You are on page 1of 27

Preprint typeset in JHEP style - HYPER VERSION

Classical Mechanics II

Ashoke Sen
Harish-Chandra Research Institute Chhatnag Road,
Jhusi, Allahabad 211019, India

Abstract: This note is based on a handwritten note by Prof Sen. This note is typed by
Sampath Mukherjee and Arnab Rudra under 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 II: Ashoke Sen Padakshep Open Teaching Project

Contents

1. Lagrangian dynamics: 1

2. Hamiltonian System (II) 14

1. Lagrangian dynamics:

1.1 Definition: Legendre transformation of the Hamiltonian


( )
uα ≡ ∂p∂H
α
=⇒ Gives uα as a function of pα & q α
p
⃗,⃗
q
Assume that these relation are invertible, So that this allows us to solve pi as function of
α
u & qα


n
L(⃗q, ⃗u) = pα uα − H(⃗
p, ⃗q) (1.1)
α=1

pα express as function of q α and uα


N ∑
N N (
∑ )
∂H ∂H
dL = dpα u −
α
( )p⃗,⃗qdpα − dq α (1.2)
∂pα ∂q α p⃗,⃗q
α=1 α=1 α=1


N
∂H α
= (pα duα − dq ) (1.3)
∂q α
α=1
( ) ( ) ( )
∂L ∂L ∂H
⇒ q⃗,⃗
u = pα , q⃗,⃗
u =− (1.4)
∂uα ∂q α ∂q α p
⃗,⃗
q

Equation of motion:
( )
∂q α ∂H
= = uα (1.5)
∂t ∂pα p
⃗,⃗
q

( ) ( ) ( )
d ∂L ∂H ∂L

q ,⃗
u =− = (1.6)
dt ∂uα ∂q α p
⃗,⃗
q ∂q α p
⃗,⃗
u

Under what condition is a transformation q α &uα ************text missing ***** of the


equation of motion.

1
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

1.2 Lagrangian : Alternate formulation

Dynamical Variables.
q α , uα 1≤α≤N (1.7)

Equations of motion:

dq α
= uα (1.8)
dt
( ) ( )
d ∂L ∂L
q⃗,⃗
u = (1.9)
dt ∂uα ∂q α ⃗
q ,⃗
u

Unlike the Hamilton’s equation, these equations do not have nice tensorial form,and hence
it is not clear what condition one must impose on the Lagrangian so that equation of motion
are preserved under a transformation(⃗q, ⃗u) −→ (⃗q˜, ⃗u
˜)
First take a brute force approach
If you take Lagrangian equation as 2N dimensional vector equation then q α & uα are not
in equal footing. So its covariance cannot be demanded so easily **********missing text******
First take
q α −→ qeα = f α (⃗q) (1.10)

df α (⃗q) ∂f α (⃗q) β ∂ qeα β


uα −→ u
eα = = u = u (1.11)
dt ∂q α ∂q β
This gives

de

=u (1.12)
dt
( )
e e
Define L e ≡ L (⃗q, ⃗u) (Justified later in Problem -3)
⃗q, ⃗u

d ∂L d ∂L e ∂e
uβ ∂L e ∂ qeβ
( α) = ( β α + β α) (1.13)
dt ∂u dt ∂e
u ∂u ∂ qe ∂u

d ∂L e ∂ qeβ (⃗q)
= ( β ) (1.14)
u ∂q α
dt ∂e

d ∂L e ∂ qeβ ∂Le ∂ 2 qeβ


= ( β ) α + β r α ur (1.15)
dt ∂e
u ∂q ∂e
u ∂q ∂q

∂L ∂L e ∂ qeβ ∂Le ∂e

=⇒ = + (1.16)
∂q α ∂ qeβ ∂q α ∂e uβ ∂q α

∂L e ∂ qeβ ∂Le ∂ 2 qeβ


= + β α r ur (1.17)
∂ qe ∂q
β α u ∂q ∂q
∂e

2
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

1.3 Equation of motion:


d ∂L e ∂ qeβ ∂Le ∂ 2 qeβ ∂L e ∂ qeβ ∂Le ∂ 2 qeβ
( β ) α + β r α ur = β α + β α r ur (1.18)
dt ∂e
u ∂q ∂e
u ∂q ∂q ∂ qe ∂q ∂e
u ∂q ∂q

d ∂L e ∂ qeβ ∂q α ∂L e ∂ qeβ ∂q α
=⇒ ( β) α r = β α r (1.19)
dt ∂e
u ∂q ∂ q̃ ∂ qe ∂q ∂ q̃

∂ qeβ ∂q α ∂ qeβ ∂uα ∂ qeβ


=⇒ + = = δrβ (1.20)
∂q α ∂ q̃ r ∂uα ∂ q̃ r ∂ qer

d ∂L e ∂L e
=⇒ ( β )δrβ = β δrβ (1.21)
dt ∂e
u ∂ qe

d ∂Le ∂L e
=⇒ ( r)− r = 0 (1.22)
dt ∂e
u ∂ qe
Thus the equation of motion have the same form in the new coordinate system if
( ) ( )
e e
L e ≡ L ⃗qe, ⃗u
⃗q, ⃗u e (1.23)
i.e. if ( )
e e
L e ≡ L (⃗q, ⃗u)
⃗q, ⃗u (1.24)
Associated with such a continuous symmetry transformation there is a conserved quantity.
Infinitesimal transformation:

q α −→ qeα = q α + ϵϕα (q) (1.25)

∂ϕα β
uα −→ u
eα = uα + ϵ u (1.26)
∂q β
( ) α
e e
L e ≡ L (⃗q, ⃗u) + ϵ ∂L ϕα + ϵ ∂L ∂ϕ uβ
⃗q, ⃗u (1.27)
∂q α ∂uα ∂q β

∂L α ∂L ∂ϕα β
=⇒ ϕ + u =0 (1.28)
∂q α ∂uα ∂q β
Conserved quantity:

∂L α
F = ϕ (1.29)
∂uα
Proof :

dF d ∂L d ∂L α ∂L ∂ϕα β
= ( α ϕα ) = ϕ + u (1.30)
dt dt ∂u dt ∂uα ∂uα ∂q β

∂L α ∂L ∂ϕα β
= ϕ + u =0 (1.31)
∂q α ∂uα ∂q β
=⇒F remains constant as ⃗q, ⃗u evolve according to the equations of motion.

