You are on page 1of 34

Preprint typeset in JHEP style - HYPER VERSION

Classical Mechanics IV

Ashoke Sen
Harish-Chandra Research Institute Chhatnag Road,
Jhusi, Allahabad 211019, India
padakshep@gmail.com

Abstract: This note is based on a handwritten note by Prof Sen. This note is typed Sampath
Mukherjeeunder Padakshep Open Teaching Project.Padakshep is an NGO which is working to
in education sector in West Bengal, India. Padakshep Open Teaching Project is an initiative to
provide free education for all. More information can be found at www.padakshep.org
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

Contents

1. Test of integrability of a system 12

Symmetric rigid body under the action of


( )
1 dSe dSeT
L = Tr Kd − M gρ(0) S̃33 (0.1)
2 dt dt
3 : Symmetry axis of the body.
Z: The line of action of gravity.
M :Mass.
ρ(0) : Distance of the centre of the mass of body from the fixed point.
Parametrization

S̃ = R3 (ϕ) R1 (θ) R3 (ψ) (0.2)

 
cos θ sin θ 0
 
R3 (θ) ≡  − sin θ cos θ 0  (0.3)
0 0 1
 
1 0 0
 
R1 (θ) =  0 cos θ sin θ  (0.4)
0 − sin cos θ
 
cos θ 0 − sin θ
 
R1 (θ) =  0 1 0  (0.5)
sin θ 0 cos θ
ϕ, ψ Should be cyclic coordinates.
Lagrangian in this coordinate system
   
− sin θ cos θ 0 0 10
dR3 (θ)    
=  cos θ − sin θ 0  = R3 (θ)  −1 0 0  (0.6)

0 0 0 0 00

= R3 (θ) T3 or T3 R3 (θ) (0.7)


In general :

dR3 (θ)
= Ra (θ) T a or T a Ra (θ) (0.8)

1
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

dS̃ d
= (R3 (ϕ) R1 (θ) R3 (ψ)) (0.9)
dt dt

( )
= R3 (ϕ) T3 ϕ̇R1 (θ) R3 (ψ) + R3 (ϕ) R1 (θ) T1 θ̇R3 (ψ) + R3 (ϕ) R1 (θ) T3 ψ̇R3 (ψ) (0.10)

dS̃
Ω = S̃ −1 (0.11)
dt

( )
= R3−1 (ψ) R1−1 (θ) R3−1 (ϕ) R3 (ϕ) T3 ϕ̇R1 (θ) + R3 (ϕ) R1 (θ) T1 θ̇ + R3 (ϕ) R1 (θ) T3 ψ̇ R3 (ψ)
(0.12)
{ }
= R3−1 (ψ) R1−1 (θ) T3 R1 (θ) ϕ̇ + T1 θ̇ + T3 ψ̇ R3 (ψ) (0.13)

R1−1 (θ) = R1 (−θ) Check

R1 (−θ)T3 R1 (θ) = T3 cos θ − T2 sin θ (0.14)

{ ( ) }
Ω̃ = T3 cos θϕ̇ + ψ̇ + T1 θ̇ − T2 sin θϕ̇ (0.15)

( )
1 e dSeT
−1 dS
T = T r S̃ Kd S̃ (0.16)
2 dt dt

1 ( ) 1 ( )
= T r ΩKd ΩT = − T r Ω2 Kd (0.17)
2 2

1 ( )
= − T r R3−1 (ψ) Ω̃R3 (ψ) R3−1 (ψ) Ω̃R3 (ψ) Kd (0.18)
2

R3 (ψ) Kd = Kd R3 (ψ) (0.19)

1 ( )
=⇒ T = − T r R3−1 (ψ) Ω̃2 Kd R3 (ψ) (0.20)
2

1 ( )
= − T r Ω̃2 Kd (0.21)
2
( )2
( ( ) )2 ∑
Ω̃2 = T3 cos θϕ̇ + ψ̇ + T1 θ̇ − T2 sin θϕ̇ ≡ αa T a (0.22)
a
( )2
∑ ∑∑
a
αa T = αa αb T a T b (0.23)
a a b

2
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

1 ∑∑ ( )
T = Tr T r Kα T a T b αa αb (0.24)
2 a b
( )
For a ̸= b , Kα T a T b has no diagonal term.

⇒ T r Kα T a T b = 0 (0.25)
( )
For a = b, T r Kα (T a )2 is non zero.
  
λ1 000
( )2   
T r Kα T ′ = −T r  λ2 0 1 0 (0.26)
λ3 001

= − (λ2 + λ3 ) = −I1 (0.27)

( )2 ( ( )2 )
T r Kα T l = − (λ1 + λ3 ) = −I2 T r Kα T 3 = − (λ1 + λ2 ) = −I3 (0.28)

1( )
⇒T = I1 (α1 )2 + I2 (α2 )2 + I3 (α3 )2 (0.29)
2
[ ( ) ( )2 ]
1 2 2 2
= I1 θ̇ + sin θϕ̇ + I3 ψ̇ + cosθϕ̇ (0.30)
2

V = +M gρ(0) S33 (0.31)

= +M gρ(0) (R3 (ϕ) R1 (θ) R3 (ψ))33 (0.32)

= +M gρ(0) (R3 (ϕ))3i (R1 (θ))ij (R3 (ψ))j3 (0.33)

= +M gρ(0) R1 (θ)33 (0.34)

= +M gρ(0) cosθ (0.35)

1 ( ) 1 ( )2
⇒ L = T − V = I1 θ̇2 + sin2 θϕ̇2 − M gρ(0) cosθ + I3 ψ̇ + cosθϕ̇ (0.36)
2 2
L does not depend on ϕ and ψ =⇒ϕ and ψ are cyclic coordinates as expected.
Conserved quantities:

∂L ( )
pϕ = = I1 sin2 θϕ̇ + I3 cosθ ψ̇ + cosθϕ̇ (0.37)
∂ ϕ̇
∂L ( )
pψ = = I3 ψ̇ + cosθϕ̇ (0.38)
∂ ψ̇

