Electromagnetism Notes, NYU Spring 2018

March 28, 2018

1 Conservation Law 1
1.1 Symmetries
As we discussed already, symmetries place an important role in EM. e.g. we use them in supplementing
Gauss’ and Ampere’s Laws to compute static EM fields, in discussion of separation of variables, solutions
to Laplace and Poisson equations, etc. As we know from mechanics, there is a deep connection (Noether’s
theorem) between symmetries and conservation laws.
So far we only discussed charge conservation (described by the continuity equation) as far as conservation
laws are concerned. The symmetry related to it is gauge invariance, as we mentioned, but we’ll have to wait
a bit until we develop the action principle for EM to see this explicitly. We thus start from the simple
symmetries such as translations (in t, ~r and spatial rotations that lead to energy, momentum and angular
momentum conservation of EM fields) and consider other examples later and revisit some of what we do
here as well once we develop the relativistic framework.
Since the EM equations are written in vector form, they are automatically rotational invariant, i.e. the
laws are the same for observers who decide to orient their reference frames differently. The same is true for
translations of space and time, i.e. the EM laws are invariant under translation of ~r and t. These symmetries
are ::::::::::
continuous, i.e. the rotations or translations can be done by some arbitrary angle or displacement that
is a continuous parameter (as opposed to discrete).
In addition to those, the EM laws are invariant under discrete
:::::::
symmetries of space-time: parity (mirror
reflection) and time (or motion) reversal.
Parity (P for short) is equivalent to space inversion due to rotational invariance, i.e. reflection in the x-y
plane gives
~r = (x, y, z) → (x, y, −z) (1)
but adding a 180◦ rotation about the z-axis, we have

~r = (x, y, z) → (−x, −y, −z) = −~r (2)

Since it’s easier to consider spatial inversion ~r → −~r instead of P , we will do so, since they are equivalent.
We have then that under ~r → −~r, ~v → −~v (and p~ → −~ p) but L~ = ~r × p~ → L ~ = (−~r) × (−~ ~ is a
p) (so L
pseudo vector).
For time reversal t → −t, ~r → −~r but p~ → −~ p and thus L~ → −L.~ Both ~r and t reversal are symmetries
of Newton’s 2nd law:
d~
p
= −∇U (3)
dt
where U is a scalar potential.

1
 i.e. under spatial inversion, we have p~ → −~
p, ∇ → −∇, then
d(−~p)
= −(−∇U ) X (U is unchanged since it’s a scalar) (4)
dt
and under time reversal p~ → −~
p and t → −t, so
d(−~
p)
= −∇U X (5)
d(−t)

How about E ~ & B?~ Consider E ~ first, and for simplicity take a static charge distribution ρ(~r). Under
parity, or space inversion, we have ρ → ρ [i.e. ρ0 (~r) = ρ(−~r)] and the inverted distribution produces an
inverted field:

Then E~ → −E ~ [i.e. E
~ 0 (~r) = −E(~
~ r)]
Now, under T reversal, nothing happens to ρ so E ~ → E.
~ For B, ~ consider a steady current in a circular
~ ~ ~ 0 ~
loop. Under ~r → −~r, j → −j, [i.e. j (~r) = −j(−~r)] (same as ~v ) i.e.

Then B ~ → B.
~ But under times reversal, the current just changes sign so that B
~ → −B.
~ Therefore, B
~
~
behaves exactly as the angular momentum L. All these relations can be summarized as:
Let’s consider the Lorentz force equation of motion
 
d~p ~ + ~v × B
F~ = =q E ~ (6)
dt c
under space inversion:  
d(−p) ~ + −~v × B
= q −E ~ X (7)
dt c
Thus the same law is recovered. Under time reversal
 
d(−~p) ~ −~v ~
=q E+ × (−B) X (8)
d(−t) c

2
 space inversion time reversal
P T
~r −~r ~r
p~, ~v −~
p, −~v −~
p, −~v
ρ ρ ρ
~j −~j −~j
~
L ~
L −L ~
E ~ −E ~ E ~
B ~ ~
B −B ~

also works out. Same for Maxwell’s equations:

(  
−∇ · −E ~ = 4πρ P X
~
∇ · E = 4πρ ⇒ (9)
∇·E ~ = 4πρ TX
  
~  −∇ × −E ~ = − ∂ B~ P X
∇×E ~ = − ∂B ⇒ ∂t
(10)
~
∂t  ∇×E ~ = − ∂(−B) TX
∂(−t)
~
(
−∇ · B = 0 PX
∇·B ~ =0⇒ (11)
∇ · −B ~ = 0 TX
( ~
4π 1 ∂ ~
E −∇ 
×B ~ = 4π (−~j) + 1 ∂(−E) PX
∇×B = ~ ~j+ ⇒  c c ∂t
(12)
c c ∂t ∇ × −B ~ = 4π (−~j) + 1 ∂ E~ TX
c c ∂(−t)

Note that under these relationships we have that the EM energy density (be to discussed below) and the
Poynting vector S (also defined below), are transformed under P , T :

~2 + B
E ~ 2 P or T E~2 + B~2
−−−−→ (13)
8π 8π
~ e ~ ~ P or T ~
S= E × B −−−−→ −S (14)
4π

1.2 Energy Conservation

Let’s now discuss EM field energy. We start for the E ~ field from the electrostatic energy for a collection
of charges we obtained when we discussed conductors (Notes #6):
Z
1 1
UE = ρ(~r)φ(~r) d3 r = − (∇2 φ)φ d3 r (15)
2 8π
~ field as follows. Since
which we can rewrite in terms of E

φ∇2 φ = ∇ · [φ∇φ] − (∇φ)2 (16)

