Last Latexed: April 6, 2010 at 10:40

Physics 504, Lecture 18

April 5, 2010

504: Lecture 18

Last Latexed: April 6, 2010 at 10:40

~ A
~ is zero. Working accurate to order (v/c)2 we may drop the
The
~ is already order (v/c)1 . Thus we may take
term, as A

c
Copyright 2010
by Joel A. Shapiro

~=
2 A

Darwin and Proca Lagrangians

Newtonian classical mechanics is based on the idea that motion is described

by the positions ~xj of particles at a given time t, and these evolve according to
forces given by the positions of all other particles ~xk (t) at that instant. When
these forces are conservative we may specify the forces by potential energies
which are again functions of the several positions ~xj (t) at that instant. This
notion violates the constraint of special relativity that information cannot
travel faster than light, so there is no way the particle at ~xj (t) can feel a force
that depends on where particle k is at this instant, unless ~xj (t) = ~xk (t). So
the only potential consistent with relativity is a delta function!
Nonetheless we know that Lagrangians with potential energies are a very
effective way of describing physics if the relevant velocities remain small
compared to c.
Let us see what we can do for charged particles interacting electromagnetically. We learned as freshmen how to do the lowest order (c ):
P
V (~xj , ~xk ) = qj qk /|~xj ~xk | and T = 12 mj ~vj2 . This encapsulates the effect
of the electric field produced by one charge on the motion of the other. In our
relativistic treatment
the interaction lagrangian (of charges with the fields)

P
~ xj ) , and (~xj ) = Pk qk /|~xj ~xk | is the
Lint = j qj (~xj ) + 1c ~uj A(~
c limit for the scalar potential. Magnetic forces are only produced to
next order in v/c, and these produce forces only proportional to the velocity
of the second particle, so these only enter to order v 2 /c2 . At that order, the
~ will depend on the choice of gauge, and it is useful
expressions for and A
~ A
~ = 0, because
here to use not the Lorenz gauge but the Coulomb gauge
R 3 0
1
2
0
in that gauge = 4, and (~r, t) = d r (~r , t)/|~r ~r 0 | really is
instantaneous. From F j = 4J j /c we have
!

1 2 ~
~ A
~ = 4 J/c.
~+
~ 1 +
~
A 2 A
2
2
c t
c t
Gaussian units.

1 2 ~
A
c2 t2

4 ~ 1 ~
J + .
c
c t

~ x 0 ) is qj ~vj 3 (~x 0 ~xj ). Its contribution

The contribution of particle j to J(~
0
0
0
to (~x ) is qj /|~x ~x j |, so it contributes qj ~vj (~x 0 ~xj )/|~x 0 ~xj |3 to /t.
As the Greens function for Laplaces equation is 1/|~x ~x 0 |, we have

Darwin Lagrangian

d3 x0
|~x ~x 0 |
Z
d3 x0
=
|~x ~x 0 |
qj ~vj
+
=
c|~x ~xj |

~ x) =
A(~

1~ 0
1 ~0
J(~x )
(~x 0 )
c
4c t
!!
qj ~ 0 ~vj (~x 0 ~xj )
qj vj 3 0
(~x ~xj )

c
4c
|~x 0 ~xj |3
!
Z
qj
~vj (~x 0 ~xj ) ~ 0
1

d3 x0
0
3
4c
|~x ~xj |
|~x ~x 0 |

where we have integrated by parts and thrown away the surface at infinity.
~ action on a function of ~x ~x 0 , so we can pull
~ out
~ 0
The gradient
0
of the integral. Let ~r = ~x ~xj and ~y = ~x ~xj . Then
Z

~ d3 y ~vj ~y 1
~ x) = qj ~vj qj
A(~
c|~r|
4c
|~y|3 |~y ~r|
Z
Z
qj ~
qj ~vj
2

y dy
d sin
=
c|~r|
4c
0
0
Z 2
y(cos vjz + sin cos vjx )
1
2
d
3
2
y
y + r 2yr cos
0
where we have chosen z in the ~r direction and ~vj in the xz plane. The
integral kills the vjx term and, writing vjz = ~vj ~r/r, we have
"

~ ~vj ~r C ,
~ r ) = qj ~vj 1
A(~
c |~r| 2
r
where the integral giving C is
C=

Z
0

dy

Z 1
1

u
= 1,
du 2
y + r 2 2yru

504: Lecture 18

Last Latexed: April 6, 2010 at 10:40

though this integral is not as straightforward as Jackson claims. Then

"

~ j (~xk ) =
A

1X
1 X
mj vj2 + 2
mj vj4
2 j
8c j

X qj qk  1
1
+
+ 2 [~vj ~vk + (~vj rjk )(~vk rjk )] ,
2 4c
j6=k rjk

Proca Lagrangian

As we saw, our lagrangian density for electromagetic fields,

1
1
F F J A ,
16
c
gives equations of motion which do not completely determine the evolution
of the fields A . Let us consider adding a term proportional to A2 :