3
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

1 ( ) 1 ( )2
E = T + V = I1 θ̇2 + sin2 θϕ̇2 + M gρ(0) cosθ + I3 ψ̇ + cosθϕ̇ (0.39)
2 2
1 1 2
p2ψ
= I1 θ̇2 + (pϕ − p ψ cosθ) + + M gρ(0) cosθ (0.40)
2 2I1 sin2 θ 2I3

cos θ = u sin θθ̇ = u̇ (0.41)

u̇2 u̇2
θ̇2 = = (0.42)
sin2 θ 1 − u2

1 u̇2 1 2
p2ψ
=⇒ E = I1 + (p ϕ − p ψ u) + + M gρ(0) u (0.43)
2 1 − u2 2I1 (1 − u2 ) 2I3
( )( )
2 1 − u2 p2ψ 1
2
=⇒ u̇ = E− − M gρ u − 2 (pϕ − pψ u)2 = f (u)
(0)
(0.44)
I1 2I3 I1

f (u) −→ ∞ as u −→ ∞ (0.45)

f (u) −→ −∞ as u −→ −∞ (0.46)

f (u) < 0 at u = ±1 (0.47)

⇒ f (u) has one zero f or u > 1 (0.48)


⇒Must have two other zeros between -1 and +1

Physically f (u) > 0 =⇒ u is restricted between u1 and u2


|u̇| > 0 for u1 < u < u2 =⇒ u̇ cannot vanish in this range and hence cannot change sign.

4
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

=⇒u increases monotonically from u1 to u2 and then decreases monotonically from u2 to u3

u1 = cos θ1 u2 = cos θ2 (0.49)


ϕ̇ = pϕ − pψ cos θ =⇒ ϕ˙ vanishes at u = ≡ u′ (0.50)

I. If u′ is outside the range (u1 , u2 ), then ϕ̇ has always the same sign.
=⇒ ϕ increases continuously or decreases continuously.

If u′ is in the range (u1 , u2 )then ϕ̇ changes sign whenever u = u′


If u′ = u or u2 then ϕ̇vanishes at θ = θ2
( )( )
2 1 − u2 p2ψ 1 2
2
u̇ = f (u) = E− − M gρ u −
(0)
2 (pϕ − pψ u) (0.51)
I1 2I3 (I1 )


u′ = ⇒ ϕ̇ = 0 at u = u1 (0.52)

5
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

Is it possible to have u′ = u1 ? (0.53)

pϕ − pψ u 1 = 0 (0.54)

Since f (u1 ) = 0
p2ψ
E− − M gρ(0) u = 0 (0.55)
2I3

p2ψ
=⇒ F or u > u1 E− − M gρ(0) u < 0 (0.56)
2I3

=⇒ f (u) < 0 −→ Contradiction (0.57)

=⇒ u′ cannot be equal to u1
General procedure for eliminating cyclic variables.
Routh
System with N degrees of freedom

q i , pi (1 ≤ i ≤ N ) (0.58)

6
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

M of the coordinates are cyclic

qi : 1 ≤ i ≤ M (0.59)

7
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

∂H
= 0, 1≤i≤M (0.60)
∂q i
=⇒ pi (1 ≤ i ≤ M ) are constants of motion.
Cyclic coordinates and momenta

Qα , Pα : (1 ≤ α ≤ M ) (0.61)
Other coordinates and momenta

qei , pei M +1≤i≤N (0.62)

(−
→ −→ − →)
H = H P , qe , pe (0.63)


P −→M dimensional vector

→ − →
qe , pe −→N-M dimensional vector
Pα are constants of motion.
Non trivial equations of motion

qi
de ∂H de
pi ∂H
= =− i (0.64)
dt ∂ pei dt ∂ qe
Pα :Parameters appearing in H
Routhian

(−
→ −→ → −) ∑
N
R P , qe , u
e = ei pi − H
u (0.65)
i=M +1

∂H
ei ≡
u (0.66)
∂ pei
Analogous to the Lagrangian, except that we use only the non-cyclic coordinates to perform
the Legendre transformation.
Equation of motion

deqi
=uei (0.67)
dt
( )
d ∂R ∂R
= i (0.68)
dt ∂e u i ∂ qe
Derivation: 1 dentical as in the Lagrangian Case.
=⇒A system with N-M degrees of freedom.
Normally we start with the Lagrangian.
( )
L q i , ui 1 ≤ i ≤ N, ui = q̇ i (0.69)
For Cyclic coordinates :

8
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

( )
L =⇒ L qei , u
ei , U α U α = Q̇α (0.70)
Pα = ∂L
∂U α −→Constants of motion


M ∑
N
H= Pα U + α
pei u
ei − L (0.71)
α=1 i=M +1


N
R= pei u
ei − H (0.72)
i=M +1


M
=L− Pα U α (0.73)
α=1
( )
=⇒ R qei , u
ei − Pα (0.74)
Example : 3-d particle in an axially symmetric potential.


3
( )2
m ẋi − V (−

1
L= x) (0.75)
2
i=1

Use cylindrical polar coordinates (ρ, z, ϕ)

x1 = ρ cos ϕ (0.76)

x2 = ρ sin ϕ (0.77)

x3 = z (0.78)

1 ( )
L = m ρ̇2 + ρ2 ϕ̇2 + ż 2 − V (ρ, z) (0.79)
2
ϕ : Cyclic coordinate

∂L
pϕ = = mρ2 ϕ̇ = Conserved (0.80)
∂ ϕ̇

p2ϕ1 ( 2 )
R = −pϕ ϕ̇ + L = − + m ρ̇ + ż 2
− V (ρ, z) (0.81)
2mρ2 2

1 ( )
= m ρ̇2 + ż 2 − Ve (ρ, z) (0.82)
2
=⇒A Thory with two degrees of freedom ρand z
pϕ : Can be treated as a parameter.
Caution: One shouldn’t simply substitute ϕ̇in terms of pϕ in L,but actually performed the
Legendre trnsform to get the Routhian.

