
the difference remains finite. We can inte-
Green function
grate using a substitution:

e i (ω−a)ξ
Z ξ
e i ωξ /µ
Z
dω 0
G (ξ) = =i d ξ 0 e i (ω−a)ξ
2π (−ω 2 + 2iωγ + ω02 ) ω−a

+∞
e i ωξ
Z
dω =
1 e i ωξ (ω − a)
Z
−∞
G (ξ) = − dω   
2πµ (+) (−) Z ξ Z +∞
ω − ω0 ω − ω0 0
= ie iaξ d ξ0 d ωe i (ω−a)ξ =
−∞
q
(±)
Z ξ
ω0 = iγ ± ω02 − γ2 = ie iaξ d ξ 0 2πδ(ξ 0 ) =

e iaξ , ξ > 0

e i ωξ 1 = 2πi = 2πiθ(ξ)e iaξ
   =  × 0, ξ < 0
(+) (−) (+) (−)
ω − ω0 ω − ω0 ω0 − ω0 (+) (−)
θ(ξ) e i ω0 ξ − e i ω0 ξ
  G (ξ) =   =
e i ωξ e i ωξ iµ ω
(+)
−ω
(−)
× 
 −  0 0
(+) (−)
ω − ω0 ω − ω0 θ(ξ) exp (−γξ)
 q 
= q sin ξ ω02 − γ 2
µ ω02 − γ 2
Green function evaluation* (evaluation using Cauchy’s
theorem)

Green function G (ξ > 0) =

e i ωξ /µ
Z
dω 
(+) (−)

G (ξ) = 1  e i ω0 ξ e i ω0 ξ
2π (−ω + 2iωγ + ω02 )
2
 + 
iµ (+) (−) (−) (+)
ω0 − ω0 ω0 − ω0
(+) (−)
1 e i ω0 ξ − e i ω0 ξ
e i ωξ =
Z
1 
iµ ω (+) − ω (−)

G (ξ) = − dω    0 0
2πµ (+)
ω − ω0
(−)
ω − ω0  q 
exp (−γξ)
= q sin ξ ω02 − γ 2 .
µ ω02 − γ 2
q
(±)
ω0 = iγ ± ω02 − γ 2
(±)
Both roots have Im ω0 > 0. G (ξ < 0) = 0
Choice of contour integration: for ξ > 0
since there is no poles inside the contou
close contour of integration in the superior
so
plane, for ξ < 0 in the inferior plane, and
θ(ξ) exp (−γξ)
 q 
use Cauchy’s theorem for evaluation of in- G (ξ) = q sin ξ ω02 − γ 2 .
µ ω02 − γ 2
tegral via residues:
Green function evaluation (summary)

Green function We do not need to make transformation

to Fourier space and backwards, just
e i ωξ /µ calculate a single integral (1) and get the
Z
dω
G (ξ) = =
2π (−ω 2 + 2iωγ + ω02 ) particular solution xpart (t) !!!
 q 
θ(ξ) exp (−γξ)
= p sin ξ ω02 − γ 2
µ ω02 − γ 2 The mass m is attached to the massless
spring with elasticity k and rests in its
q equilibrium. In the moment t = 0 the
(±)
ω0 = iγ ± ω02 − γ 2 constant external force F (t) starts acting
on the system and acts during the interval
Particular solution of time T . Evaluate the oscillations of
the system x(t). Solve the equations
Z t directly and using general formalism and
xpart (t) = d τ G (t − τ )F (τ ), (1) demonstrate that results coincide
−∞

θ(ξ) reflects casuality: the force acting at

the moment τ can only affect the solution
xpart at t > τ .
Forced oscillations
Direct method:

The mass m is attached to the massless  0, t<0
F
spring with elasticity k and rests in its x(t) = k
(1 − cos ωt) , t ∈ (0, T ]
equilibrium. In the moment t = 0 the A sin ωt + B cos ωt, t > T

constant external force
F (t) = F θ (0 ≤ t ≤ T ) starts acting on Use continuity:
the system and acts during the interval of xt<0 (0) = xt>0 (0), ẋt<0 (0) = ẋt>0 (0),
time T . Evaluate the oscillations of the
system x(t). Solve the equations directly xt<T (T ) = xt>T (T ), ẋt<T (T ) = ẋt>T (T ),
and using general formalism and F
A sin(ωT ) + B cos(ωT ) = (1 − cos ωT ) ,
demonstrate that results coincide. Assume k
for the sake of simplicity that damping F
γ = 0. A cos(ωT ) − B sin(ωT ) = sin (ωT ) ,
k
F sin ωT F
⇒A= , B = − (1 − cos ωT )
k k
Case t > T :