F + 2 A =

1
2
1
F F + A A J A ,
16
8
c
known as the Proca Lagrangian, which as we shall see describes a field which
has quanta of mass rather than the massless photons whose classical limit
is Maxwell theory. We still mean F to be shorthand for A A rather
than an independent field, so the homogeneous Maxwells equations still hold,
as they are consequences of F = dA. The extra term does not contribute
to P , as it depends only on A and not on derivatives thereof, so the extra
contribution to the equations of motion is just from L/A = (2 /4)A ,
and
4
J .
F + 2 A =
c

4
J .
c

The first term is identically zero by symmetry, and if the current density still
represents a conserved charge, the right hand side is also zero, so the Lorenz
condition A = 0 is now a consequence of the equations of motion and not
an arbitrary choice. As a consequence, we now have F = A , so


+ 2 A =

4
J .
c

In the absence of sources, this has solutions as before,

X
~k

A~k + eik~xi~k t + A~k eik~x+i~k t ,

~ h
~k
but with 2 = c2 (~k 2 + 2 ). Quantum mechanically we know ~p = ih
2
and E = i
h/t =
h, so this field represents particles for which E =
2 c2 . Of course quantum field theoriests take h
= 1 and c = 1, so
P 2 c2 + 2 h
this represents a massive photon with mass .
If we consider a point charge at rest and look for the static field it would
generate, we need to solve

LEM =

LProca =

One consequence comes from taking the 4-divergence of this equation:

qj
(~xk ~xj )~vj (~xk ~xj )
~vj +
.
2c|~xj ~xk |
|~xk ~xj |

where of course ~rjk := ~xj ~xk , rjk := |~rjk |, and rjk = ~rjk /rjk .
We mostly experience slightly relativistic particles in atomic physics,
though the electrons are best described by the Dirac formalism, so the velocities are replaced by
~ . It is also of use in plasma physics.

Last Latexed: April 6, 2010 at 10:40

Multiplying by qk~vk /c to get the appropriate contribution to Lint , and correcting the free-particle Lagrangian, mc2 1 + mc2 12 mv 2 + 18 mv 4 /c2 , we
get the Darwin Lagrangian
LDarwin =

504: Lecture 18

2 + 2 = 4q 3 (~r)
or

r2
+ r 2 2 = q(r).
r
r
Away from r = 0 this clearly requires r(r) = Cer and Gausss law tells
us
Z
4C,
4q = 4R2 d/dr|R + 2
r<R

so C = q and

R0

er
,
with r = |~x|.
r
This is the well-known Yukawa potential, which nuclear physicists had found
was a good fit to the binding of nucleons in a nucleus, leading Yukawa to
propose the existance of a massive carrier of the nuclear force, which we now
know to be the meson.
(~x) = q

504: Lecture 18

2.1

Last Latexed: April 6, 2010 at 10:40

Superconductors

In the BCS theory of superconductivity, electrons form pairs, and each pair
acts like a boson. So the quantum mechanical state that each pair is in
can be multiply occupied, and superconductivity occurs when states develop
macroscopic occupation numbers,  1. The wave function (~x) describing
these particles is a complex function, with the density of particles n(~x) =
, so = n(~x)ei(~x) . We may approximate n(~x) as being roughly constant.
The velocity of these particles is related to the canonical momentum by


~v =

1 ~ q~
P A
m
c

which can be viewed as an operator acting between and . It is the

~ Thus the current density is
canonical momentum P~ which acts like i
h.


nq
q~
h
A
.
J~ = q ~v =
m
c

(1)

~
~ = 0. This equation doesnt quite say
as
2

nq ~
A,
J~ =
mc

(2)

but it does say, in a simply connected region, that the difference is the gra~ by a gauge
dient of something, and as such a gradient could be added to A
transformation, we might as well assume (2), which is known as the London
equation. This gauge is still compatible with Lorenz (which can be viewed
as determining A0 ), so we have
~
2 A

Last Latexed: April 6, 2010 at 10:40

~
1 2A
4 ~ 4nq 2 ~
J=
A,
=

c2 t2
c
mc2

which is the Proca equation with 2 = 4nq 2 / mc2 .

At the boundary of the superconductor, if no current is crossing the
~ = 0. If we look for a static solution for a
boundary, we must have ~n A

planar boundary z, uniform along the boundary, we have A ez . The

London penetration depth is
1
L := =

mc2
.
4nq 2

With q = 2e and m = 2me for the electron pair, and taking n as the
density of valence electrons, the penetration depth is of the order of tens
of nanometers. As the A field is not penetrating further than that into the
medium, any external magnetic field has been excluded.
But magnetic field lines can enter the medium if our assumption of being
~ A
~ by a gauge transformation is not correct. That
able to do away with
could happen if the region of the superconductor is not simply connected
that is, a flux line could enter and destroy the superconducting region around
which is incremented by a multiple of 2. This is called a vortex line, and
corresponds to a quantized amount of flux, as
I

If we take the curl of both sides, we get

2
2
~ A
~ = nq B,
~
~ J~ = nq

mc
mc

504: Lecture 18

~ d` = 2N
A
hc/q =

Z
S

~ A
~ = B ,

with N Z. With q = 2e, the quantum of flux is hc/2e.