Z Z
1 ~+ 1
UE = − φ∇φ · dS (∇φ)2 d3 r (17)
8π S 8π

3
 Since φ ∼ 1r , ∇φ ∼ 1 ~ ∼ r2 , the first term vanishes when r → ∞
r2 , dS
Z
1 ~ 2 d3 r
UE = E (18)
8π
2
Thus uE = E ~ field.
8π is the energy density if the E
::::::::::::
In presence of material media, by same argument given in Notes #6, UE = 12 ρfree (~r)φ(~r) d3 r, where
R

φ is the Rrelevant potential in e.g. presence of dielectric. Following the derivation above, then we have
UE = 8π 1 ~ ·D
E ~ d3 r.
Let’s now do the same for the magnetostatic case, which we did not discuss before. Again, we want to build
a current distribution ~j(~r) from nothing and we calculate the corresponding energy for such configuration
by adding the work required to build it. The tricky part, compared to the electrostatic calculation, is that
from the Lorentz force, it follows that electric fields do work on charges, but magnetic fields do not, since
~v × B~ is perpendicular to ~v .
Thus, in the electrostatic case as we built the charge configuration, we calculated the work from the change
~ (and while this would induce B,
in E ~ it cannot contribute to the work); now as we build ~j the changes ∆~j
~
in times ∆t changes B (but does no work) and E ~ through Faraday, which provides the contribution to the
energy we are looking for.
~ induced by Faraday as ~j → ~j + ∆~j in ∆t gives by Lenz’s law a force on the moving charges (~j)
So, the E
that oppose the charge ∆~j. The work required to overcome this opposition is
Z
∆W = −∆t q~v · E ~ d3 r (19)

~ =
but, ~j = q~v and since ∇ × B 4π ~
c j
we can write
Z Z
c∆t ~ ·E~ d3 r = − c∆t B ~ ×∇×E
~ d3 r
∆W = − ∇×B (20)
4π 4π
where we use integration by parts.
Z Z
1 ~ d3 r = ∆ 1
~ ·B ~ 2 d3 r
⇒ ∆B B (21)
4π 8π
Thus, adding all contributions, we have:

1
Z ~2
B
UB = ~ 2 d3 r
B or uB = magnetic energy density (22)
:::::::::::::::::::::
8π 8π

~ ~
Again, it is simple enough to generalize this to material media, in which case uB = B· H
8π , analogously to
the electric case.
Now we are ready to discuss local energy conservation in the full, time-dependent case. Since the energy
is a scalar, its flux is a vector and thus we expect that the electromagnetic energy density obeys a continuity
equation of the form:
∂u ~
= −∇ · S (23)
∂t
with S~ the energy flux density. However, this equation makes an important assumption (not always correct):
::::::::
that the EM fields are not interacting with charges which can also of course change the energy in EM fields.

4
Thus, to include the most general case with charges present we add to the continuity equation above a term
W that is the energy transferred from field to charges per unit time and per unit volume:
∂u ~ −W
= −∇ · S (24)
∂t
Of course we can just calculate W from the Lorentz force by just doing power (energy per unit time) per
unit volume:
F~ · ~v  q  ~ ~v
 
W = = E+ ×B ~ · ~v = ρ~v · E
~ = ~j · E
~ (25)
V V c
Then we have:
∂u ~ + ~j · E
~ =0
+∇·S (26)
∂t
or, in integral form
Z ZZ Z
d ~ · dA
~+ ~ · ~j d3 r
− u d3 r = S E (27)
dt V S(V ) V

Let us now go back to Maxwell’s equations:

~
~j = c ∇ × H ~ − 1 ∂D (28)
4π 4π ∂t
c
Z
c
Z
1
Z ~
~ · ∂ D d3 r
 
~ ~
⇒E·jd r =3 ~ ~
H ·∇×Ed r− 3 ~ ~
∇· E×H d r− 3
E (29)
4π V 4π V 4π V ∂t
 
where we use ∇ · E~ ×H~ =H ~ ·∇×E ~ +E ~ ·∇×H ~
~ = − 1 ∂ B~ :
But using ∇ × E c ∂t
!
Z Z ~ ~ ZZ
~ · ~j d3 r = − 1
E ~ ∂B + E
H ~ ∂D −
c ~ × H)
(E ~ · dA
~ (30)
V 4π V ∂t ∂t 4π S(V )

 
~ · ~
∂B 1 ∂ ~ ~ E ~· ~
∂D 1 ∂ ~ ·D
~ , then we have (using a volume
Now, assuming linear media: H ∂t = 2 ∂t (H · B), ∂t = 2 ∂t E
V arbitrary, fixed in time, as we did above)

d
Z ~ ·B
H ~ +E ~ ·D
~ ZZ
c ~
Z
− d3 r = ~ · dA
(E × H) ~+ ~ · ~j d3 r
E (31)
dt V 8π S(V ) 4π V

which now comparing to the last boxed equation above, we can identify:

~ ·D
E ~ +B ~ ·H
~
EM energy density: u≡ (32)
8π

EM flux density: ~ = c (E
S ~ × H)
~ “Poynting vector” (33)
4π
2
d p p
~ d~
p
= ~v · F~ , we can also write ~ d3 r =
~j · E dK
R
Noting that dt 2m = m dt V dt , where K is the kinetic energy of
the charges.

5
 Then we can write ZZ
d ~ · dA
~
(K + U ) = − S (34)
dt S(V )

One may ask under which situations the “total energy” K + U in particles + fields is conserved. We
need the RHS to vanish, this can be achieved (for a volume V e.g. spherical with R → ∞) if |S| ~ decreases
1 2
more rapidly than r2 (since A ∼ r ). This is true for the electric and magnetic fields that we discussed so
far, but it :::
will::::
not be true for radiation fields that we will discuss later.