3
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

Example:

3
(pα )2
H= + V (r) (1.32)
2m
α=1

∂H pα
uα = ( )p⃗,⃗q = (1.33)
∂pα m
∑ ∑ 1 ∑ α α
L= u α pα − H = m uα uα − m u u − V (r) (1.34)
α α
2 α

1 ∑ α α
= m u u − V (r) (1.35)
2 α

Invariant under:

q α −→ qeα = Rαβ q β RRT = I (1.36)

uα −→ u
eα = Rαβ uβ (1.37)
Infinitesimal Transformations:

Rαβ = δαβ + ϵωαβ (1.38)

qeα = q α + ϵωαβ q β (1.39)


where ϕα (q)

∂L α
=⇒ F = ϕ (q) = muα ωαβ q β (1.40)
∂uα

ω = a1 T1 + a2 T2 + a3 T3 (1.41)

F = a1 F1 + a2 F2 + a3 F3 (1.42)

F1 = m(u2 q 3 −u3 q 2 ), F2 = m(u3 q 1 −u1 q 3 ), F3 = m(u1 q 2 −u2 q 1 ) −→ Angular Momenta (1.43)

Transformation Laws of uα are determined from the beginning in terms of those of q α .


We do not need to check if the transformations are canonical.
More general transformations :

q α −→ qeα = f α (⃗q, ⃗u) (1.44)


de ∂f α β ∂f α β
uα −→ u
eα = = u + u̇ (1.45)
dt ∂q β ∂uβ

4
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

Figure 1: Type I Dualities.

−→Cannot be viewed as a transformation of phase space coordinates:


Pe is not determined in terms of P .
Passive view point seems to be lost.
β
Way out: Equation of motion can be used to express du
dt in terms of ⃗
u & ⃗q

eα = hα (⃗q, ⃗u)
=⇒ u (1.46)

The new transformation Laws agree with the old ones while acting on solutions the equation
of motion.
Must do this in order to write the transformation laws of p⃗, ⃗q in hamiltonian formalism.
We can stick to the original transformations in the Lagrangian formulation.
Active point of view still holds.
A solution of equation of motion: ⃗q0 (t) =⇒all derivatives of ⃗q0 (t)
qe0α (t) & u
eα0 (t) are defined.
The transformation (q α , uα ) −→ (e eα ) is a symmetry if it takes all solutions of the
qα, u
equation of motion to solutions.
We would like to determine what kind of transformations satisfy this property and call them
symmetry transformations.
A related problem
( )
e ⃗qe, ⃗u
L e = L(⃗q, ⃗u) (1.47)

∂f α β ∂f α β
=⇒ qeα = f α (⃗q, ⃗u), u
eα = u + u̇ (1.48)
∂q β ∂uβ
( )
L ⃗qe, ⃗u
e = f unction of q α , uα & u̇ (1.49)

5
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

( )
e ⃗qe, ⃗u
L e = L(⃗q, ⃗u) : F unction of q α , uα (1.50)

e cannot be equal to L in a generic Case.Change point of view.


=⇒L
Fundamental Dynamical Variables.
Functions ⃗q(t) & not ⃗q, ⃗u
Actually here fundamental variable is path which is varying.
Instead of using the Lagrangian L, use

∫t2
S({⃗q(t)}) = L(⃗q(t), ⃗q˙(t))dt (1.51)
t1

where q(t1 ) = q1 , q(t2 ) = q2 =Boundary conditions.

S: Functions of function ⃗q(t)=⇒Functionals.


For given ⃗q(t),we have a definite rule for computing S.
If ⃗q(t) is a classical solution then S always increases under a local deformation of the path
away from the boundaries.

6
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

∫t2
∂L ∂L
δS = ( α δq α + α δ q̇ α ) (1.52)
∂q ∂u
t1

∫t2
∂L d ∂L
= ( α δq α − ( α )δq α ) = 0 (1.53)
∂q dt ∂u
t1

Noticed this fact S be extremum for classical path we will use it get SYMMETRY CONDI-
TION

H(⃗q, p⃗) −→ L(⃗q, ⃗u) (1.54)


A change of coordinate

q α −→ qeα = f α (⃗q, ⃗u) (1.55)

df α (⃗q, ⃗u)
uα −→ u
eα = (1.56)
dt
makes sense not on phase space coordinates but on functions q α (t)
Paths:

Reverse is also true.


If S always increases under a local deformation of a path away from the boundary, then the
path describes a solution of the classical equation of motion.

