1 Lagrangian for a continuous system

Lets start with an example from mechanics to get the big idea. The physical
system of interest is a string of length and mass per unit length fixed at
both ends, and under tension Choose axis along the unperturbed string,
and axis perpendicular to it. When the string is vibrating, its kinetic energy
is:
Z Z 2 Z
1 2 1 1
= = = ()2
0 2 0 2 0 2
To get the potential energy, we use the method of virtual work. The net force
on a string segment has components:

= cos 1 cos 2 ' 0

and

!

= sin 1 sin 2 tan 1 tan 2 =
+
2
=
2
Then the virtual work is
Z Z
2
= =
0 0 2

Now integrate by parts, and make use of the fixed end condition:
" Z # Z
0 0 0 h 0 2i
= |0 ( ) = ( )
0 2 0
R
Then if =
then = and = Here
Z Z
2
= = ( 0 )
2 0

The Lagrangian for the string is:

Z
1h 2 2
i
= = () ( 0 )
0 2

where
1h 2 2
i
L= () ( 0 ) (1)
2
is the Lagrangian density for the string.

1
 The action is
Z Z Z
2
1h 2 2
i
= = () ( 0 )
1 0 2
Taking the variation of the action, we get
Z 2 Z
L L 0 L
= + 0 +
1 0
Integrating by parts gives:
Z 2 Z
L L L
= +
1 0 0
Thus for the action to be an extremum, we need
L L L
0
+ =0 (2)

Using equation (1), we find:

(2) + (2 0 ) 0 = 0

or

00 = 0
which is the wave equation for the string.
An alternative approach is to write the string displacement as a sum over
normal modes: X
= () sin


Then the Lagrangian density (1) is
XX 2
L= sin sin 2 cos cos

and then the Lagrangian is Z

= L

When we integrate over the only terms that survive are those with =

X 2 2
= ( ) 2 2 2 (3)
2

The mode amplitudes act as the generalized coordinates for the string. Then
Lagranges equations are
2
= 2 2 = 0

p
which is the harmonic oscillator equation with frequency = ()

2
2 Lagrangian for the electromagnetic field
Now we want to do a similar treatment for the EM field. We want a Lagrangian
density such that the action
Z
= L4

is a Lorentz invariant, and where L is a function of the fields. The "obvious"

invariant to try is
Lguess =
(Recall this is proportional to 2 2 an "energy-like" thing.) Here the com-
ponents of the potential are the "normal modes" they behave like the
in the previous section. Then Lagranges equations (2) are:
L L
=0 (4)

To evaluate this, note that

Lguess =

and so
Lguess
=





+





( )

= ( ) + ( )
= ( ) ( ) + ( ) ( )
= 4 ( ) = 4

while
Lguess
=0

So equations (4) become
= 0
which are Maxwells equations in the absence of sources. We can fix up the
Lagrangian by adding the interaction term 1 Thus
1 1
L=
16
With this Lagrangian density
L 1
=

and Lagranges equations become
1 1
+ = 0
4

3
or
4
=

which are the two Maxwell equations that include sources.

3 The Hamiltonian
Now we form the Hamiltonian. First lets look at the string. Using equation
(3):
X L
= L



X 2 2 2
= 2 ( ) ( ) 2 2 2
2
X 2 X
= 2 ( )2 + 2 2 2 =
2

where is the total (kinetic plus potential) energy per mode. By analogy, we
get for the EM field system without sources
L
= L
( )

1 1
=
4 16

1 1
= +
4 4

This tensor is not symmetric, because the first term contains only one half of
the field tensor: rather than The conservation laws require that the
energy tensor be symmetric, so we have to modify the result.

4 The energy-momentum tensor

1
2
Recall that the field energy density (non-relativistic) is 8 + 2 and the
Poynting theorem may be written
1 2

+
=0
+ 2 + (5)
8 4

4
Wed like to express this result in covariant form. We obviously need something
quadratic in the fields. For example:

0 0
0
= 0
0 0
0 0
2
+ 2 + 2
2 + 2 + 2
= 2 + 2 + 2


2 + 2 + 2

2



2
+ 2
+ 2


=

2 + 2 + 2

2 + 2 + 2

Now wed like the (0,0) component to be the energy density. We can get that
if we add the tensor 14 = 12 2 2 Then

1 1
= +
4 4
2 2

2 +



2

2 2
+ + 2 2 2 2

1



= 2 2
4 2 2 2
+ + 2


2 2

2 + 2 + 2
2

2 + 2



2

2 + 2
2
2

1
2

= 2 2
4 +
2
2



2
2 2

+
2
2

2

Then !
2 + 2

0
= +
8 4
= 1 0
which is part of equation (5). On the right hand side we need 1
Thus we have the relation
1
0 = 0

5
and so we guess that the full set of conservation laws are given by:
1
=

I leave it to you to show that the = components give momentum conservation
(Jackson equation 6.122).

