SPA5304 Physical Dynamics Lecture 26-27

David Vegh

13 March 2019

1 Hamiltonian mechanics

1.1 Example: Particle in a potential V (~r)

m ˙2
L= ~r − V (~r)
2
∂L
p~ = = m~r˙
∂~r˙
We need to express ~r˙ = ~r˙ (~r, p~, t). We get
p~
~r˙ =
m
 2 !
p~ 1 p~ p~2
H = p~ · ~r˙ − L = p~ · − m − V (~r) = + V (~r)
m 2 m 2m

Hamilton’s equations: 
 ~r˙ =
∂H p
~
 ∂~
p = m

 p~˙ = − ∂H = − ∂V

∂~
r ∂~
r

In particular, for the free particle (V = 0), ~r is cyclic ⇒ p~˙ = 0 ⇒ p~ is constant.

1.2 Example: Particle in a central potential in polar coordinates

m 2
L= (ṙ + r2 φ̇2 ) − V (~r)
2
∂L ∂L
pr = = mṙ pφ = = mr2 φ̇
∂ ṙ ∂ φ̇
  p 2 
pr pφ m  p r 2 2 φ
H = pr ṙ + pφ φ̇ − L = pr + pφ − +r + V (r)
m mr2 2 m mr2

p2r p2φ
H= + + V (r)
2m 2mr2

Hamilton’s equations for r:

pr
 ∂H

 ṙ = ∂pr = m




 " #
p2φ p2φ
ṗr = − ∂H
∂r = mr 3 − V 0 (r) = − ∂r
∂
V (r) +
2mr2





 | {z }

Veff (r)

1
 and for φ:

pφ

∂H
 φ̇ =
 ∂pφ = mr 2

 ṗ = − ∂H = 0

⇒ pφ = mr2 φ̇ = const.
φ ∂φ

1.3 Time evolution in Hamiltonian mechanics

A = A(~q, p~, t) : Some physical quantity of the system.
How does it evolve?
dA ∂A ∂A ∂A
Ȧ = = q̇i + ṗi +
dt ∂qi ∂pi ∂t
Using Hamilton’s equations this is
∂A ∂H ∂A ∂H ∂A
Ȧ = − +
∂qi ∂pi ∂pi ∂qi ∂t

If we define the Poisson bracket by

∂X ∂Y ∂X ∂Y
{X, Y } ≡ −
∂qi ∂pi ∂pi ∂qi

then
∂A
Ȧ = {A, H} +
∂t

2
1.4 Aside: Poisson brackets and quantum mechanics
There is a special {, }, namely:
∂x ∂p ∂x ∂p
{x, p} = − =1
∂x ∂p ∂p ∂x
The canonical commutation relation of quantum mechanics is

[x̂, p̂] ≡ x̂p̂ − p̂x̂ = i~

(In the basis where the operator x̂ means multiplying the wavefunction by x, while p̂ is nothing but the
∂
derivative: p̂ = −i~ ∂x .)
The similarity between the Poisson bracket and the commutator suggests that we can go from classical
mechanics to quantum mechanics by replacing
1
{X, Y } → [X̂, Ŷ ]
i~
where the hatted quantities are operators.
Using this “dictionary” (whenever it works), the time evolution of a quantum mechanical operator Â
should be
˙ 1 ∂ Â
 = [X̂, Ĥ] +
i~ ∂t
This is the so-called Heisenberg equation in quantum mechanics.

1.5 Properties of Poisson brackets

(i)
{A, B} = −{B, A}

(ii)
{A, A} = 0

(iii)
{A, c} = 0
if c =const.

(iv)
{A + B, C} = {A, C} + {B, C}

(v)
{A, BC} = B{A, C} + {A, B}C

(vi) The Jacobi identity

{A, {B, C}} + {B, {C, A}} + {C, {A, B}} = 0

3
1.6 A consequence of the Jacobi identity
If I˙1 = I˙2 = 0, i.e. they are both conserved, then {I1 , I2 } is also conserved (Poisson’s theorem). This way
we can generate a new conserved quantity (unless it is just a combination of the old I1 and I2 ).
Proof:
d ∂
{I1 , I2 } = {{I1 , I2 }, H} + {I1 , I2 }
dt ∂t
using the Jacobi identity we get
∂I1 ∂I2
= −{{I2 , H}, I1 } − {{H, I1 }, I2 } + {
, I2 } + {I1 , }
∂t ∂t
   
∂I2 ∂I1
= −{I2 , H} − , I1 + −{H, I1 } + , I2 = 0
∂t ∂t
Q.E.D.

1.7 A charged particle in an electromagnetic field

The Lagrangian of a particle of mass m and charge e in an external electromagnetic field is

1 ˙2
L= m~r − V (~r, ~r˙ )
2

with

V (~r, ~r˙ ) = eφ − e~r˙ · A

~

~ r, t) are the scalar and vector potentials related to electromagnetic fields E

where φ(~r, t) and A(~ ~ and B
~ via

~ = −∂t A
E ~ − ∇φ
~

~ =∇
B ~ ×A
~

(This is our first encounter with a velocity-dependent potential. In this case, force is not given by − ∂V
∂q ; see
Goldstein §1.5)
We prove the above by showing that the Euler-Lagrange equation derived from the above Lagrangian is
the familiar

mẍi = eEi + e(~r˙ × B)

~ i

where i = 1, 2, 3. Note that we have set c = 1.

