Lecture 17

M. Siddikov

January 2, 2020
Outline
Where we are Plan for today
Minimal action principle (revisit).
Systems with constraints
Hamiltonian formulation of classical
mechanics

For Zoom users

F In case of connection problems,
please check regularly your email for
instructions/comment
F I’ve uploaded the .PDF- file of
lecture to Aula

Symon→Marion→Fetter, Walecka, Landau

Announcements
I’ve uploaded to Aula Virtual the homework №2 (“Oscillations”, deadline: Sunday,
05/01/2020) and №3 (“Scattering”, deadline: Sunday, 12/01/2020). Please check
Aula Virtual for details.

Some students commented me that cannot download recordings of early sessions.

This is because I had to erase them due to space limitations in Zoom Cloud (1 GB,
approx. 4 lectures for Zoom Pro accounts). But I have uploaded them to Dropbox
folder (In the future, I will periodically move old (=older than one week) lectures from
ZOOM Cloud to Dropbox)

Link of the shared Dropbox folder:

https://www.dropbox.com/sh/jta8pvrsed39jo3/AADoeJTwim1xRJPxHGHsF_XRa?dl=0
Different formulations of classical mechanics
The first law of Newton Minimal action principle (Hamilton)
If the net force (the vector sum of all forces For each dynamical system there is a
acting on an object) is zero, then the velocity functional called “action”
of the object is constant. Z t2
S= L (qi , q̇k , t) dt
t1
The second law of Newton
such that real trajectories correspond
d~p d~v to the minimum of the action. The
=m ~ (~r , ~v , t)
= m ~a = F (1)
dt dt function L is called Lagrangian of the
system.
The third law of Newton 3
X m q̇a2
L=T −U = − U (qa )
All forces between two objects exist in equal 2
a=1
magnitude and opposite direction: if one
object A exerts a force FA on a second object
B, then B simultaneously exerts a force FB Euler-Lagrange equations
on A, and F ~ A = −F
~B .
 
δF ∂L d ∂L
0= = − , a = 1, ..
δqa ∂qa dt ∂ q̇a
∂L
pa ≡ = m q̇a
∂ q̇a
 In some systems variables are not in-
Minimal action principle dependent:
t2
Case of holonomic constraints: can
Z
S= L (qi , q̇k , t) dt = min
t1 express one variable in terms of others

q1 = f1 (q2 , ...qN ) ,

substitute into lagrangian and get a sys-

tem with independent degrees of free-
dom
Case of non-holonomic constraints
(e.g. contact of two rigid bodies): can-
not solve
The action is minimized by a curve
X
ci (q1 , ...qN ) q̇i = 0
 
δF ∂L d ∂L
0= = −
δqa ∂qa dt ∂ q˙a Our goal
-”Euler-Lagrange” equations, correspond to the Reformulate the EOM for the case
second law of Newton when constraints cannot be resolved
In derivation we assumed that all coordinates explicitly
qa are independent
Lagrange’s multipliers method
We’ve analyzed the minimum of the
function f (x, y ) = x y on the surface
Assume that the function f (x, y , z) has a
local minimum P on a constrained surface x 2 + y 2 = 1 and demonstrated that
method of the Lagrange’s multipliers
g (x, y , z) = c (1)
gives the same result as direct mini-
 On the constained surface mization result.

∇f (x, y , z) = const ∇g (x, y , z) , (2)

The constaine minimum of a function

f (x, y , z) coincides with minimum of function

L(x, y , z) = f (x, y , z) − λ g (x, y , z)

where λ is unknown coefficient (Lagrange’s

multiplier), and we should fix it solving
together (1) and (2)
Lagrange’s multipliers method
Assume that index (a) numerates all possile Case of non-holonomic constraints
constraints given by
Variation of the action: X (a)
Z    ci (q1 , ...qN ) q̇i = 0
∂L d ∂L
δS = dt − δqa
∂qa dt ∂ q̇a Recall that q̇i ≡ δqi /dt, so
Existence of constraints⇒some variations X (a)
ci (q1 , ...qN ) δqi = 0
δqi not independent!
Case of holonomic constraints given by ⇒ instead of (2) add constraints to δL,
ga (q1 , ..., qN ) = 0: P (a)
i.e. δL → δ L̃ = δL + λa δqi ci (q1 )
 we should minimize
  X
Z ∂L d ∂L (a)
S̃ = dt L̃ (1) ⇒ − = λa ci (q1 )
∂qa dt ∂ q̇a
X
L̃ = L + λa ga (q1 , ..., qN ) , (2)
a