∫t2
∂L ∂L
0 = δS = ( α δq α + α δ q̇ α ) (1.57)
∂q ∂u
t1

∫t2
∂L d ∂L
= δq α (t)( − ( )dt (1.58)
∂q α dt ∂ q̇ α
t1

δq α arbitrary. Take δq α = δαβ δ(t − t′ )(say)


at t′ we have

7
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

∂L d ∂L
=⇒ β
− ( β) = 0 (1.59)
∂q dt ∂q
−→valid for allβand all t′
=⇒A Classical solution

⃗q, ⃗u −→ ⃗qe, ⃗u
e (1.60)


de
=⇒ qeα = f α (⃗q, ⃗u), u
eα = (1.61)
dt

e ⃗qe, ⃗u ˙ ¨
L(⃗q, ⃗u) ≡ L( e, ⃗u
e, ⃗u
e.....) (1.62)

∫t2
e ⃗qe, ⃗u ˙ ¨
= L( e, ⃗u
e, ⃗u
e.....) (1.63)
t1

∫t2
= L(⃗q, ⃗u) = S({⃗q(t)}) (1.64)
t1

If S is minimized for the path ⃗q(t), then S̃ is minimized for the path ⃗q̃(t)
=⇒Equation of motion of ⃗qe is obtained by minimizing S. e
We want to ask under what condition is this equation has the same form as the equation of
motion for ⃗q(t).
e ⃗qe, ⃗u ˙ ¨ ˙ ¨
Ans. : If L( e, ⃗u
e, ⃗u
e.....) = L(⃗qe, ⃗u
e) + d K(⃗qe, ⃗u
dt e, ⃗u
e, ⃗u
e.....) for some K
Proof : In this case

∫t2
Se = L(⃗qe, ⃗u
e) + K(qe⃗2 , ⃗u
e2 , ...) − K(⃗qe1 , ⃗u
e1 , ...) (1.65)
t1

Consider a variation δe
q (t) of the path away from the boundaries.

8
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

∫t2
∂L α ∂L
δ Se = dt[ δe q α )]
q + α ∂t (δe (1.66)
∂ qeα ∂e
u
t1

∫t2
∂L d ∂L
qα[
dtδe − ] (1.67)
∂ qeα dt ∂e

t1

Physical significance: Suppose ⃗q(t) = ⃗q0 (t) is a solution of the equations of motion.
Then q α (t) = f α (⃗q0 (t), ⃗u0 (t) = d⃗qdt
0 (t)
) is also a solution of the equations of motion.
=⇒A Symmetry in the active senses.
Conservation Law
If under an infinitesimal transformation

q α −→ qeα = q α + ϵϕα (⃗q, ⃗u) (1.68)

e ⃗qe, ⃗u d
L( e) = L(⃗qe, ⃗u
e) + ϵ N (⃗q, ⃗u) (1.69)
dt
Then
∂L(⃗q, ⃗u) α
Q= ϕ (⃗q, ⃗u) + N (1.70)
∂uα
is conserved
Proof :

e ⃗qe, ⃗u
L( e) ≡ L(⃗q, ⃗u) (1.71)

∂L α ∂L dϕα dN
L(⃗q, ⃗u) = L(⃗q, ⃗u) + ϵ α
ϕ +ϵ α +ϵ (1.72)
∂q ∂u dt dt

dQ d ∂L dN
=⇒ = ( α ϕα ) + (1.73)
dt dt ∂u dt

∂L α ∂L dϕα dN
=⇒ ϕ + + =0 (1.74)
∂q α ∂uα dt dt
=⇒Q is conserved along the motion
Example
Time translation Symmetry
Given a solution q α (t), q α (t + a) is also a solution.
For small a = ϵ, q α (t + ϵ) = q α (t) + ϵuα (t) where uα (t) ≡ ϕα (⃗q, ⃗u)

uα (t) −→ uα (t) + ϵu̇α (t) (1.75)

e ⃗qe, ⃗u
L( e) ≡ L(⃗q, ⃗u) = L(e eα + ϵu̇α )
q α + ϵuα , u (1.76)

9
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

∂L ∂L duα
= L(⃗qe, ⃗u
e) − ϵ α uα − ϵ α (1.77)
∂q ∂u dt

dL
= L(⃗qe, ⃗u
e) − ϵ (1.78)
dt

=⇒ N = −L (1.79)

∂L α
Q= ϕ +N (1.80)
∂uα

∂L α
= u −L (1.81)
∂uα

∂L
= pα =⇒ Q = pα uα − L = H (1.82)
∂uα
We are checking the mathematical consistency of two formalism.
Compare with the transformations generated by the conserved charge H in the Hamiltonian
formalism :
{ }
δxi = ϵ xi , H (1.83)

∂H
δq α = ϵ {q α , H} = ϵ = ϵuα (1.84)
∂pα

∂H
δpα = ϵ {pα , H} = −ϵ (1.85)
∂q α
( ) ( )
∂uα ∂uα
=⇒ δuαH = β
δp + δq β (1.86)
∂pβ p
⃗,⃗
q ∂q β p
⃗,⃗
q

Compare with

duα
δuαL = ϵ (1.87)
dt

∂H
uα = (1.88)
∂pα

duα ∂ 2 H dq β ∂ 2 H dpβ
=⇒ = α β + α β (1.89)
dt ∂p ∂p dt ∂p ∂p dt

∂2H ∂ 2 H ∂H
= − (1.90)
∂pα ∂q β ∂pα ∂pβ ∂q β
-
( ) ( )
duα ∂uα β ∂uα
=⇒ ϵ = δp + δq β (1.91)
dt ∂pβ ∂q β

10
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

=⇒The transformation agree after using equations of motion


L0 (⃗q, p⃗): Some given Lagrangian.
We want to study system with constrains:ϕi (⃗q) = 0 i = 1, 2, ..., m

1 ∑
m
L(⃗q, p⃗) = L0 (⃗q, p⃗) − (ϕi (⃗q))|2 (1.92)

i=1

For η −→ 0 this will constrain the system to move along ϕi (⃗q) = 0


We change coordinates such that ϕ1 (⃗q), ....ϕm (⃗q) are the first m coordinates and the rest os
(n-m) coordinates are some functions of ⃗q′ s.
How do the equation change under a coordinate transformation?

qeα = f α (⃗q) (1.93)

∂f α (⃗q)uβ
ũα = ? (1.94)
∂q β


de
eα automatically
=u (1.95)
dt
e ⃗qe, ⃗u
Define L( e) = L(⃗q, ⃗u)