9
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

Time dependent problem


General sysytem :

dxi
= F i (−

x , t) 1≤i≤n+1 (0.83)
dt
Introduce an (n + 1)th variable xN +1 :

dxN +1
=1 (0.84)
dt

dxi (→ N +1 )
= Fi −
x,x (0.85)
dt
Autonomous system with one more dgrees of freedom.
Hamiltonian System

dq α ∂H (−

p ,−→q , t)
= α
1≤α≤N (0.86)
dt ∂p

dpα ∂H (→

p ,−→q , t)
= α
1≤α≤N (0.87)
dt ∂q
Introduce new coordinates q N +1 and conjugate momenta pN +1

(− − )
H′ = H →
p ,→
q , q N +1 + pN +1 (0.88)

(→ − ) (− − ) (→ − )
dq N +1 ∂H ′ dq α ∂H −
p ,→
q , q N +1 dpα ∂H →
p ,→
q , q N +1 dpN +1 ∂H −
p ,→q , q N +1
= N +1 = 1 = =− =− (0.89)
dt ∂p dt ∂pα dt ∂pα dt ∂q N +1

Small Oscillation
Consider a general system

L=T −V (0.90)

T −→ 12 α,β fαβ (→
−q ) q̇ α q̇ β


V −→ V ( q )


q (0) an extrimum of V

∂V
=⇒ =0 at q i = q (0)i (0.91)
∂q i
We want to study small oscillation around qei = q i − q (0)i
Expand L to quadratic order in qei
(−→)
1∑ ḟ 1 ∑ ∂2V
=⇒ L = fαβ q (0) qḟ
αq β − |−
→(0) (0.92)
2 2 ∂q α ∂q β q
α,β α,β

10
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

 
m1
 
m = U T mα U U T U = 1, md =  ...  (0.93)
mn

1∑ α T 1∑ α
L= qḟUαβ (md )βγ Uγδ qeδ − qfkαβ qf
β (0.94)
2 2
α,β α,β

( ) ( )β
qbα = Uαβ q̃ β q̃ β = U −1 q̂ = U T q̂ = Uαβ q̂ β (0.95)

1 ∑ dq̂ α dq̂ γ 1∑
L= (mα )βγ − Uαβ q̂ γ Kαβ Uβδ q̂ δ (0.96)
2 dt dt 2
βγ αβ

( )2
1∑ dq̂ β 1∑ γ( T )
= mβ − q̂ U KU γδ q̂ δ (0.97)
2 dt 2
β γ,δ


→ √
q β = mβ q̂ β mβ > 0 since mαβ is positive def inite (0.98)

( )2
1∑ dq̂ β 1∑ γ 1 ( T )
L= − q √ U KU γδ q̂ δ (0.99)
2 dt 2 mγ mδ
β γ,δ
 
K1
 
 .... 
 
K = V T Kd V V TV = 1 Kd =  ....  (0.100)
 
 .... 
KN
( )2
1∑ dq̂ β 1∑ γ T
= − q Vγα (Kd )αβ Vβδ q δ (0.101)
2 dt 2
β γ,δ

( )δ ( )δ
Qβ = Vβδ q δ , q δ = V −1 Q = V T Q = Vβδ Qα (0.102)

1 ∑ dQα dQγ 1∑ α
L= Vαβ Vγβ − Q (Kd ) (0.103)
2 dt dt 2
β,α,γ αβ

1 ∑ dQα dQα 1 ∑
= − Kα (Qα )2 (0.104)
2 dt dt 2
=⇒A set of decoupled harmonic oscillator with time period √2π Kα
If any of the Kα′ s are -ve the system is unstable.
=⇒Maximum
−→ of V instead of a minimum in that direction.
2V
If q is aminimum then ∂q∂α ∂q β |→
(0) −
q (0) is a positive definite matrix.
=⇒All Kα′ s are positive.

11
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

1. Test of integrability of a system

A system with 2 degrees of freedom (after elimination of cyclic variables)


( )
H p1 , p 2 , q 1 , q 2 (1.1)
( )
Is there a second conserved quantity F p1 , p2 , q 1 , q 2
Suppose that there is such a quantity.
What will be the consequences??
Phase Curves: Lie on a 2-d surface in the 4-d phase space.

H (⃗
p, ⃗q) = E (1.2)

F (⃗
p, ⃗q) = C (1.3)
A torous if the motionis bounded.
Consider the intersection of this 2-d surface with a 3-d surface(say q ′ = 0)
The intersection will be a 1 d curve.
If there is no second conserved quantity F then phase curve lies inside a 3-d voume in the
4d phase space.
( )
Intersection of their volume with the 3-d surface q 1 = 0 will be a plane.
( )
H p1 , p2 , q 1 , q 2 = E (1.4)

( )
=⇒ p1 = f p2 , q 1 , q 2 (1.5)

The phase space coordinates of a particle is fixed once p2 , q 1 , q 2 are given and possibly the
sign of p

Study it whenever q ′ = 0, p′ > 0. The intersection of the phase curve with the q 2 , p2 plane
will be scattered set of points.

12
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

( )
On the other hand, Suppose there is a second conserved quantity F p1 , p2 , q 1 , q 2
( )
F p1 , p2 , q 1 , q 2 = C (1.6)

( ( ) ) ( )
=⇒ F f p2 , q 1 , q 2 , p2 , q 1 , q 2 = C −→ G p2 , q 1 , q 2 = C (1.7)

The Phase curves lies on the surface.


The q ′ = 0 plane intersects the surface doing a curve C.
( )
If follow thephase curve and record the value of p2 , q 2 whenever q ′ = 0 and p′ > 0 the
resulting set of points will all lie on the curve C.
Procedure

• Start evolving the system from a given point in phase space.

13
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

• During hte evolution, whenever q ′ = 0and p > 0, record the values of p2 and q 2 −→Poincare

• Plot them on the p2 − q 2 plane.

• If they lie on a curve, then there is a secomd conserved quantity.

• If they scatter over the plane then there is no second conserved quantity.

Do it for different initial; conditions with

• The Same Energy.

• Different Energies.