where λa are unknown (free) coefficients which

should be fixed solving Euler-Lagrange equa-
tions together with constraints
Lagrange’s multipliers method The particle of mass m slides without
Case of holonomic constraints friction over the internal surface of the
ga (q1 , ..., qN ) = 0 cone, as shown in the Figure. The axis
Z of the cone is vertical, and the
S → S̃ = dt L̃ gravitational field is directed
downwards. The apex half-angle of the
cone is α. Integrate the equations of
X
L̃ = L + λa ga (q1 , ..., qN )
a motion using the method of Lagrange
multipliers.
where λa are unknown (free) coefficients
which should be fixed solving Euler-Lagrange
equations together with constraints
Case of non-holonomic constraints:
  X
∂L d ∂L (a)
− = λa ci (q1 )
∂qi dt ∂ q̇i
 Method of Lagrange’s multipliers:
g (r , z, φ) = r − z tan α = 0
L̃ = L + λ (r − z tan α)
m 2 
L= ṙ + r 2 φ̇2 + ż 2 − m g z
2 m
r̈ = m r φ̇2 + λ (2)
Direct method (simple constaraint, can solve 2
directly): m
z̈ = −mg − λ tan α (3)
2
1 1
z → r / tan α, 1+ = r = z tan α ⇒ z̈ = r̈ cot α
tan2 α sin2 α
(2) + (3) × cot α = (1) as expected
 equation for φ̇ is the same in both
ṙ 2
 
L=
m
+ r 2 φ̇2 − m g r cot α approaches
2 sin2 α
⇒Both methods give equivalent
Equations of motion:
equations of motion
m
r̈ = −mg cot α + m r φ̇2 (1)
2 sin2 α
 
d m r 2 φ̇ dMz
= =0
dt dt
More rapid method: use integrals of motion
Mz = mr 2 φ̇, E = T + U
Hamiltonian formulation (intro)

Equation of motion for a single particle in Lagrangian formulation (2nd law of

Newton):
~p˙ = F
~ (q, q̇, t) → ~p˙ = F
~ (q, p, t), ~p = m ~q˙
2nd order differential equation with respect to variable q(t)
Can consider as a pair of 2 1st order differential equation with respect to
variables (p(t), q(t)).
In general can describe evolution in terms of first-order system of equations
(pa , qa ), where pa are momenta and qa are coordinates
 Trivial for a single particle, but not trivial in more general case
Allows to study much wider class of transformations which mix coordinates and
momenta
Serves as a base for powerful Hamilton-Jacobi formalism (we’ll see later)
In quantum mechanics You’ll learn links of classical hamiltonian approach with
operator formulation.
Below we’ll derive the approach more formally
Hamiltonian formulation
Alternative to Lagrangian formulation Hamiltonian H - related by Legendre
Describes evolution in terms of first-order transform to Lagrangian:
system of equations (pa , qa ), where pa are mo- X
menta and qa are coordinates H≡ pa q̇a − L

L = L (q, q̇)
  Hamilton’s (Canonical) equations
X ∂L ∂L
dL = dqa + d q̇a
a
∂qa ∂ q̇a ∂H (p, q)
ṗa = − ,
∂L ∂L ∂qa
pa = , ṗa =
∂ q̇a ∂qa ∂H (p, q)
q̇a = .
⇒ dL =
X
(ṗa dqa + pa d q̇a ) ∂pa
a

Control question
X X X
pa d q̇a = d (pa q̇a ) − q̇a dpa
a a a Evaluate the hamiltonian for the
harmonic oscillator and demonstrate
X X
dH = − ṗa dqa + q̇a dpa (1)
a a that canonical equations are equivlent
to Euler-Lagrange equations
Hamiltonian formulation
Alternative to Lagrangian formulation Hamiltonian H - related by Legendre
Describes evolution in terms of first-order transform to Lagrangian:
system of equations (pa , qa ), where pa are mo- X
menta and qa are coordinates H≡ pa q̇a − L

L = L (q, q̇)
  Hamilton’s (Canonical) equations
X ∂L ∂L
dL = dqa + d q̇a
a
∂qa ∂ q̇a ∂H (p, q)
ṗa = − ,
∂L ∂L ∂qa
pa = , ṗa =
∂ q̇a ∂qa ∂H (p, q)
q̇a = .
⇒ dL =
X
(ṗa dqa + pa d q̇a ) ∂pa
a