5 Angular momentum
Cross products are not proper vectors. They are pseudo-vectors because they
do not transform properly under reflections. Thus it is usually better to express
quantities such as angular momentum of a particle ( = ) as antisymmetric
tensors. For example the tensor
X
= ( )
particles

has three independent compomemts: the components of the vector
Extending this idea, lets look at the tensor
X
=
particles

The 33 spacelike part is the tensor and thus represents the angular
momentum of the system. In addition:
X
0 =

where is the energy of the particle. Conservation of angular momentum for
the system is expressed as = constant. Thus we conjecture that the full
conservation law is = constant. (Or equivalently = 0) This gives
for the ( 0) component:
X
0 = = constant

P
Now if we divide through by we get
P P

P = 2 P

The term on the left hand side is the position of the center of mass,
P

= P

while the term on the right hand side is the CM velocity times
P

= P

6
an eminently sensible result.
To get the equivalent result for the EM field we form the tensor

and then the conservation laws should be given by:

= 0
Taking = 0 gives the CM motion as above.

6 The Darwin Lagrangian

The analysis above is for source-free fields. We might attempt to add the free-
particle Lagrangian to get a complete description of the particle-plus-field sys-
tem, but this approach fails because of retardation eects. (The fields propagate
at the speed of light.) We can calculate a complete Lagrangian in a single ref-
erence frame, inlcuding relativistic eects up to order 2 = ()2
Lets start with a 2-particle system. Both particles produce fields and both
can move under the influence of those fields. The interaction term for charge 1
interacting with the fields due to 2 is

1 1 1
1
2 = (
1 ) 2 2 = 1 2 2 (6)

If we now work in a single reference frame and use the coordinate time rather
than proper time as our time variable, we should drop the factor We want to
evaluate this expression to second order in If we work in Coulomb gauge,
to first order since it appears
the potential 2 = 2 is exact. We only need
in combination with 1 This means we can ignore retardation eects. Then:
Z
2 ' 1
0
| 0 |

where the transverse current is

Z 0
1 22 (0 2 ) 0
= = 22 ( 2 )
4 | 0 |
2 2 ( 2 )
= 22 ( 2 )
4 | 2 |3

Therefore Z 0
= 2 2 2

1 0 2 ( 2 ) 0

4 | 0 | 3
|0 2 |

7
Lets look at the integral. First make a change of origin. Let = 0 2 and
with = 2 we get
Z 0 Z
1 0 2 ( 2 ) 0 = 1 2 3
0
3 3
| | 0
| 2 | | |
Z
1 2 1 2 3
=
| | 3 at 3
Z
1 2 3
=
| | 2
Z X

= () 2
=0
+1
where we have put the polar axis for along Now
2 = 2 ( cos + sin ( cos + sin ))
Integration over renders the and components zero.
Next we make use of the orthogonality of the () noting that cos =
1 () Only = 1 survives the integration over We obtain:
Z
2
integral = 2 2 32
0
Z Z
4
= 2 +
3 2 2
0

4 1
= 2 + 1 = 2 2
3 2
And thus

2 2 2 2
= 2 2
4

2 2 1 2 2
= 2
2
2
= [2 + (2 ) ]
2
Then the interaction term for 2 particles (equation 6 with the dropped) is:
!
2
1 2 1 2
1 2 = 1 [2 + (2 ) ]
2

2 1
= 1 1 2 [1 2 + (2 ) (1 )]
2
Adding this term to the kinetic energy (Lagrangian notes pg 5), we have the
Darwin Lagrangian for a collection of charged particles:
r
1X 2 2 X 1
= 1 2 1 2 [ + ( ) ( )]
2
2

8
To be consistent, we should evaluate the first term to second order in i.e.
12 2 4
1 2 2 ' 1 12 2 + 12 12 2 4 Finally, dropping the constant leading

term which is irrelevant, we have

1X 1 X X 1

2 4
= + 2 1 2 [ + ( ) ( )]
2 8
2

correct to second order in