It may so happen that for some initial conditions the points lie on a curve, whereas for others
the points scatter.
Try ẍ + ẋ + sin x = cos t
Example : Henon - Heiles system :
( )
1( 2 ) 1 ( 1 )2 ( 2 )2 ( 1 )2 2 2 ( 2 )3
H= 2
p + p2 + q + q +2 q q − q (1.8)
2 1 2 3
( )
E = 0.08333 : Most trajectories intersect the p2 , q 2 plane along a curve.
( )
E = 0.12500 : Some trajectories intersect the p2 , q 2 plane along a curve, but the others
scatter over the plane.
( )
E = 0.16667 : Most trajectories intesect the p2 , q 2 plane at scattered points.
Why does this happen?
Perturbation theory (Conventional)
Suppose we have an integrable sysytem:
( )
H0 J⃗ θi 1≤i≤N (1.9)


where J −→ N Action Variables, θ−→N Angle variable.
Consider a perturbed Hamiltonian.
(−
→ →−) (−
→) (→
− −→)
H J , θ = H0 J + ϵV J , θ (1.10)

Periodic in each θi with period 2π


(−
→) ∑ (−
→) ( )
= H0 J + ϵ Vni J exp i ni θi (1.11)
ni

( ) (−
→)

∑ (→
−) i

= H0 J + ϵV0,0,0 J + ϵ Vn1 n2 ,.. J ei(ni θ ) (1.12)
n1 ,n2

Try to construct new action angle variables Jei , θei such that H is a function of Jei only to
order ϵ (→− − )
e →
Generating function F3 J , θ

14
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

(−
→ − ) (−
→ → )
e → e −
∂F3 J , θ ∂F3 J , θ
Ji = , θei = (1.13)
∂θi ∂Ji
 (−
→ − ) 
e →
(−
→)
 ∂F3 J , θ 

H , θ  = H̃ Je
i
(1.14)
∂θi

Try:
(→
− − ) ′
∑ (−
→)
e → e gn1 ,n2 ,.. Je ei(ni θ )
i i
F3 J , θ = Ji θ + ϵ (1.15)
n1 ,n2 ,...

(→ ( )
−) ′
∑ (→
−)
e Je = H Jei + ϵ
=⇒ H (ini ) gn1 ,n2 ,... Je ei(ni θ )
i
(1.16)
n1 ,n2 ,..

( (− )
→) ( ) ( )
′ ′
∑ ∑
Jei0 + ϵ e e Vn1 n2 ,.. Jei ei(ni θ )
i
i(ni θ ) i
= H0 (ini ) gn1 ,n2 ,... J e + ϵV0,0,0 Ji + ϵ
n1 ,n2 ,.. n1 ,n2
(1.17)

(−
→) (−
→) ′ (−
→)
′ (−
→)
e0 e e ∂H0 ∑ e

i(ni θi )
Vn1 n2 ,.. Je ei(ni θ )
i
=H J + ϵV0,0,0 J + ϵ (ini ) gn1 ,n2 ,... J e +ϵ
∂ Jei n1 ,n2 ,.. n1 ,n2
(1.18)
Choose :
(−
→) ∂H0 (−
→)
ini gn1 ,n2 ... Je + Vn1 ,n2 ... Je = 0 (1.19)
∂ Jei
(→
−)
(−
→) Vn1 ,n2 ,.... Je
=⇒ gn1 ,n2 ,.. Je = i (→
−) (1.20)

ni ωi Je i

This way we can generate canonical transformations thatconstructs a new triangle Hamil-
tonian from the old one. (−→)

This procedure breaks down if i ni ωi Je = 0 for some set of {ni }
Could be the reason why the perturbation theory breaks down for some values of E and
good for other values.
The situation is more subtle.

∑ (−
→)
ni ωi Je = 0 (1.21)
i

15
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

whenever ωωji = rational for some pair (i, j) NB: Rational−→dense in the set of real numbers.
=⇒ The set of points wher the perturbation theory is expected to break down are dense on
the E axis.
The situation is even worse.
Consider an arbitrary point on the E axis.

ωi
= Irrational (1.22)
ωj
But ωωji can be made arbitrarily close to some rational number.
=⇒It is possible to find some large (ni , nj )such that ωωji ........ nnji + arbitrary small number
seem to indicate that perturbation theory breaks down for all.
Start with a Hamiltonian:

H (→

p ,−

q ) = H0 (−

p ,−

q ) + ϵV (−

p ,−

q) (1.23)
⃗ θ⃗
Action angle variables for H0 : J,

−)
(→ → (→) (→
− −→)
e −
H e0 −
J, θ =H J + ϵVe J , θ (1.24)
Look for canonical transformations

→ −
− → → →
− −
J , θ −→ Jˆ , θ̂ (1.25)

(→
− − ) (→
− − )
ˆ → ˆ →
∂F3 J , θ ∂F3 J , θ
Ji = , θ̂i = , (1.26)
∂θi ∂ Jˆi
(− (−
→)
e −)
→ → ( )
s.t. H J , θ = Ĥ Jˆ + 0 ϵ2 (1.27)

Answer

( → ) ∑ ′
∑ (−
→) −
⃗ˆ − ˆ →− →
F3 J, θ = i
Ji θ + ϵ g−

n Jˆ ei n . θ (1.28)
i −

n
(−→)
(→
−) V−

n Je
=⇒ g−

n Jˆ = i ∑ (1.29)
i ni ω̂i
(−
→) (−
→) (−→)
H J = H0 J + ϵV0,0,.. Jb
b ˆ f ˆ (1.30)

Problems :
∑ (−
→)
ni ω̂i = 0 =⇒ g−

n Jˆ is divergent (1.31)
i
ω̂i
If ω̂j is rational, then we can find

16
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

ni , nj such that ni ω̂i + nj ω̂j = 0


=⇒ There is problem whenever ω̂ω̂ji is a rational number for any pair (i, j)
Set of rational numbers are dense in the −set of real numbers.

=⇒Set of singular points is dense in the Jˆ space.


Given any point in the Jˆ space, we can find a point arbitrarily close to it where the
perturbation expansion diverges.


Consider an arbitrary point in the Jˆ space where the ratios of the ω̂i are all irrational
numbers.

