2c.

hopefully, we can use to simplify our problems.

Lecture 2c
The vector A is called a vector potential. We will introduce
another vector potential below, so let's be more precise and
Potentials call A the magnetic vector potential since the magnetic field is
obtained from it by a differential operation. The scalar  is
Introduction often called the scalar potential, although it need not have
In general solving a vector wave equation is a very difficult physical significance.
undertaking, and we would like to avoid working with vector The magnetic vector potential is proportional to the
equations whenever possible. In our undergraduate physics electric field plus a "wiggle term."
courses we saw that an electrostatic vector field can be derived
from a scalar potential Solving for E we have

E=−∇ (1) E=− j  A −∇ (6)

This allows us to work out many electrostatics problems using The addition of the  term does not change the value of
scalar equations. For example, we can calculate the potential ∇×A since ∇ ×∇ ≡0 . The curl of A is what has physical
as an integral over the distribution of charge: significance through (4).
1 q r' The divergence of A is
  r=
4  0
∭ ∣r−r'∣
dV ' (2)
j
If we need the vector field E we can then apply the necessary
∇⋅A=

∇⋅E∇2   (7)
differential operator to derive it from the scalar potential.
E is a physical quantity so ∇⋅E is determined by physics.
When dealing with non-static fields we cannot, in general, However, we can take ∇2  to be an arbitrary function of
derive the vector fields from a scalar potential. However, we 2
will see that we can derive them from a vector potential that position. For example, if we set ∇  r =−q r / 0 we can
can be calculated from the current distribution using a integral interpret qr  as a (fictitious) charge density and   r as
relation similar to that above. And there will be special cases the resulting electrostatic field. This is simply Poisson's
where the vector potential effectively reduces to a scalar field. equation and we know that it has a solution for any function
qr  , namely (2).
Magnetic vector potential Therefore, we can set the divergence of A to be an arbitrary
function of position. On the other hand, the electric field E is a
Faraday's law ∇×E=− j  H can be rearranged to give
physical quantity fully constrained by physics. We can think
j of the magnetic vector potential A as a mathematical device,
H= ∇×E (3) related to E, but with a degree of freedom that will prove very

useful for simplifying analysis.
This shows that H is completely determined by E. For now,
The curl of A is physically constrained but the
let's call A= jE / . We then have divergence of A is arbitrary.
1 In general we have
H= ∇×A (4)

1
Therefore H can be written as the curl of a vector. This is just H= ∇× A

a restatement of Faraday's law with a redefinition of the (8)
1
electric field times a constant as a new vector A. But here is E=  ∇×∇× A− J 
j
the “trick.” Recall that the curl of the gradient of a scalar field
is identically zero, that is ∇×∇ ≡0 . Therefore, we can add The second equation is from substituting the first equation into
the gradient of an arbitrary scalar field into our expression for Ampere's law. In a source-free region these reduce to
A without changing the value of the curl. That is, we can take
1
H= ∇×A
j 
A =  E∇   (5) (9)
 1
E= ∇×∇×A
j  
where  is any scalar field, and (4) will still be true. The
scalar field  need not have any physical significance; it can If we take A to have only a z component then we have, for a
be anything we want1. It is a mathematical “wiggle term” that, source-free region

1 It does have to be twice differentiable so that we can take the

gradient followed by the curl.

EE518: Advanced Electromagnetic Theory Scott Hudson 2015-02-02

 2c.2

1 ∂ A  =∇⋅B
H x= (16)
 ∂y z A= ∇×B
1 ∂ A Then
H y =−
 ∂x z
H z =0 F=−∇ ∇×A (17)
j ∂ ∂ (10) This is the Helmholtz theorem. We call  the scalar
E x =− A
 ∂x ∂z z potential and A the vector potential.
j ∂ ∂ A
E y =− What is the advantage of this? Consider the divergence and
  ∂ y ∂ z z curl of F
j
[ ]
2 2
E z= ∂ A ∂ A
2 z z ∇⋅F=−∇ 2 
 ∂ x ∂ y2 (18)
∇ ×F= ∇ ×∇ ×A
Since H has no z component we will refer to this as a TM z
This shows that the divergence and curl of a vector field are
field (the magnetic field is transverse to the z direction). It
independent in the sense that they can be separately specified.
turns out that any field that has H z ≡0 can be expressed in If the curl is fixed (A specified), the divergence can be set to
terms of the scalar field A z . So, we reduced the problem of an arbitrary function (specify  ) and conversely. This will be
finding the five field components E x , E y , E z , H x , H y to that quite useful to use later on.
of finding the single component A z . This is a tremendous
simplification for analytical purposes. Electric vector potential
In a source-free region ( J =0 ) Ampere's law becomes
Helmholtz theorem
∇×H= j   E (19)
The ideas we have been discussing are formalized in the
Helmholtz theorem. The Helmholtz theorem says that any We can write
(physically plausible) vector field F can be represented as the
1
combination of the gradient of a scalar field and the curl of a E=− ∇× j H /  (20)
vector field. Mathematically we can write 