F sin ωT F
sin ωt − (1 − cos ωT ) cos ωt =
k k
F
= [cos (ω(t − T )) − cos(ωt)]
k
Forced oscillations
General solution:
Z t
The mass m is attached to the massless
xpart (t) = d τ G (t − τ )F (τ ) =
spring with elasticity k and rests in its −∞
equilibrium. In the moment t = 0 the 
0, t < 0
constant external force
 Rt
= F 0 d τ G (t − τ ), t < T
F (t) = F θ (0 ≤ t ≤ T ) starts acting on  RT
F 0 d τ G (t − τ ), t > T
the system and acts during the interval of
time T . Evaluate the oscillations of the
system x(t). Solve the equations directly θ(ξ)exp (−γξ)
 q 
and using general formalism and G (ξ) = q sin ξ ω02 −γ 2
demonstrate that results coincide. Assume µ ω02 −γ 2
for the sake of simplicity that damping
Z T cos (ω(t − T )) − cos(ωt)
γ = 0.
dτ G (t − τ ) =
0 k

Z t 1 − cos(ωT )
dτ G (t − τ ) =
⇒
0 k

 0, t<0
F
⇒ x(t) = k
(1 − cos ωt) , t ∈ (0, T ]
 F
k
[cos (ω(t − T )) − cos(ωt)] , t

in agreement with direct solution!

Forced oscillations
.

The mass m is attached to the massless

spring with elasticity k and rests in its
equilibrium. In the moment t = 0 the
constant external force
F (t) = F θ (0 ≤ t ≤ T ) starts acting on
the system and acts during the interval of
time T . Evaluate the oscillations of the
system x(t). Solve the equations directly Note that arbitrary force might be ap-
and using general formalism and proximated as a sum of short constant im-
demonstrate that results coincide. Assume pulses,
for the sake of simplicity that damping X
F (t) ≈ F (tn ) θ (tn ≤ t ≤ tn+1 ) ,
γ = 0.
n

so we could use “direct” solution which

R we
found earlier to derive our result G (t −
τ )F (τ )d τ (Symon, 2-11); we have to make
substitution

t → t − tn , T → tn+1 − tn , F → F (tn )
(n)
in each of our particular solution xpart (t).
Forced oscillations

Green function The mass m is attached to the massless

spring with elasticity k and rests in its
e i ωξ /µ
Z
dω
G (ξ) = = equilibrium. In the moment t = 0 the
2π (−ω 2 + 2iωγ + ω02 )
 q  constant external force
θ(ξ) exp (−γξ)
= p sin ξ ω02 − γ 2 F (t) = F0 exp (−αt) starts acting on the
µ ω02 − γ 2
system and acts during the interval of
q time T . Evaluate the oscillations of the
(±)
ω0 = iγ ± ω02 − γ 2 system x(t).

Particular solution
Z t
xpart (t) = d τ G (t − τ )F (τ ), (1)
−∞

θ(ξ) reflects casuality: the force acting at

the moment τ can only affect the solution
xpart at t > τ .
Forced oscillations - N dimensions
General case:
Z X
Assume that in addition to the elastic xpart.i (t) = d τ Gij (t − τ )Fj (τ ),
force there is an external “driving” force j
Fi (t),
Green function
mij ẍj = −kij xj + Fi (t).
Z
dω
−1 i ωt
Integrate the equations of motion in this Gij (t) = −ω 2 mij + kij e
2π
case.
Note that
Formal solution:
X d2
X X 
mij ẍj + kij xj = Fi (t). mij + kij Gjk (t) = δik δ(t)
j
dt 2
j j
Z
d ω i ωt This is general solution. But if N is not
xpart,i (t) = e x̃i (ω)
2π large, it might be easier to rewrite problem
in terms of normal coordinates Qn (t) and
Z
d ω i ωt
Fi (t) = e F̃i (ω) solve N decoupled 1D problems.
2π
X
−1
x̃i (ω) = −ω 2 mij + kij F̃j (ω).
j
Example - forced oscillations in coupled system
We’ll work with normal coordinates to
simplify consideration
X (n) i ω t X (n)
xj = Aj e n = Aj Qn (t)
n n