ω̂i
= Irrational (1.32)
ω̂j

In considerthose terms for which n3 = n4 = ....... = 0

V→
−n V−
→ 1
g−

n =i = i n . ω1 n1 (1.33)
n1 ω1 + n2 ω2 n2 ω2 ω2 + n2
ω1
ω2 is a real number
For a fixed value of ωω12 it is possible to choose nn12 in such a way that | ωω21 + nn12 | < ϵfor any ϵ
=⇒The denominator can be made arbitraliy small.
=⇒Even for non rational values of ωω12 we cannot avoid small denominators.
Way out : For given ωω21 it is possible to choose a rational number nn21 arbitrarily close to it
but this will typically require large values of n1 and n2
If we assume that V⃗n decreases sufficient rapidly for large |ni | , then the sum over − →n may
still be a finite answer for
′ ′
∑ −
→ ∑ V⃗n →− →
i−

n.θ ∑

gn e =i ei n . θ (1.34)

→ −
→ ni ωi
n n

N = 2 Case (1.35)

( )
( → ) ⃗
∑ ∑ V Jˆ
⃗ˆ − n ,n
1 2
g J, θ ≡ gn1 ,n2 ei(n1 θ1 +n2 θ2 ) = i ei(n1 θ1 +n2 θ2 ) (1.36)
n ,n
n1 b
ω 1 + n 2 b
ω 2
1 2

x≡ ω̂1
ω̂2 E are two independent conserved quantities.
( − ) ∑ fn ,n (x, E)
⃗ˆ →
g J, θ =i 1 2
n2 ei(n1 θ1 +n2 θ2 ) (1.37)
n +n
x + n1
1 2
( )

Vn1 ,n2 Jˆ
fn1 ,n2 (x, E) = (1.38)
ω2 n1
x + nn21 can be made small by choosing sufficiently large n1 and n2
Suppose we consider only those values of x for which

n2 η
|x + | > (1.39)
n1 (|n1 | + +n2 |)α

17
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

η asmall but fixed number


αsome positive number
η n1
i.e., we exclude a region of with (|n1 |++n2 |)α
around every rational number n1

( − ) ∑ |fn ,n (x, E) |
⃗ˆ →
|g J, θ | < 1 2
η (1.40)
n1 ,n2 (|n1 |++n2 |)α

1 ∑
= fn1 ,n2 (x, E) ((|n1 | + +n2 |)α ) (1.41)
η n ,n
1 2

If for large n1 , n2

1
|fn1 ,n2 (x, E) | < (1.42)
(|n1 | + +n2 |)β
then
( → ) ∑
⃗ˆ − 1
|g J, θ | < = F inite f or β > α (1.43)
n1 ,n2 (|n1 | + +n2 |)β−α
( → )
⃗ˆ −
=⇒ g J, θ will be well def ined (1.44)
(→
− − ) (−
→)
→ −

Typically if V Jˆ , θ is a smooth function of θ , then Vn1 ,n2 Jˆ decays exponentially
for large |n1 |,|n2 |
=⇒The bound on V is satisfied.
Puzzle: The result makes sense only for those values of x which are away from the rational
n1 1
no n2 bya distance more than |n1 |+|n 2|
Take the real axis

Exclude a finite segment of length |n1 |+|n2|
α around each rational number nn21
Infinite number of rational numbers(dense on the real axis)
Calculate the total excluded volume between 0 and 1
∞ ∑
∑ n1 ∞

2η n1
α < 2η (1.45)
|n1 | + |n2 | (n1 )α
n1 =1 n2 =1 n1 =1


∑ 1 π2 π2
= 2η = 2η = η −→ 0 as η −→ 0 (1.46)
n1 =1
(n1 )α−1 6 3

Thus by choosing a small η we can ensure that the first order term in the perturbation
theory will be convergent on most of the real axis.
However the points at which the perturbation expanion diverges are dense on the real axis.
Higher order terms :
KAM (Kolmogrov, Arnold, Moser )
If the Hamiltonian stisfies certain conditions then the same feature holds even for the higher
order terms.
But areorganisation of perurbation theory is necessary.

18
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

Unperturbed Hamiltonian H0 : The phase curve lies on torous 0 ≤ θi ≤ 2π




Different values of Je =⇒ Different tori embedded in the
( )2N dimensional phase space.


e0
∂H Je
Each torus −→ characterised by frequencies ωei = ∂Ji

Resonant tori : e n = 0 for some →
ω −
n
∑i i
Non-resonant tori: ω ei ni ̸= 0 for any −

n
KAM: Shows that under certain conditions on H, non resonant tori are not destroyed by
perturbation although they are deformed.
Resonant tori on the other hand are destroyed by perturbationi.e., the phase curves do not
lie in the subspace resembling an N dimensional torus or any N dimensional torus.
Resonant tori dense in the phase space.
However it is possible that in certain region of phase space fraction of the volume occupied
by stable tori is finite or even 1
In other parts fraction of volume occupied by stable tori may be 0
Γ (E): Fraction of phase space volume at energy E for which the KAM tori are not destroyed.
−→A measure of chaos.
The question
How fast do the two neighbouring trajectories diverge ?
If they diverge exponentially with time, then the motion is chaotic.
If they do not diverge exponentially with time, then the motion is regular.
Integrable system
Phase curve : lie on tori

(0)
Ji = constant, θi = ωi t + θi (1.47)

Consider two neighbouring trajectoris.

(0)
△Ji = △Ji (1.48)

(0)
△θi = (△ωi ) t + △θi (1.49)

Neighter △Ji nor △θi diverges exponentially with time.


=⇒ We expect the motion to be regular
Formal Analysis
Stability matrix:

xi (t) : a trajectory 1≤i≤N (1.50)

∂xi (t)
Mij (−

x (0) , t) = (1.51)
∂xj (0)
Two trajectories with infinitesimal separation:

δxi (t) = Mij (−



x (0) , t) δxj (0) (1.52)

19
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

Mij (−

x (0) , t) measures how fast two trajectories diverge

d i ∂H (−

x (t))
x (t) = Ωij (1.53)
dt ∂xj

d ( i ) ∂H (−→y (t))
x (t) + δxi (t) = Ωij |→
−y (t)=−
→x (t)+δ −

x (t) (1.54)
dt ∂y j
( )
d ( i ) ∂H ∂H
=⇒ δx (t) = Ωij |−→ −
→ −
→ − (1.55)
dt ∂y j y (t)= x (t)+δ x (t) ∂xj

∂ 2 H (→−
x (t)) k
= Ωij δx (t) (1.56)
∂x ∂xk
j

where δxk (t) = Mkl (t) δxl (0) (1.57)


On the other hand,
( )
d ( i ) d −

δx (t) = Mik ( x (0) , t) δx (0) (1.58)
dt dt
Compare ( )
∂ 2 H (−

(Mil (−
→ Mkl (−

d x (t))
=⇒ x (0) , t)) = Ωij x (0) , t) (1.59)
dt ∂xj ∂xk

|δx (t) |2 = δxi (t) δxi (t) = Mik δxk (0) Mil δxl (0) (1.60)

( )
= δxk (0) M T M kl δxl (0) (1.61)
Thus how fast the neighbouring trajectories diverge is detrmined by how fast the eigenvalues
of M T M diverges.