F=−∇ ∇×A (11) or

1
We know from Poisson's equation of electrostatics that if E=− ∇ ×F (21)

2
∇  r =−q r  (12)
where F= j H / is the electric vector potential. We can add
then the gradient of a scalar function to F without changing E
1 q r ' 
 r =
4
∭∣r−r '∣
dV ' (13)
F=
j
 H∇ e  (22)

One thing this tells us is that any scalar field q can be written
as the Laplacian of another scalar field  , and it even gives Again, this gives us a wiggle term to use for simplifying
us a formula for computing  from q. problems.

Now, let F be any vector field. Each of the three components Keep in mind that unlike the magnetic vector potential
of F is a scalar field, so we can use our previous result to write concept, which is always valid,
2
F x =−∇ B x where B x is some other scalar field. We can do the electric vector potential concept is only applicable
likewise for the y and z components of F. In vector notation we in a source-free region.
have When F is used we have
2
F=−∇ B (14) 1
E=− ∇×F
This is the vector counterpart to (12). Using a vector identity 
(23)
we write 1
H=− j ∇×∇×F
−∇ 2 B=−∇  ∇⋅B ∇× ∇×B (15) 

Now, define the scalar and vector fields In a source-free region with F= a z F z

EE518: Advanced Electromagnetic Theory Scott Hudson 2015-02-02

 2c.3

1 ∂ Wave equations for A and F

E x =− F
 ∂y z Recall that we started with Faraday's law in our development
1 ∂ of the magnetic vector potential concept. So, Faraday's law is
Ey= Fz
∂x already "built in" to our analysis. Let's now consider Ampere's
E z =0 law
j ∂ ∂ (24) ∇×H=J  j  E (28)
H x =− F
  ∂ x ∂ z z
Substituting (4) and (6) results in
j ∂ ∂ F
H y =−
 
z
 ∂ y ∂z 1
∇× ∇×A =J  j    − j A− ∇   (29)
H z=
j
[∂2 F  ∂2 F
   ∂ x2
z
∂y
2 z
] 

Remember that  is an arbitrary scalar field, our "wiggle

Since E has no z component we will refer to this as a TE fieldz term." Let's assume =const . This allows us to write
(the electric field is transverse to the z direction). Any field in
∇×∇× A= J j     − j  A− ∇   (30)
a source-free region that has E z ≡0 can be expressed in terms
of the scalar field F z . So, we have reduced the problem of Using a vector identify for ∇×∇× A we have
finding the five field components E x , E y , H x , H y , H z to that ∇ ∇⋅A −∇ 2 A= J j     − j A− ∇   (31)
of finding the single component F z .
Rearranging gives us
Using a combination of A and F 2 2
∇ A +ω μ ϵ A=−μ J +∇ ( ∇⋅A )+ j ωμ ϵ ∇ ψ (32)
Although we can fully specify both E and H using the The "ugliest" term is ∇ ∇⋅A  . If
magnetic vector potential A alone, or using the electric vector
potential F alone (in the case of a source-free region), we j ωμ ϵ ∇ ψ=−∇ (∇⋅A) (33)
sometimes find it useful to use A and F at the same time. This then this will go away. If both m and e are constants then
gives us more degrees of freedom that we can use to simplify
analysis. When using both A and F we have j
 =− ∇⋅A  (34)
1 1  
H= ∇×A − j ∇ ×∇×F
   achieves this for us. We have had to assume that both of  , 
(25)
1 1 are constants, so we are dealing with a simple medium at this
E=− ∇×F− j ∇× ∇×A
  point.
As you will show in the homework, an arbitrary plane wave We now have the result, valid for any simple medium,
can be represented using just the A z , F z components. Just as ∇ 2 A 2 A=− J (35)
any (physically plausible) function of time can be represented
as a superposition of complex sinusoids (inverse Fourier Compare this to the simple-medium equation we derived for E
transform), any (physically plausible) electromagnetic field j
2 2
can be represented as a superposition of plane waves. It ∇ E E= j  J  ∇  ∇⋅J  (36)
follows that any EM field can be represented by specifying 
A z , F z as functions of position. This is a very powerful fact Our equation in A is much more useful, in particular because
for analytic work. each component of A depends only on the corresponding
component of J. For example,
Hertz vectors ∇ 2 Az  2 A z =− J z (37)
Although we will not use them, an alternate approach is to
The ∇ ∇⋅J  term in the E equation, on the other hand,
employ the so-called Hertz vectors  ,  m These are related
causes each component of E to depend on each component of
to the vector potentials by
J.
A= j     (26) In a source-free simple medium we have
F= j     m (27) ∇ 2 A 2 A=0
2 2 (38)
You may see these in the literature and in texts. For example, ∇ F F=0
Ishimaru uses this approach.
The second equation can be derived in a similar manner to the
first. We have the interesting result that in a source-free simple