1 X
L= [mij ẋi ẋj − kij xi xj ] + xi Fi (t) =
2
1 X h i
= Q̇n2 − ωn2 Qn2 (t) +
2 n
X X (n)
+ Qn (t) Aj Fj (t)
n j
X (n)
Q̈n + ωn2 Qn = Aj Fj (t) = fn (t)
j

-independent equations for forced oscilla-

tions in each normal coordinate !!!
  Z τ
i ωt
⇒ Qn = Re A e + Gn (t − τ ) fn (τ )
−∞

θ(ξ)
Gn (ξ) = sin (ωn ξ)
ωn
Example - forced oscillations in coupled system

The system consists of two masses m

connected with elastic springs as shown in
the Figure. The external force acting on
the first mass is F1 (t), and the external
force acting on the second pendulum is
F2 (t) . Evaluate the amplitude of forced
oscillations in the system. For simplicity
assume that there is no damping in the
system, and assume that k1 = k3
(symmertic case).
Example - forced oscillations in coupled system
. We’ll work with normal coordinates to
simplify consideration
X (n) i ω t X (n)
xj = Aj e n = Aj Qn (t) =
n n
   
x1 1 1
= √ Q1
Symmetric case k1 = k3 : Ω1 = Ω3 x2 2 −1
 
1 1
2 k1 + 2 k2 k1 k2 +√ Q2 =
ω1, = Ω21 + 2Ω22 = = + 2 1
m m m/2  
k1 1 Q1 + Q2
ω22 = Ω21 = = √
m 2 −Q1 + Q2

~ (1) = √1 (1, −1)

⇒A 1 X
2 L= [mij ẋi ẋj − kij xi xj ] + xi Fi (t) =
2
~ (2) = √1 (1, 1) 1X 2 h i
⇒A = Q̇n − ωn2 Qn2 (t) +
2 2 n
-oscillations in the opposite and the same
X (1)
+ Q1 (t) Aj Fj (t)
directions j
X (2)
+ Q2 (t) Aj Fj (t)
j
Example - forced oscillations in coupled system
. Symmetric case k1 = k3 : Ω1 = Ω3
k1 + 2 k2 k1 k2 1 X
2
ω1, = Ω21 + 2Ω22 = = + L= [mij ẋi ẋj − kij xi xj ] + xi Fi (t) =
m m m/2 2
1X 2 h i
k1 = Q̇n − ωn2 Qn2 (t) +
ω22 = Ω21 = 2 n
m
X (1) X (2)
+ Q1 (t) Aj Fj (t) + Q2 (t) Aj Fj (t)
~ (1) = √1 (1, −1)
⇒A j j
2
X (1) F1 − F2
~ (2) = √1 (1, 1)
Q̈1 +ω12 Q1 = Aj Fj (t) = √ ≡ f1
⇒A 2
2 j

-oscillations in the opposite and the same X (2) F1 + F2

directions Q̈2 +ω22 Q2 = Aj Fj (t) = √ ≡ f2
We’ll work with normal coordinates to j
2
simplify consideration
-independent equations for forced oscilla-
tions in each normal coordinate Q1 , Q2 !!!
X (n) i ω t X (n)
xj = Aj e n = Aj Qn (t)
n n Z τ
(part)

x1

Q1

1

Q2

1
 ⇒ Qn =+ Gn (t − τ ) fn (τ )
= √ +√ −∞
x2 2 −1 2 1
1

Q1 + Q2
 θ(ξ)
= √ Gn (ξ) = sin (ωn ξ)
−Q1 + Q2 ωn
2
Anharmonic terms

1D case, take into account O(x 3 ) terms:

1 1 λ x3
mẋ 2 − k x 2 − + O x4


L≈
2 2 3
m ẍ + k x = −λx 2
-nonlinear differential equation.

How can we solve this equation ?