MT M : +ve def inite matrix (1.62)


=⇒has positive eigenvalues
d2α (t)Eigenvalues of M T M
Liapunov function associated with the point ⃗x (0):

1
λα (t) = ln dα (t) (1.63)
t
Liapunov exponent :

1
λα = lim ln dα (t) (1.64)
t−→∞ t

=⇒As t −→ ∞,dα (t) ∼ eλα t


We shall now study the properties of t
Liapunov exponents
Initial conditions on M:

20
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

Mij (→

x (0) , t = 0) = δij (1.65)
dxi { i }
= x ,H (1.66)
dt
{ }
=⇒ xi (t + δt) = xi (t) + δt xi (t) , H (1.67)
=⇒ xi (t + δt)is related to xi (t)by an infinitesimal Canonical transformation
Repeating this,
xi (t)is related to xi (0)by a finite canonical transformation

∂xi (t) kl ∂xj (t)


Ω = Ωij (1.68)
∂xk (0) ∂xl (0)

Mik Ωkl Mjl = Ωij (1.69)

=⇒ M ΩM T = Ω M T ΩM = Ω (1.70)
M M T : A +ve definite symmetry matrix.
=⇒M M T = V DV T , V TV = I
d2(α) : eigenvetors of M T M
( ) j
MT M ij
v(α) = d2(α) v(α)
i
(1.71)

 
1 v1 · · · · · · v1
v(1) (1) (1)
 . .. 
 .. ..
. 
 . 
 . 

V = .. .
. . 
.
. n = 2N (1.72)
. 
 . .. 
 .. ..
. 
 . 
1 v1 · · · · · · v1
v(1) (1) (1)
( ) ( )−1
M T M v(α) = d2(α) v(α) =⇒ M T v(α) = d−2
(α) M v(α) (1.73)

( ) ( )( )−1
M T M Ωv(α) = d2(α) v(α) = M T M ΩM T M T v(α) (1.74)

( )−1
= MT Ω MT v(α) (1.75)

= d−2 T T −2
(α) M ΩM v(α) = d(α) Ωv(α) (1.76)

=⇒ Ωv(α) is an eigenvector of M T M with eigenvalue d−2


(α)
=⇒If d2(α) is an eigenvalue of M T M so is d−2
(α)

1
λ(α) (t) = ln dα (t) (1.77)
t

21
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

=⇒ If λ(α) (t) is a Liapunov function so is −λ(α) (t)


=⇒ If λ(α) is a Liapunov function so is −λ(α)
Thus for Hamiltonian system the Liapunov exponents come in equal nd opposite
pairs.Existence of the limit

d (2) ∂2H
M = ΩH (2) M Hij = (1.78)
dt ∂xj ∂xk
d T
M = −M T H (1) Ω (1.79)
dt
d
M M T = ΩH (2) M M T − M M T H (1) Ω (1.80)
dt
d {( )p } ( ( )p ) (( )p )
Tr MMT = T r ΩH (2) M M T − nT r M M T H (1) Ω (1.81)
dt
d ∑ ( 2 )p ∑ ( )p {( ) ( ) }
dα = p d2α ΩH (2) − H (1) Ω (1.82)
dt α α
αα αα

d ∑ ( 2 )p ∑ ( )p−1 d ( )
dα = p d2α d2α (1.83)
dt α α
dt
d ( 2) ( ) ( )
=⇒ dα = ΩH (2) − H (1) Ω d2α = 2 ΩH (2) d2α (1.84)
dt αα α
( )
=⇒ ln d2α (t) = 2Int.dt′ ΩH (2) (t) dt′ (1.85)
αα

( ) 1 2
ΩH (2) (t) If it is bounded f rom above, then lim dα (t) exists (1.86)
αα t−→∞ t

∂2H
This is the case if ∂xj ∂xk
is bounded everywhere.

dxi
= F i (−

x (t)) (1.87)
dt

=⇒ xi (t) = Φi (−

x (0) , t) (1.88)

dxi (t)
≡ Mij (t, −

x (0)) (1.89)
dxj (0)
d2α (t) : Eigenvalue of M T M
−→Measure of the rate at which the distance between neighbouring trajectories increase

1
λα (t) = ln dα (t) (1.90)
t

λα = lim λα (t) (1.91)


t−→∞
For Hamiltonian system :

22
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

∂2H
• λα are well defined if ∂xi ∂xj
is finite everywhere in the phase space for all v

• λα ’s come in equal and opposite pair λ2α = −λ2α

Based of M ΩM T = Ω −→ Proved
Also used M T ΩM = Ω
Proof :
M ΩM T ΩM = ΩM = −M (1.92)

=⇒ ΩM T ΩM = −1 (1.93)

=⇒ M T ΩM = Ω (1.94)

For systems with N degrees of freedom and N conserved quantities with vanishing Poisson
Bracket, the Liapunov exponents all vanish.

• First show that the result holds in the action angle coordinates.

• Then show that the result is coordinate independent.

dM (2) ∂2H
= ΩH (2) M Hij = (1.95)
dt ∂xi ∂xj
 
0 1 ···
 
 −1 0 
 
 0 1 
Ω=  (1.96)
 .. 
 . −1 0 
 
..
.
Action
(− angle variable :Jα ,θα
→)
H J

∂2H ∂2H ∂2H


̸= 0, = 0, =0 (1.97)
∂Jα ∂Jβ ∂Jα ∂θβ ∂θα ∂θβ

∂Jα ∂Jβ −→ Constant (1.98)

dM (2) t
= ΩH (2) M, M = I at t = 0 =⇒ M = eΩH I (1.99)
dt

ΩH (2) −→ T ime independent matrix (1.100)

( )2
ΩH (2) = ΩH (2) ΩH (2) = 0 (1.101)
Proof:

23
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

 
x1  1 
 2  θ
 x   
 .   J1 
 .   2 
 .  θ 
 .   
 .  J 
 . = 2  (1.102)
 .   . 
 .   .. 
 .   
 .   N 
 .  θ
 . 
JN
x2N

x2α−1 = θα , x2α = Jα , Ω2α−1,2α = 1, Ω2α,2α−1 = −1 (1.103)

( ) ∂2H ∂2H
(2)
ΩH (2) Ω = −H2α,2β = − − ̸= 0 (1.104)
2α−1,2β−1 ∂x2α ∂x2β ∂Jα ∂θβ
( ) ∂2H
(2)
ΩH (2) Ω = H2α,2β−1 = =0 (1.105)
2α−1,2β ∂Jα ∂θβ
( ) ∂2H
(2)
ΩH (2) Ω = H2α−1,2β = =0 (1.106)
2α,2β−1 ∂θα ∂J β
( ) ∂2H
(2)
ΩH (2) Ω = −H2α−1,2β−1 = − =0 (1.107)
2α,2β ∂θα ∂Jθβ
( ) ( )
(2)
ΩH (2) ΩH (2) = ΩH (2) Ω Hkj (1.108)
ij ik

( ) ( )
(2)
ΩH (2) ΩH (2) = ΩH (2) Ω H2β−1,j = 0 (1.109)
2α−1,j 2α−1,2β−1

(2) t
=⇒ M = eΩH = 1 + ΩH (2) t (1.110)

( )( )
M T M = 1 − H (2) Ωt 1 + ΩH (2) t (1.111)

( ) ( )2
= 1 + ΩH (2) − H (2) Ω t + H (2) t2 (1.112)
∂2H
rα : Eigenvalues of ∂Jα ∂Jβ
Consider the basis in which H (2) is diagonal.
   
0 0 1
   −1 0 
 r2   
   
UH U = 
(2) T

0 
 U ΩU = 
T

0 1 
 (1.113)
 r2   −1 0 
   
.. ..
. .

24
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

 
0 r1
r 0 
( )  1 
 
U ΩH − H Ω U = 
(2) (2) T

0 r2 
 (1.114)
 r2 0

 
..
.
 
1 r1 t
 r t 1 + r 2 t2 
 1 
 1

T T 
=⇒ U M M U =  1 r2 t  (1.115)

 r2 t −r2 t2 
 
..
.
[( 2 )( ) ( )]
Eigenvalues :πα d − 1 d2 − rα2 t2 − 1 − rα2 t2 = 0
[( ) ( ) ]
2
πα d2 − rα2 t2 + 2 d2 + 1 = 0 (1.116)

(2) T
H2α = Vαr (D)x−δ Uδβ (1.117)

 
r1
 
 r2 
 .. 
 . 
D=  V TV = 1 (1.118)
 .. 
 . 
 
..
.
Choose U such that

U2α,2β−1 = U2α−1,2β = 0 U2α−1,2β−1 = U2α,2β = Vαβ (1.119)

( )
(2) T (2)
Uij Hjk Ukl = Hα check (1.120)
il

 
0
 
 r1 
 
Hα(2) =

0 
 (1.121)
 r2 
 
..
.

Uij Ωjk Ukl


T
= Ωil (1.122)

{ √ }
1 2
d2α± = 2 + rα t ± (rα t + 2) − 4
2 2 2 2 (1.123)
2

25
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

1{ √ }
= 2 + rα2 t2 ± rα4 t4 + 4rα2 t2 1≤α≤N (1.124)
2

=⇒ 2N eigenvalues (1.125)

Product of d2α+ & d2α− = 1


=⇒ Agres with the general result

1
λα± (t) = ln d2α± (t) (1.126)
t
[ { √ }]
1 1
= ln 2 + rα t ± rα t + 4rα t
2 2 4 4 2 2 (1.127)
t 2
As t −→ ∞
[ { √( )}]
1 4
= 2 + rα t ± rα t
2 2 2 2
1+ 2 2 (1.128)
2 rα t
[ { ( )}]
1 2
= 2 + rα t ± rα t 1 + 2 2
2 2 2 2
(1.129)
2 rα t

1 ( )
=⇒ lim λαt (t) = lim ± ln 2 + rα2 t2 = 0 (1.130)
t−→∞ t−→∞ t
=⇒ All the Lipunov exponents vanishCoordinate dependence

xi −→ y i (1.131)

δxi (t) = Mij (t, −



x (0)) δxj (0) (1.132)

∂y i k
δy i (t) = δx (k) (1.133)
∂xk

∂xl (θ) j
where δxk (k) = Mkl (t, →

x (0)) δxΩ (0) = δy (0) (1.134)
∂y j (0)

∂y i →
− ∂xl
= | M
t kl (t, x (0)) |t=0 δy j (0) (1.135)
∂xk ∂y j

fij (t, −
→ ∂y i −
→ ∂xl
N ow M y (0)) = | M
t kl (t, x (0)) |t=0 (1.136)
∂xk ∂y j

∂y i ∂xl
=⇒ Aik (−

x)≡ =⇒ A−1 −
lj ( →
x ) = (1.137)
∂xk ∂y j

fij (t, →
M −
y (0)) = Aik (→

x (t)) Mkl (t, −

x ) A−1 −

lj ( x (0)) (1.138)

26
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

( ) →
f = A−1 T (−
fT M
M x (0)) M T AT (−

x (t)) A (−

x (t)) M A−1 (−

x) (1.139)
Convert a basis where H (2) is diagonal

(2)
H (2) = U Hd U T (1.140)

( )
(2)
M = U 1 + ΩHd (t) U (1.141)

( ) → ( ) ( ) ( ) −
f = A−1 T (−
fT M
M x (0)) U 1 + ΩHd t U T AT (−
(2) T →
x (t)) A (−

x (t)) U 1 + ΩHd t U T A−1 (→
(2)
x (0))
(1.142)
 
1 r1 t
0 1 
( ) 



1 + ΩHd t =  
(2) 1 r2 t
  (1.143)
 0 1 
 
..
.
Change of coordinates : non-singular everywhere

∂y i ( −1 ) ∂xi
Aij = , A ij
= are f inite (1.144)
∂xj ∂y j
Components of U: finite
Components of M, M T : increases linearly with t
fT M
Components of M f: increases at most quadratically with time.