∂L ∂L e ∂ qeα ∂Le ∂e

= + (1.96)
∂uβ ∂ qeα ∂uβ uα ∂uβ
∂e
=
∂Le ∂f α (⃗q)
= (1.97)
uα ∂q β
∂e

d ∂L d ∂L e ∂f α
( β) = ( α β ) (1.98)
dt ∂u dt ∂e
u ∂q

d ∂L e ∂f α ∂Le ∂2f α
= ( α ) β + α β r ur (1.99)
dt ∂e
u ∂q ∂e
u ∂q ∂q

∂L ∂L e ∂ qeα ∂Le ∂e

= + (1.100)
∂uβ ∂ qeα ∂q β uα ∂q β
∂e

∂L e ∂f α ∂Le ∂2f α
= + ur (1.101)
∂ qeα ∂q β uα ∂q β ∂q r
∂e

d ∂L ∂L
=⇒ ( )− β =0 (1.102)
dt ∂uβ ∂q

d ∂L e ∂L ∂f α
=⇒ ( ( α) − α) β = 0 (1.103)
dt ∂e
u ∂ qe ∂q

d ∂L e ∂L
=⇒ ( α) − α = 0 (1.104)
dt ∂e
u ∂ qe

11
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

e ⃗qe, ⃗u
L( e) is the new Lagrangian.
Now go back to the problem at hand

f i (⃗q) = ϕi (⃗q) = 0 i = 1, 2, ..., m (1.105)


f i (⃗qarbitrary for i = 1, 2, ..., m
e 0 (⃗qe, ⃗u
L e) ≡ L0 (⃗q, p⃗)

1 ∑ 2
m
e ⃗ ⃗
L(qe, u e ⃗ ⃗
e) = L0 (qe, u
e) − qei (1.106)

i=1

e ⃗qe, ⃗u
d ∂ L( e) e ⃗qe, ⃗u
∂ L( e)
=⇒ ( ) − =0 (1.107)
dt ∂eu α ∂ qeα

d ∂L e 0 (⃗qe, ⃗u
e) e 0 (⃗qe, ⃗u
∂L e) 1
=⇒ ( ) − − qe = 0 i = 1, ..., m (1.108)
dt ∂eu i ∂ qe i η

d ∂L e 0 (⃗qe, ⃗u
e) e 0 (⃗qe, ⃗u
∂L e)
=⇒ ( ) − = 0 f or α = m + 1, ..., n (1.109)
dt ∂e u α ∂ qe α

η −→ 0 qei = 0 ei = 0
=⇒ u (1.110)
Define :

Lnew (⃗qe, ⃗u e 0 (⃗qe, ⃗u


e) = L e)|qei =0,eui =0 i = 1, ...., m (1.111)

q1 , ..., qen−m ) = (e
(e q m+1 , ..., qen )|qei =0,eui =0 (1.112)

(e en−m ) = (e
u1 , ..., u en )
um+1 , ..., u (1.113)

d ∂Lnew (⃗qe, ⃗u
e) ∂Lnew
=⇒ − =0 (1.114)
dt ∂e
u k ∂ qek
1.3.1 Force due to constraint :
1 d ∂L0 (⃗qe, ⃗u
e) ∂L(⃗qe, ⃗u
e)
− qei = +( − ) (1.115)
η dt ∂e u i ∂ qei

Prescription: Simply eliminate m,⃗qvariables using the constraints & express the Lagrangian
in terms of remaining ⃗q, ⃗u .
=⇒The New Lagrangian .
Once we have the New Lagrangian we can use Action Principle,Hamiltonian Formalism etc.
What about ϕi (⃗q) = 0 i = 1, 2, ..., m
See M.R.Flannory, Am.J.Physics, Vol.73,p 265.
If the constrain equation contains terms higher than 1st order,then there is a sign ambigu-
ity.One can choose any sign and solve .But one has to check whether the solution approaches
the boundary as it evolves in time.

12
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project


i.e.,suppose constrain equation is x2 + y 2 + z 2 = 1 ⇒ z = ± 1 − x2 − y 2 one can put

z = + 1 − x2 − y 2 (upper Hemisphere) and checks whether the solution approaches z = 0 (at
the boundary).It approaches the boundary in finite time then one has to deal with both the
cases.
i.e., locally the sign ambiguity has no meaning.It takes meaning only in global sense.
Symmetry of the equations of motion underq α −→ qeα = f α (⃗q, ⃗u) requires

d
L(⃗q, ⃗u) = L(⃗qe, ⃗u
e) + k(⃗q, ⃗u) (1.116)
dt
for some k
Infinitesimal transformation

q α −→ qeα = q α + ϵϕα (⃗q, ⃗u) (1.117)

d
L(⃗q, ⃗u) = L(⃗qe, ⃗u
e) + ϵ N (⃗qe, ⃗u
e) + 0(t2 ) (1.118)
dt

∂L α ∂L dϕα dN
=⇒ ϕ + + =0 (1.119)
∂q α ∂uα dt dt
Associated conserved quantity

∂L α
Q= ϕ (⃗q, ⃗u) + N (1.120)
∂uα
dQ
dt = 0 as a result of 1.Symmetry 2.Equation of motion
Caution : While checking a given transformation is a symmetry,never use equations of
motion
Otherwise every transformation will appear to be a symmetry

∂L α ∂L dϕα d ∂L α ∂L dϕα d ∂L
α
ϕ + α
= α
ϕ + α
= ( α ϕα ) (1.121)
∂q ∂u dt dt ∂u ∂u dt dt ∂u
αϕ − N = 0
∂L α
Corresponding conserved quantity ∂u
Reason for this “symmetry”
⃗ for the classical path
δS = 0to first order in δq α for any δq
({ })
S ({⃗q}) = S ⃗ q , ⃗u)
⃗q + ϵϕ(⃗ (1.122)


for any ϕfor the classical path.By using equation of motion we are restricting our checking
to classical paths only.