EE518: Advanced Electromagnetic Theory Scott Hudson 2015-02-02

 2c.4

 
medium, E, H, A and F all satisfy the Helmholtz equation. − J 0 − j r0 1 j  r 0
∯S ∇ A z⋅dS= 4
e
r 02
4  r 20
(48)
Relation between A and J
− J 0
Let's go back to (35) and see if we can derive a solution for A
in terms of J. Let's consider only the z component of (35) and So, we have shown that
let J z be a "point current" at the origin of the form J 0  r , − j r
 e
Az= J0 (49)
∇ 2 Az  2 A z =− J 0  r (39) 4 r

Notice that if =0 then ∇ 2 Az =− J 0  r which looks is the solution to
like Poisson's equation for a point charge at the origin. The ∇ 2 Az  2 A z =− J 0  r (50)
solution would be
Now let J z be any scalar field. We can write
 1
Az= J (40)
4 0r J z r=∭ J z  r'  r−r ' dV (51)
We will now show that for ≠0 the solution is The potential produced by J z r'   r−r'  will be
− j βr
μ e μ e
− j β∣r−r'∣
Az= J (41) J (r ' ) (52)
4π 0 r 4π z ∣r−r'∣
The Laplacian for a function or r only (in spherical Therefore
coordinates) is − j  ∣r−r'∣
 e
2 d
2
2 d A z r=
4
∭V z
J r '
∣r−r'∣
dV ' (53)
∇ Az = 2 Az Az (42)
dr r dr
We could repeat exactly the same steps for the x and y
It is left as an exercise to show that (41) satisfies coordinates. The result is the vector equation
∇ 2 Az  2 A z =0 for r ≠0 . We then need to show that
 e− j ∣r−r '∣
∇ 2 Az  2 A z behaves like − J 0  r in the sense that A r=
4
∭V
J r' 
∣r−r '∣
dV ' (54)

∭V ∇ 2 A z  2 Az dV =− J 0 (43)

This is a very powerful result. We cannot derive such a simple
over any volume V containing the origin. Consider a sphere of relationship between J and either E or H. This tells us that if,
radius r 0  0 . In spherical coordinates we have for example, J has only an x component, then A will have only
an x component. This would apply to a wire antenna parallel to
dV =r 2 sin  dr d  d  .
the x axis. All six components of E and H may be derived
− j r from the single component of A.
∭V A z dV = 4  J 0 ∭V e r
 2
r sin  dr d  d 
(44)
0
References
1. Ishimaru, A., Electromagnetic Wave Propagation,
since r 0  0 and the integrand has a factor of r. Therefore the Radiation, and Scattering, Prentice Hall, 1991, ISBN 0-
2
β A z term of (43) does not contribute and we must have 13-249053-6.
2. Balanis, C. A., Advanced Engineering Electromagnetics,
∭V ∇ 2 Az dV =− J 0 (45)
Wiley, 1989, ISBN 0-471-62194-3.
The divergence theorem gives us

∭V ∇ 2 Az dV =∯S ∇ A z⋅dS (46)

d
Using ∇ A z = a r A we obtain
dr z

∇ A z=
4 
− J 0 − j  r 1 j  r
e
r
2 a r
 (47)

Then

EE518: Advanced Electromagnetic Theory Scott Hudson 2015-02-02