Anharmonic terms
1D case, take into account O(x 3 ) terms:

1 1 λ x3 Assume that the parameter λ is small, so

mẋ 2 − k x 2 − + O x4


L≈
2 2 3 that λx 2  k x. Develop a perturbation
theory
m ẍ + k x = −λx 2
-nonlinear differential equation. x(t) = x0 (t) + λ x1 (t) + ...
In 1D case we know exact (formal) solution
for any potential, and write out explicitly the equations for
x0 (t) and x1 (t).
mẋ 2 kx 2 λx 3
 
E = +U(x), U(x) = +
2 2 3
Z
dx
⇒t= q
2
m
(E − U(x))
but only a few potentials allow explicit eval-
uation of dependence x(t). In higher di-
mensions application is more complicated,
so we’ll develop an approximate method
which gives explicit x(t)-dependence and
can be generalized to any number of di-
mensions.
Anharmonic terms
1D case, take into account O(x 3 ) terms:
2
1 1 λ x3 x02 (t) = A20 e 2i ω0 t + (A∗0 ) e −2i ω0 t
mẋ 2 − k x 2 − + O x4


L≈
2 2 3 + 2 A∗0 A0
m ẍ + k x = −λx 2
-nonlinear differential equation. Solve explicitly the equation (1)
λx 2  k x, develop a perturbation theory
assuming that λ is small:

x(t) = x0 (t) + λ x1 (t) + ...

Terms ∼ O λ0 :

r
k
m ẍ0 + k x0 = 0, ω0 =
m
 
x0 = 2Re A0 e i ωt = A0 e i ω0 t + A∗0 e −i ω0 t

Terms ∼ O λ1 :

m ẍ1 + k x1 = −x02 (1)

Anharmonic terms
1D case, take into account O(x 3 ) terms:
2
1 1 λ x3 x02 (t) = A20 e 2i ω0 t + (A∗0 ) e −2i ω0 t
mẋ 2 − k x 2 − + O x4


L≈
2 2 3 + 2 A∗0 A0
2
m ẍ + k x = −λx  
⇒ x1 (t) = Re A1 e ±i ω0 t + x1, part (t)
-nonlinear differential equation.
λx 2  k x, develop a perturbation theory
A20 e 2i ω0 t A∗2 e −2i ω0 t

assuming that λ is small: 1
x1, part (t) = − 2 2
+ 02
m ω0 − 4ω0 ω0 − 4ω02
x(t) = x0 (t) + λ x1 (t) + ... 
2 A∗0 A0 0 i ω0 t 
Terms ∼ O λ0 :

+ e .
ω02 | {z }
=1
r
k Homogeneous solutions:
m ẍ0 + k x0 = 0, ω0 =
m
 
x0 = 2Re A0 e i ωt = A0 e i ω0 t + A∗0 e −i ω0 t x0 + x1, homog + ... =
   
= Re A0 e i ω0 t + λ Re A1 e i ω0 t + ... =
Terms ∼ O λ1 :


h i h
= Re (A0 + λA0 + ...) e i ω0 t = Re A e i ω0 t
m ẍ1 + k x1 = −x02
-in higher orders ignore homogeneous solu-
tions, they are corrections to LO amplitude.
Anharmonic terms
1D case, take into account O(x 3 ) terms: .

1 1 λ x3
mẋ 2 − k x 2 − + O x4


L≈
2 2 3
m ẍ + k x = −λx 2

x(t)
-nonlinear differential equation. x0 (t)
λx 2  k x, develop a perturbation theory x0 (t)+λ x1 (t)

assuming that λ is small:

t

x(t) = x0 (t) + λ x1 (t) + ... Inclusion of small perturbation deforms

the x(t), yet the motion remains periodic
in this order with the same period !
  (You may guess that in higher order n we
x0 = 2Re A0 e i ω0 t = A0 e i ω0 t + A∗0 e −i ω0 t get harmonics up to e ±in ω0 t !!!)

1 A20 e 2i ω0 t A∗2 e −2i ω0 t

x1, part (t) = − 2 2
+ 02
m ω0 − 4ω0 ω0 − 4ω02

2 A∗0 A0 0 i ω0 t
+ e .
ω02 | {z }
=1