Example : Time translational Symmetry


q α −→ qeα = q α (t + ϵ) = q α (t) + ϵuα (t) (1.123)

duα
uα −→ u
eα = uα (t) + ϵ (1.124)
dt

13
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

Conserved charge Q = H
Compare with the transformation laws in the Hamiltonian framework.
{ }
Conserved quantity ψ↔transformation lawδxi = ϵ xi , ψ
ψ=H

∂H
δq α = {q α , H} = ϵ = ϵuα (1.125)
∂pα

∂H
δpα = {pα , H} = −ϵ( )|⃗q,⃗p (1.126)
∂q α

∂H
uα = ( )|p⃗,⃗q (1.127)
∂pα

∂ 2 H ∂H ∂ 2 H dq β ∂ 2 H dpβ
(δuα )H = −ϵ + ϵ + ϵ (1.128)
∂pα ∂pβ ∂q β ∂pα ∂q β dt ∂pα ∂pβ dt

∂ 2 H ∂H ∂ 2 H ∂H
= −ϵ + ϵ (1.129)
∂pα ∂pβ ∂q β ∂pα ∂q β ∂pβ
Thus the two transformations agree after using equations of motion:
−→Do constrained dynamics

1. L −→ L − 1ϵ (ϕi (⃗q)2
2. Variational Principle
3.ϕi (⃗q, ⃗u) =non inertial normalization
Lagrangian and action formalism :
More efficient to discover symmetries & hence conservation laws.
After we discover them,how do we use them to analyze the system?
−→Go back to the Hamiltonian formalism.

2. Hamiltonian System (II)

2.1 Conserved charges: Integrability


I0 ∈ H, I1 , I2 , ....Ip (2.1)

2.1.1 How many do we need to get a solvable theory of Canonical Transformations.


Canonical Transformations:

xi −→ y i

such that

∂y i ∂y j kl
Ω = Ωij (2.2)
∂xk ∂xl

14
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

In other words, If we define

∂y i
Mik = (2.3)
∂xk
then

M ΩM T = Ω (2.4)

=⇒ Ω = M −1 Ω(M T )−1 (2.5)

∂y i ∂y j
=⇒ ω = M T ωM ⇒ ωij = ωkl (2.6)
∂xk ∂xl
Infinitesimal Canonical transformation:
{ }
xi −→ y i = xi + ϵ xi , ψ(⃗x) (2.7)

where ψ(⃗x) is the generator of the Canonical Transformation.


Now we shall look for generators of finite canonical transformation instead of the hit and
miss approach.
So far we looked for a transformation such that functional form of Hamiltonian remains
the same and check whether it is Canonical transformation or not.——-This is hit and miss
approach.
Phase Space Coordinates:
 
x1
 .. 
 . 
  where n = 2N
 .. 
 . 
xn

where,
x1 = q 1 , x2 = p1 , x3 = q 2 , x4 = p2 , · · · · · · · · ·
and
 
0 1
 
 −1 0 
Ω=  0 1
,
 (2.8)
 
..
−1 0 .
 
0 −1
 
1 0 
ω=Ω = −1   (2.9)
0 −1 
 
..
1 0 .
Consider a covariant vector Ai

15
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

   
A1 p1
    0
   
   
 A2    p2
   
    0
   
 .. =  (2.10)
 .   
   .. 
   . 
   
 ..   
 .   
..
A2N .

∂i Aj− ∂j Ai = +ωij (2.11)


Canonical Transformation xi −→ x̃i Ωij −→ Ω̃ij = Ωij and ωij −→ ω̃ij = ωij

x̃1 = q̃ 1 , x̃2 = p̃1 , ......... (2.12)


Define a covariant vector B̃i in the new coordinate system:∂˜i B̃j −∂˜j B̃i = ωij where ∂˜i = ∂
∂ x̃i
   
B̃1 p̃1
   
   0 
 .   
 ..  =  p̃2  (2.13)
   
   .. 
   . 
B̃N p̃N
Bi : Components of the vector B̃ in the {⃗x}coordinate system.

∂ x̃k
Bi = B̃k (2.14)
∂xi

∂ ∂ x̃k ∂ ∂ x̃k
∂i Bj− ∂j Bi = i
( j B̃k ) − j ( i B̃k ) (2.15)
∂x ∂x ∂x ∂x

∂ x̃k ∂ B̃k ∂ x̃k ∂ B̃k


= − (2.16)
∂xj ∂xi ∂xi ∂xj

∂ x̃k ∂ x̃l ˜ ∂ x̃k ∂ x̃l ˜


= ∂ l B̃k − ∂l B̃k (2.17)
∂xj ∂xi ∂xi ∂xj

∂ x̃k ∂ x̃l ˜
= (∂k B̃l − ∂˜l B̃k ) (2.18)
∂xi ∂xj

= ωij (2.19)

Ci = Ai − Bi (2.20)

∂i Cj − ∂j Ci = 0 (2.21)

16
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

∂F1
=⇒ Ci = (2.22)
∂xi
In any arbitrary coordinate system x̂i

∂xj ∂F1 ∂F1


Ĉi = = (2.23)
∂ x̂i ∂xj ∂ x̂i

δF = Ĉi δ x̂i = Âi δ x̂i − B̂i δ x̂i (2.24)

∂xk i ∂ x̃k i
= Ak δ x̂ − B̃k δ x̂ (2.25)
∂ x̂i ∂ x̂i

Ak δxk − B̃k δ x̃k (2.26)

δF1 = pα δq α − p̃α δ q̃ α (2.27)