Example: Harmonic oscillator

X X X
pa d q̇a = d (pa q̇a ) − q̇a dpa
a a a

dH = −
X
ṗa dqa +
X
q̇a dpa (1) p2 k q2
H= +
a a 2m 2
ṗ = −k q q̇ = p/m

in agreement with Euler-Lagrange

equations
Routhian of the system

Hamiltonian H - related by Legendre transform In Hamilton’s approach we should re-

to Lagrangian: place each pair (qa , q̇a ) → (qa , pa )
X Note that we can make transition only
H≡ pa q̇a − L over some of the variables
In this way we get Routhian
R (q1 , p1 , qs , ps , qs+1 , q̇s+1 , ...) -
Hamilton’s (Canonical) equations hybrid between Lagrangian and
Hamiltonian approaches
ṗa = −
∂H (p, q)
, Should write Canonical equations
∂qa over variables 1...s
∂H (p, q) Should write Euler-Lagrange equa-
q̇a = . tions over variables s + 1...N
∂pa
In some problems might be more con-
venient for analysis than lagrangian
 X  ∂L ∂L

Formulation of Routh dL =
∂qa
dqa +
∂ q̇a
d q̇a
In some problems we may benefit from ex- a

cluding some (e.g. cyclical) variables from Leg-


X ∂L 
∂L ˙
+ d ξa + d ξa
endre trasformation: ∂ξa ∂ ξ˙a
a
Routhian R (q1 , p1 , qs , ps , qs+1 , q̇s+1 ) -
hybrid between Lagrangian and Hamiltonian ∂L ∂L
pa = , ṗa =
approaches ∂ q̇a ∂qa
s
X X
R= pa q̇a − L, s<N ⇒ dL = (ṗa dqa + pa d q̇a )
a=1 a

Canonical equations over variables

X X X
pa d q̇a = d (pa q̇a ) − q̇a dpa
(qi , pi ) , i ≤ s; Euler-Lagrange equations for a a a
i > s:
X X
∂R ∂R dR = − ṗa dqa + q̇a dpa (1)
ṗi = − , q̇i = , i ≤ s, a a
∂qi ∂pi
  X  ∂R ∂R ˙

∂R d ∂R + d ξa + d ξa
− = 0, i >s ∂ξa ∂ ξ˙a
∂qi dt ∂ q̇i a

Let’s use q, q̇ for variables included in Legen- ∂L

≡−
∂R
,
∂L
≡−
∂R
dre transformation, and ξ, ξ˙ for variables not ∂ξa ∂ξa ∂ ξ˙a ∂ ξ˙a
included  
L = L q, q̇, ξ, ξ˙
Formulation of Routh
In some problems we may benefit from ex- Example: motion in the central field
cluding some (e.g. cyclical) variables from Leg- (θ = π/2 plane, 2 degrees of freedom):
endre trasformation:
m 2 
Routhian R (q1 , p1 , qs , ps , qs+1 , q̇s+1 ) - L= ṙ + r 2 φ̇2 − U(r )
2
hybrid between Lagrangian and Hamiltonian
approchaes Hamiltonian (2 degrees of freedom):
s
!
X pr2 pφ2
R= pa q̇a − L, s<N H= + + U(r )
a=1
2m 2m r 2

Canonical equations over variables Routhian:

(qi , pi ) , i ≤ s; Euler-Lagrange equations for
i > s: pφ2 mṙ 2
R= + U(r ) −
|2m r {z 2
2
∂R ∂R
ṗi = − i ≤ s,
}
, q̇i = , Ueff (r )
∂qi ∂pi
∂R d

∂R
 -excluded cyclic variable φ, depends
− = 0, i >s only on pφ = const
∂qi dt ∂ q̇i

Below we’ll focus on Hamiltonian only

 Maupertui’s principle
Hamiltonian formulation
If we restrict variations in configuration
∂L ∂L space to trajectories with constant
pa = , ṗa =
∂ q̇a ∂qa energy (H = E = const)
X
H≡ pa q̇a − L
Z X Z
S̃ = dt pa q̇a = 2 dt T = min
a
Hamilton’s (Canonical) equations
Single particle case:
2 √
1 d~r
 
∂H (p, q) T = ⇒ dt = d |r |/ 2T
ṗa = − ,. 2 dt
∂qa
∂H (p, q) Z √ Z
q̇a = .
p
∂pa S̃ = d |r | T = d |r | E − U(~r )

Modified Hamilton’s principle Fermat’s principle

Canonical equations might be obtained For free particles T = E = const, so
from condition that functional particle follows the path of the
Z ! shortest time
X Z t2
S̃ = dt pa q̇a − H(pa , qa , t)
a
S̃ = 2E dt = 2E (t2 − t1 ) = min
t1
achieves minimum on the trajectory
Hamiltonian formulation
Hamilton’s (Canonical) equations
Alternative to Lagrangian formulation
Describes evolution in terms of first-order ∂H (p, q)
ṗa = − ,
system of equations (pa , qa ), where pa are mo- ∂qa
menta and qa are coordinates q̇a =
∂H (p, q)
.
∂pa
L = L (q, q̇)