4
 Proof:

1 ˙2
L= m~r − eφ + e~r˙ · A
~
2
∂L
pi = = mẋi + eAi
∂ ẋi
∂L
= −e∇i φ + eẋj ∇i Aj
∂xi
d
ṗi = mẍi + e Ai (~r, t) = mẍi + e(∂t Ai + ẋj ∇j Ai )
dt
∂L
So the Euler-Lagrange equations (ṗi = ∂xi ) give

mẍi + e(∂t Ai + ẋj ∇j Ai ) = −e∇i φ + eẋj ∇i Aj

mẍi = e (−∂t Ai − ∇i φ) +eẋj (∇i Aj − ∇j Ai )

| {z }
Ei

But
(~r˙ × B)
~ i = (~r˙ × (∇
~ × A))
~ i = ijk ẋj (∇
~ × A)
~ k = ijk ẋj klm ∇l Am

Using the identity: kij klm = δil δjm − δim δjl we get

(δil δjm − δim δjl )ẋj ∇l Am = ẋj (∇i Aj − ∇j Ai )

Therefore,
mẍi = eEi + e(~r˙ × B)
~ i

Q.E.D.

1.8 Remark 1: Lorentz invariance

The interaction term Z Z
Sint = − dt V = − dt(eφ − e~r˙ · A)
~

can be written as
Z Z
Sint = − d x(ρφ − J~ · A)
4 ~ = d4 x J µ Aµ

where
~
Aµ = (−φ, A)

~
J µ = (ρ, J)

These are four-vectors of relativity. ρ is the charge density and J~ is the current density.

5
 In the above point-particle case:

ρ = eδ (3) (~r − ~r(t))

J~ = e~r˙ δ (3) (~r − ~r(t))

where the particle of charge e is moving along the trajectory ~r = ~r(t).

J µ Aµ is a scalar (invariant under Lorentz transformations), so the action is also Lorentz invariant.

1.9 Remark 2: “gauge invariance”

We have previously seen that the Euler-Lagrange equations are unchanged if the Lagrangian is changed by
a total derivative.
We have
1 ˙2
L= m~r + Lint
2
Lint = −eφ + e~r˙ · A
~

Let us consider the transformation

Aµ → Aµ + ∂µ Λ
~ and µ = 0, 1, 2, 3. Here Λ = Λ(~r, t) is an arbitrary function. The
Here we used the notation ∂µ ≡ (∂t , ∇)
above transformation means
φ → φ − ∂t Λ
and
~→A
A ~ + ∇Λ
~

It is a gauge transformation.

~ = Lint + e d Λ(~r(t), t)
Lint → L0int = Lint + e∂t Λ + e~r˙ · ∇Λ
dt
The second term is a total derivative and thus it will not affect the Euler-Lagrange equations. Hence, the
transformation is a symmetry of the theory1 and this is called gauge invariance.
• One can understand this also by noting that the physical fields E, ~ B~ do not change under gauge
transformations:
~ = −∂t A
E ~ − ∇φ
~ → ~ + ∇Λ)
−∂t (A ~ −∇~ (φ − ∂t Λ) = E
~
| {z } | {z }
~0
A φ0

and
~ =∇
B ~ ×A
~ → ~ × (A
∇ ~ + ∇Λ)
~ ~ ×A
=∇ ~+∇
~ × ∇Λ
~ =B
~
| {z }
0

1 It really is not a physical symmetry, but a redundancy in the description of the physical degrees of freedom.

6
 ~ that give the same E,
• There are infinitely many different ways to choose (φ, A) ~ B.
~ This arbitrariness
~
can be removed by “fixing the gauge”, i.e. by imposing some conditions to be satisfied by φ and A.
For instance the Coulomb (or radiation) gauge is:

~ ·A
∇ ~=0

Can we do further gauge transformations while staying in this gauge?

~ ·A
0=∇ ~0 = ∇
~ · (A
~ + ∇Λ)
~ = ∇2 Λ = 0

If we require that Λ → 0 as |~r| → ∞, the only solution is Λ = 0. Thus, the Coulomb gauge choice completely
fixes the gauge.

1.10 Hamiltonian of a charged particle

1 ˙2
L= m~r − eφ + e~r˙ · A
~
2
∂L 1
pi = = mẋi + eAi ⇒ ẋi = (pi − eAi )
∂ ẋi m
 
1 1 1 1
H = pi ẋi − L = pi (pi − eAi ) − m 2 (pi − eAi )2 − eφ + e (pi − eAi )Ai
m 2 m m
1 1
= (pi − eAi ) (pi − eAi ) − (pi − eAi )2 + eφ
m 2m

1 ~ 2 + eφ
H= p − eA)
(~
2m

• The Hamiltonian of a charged particle is obtained by

(i) Perform “minimal substitution”:

~
p~ → p~ − eA

(ii) Add the potentia lterm, eφ

7
1.11 A peek at quantum mechanics
What is the Schroedinger equation for a particle of mass m and charge e in an electromagnetic field?
1 ~ 2 + eφ
Hclassical = p − eA)
(~
2m
~ So
In quantum mechanics, p~ → −i~∇.

~ 2 → (−i~∇
p − eA)
(~ ~ − eA)
~ 2 = −~2 ∇
~ 2 + e2 A
~ 2 + i~e(A
~ ·∇
~ +∇
~ · A)
~ = −~2 ∇
~ 2 + e2 A
~ 2 + 2i~eA
~ ·∇
~ + i~e(∇
~ · A)
~

~ ·A
If we go to Coulomb gauge, then ∇ ~ = 0. So in this gauge the Schroedinger equation is

~2 ~ 2 e2 ~ 2 i~ ~ ~
 
− ∇ + A + A · ∇ + eφ Ψ = EΨ
2m 2m m