F1 : Function of 2n independent variables .
Take them to be q α and q̃ α

∂F1 α ∂F1 α
δF1 = δq + α δ q̃ (2.28)
∂q α ∂ q̃

∂F1 (⃗q, ⃗q̃) ∂F1 (⃗q, ⃗q̃)


=⇒ pα = , p̃ α = − (2.29)
∂q α ∂ q̃ α
The 1st equation gives ⃗q̃as functions of ⃗q, p⃗and 2nd equation gives ⃗peas functions of ⃗q, p⃗
Given F1 (⃗q, ⃗q̃) this gives p̃α and q̃ α as functions of ⃗q, p⃗
Generates a Canonical Transformation:
This freedom can be carried not for any Ai & B̃i satisfying

∂i Aj− ∂j Ai= ωij ∂˜i B̃j − ∂˜j B̃i = ωij (2.30)

Other examples
       
A1 0 B̃1 p̃1
 ..   q1   ..   
 .     .   0 
       
     p̃ 
 = − 0
.. ..
 .  2  . = 2 (2.31)
  q     
 ..     ..   0 
 .  ..  .  ..
A2N . B̃2N .

∂i Aj− ∂j Ai= ωij ∂˜i B̃j − ∂˜j B̃i = ωij (2.32)


Generating functions of Canonical Transformation

xi −→ x̃i

17
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

(q α , pα ) −→ (q̃ α , p̃α )
If (q α , q̃α ) can be taken to be independent phase space coordinates, than there is a function
q ,⃗ ⃗
F1 (⃗q, ⃗q̃) such that,pα = ∂F∂q 1 (⃗
α
q̃)
p̃α = − ∂F∂1q̃(⃗αq,q̃)
The 1st equation gives q̃ α as functions of ⃗q, p⃗and 2nd equation gives p̃α as functions of ⃗q, p⃗
Example :

F1 = q α q̃ α (2.33)

pα = q̃ α , p̃α = −q α (2.34)
In this case (q α , q̃ α ) can certainly be taken to be independent phase space coordinate.
Starting point of the Analysis :
   
p1 p˜1
 0   0 
   
   
A= p 
 2 B̃ =  p˜ 
 2 (2.35)
 0   0 
   
.. ..
. .

∂i Aj− ∂j Ai= ωij ∂˜i B̃j − ∂˜j B̃i = ωij (2.36)


We can try to carry out the analysis with other choices of A,B satisfying these requirements.
Example :
   
0 p˜1
q   0 
 1  
   
A = − 0
  B̃ =  p˜ 
 2 (2.37)
 q2   0 
   
.. ..
. .

∂i Aj− ∂j Ai = ωij ∂˜i B̃j − ∂˜j B̃i = ωij (2.38)


Follow exactly the same steps .

Ci = Ai − Bi satisf ies (2.39)

∂F2
∂i Cj− ∂j Ci = 0 =⇒ Ci = − (2.40)
∂xi
∂F2
Ĉi = Âi − B̂i = (2.41)
∂ x̂i
in any coordinate system.

−δF2 = Ĉi δ x̂i = Âi δ x̂i − B̂i δ x̂i (2.42)

18
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

= Ai δxi − B̃i δ x̂i = −q α δpα − p̃α δ q̃ α (2.43)


Take F2 as a function ofpα and q̃ α

∂F2 ∂F2
δF2 = δpα + α δ q̃ α (2.44)
∂pα ∂ q̃

∂F2 (⃗ p, ⃗q̃) ∂F2 (⃗p, ⃗q̃)


=⇒ q α = , p̃α = +
∂pα ∂ q̃ α
1st equation gives q̃ α as functions of ⃗q, p⃗and 2nd equation gives gives p̃α as functions of ⃗q, p⃗

F2 = +pα q̃ α (2.45)

=⇒ q α = q̃ α , p̃α = pα (2.46)

F2 = +pα q̃ α − ϵψ(⃗
p, ⃗q̃) (2.47)

∂ψ ∂ψ
q α = q̃ α − ϵ , p̃α = pα − ϵ (2.48)
∂pα ∂pα

p, ⃗q̃)
∂ψ(⃗ ∂ψ(⃗
p, ⃗q)
q̃ α = q α + ϵ ≃ qα + ϵ = q α + ϵ {q α , ψ} (2.49)
∂pα ∂pα

∂ψ(⃗p, ⃗q̃) ∂ψ(⃗


p, ⃗q)
p̃α = pα − ϵ ≃ pα − ϵ = pα + ϵ {pα , ψ} (2.50)
∂pα ∂pα
−→Standard form of infinitesimal transformations.
Other choices:
   
p1 0
 0   −q̃ 
   1
   
A= p 
 2 B̃ = 

0 
 (2.51)
 0   −q̃2 
   
.. ..
. .
⃗ ⃗q)
∂F3 (p̃,
q̃ α = −→ gives q̃ α as a f unction of (⃗
p, ⃗q) (2.52)
∂ p̃α
⃗ ⃗q)
∂F3 (p̃,
pα = −→ gives peα as a f unction of (⃗
p, ⃗q) (2.53)
∂q α
   
0 0
 −q̃ 1   −q̃ 
   1
   
A=  0  B̃ =  0  (2.54)
  
 −q̃ 2   −q̃2 
   
.. ..
. .