Control question

X ∂L 
∂L
dL = dqa + d q̇a
a
∂qa ∂ q̇a Demonstrate that
∂L ∂L dH(p, q, t) ∂H ∂L
pa = , ṗa = = =−
∂ q̇a ∂qa dt ∂t ∂t
Hamiltonian H - related by Legendre transform
to Lagrangian:
X
H≡ pa q̇a − L
Hamiltonian formulation
Derivative of any function f = f (p, q, t)

Prove the following properties of the

 
df ∂f X ∂f ∂f
= + q̇a + ṗa =
dt ∂t ∂qa ∂pa Poisson bracket (below ci = const,
a
  fi , g are some functions, qa are
∂f ∂H ∂f ∂H ∂f
coordinates, pa are momenta)
X
= + − =
∂t a
∂pa ∂qa ∂qa ∂pa
∂f [f , g ] = − [g , f ]
= + [H, f ]
∂t
[f , c] = 0
Poisson bracket [H, f ] [c1 f1 + c2 f2 , g ] = c1 [f1 , g ] + c2 [f2 , g ]

X  ∂H ∂f  [f1 f2 , g ] = [f1 , g ] f2 + f1 [f2 , g ]

∂H ∂f
[H, f ] = − ∂f ∂f
a
∂pa ∂qa ∂qa ∂pa [f , qa ] = , [f , pa ] = − ,
∂pa ∂qa
In some books notation {H, f } is used [qa , qb ] = 0, [pa , pb ] = 0,
If the quantity f is integral of motion [pa , qb ] = δab ,
and does not depend explicitly on time, then
[[f , g ] , h]+[[h, f ] , g ]+[[g , h] , f ] = 0
[H, f ] = 0
Hamiltonian formulation

[f , g ] = − [g , f ]
Poisson’s theorem
Prove that if f , g are integrals of
[f , c] = 0
motion, then h = [f , g ] is also an
[c1 f1 + c2 f2 , g ] = c1 [f1 , g ] + c2 [f2 , g ] integral of motion
[f1 f2 , g ] = [f1 , g ] f2 + f1 [f2 , g ]
Obviously number of integrals of mo-
∂f ∂f tion is finite, so after iterative applica-
[f , qa ] = , [f , pa ] = − ,
∂pa ∂qa tion of procedure, h might vanish or re-
[qa , qb ] = 0, [pa , pb ] = 0, duce to linear combination of known in-
[pa , qb ] = δab , tegrals

[[f , g ] , h] + [[h, f ] , g ] + [[g h] , f ] = 0

Hamiltonian formulation Case f , g do not depend on time:
[f , g ] = − [g , f ]  

[[f , g ] , H] + [H, f ], g  +
 
[f , c] = 0 | {z }
=0
[c1 f1 + c2 f2 , g ] = c1 [f1 , g ] + c2 [f2 , g ]  
[f1 f2 , g ] = [f1 , g ] f2 + f1 [f2 , g ]
+ [g H], f  = 0
 
∂f ∂f | {z }
[f , qa ] = , [f , pa ] = − , =0
∂pa ∂qa
[qa , qb ] = 0, [pa , pb ] = 0, ⇒ h = [f , g ]
[pa , qb ] = δab , Case ∂t f , ∂t g 6= 0:

[[f , g ] , h] + [[h, f ] , g ] + [[g h] , f ] = 0

   
d ∂f ∂g
[f , g ] = , g + f, +
dt ∂t ∂t
Poisson’s theorem
   
∂f ∂g
+ [H, [f , g ]] = , g + f,
Prove that if f , g are integrals of motion, ∂t ∂t
then h = [f , g ] is also an integral of motion + [[H, f ] , g ] + [f , [H, g ]] =
   
df dg
= , g + f, =0
dt dt
Hamiltonian formulation

[f , g ] = − [g , f ]
Assuming that qa are generalized
coordinates, pa are corresponding
[f , c] = 0 momenta, and Ma are components of
[c1 f1 + c2 f2 , g ] = c1 [f1 , g ] + c2 [f2 , g ] the angular momentum, evaluate the
[f1 f2 , g ] = [f1 , g ] f2 + f1 [f2 , g ] Poisson brackets
∂f ∂f [Ma , pb ], [Ma , qb ], [Ma , Mb ]
[f , qa ] = , [f , pa ] = − ,
∂pa ∂qa
[qa , qb ] = 0, [pa , pb ] = 0,
The particle moves in the central field
[pa , qb ] = δab , U(r ). Demonstrate that all
[[f , g ] , h] + [[h, f ] , g ] + [[g h] , f ] = 0 componentsMa are integrals of
motions