1
=⇒ lim ln dj = 0 (1.145)
t−→∞ t

=⇒Liapunov exponents are all zero.


General result for non-integrable system:
1. Liapunov exponents are independent of the choice of coordinate system, although Lia-
punov unctions are dependent on this choice.
2.If the system has M conserved quantities with vanishing Poission Bracket,then at least
2M of the Liapunov exponents.
E: Given energy

Γ (E) = Integration dN p dN qδ (E − H (−

p ,−

q )) −→ T otal phase (1.146)
Space volume at Energy E


N
K (E) = Integrate dN p dN qδ (E − H (→

p ,−

q )) λα+ (1.147)
α=1

−→Kolmogorov (KS) entropy

27
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

KSmetric entropy
K (E) ̸= 0 =⇒Chaotic motion
Flow of trajectories =⇒ K- flow
Discrete dynamics

dxi
= F i (→

x) (1.148)
dt

=⇒ xin+1 = F i (−

x n) (1.149)

F i is generating function of xn only not xn−1 , xn−2 ... because like continnum mechanics it is
local
n discretized time.
Usefulness
1. Continuous time system −→discrete time system during numerical evolution.
2. Delayed Feedback

Area of Discrete dynamics


1.When we do numerical analysis of a dynamical system
2.Annual rate of change (population , profit,..)
Start from the simplest situation
1. One variable
2.F (x) = ax + b Linear function.
b
Fixed point : x = ax + b =⇒ xF = 1−a

b
y = x − xF = x − (1.150)
1−a

b b ab b
yn+1 = xn+1 − = axn + b − = ayn + +b− (1.151)
1−a 1−a 1−a 1−a

=⇒ yn+1 = ayn (1.152)

28
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

yn = an y0 (1.153)

|yn | −→ ∞ f or n −→ ∞ if |a| > 1 (1.154)

|yn | −→ 0 f or n −→ ∞ if |a| < 1 (1.155)


Next example

F (x) = ax2 + bx + c (1.156)


Fixed point : xF = F (xF )
Roots : α, β (say)

=⇒ F (x) − x = a (x − α) (x − β) (1.157)

xn+1 = a (xn − α) (xn − β) + xn (1.158)


y =x−α

yn+1 = xn+1 − α = (xn − α) {a (xn − β) + 1} (1.159)

= yn {a (yn + α − β) + 1} (1.160)

{ }
1
=⇒ yn+1 = ayn (yn + α − β) + (1.161)
a

y
z= (1.162)
β−α− 1
a
( )
1
zn+1 = azn β − α − {zn−1 } (1.163)
a

= {1 + (α − β) a} zn (1 − zn ) (1.164)
But {1 + (α − β) a} = µ
Since (α − β) can be chosen as we please
Choose a (α − β) > 0 =⇒ µ > 1
Experimental result
Start from a given z0
Calculate z1 , z2 , z3 , ...
For 0 ≤ µ ≤ 1 zn −→ 0 as n −→ ∞
For 0 ≤ µ ≤ 3 zn −→ 1 − µ1 as n −→ ∞
µ = 3.1 Zn oscillates between .557 and .765 as n −→ ∞

29
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

P ERIOD 2 (1.165)
µ = 3.5 Zn oscillates between four fixed points of period 4
As we increase µ, the period keeps doubled.
µ = 3.6 no stable periodic motion −→Chaotic motion.
µ = 3.84 xn oscillates between 3 fixed points as n −→ ∞

P ERIOD 3 (1.166)
Need to understand this phenomenon analytically.

Analytical results: Call z, x

F (x) = µx (1 − x) xn+1 = F (xn ) (1.167)


Fixed point at x = 0, x = 1 − 1
µ
Stability analysis

xn = xF + ϵ (say) (1.168)


xn+1 = F (xF + ϵ) = xF + ϵF (xF ) (1.169)


=⇒ (xn+1 − xF ) = F (xF ) (xn − xF ) + 0 (xn − xF )2 (1.170)


=⇒ |xn+1 − xF | < |xn − xF | if |F (xF ) | < 1 (1.171)

Stable for |F (xF ) | < 1

Unstable for |F (xF ) | > 1

30
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project


F (x) = µx (1 − x) − µx − µ (1 − 2x) (1.172)


F (0) = µ (1.173)

( ) ( )
′ 1 2
F 1− =µ 1−2+ = (2 − µ) (1.174)
µ µ
0 is stable for 0 ≤ µ ≤ I
unstable for µ > I
For 0 ≤ µ ≤ 1 z−→ 0 as n −→ ∞
1 − µ1 is stable for 1 ≤ µ ≤ 3
unstable for µ ≥ 3
For 1 ≤ µ ≤ 3 z−→ 1 − µ1 as n −→ ∞
Q. What happens for µ > 3?
Introduce second generation map

f2 (x) = f (f (xF )) = f (xF ) = xF (1.175)


=⇒A fixed point of f is also a fixed point of xF
What about its stability?

′ ′ ′ ′
f2 (x) = f (f (xF )) f (xF ) = |f (xF ) |2 (1.176)

′ ′
=⇒ if |f (xF ) | < 1 =⇒ |f2 (x) | < 1 =⇒ Stable (1.177)

′ ′
=⇒ if |f (xF ) | > 1 =⇒ |f2 (x) | > 1 =⇒ U nstable (1.178)
But f2 may have other fixed points which are not fixed points of x

f2 : quadratic in x (1.179)
x = f2 (x)has 4 roots
two of them are 0 and 1 − µ1
What about the others? If they are real, then they correspond to a period of 2 cycle of f
(2)
xF A fixed point of f2
( )
(2) (2)
=⇒ f2 xF = xF (1.180)
( )
(2) (2)
But f xF ̸= xF

( )
(2)
For stability we need |f2 xF | < 1
Precisely at µ = 3two new stable fluid points of xn+1 = f2 (x) appers.
Graphical Analysis :
µ<3

31
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

µ=3
µ=3+ϵ
=⇒f2 develops two stable fixed point
=⇒Period doubling.
As we increase µthe two fixed points of f2 becomes unstable at some µ = µc , and precisely
at that point f4 develops four stable fixed points.
−→This way we can explain the phenomenon of period doubling.

32
Classical Mechanics IV: Ashoke Sen Padakshep Open Teaching Project

33

You might also like