19
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

∂F4 (⃗ ⃗
p, p̃) ∂F4 (⃗ ⃗
p,p̃)
qα = − , q̃ α = (2.55)
∂pα ∂ p̃α
Mixed choices
 
p1
 0 
 
 

A= 0  (2.56)

 −q 2 
 
..
.
( ) ( )
pα 0
In each of the N slots,we can choose or
0 −q α
=⇒2N choices for A
=⇒2N choices for B
Total number of choices 2N × 2N = 22N
The generic Canonical transformations can be represented by any of the 22N generating
functions.
Special clan of Canonical transformations may not admit 1 or more of these
2N
2 representation.
Depending on the specific problem, we may choose to use one or the other kind of generating
functions.
1

2.2 Hamilton Jacobi method for solving equations of motion


H(⃗
p, ⃗q) Original Hamiltonian.
Look for Canonical transformations

⃗ ⃗q)
∂F3 (p̃, ⃗ ⃗q)
∂F3 (p̃,
pα = , q̃ α = (2.57)
∂q α ∂ p̃α
Such that H̃(⃗pe, ⃗qe) ≡ H̃(⃗
p, ⃗q) is independent of q̃

⃗ ⃗q)
∂F3 (p̃,
=⇒ H( , qα) (2.58)
∂q α
is independent of q̃ α
⃗ q)
q̃ α dependence of H( ∂F∂q
3 (p̃,⃗
α , ⃗q) comes through q α dependence.
⃗ q)
=⇒We need H( ∂F∂q 3 (p̃,⃗
α
⃗ independent of ⃗q
, ⃗q) = H̃(p̃)
=⇒a set of partial differential equations of F3 .
=⇒Solve to get F3
Suppose we can find such an F3 .
Equation of motion in (⃗pe, ⃗qe) coordinate system:
1
Non holonomic constrains MR Flannery,Am J. Phys. Vol. 73 Page 265

20
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

dp̃α ∂ H̃
= − α = 0 =⇒ p̃α = p̃(0)
α = constant. (2.59)
dt ∂ q̃

dq̃ α ⃗
∂ H̃(p̃)
= = Constant (2.60)
dt ∂ q̃α


∂ H̃(p̃)
=⇒ q̃ α = |⃗(0) t + q̃ α(0) (2.61)
∂ p̃α p̃
Once p̃α &q̃ α are known, pα and q α are obtained by solving the set of equations

⃗ ⃗q)
∂F3 (p̃, ⃗ ⃗q)
∂F3 (p̃,
pα = , q̃ α = (2.62)
∂q α ∂ p̃α
This system has N independent conservation laws:

p̃α (1 ≤ α ≤ N ) (2.63)
H : Function of p̃α −→not an independent conserved quantity.

2.3 Action angle variable


p̃α , q̃ α q̃ α ≡ q̃ α + 2π (2.64)


H̃(p̃) (2.65)

dp̃α dq̃ α ∂ H̃
=0 = ≡ ω α (p̃)
⃗ (2.66)
dt dt ∂ p̃α

p̃α = p̃(0)
α (2.67)

q̃ α = ω α (p̃⃗(0) )t + q̃(0)
α
(2.68)

Periodic Orbit :
ω(p̃⃗(0) ) = 0 f or all α except α = β (2.69)

q̃ α = q̃(0)
α
f or α ̸= β (2.70)

β
q̃ β = ω β t + q̃(0) (2.71)

Under t → t + ω2πβ ,q̃ β ≡ q̃ β + 2π


→Same point
⇒The Motion is Periodic.
⃗ all ω α are non zero
For generic p̃,
For Periodic motion we need

21
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

⃗ = 2πnα
ω α (p̃)T f or some T (2.72)

ωα nβ
⇒ = = Rational N umber (2.73)
ωβ nβ
α
In Generic Case ωωβ is not rational
⇒The Moton is not Periodic
Example : 2-D Harmonic Oscillator
Summary
1. A Hamiltonian System with N degrees of freedom is solvable if it has N conserved
quantities.
2. Conserved quantities ←→Infinitesimal Canonical Transformations that leave the Hamil-
tonian invarient
If we find such aCanonica Transformation, we can construct the corresponding conserved
quantities & vice versa.
Sometimes Symmetries are easily visible.
Sometimes they are not →hidden symmetries.
Example:
Motion with 2 centers of attraction :

p2x p2y 1 1
H= + − k( + ) (2.74)
2m 2m r1 r2
→Solvable by Hamiltonian - Jacobi method.
⇒Two Conserved quantities:

p̃1 = C1 , p̃2 = C2 (2.75)

{C2 , H} = 0 (2.76)

⇒ δpα = ϵ {pα , C2 (⃗
p, ⃗q)} (2.77)

δq α = ϵ {q α , C2 (⃗
p, ⃗q)} (2.78)

⇒ δH = 0 (2.79)

=⇒ A new symmetry of H
Not visible →Hidden Symmetry.
Method for finding conserved quantities :
1. Use visible symmetries of H/L
2.Hamilton - Jacobi Method
3.Construction of Lax pairs.

22
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

Method

dpα ∂H dq α ∂H
= − α, = α 1≤α≤N (2.80)
dt ∂q dt ∂p
Suppose we can construct NxN matrices
Aαβ (⃗
p, ⃗q) and Bαβ (⃗
p, ⃗q) such that

Aαβ = Aβα , Bαβ = −Bβα (2.81)

with the property

dA
= BA − AB (2.82)
dt
as a consequence of equations of motion.
Then we can construct N conserved quantities.
A, B →Lax pairs.

d dA n−1 dA
(An ) = A + A An−2 + .......... (2.83)
dt dt dt


n−1
dA n−p−1
= Ap A (2.84)
dt
p=0


n−1
= Ap (BA − AB)An−p−1 (2.85)
p=0


n−1
= (Ap BAn−p − Ap+1 BAn−p−1 ) (2.86)
p=0


n−1 ∑
n−1
′ ′
= p
A BA n−p
− Ap BAn−p (2.87)
p=0 p′ =1

= (BAn − An B) (2.88)

d
T r(An ) = T r(BAn ) − T r(An B) = 0 (2.89)
dt
=⇒ T r(An ) is conserved for all n
How many of them are independent?
A: NxN symmetric matrix
Eigenvalues : λ1 ......λN
 
λ1
 ..  −1
A=S . S (2.90)
λN

23
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

An = SAα S −1 .SAα A−1 · · · · · · · · · n times (2.91)

 
(λ1 )n
  −1
= S(Aα )n S −1 = S  ..
. S (2.92)
(λN )n


N
n
=⇒ T r(A ) = (λα )n (2.93)
α=1

where the term in the bracket has N independent variables.


=⇒There are N independent functions

T r(An ) (2.94)

Take them for 1 ≤ n ≤ N =⇒N conserved charge


or we can take the conserved quantities to be λα 1≤α≤N
Example 3 body Toda lattice


m
p2 i
H= + V (⃗q) (T ake m = 1) (2.95)
2
i=1

V (⃗q) = e−(q
1 −q 2 )
+ e−(q
2 −q 3 )
+ e−(q
3 −q 1 )
(2.96)

dp1
= −e−(q −q ) + e−(q −q )
3 1 1 2
(2.97)
dt

dp2
= −e−(q −q ) − e−(q −q )
2 3 1 2
(2.98)
dt

dp3
= e−(q −q ) − e−(q −q )
3 1 2 3
(2.99)
dt

dq α
= pα (2.100)
dt

1 1 1 2 1 1 2 3 1 1 3 1 pα
a1 ≡ e− 2 (q −q ) , a2 ≡ e− 2 (q −q ) , a3 ≡ e− 2 (q −q ) , bα = (2.101)
2 2 2 2
   
b1 a1 a3 0 a1 −a3
   
A ≡  a1 b2 a2  , B =  −a1 0 a2  (2.102)
a3 a2 b3 a3 −a2 0
Example : Check that

dA
= (BA − AB) (2.103)
dt

24
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

=⇒Lax pair =⇒3 Conserved quantities

1∑
T r(A) = b1 + b2 + b3 = pα (2.104)
2
→Generates

δq α = ϵ, δpα = 0 (2.105)

 
b21 + a21 + a23 b1 a1 + a1 b2 + a2 a3 b1 a3 + a1 a2 + a3 b2
 
A2 =  a1 b1 + b2 a1 + a2 a3 a21 + b22 + a22 a1 a3 + b2 a2 + a2 b3  (2.106)
a3 b1 + a2 a1 + b3 a3 a3 a1 + a2 b2 + b3 a2 a23 + a22 + b23

=⇒ T r(A2 ) = b21 + b22 + b23 + 2(a21 + a22 + a23 ) (2.107)

1
= H (2.108)
2

T r(A3 ) = (b21 + a21 + a23 )2 + 2(a1 b1 + b2 a1 + a2 a3 )2 + 2(a3 b1 + a2 a1 + b3 a3 )2 (2.109)

+(a21 + b22 + a22 )2 + (a23 + a22 + b23 )2 + 2(a1 a3 + b2 a2 + a2 b3 )2 (2.110)


−→Generates a hidden symmetry
The System is solvable by virtue of the three conserved charges.
Introduce Routhian
Some comments on symmetries and conservation Laws :
1.The Symmetry generated by conserved quantity take simple form in the action angle
variable.
Conserved quantities Qα = p̃α
ϵβ Qβ generates

δ p̃α = {p̃α , ϵβ Qβ } = {p̃α, ϵβ p̃β } = 0 (2.111)

δ q̃α = {q̃α , ϵβ Qβ } = {q̃α, ϵβ p̃β } = ϵα (2.112)


=⇒The conserved quantities generate translation symmetry of q̃ α
2.If Q1 and Q2 are conserved and {Q1 , Q2 } = 0 then the infinitesimal symmetries generated
by Q1 and Q2 commutes .
Define :

F̃ (x̃) = F (x), δF = F̃ (x) − F (x) (2.113)

δϵ1 F = {F, ϵ1 Q1 } (2.114)

25
Classical Mechanics II: Ashoke Sen Padakshep Open Teaching Project

{ }
δϵ2 G = G, ϵ2 Q2 (2.115)
First apply δϵ1 : F −→ F + ϵ1 {F, Q1 } = G
( )
Then apply δϵ2 : G −→ G + ϵ2 {G, Q2 } = F + ϵ2 {F, Q2 } + ϵ1 {F, Q1 } + ϵ2 {{F, Q1 } , Q2 }

⇒ (1 + δϵ2 ) (1 + δϵ1 ) F = F + ϵ2 {F, Q2 } + ϵ1 {F, Q1 } + ϵ1 ϵ2 {{F, Q1 } , Q2 } (2.116)

Reverse Transformation

⇒ (1 + δϵ1 ) (1 + δϵ2 ) F = F + ϵ1 {F, Q1 } + ϵ2 {F, Q2 } + ϵ1 ϵ2 {{F, Q2 } , Q1 } (2.117)

⇒ δϵ2 δϵ1 F − δϵ1 δϵ2 F = ϵ1 ϵ2 ({{F, Q1 } , Q2 } − {{F, Q2 } , Q1 }) (2.118)


Jacobi Identity
Example :

{{A, B} , C} + {{B, C} , A} + {{C, A} , B} = 0 (2.119)

{{F, Q1 } , Q2 } − {{F, Q2 } , Q1 } = − {{Q1 , F } , Q2 } − {{F, Q2 } , Q1 } = {{Q2 , Q1 } , F } (2.120)

Thus if {Q2 , Q1 } = 0 then δϵ2 δϵ1 F − δϵ1 δϵ2 F = 0


=⇒The two symmetries commute.
[Lx , Ly ] ̸= 0can be traced to non-commtate of Rotation about X and Y axis.
   
1 0 0 cos ϕ 0 sin ϕ
   
Rx (θ) =  0 cos θ sin θ  Ry (ϕ) =  0 1 0  (2.121)
0 − sin θ cos θ sin ϕ 0 cos ϕ

Rx (θ) Ry (θ) ̸= Ry (ϕ) Rx (θ) (2.122)

26

You might also like