Propagation of vectorial laser beams

Peter Muys
Laser Power Optics, Hooiland 3, B-9030 Gent, Belgium (peter.muys@gmail.com)
Received December 19, 2011; accepted January 23, 2012;
posted January 27, 2012 (Doc. ID 159962); published April 18, 2012
The angular spectrum of a vectorial laser beam is expressed in terms of an intrinsic coordinate system instead of
the usual Cartesian laboratory coordinates. This switch leads to simple, elegant, and new expressions, such as for
the angular spectrum of the Hertz vectors corresponding to the electromagnetic fields. As an application of this
approach, we consider axially symmetric vector beams, showing nondiffracting properties of these beams, without
invoking the paraxial approximation. 2012 Optical Society of America
OCIS codes: 260.1960, 260.5430, 260.2110.
1. INTRODUCTION
In many branches of physics, it is necessary to solve the
vectorial wave equation in three dimensions. One of the best
known and most general methods for the vectorial Helmholtz
equation is the expansion in plane waves by representing the
solution as a three-dimensional (3D) Fourier integral over the
3D reciprocal wavevector space. Especially in laser physics,
the relevant electric and magnetic fields are in fact solutions
having a directed-energy beam-character. That is, the fields
propagate along a certain axis, called the optical axis and
which we choose to be the z axis of the Cartesian coordinate
system linked with the beam. A number of mathematical
methods are known to treat this case efficiently [1,2]. In this
paper, we consider three methods of solution, the angular
spectrum representation of the beam, the TE/TM decom-
position of the beam and finally the Hertz vectors of the beam.
The last method is the least well known, despite the fact
that it shows a major advantage by generating particularly
simple expressions for the fields. We shall indicate the
interrelations between the three methods. In a first step we
show how to transform the Cartesian axes to an intrinsic
coordinate system linked with the propagation constants of
the beam.
Physically, the new base vectors are linked with the clas-
sically known p and s components of polarization optics,
but their expression in a Cartesian or cylindrical coordinate
system is not evident.
In a second stage, we determine the angular spectrum not
of the electric and magnetic fields E and B, but of the Hertz
potential vectors P
e
and P
m
corresponding to these vectors.
Then we use the results of the first phase and switch to
the new coordinates. New and simple formulas are obtained.
This leads to an alternative angular spectrum representation
of the solution of the vectorial Helmholtz equation.
As an application of the usefulness of this representation,
we will briefly discuss diffraction-free vectorial beams. We
will do this without the restriction to reduce the Maxwell
equations to their paraxial form. The diffraction-free solution
is hence a rigorous solution and not an approximate one,
satisfying in particular the zero divergence Maxwell equation
for the electric field, which a paraxial solution does not
satisfy.
2. INTRINSIC COORDINATE SYSTEM
Let k be the wave vector of a plane wave
k = k
x
e
x
k
y
e
y
k
z
e
z
= k
t
e
r
k
z
e
z
= k
t
cos e
x
k
t
sin e
y
k
z
e
z
(1)
and its magnitude is given by
k
2
= k
2
x
k
2
y
k
2
z
= k
2
t
k
2
z
.
In Eq. (1) we have denoted the split of the wave vector into its
transverse (subscript t) and longitudinal part (subscript z).
The k vector forms together with the z axis the meridional
plane [3]. In optical terminology, we would call it the plane
of incidence (on a mirror for example, where z is the normal
to the mirror). The angle between the z axis and the k axis is
the polar angle . The cylindrical coordinates are defined
through their unit vectors e
r
, e
z
, and e

, being the azimuthal

angle of the k vector in the transverse (x; y) plane.
The 3D locus vector R is defined by
R = xe
x
ye
y
ze
z
= re
r
ze
z
= r cos e
x
r sin e
y
ze
z
; (2)
where we indicated the split of the locus vector in its trans-
verse and its longitudinal part.
The key element in our approach is the point brought up in
[2] and [4], which is that we should consider an intrinsic co-
ordinate system. It contains three unit vectors: two of them
are respectively perpendicular and parallel to the meridional
plane, the third being the unit vector in the k direction:
s = e

p =
k
z
k
e
r

k
t
k
e
z
m = kk. (3)
990 J. Opt. Soc. Am. B / Vol. 29, No. 5 / May 2012 P. Muys
0740-3224/12/050990-07$15.00/0 2012 Optical Society of America
These unit vectors, taken in the order s, p, m, form a right-
hand sided coordinate system.
Note that s does not possess a z component. s is perpendi-
cular to the meridional plane, whereas p is coplanar with it,
just like m. s has only a transverse component, its azimuthal
part, whereas p both has a transverse component, its radial
part, and a longitudinal component. Although the definition
of the s and p vector are classically known in the literature
on optical polarization, their relation [Eq. (3)] with the cylind-
rical base vectors is not evident. An alternative but equivalent
definition of the intrinsic coordinates is
s =
e
z
k
[e
z
k[
=
e
z
k
k
t
p =
s k
k
= s m.
The transformation of the Cartesian coordinates to the intrin-
sic coordinates is given by
s =
k
y
k
t
e
x

k
x
k
t
e
y
p =
k
z
k
x
kk
t
e
x

k
z
k
y
kk
t
e
y

k
t
k
e
z
m =
k
x
k
e
x

k
y
k
e
y

k
z
k
e
z
. (4)
The inverse transformation from the intrinsic coordinate
system to the Cartesian laboratory coordinate system is, after
lengthy but straightforward calculations, given by
e
x
=
k
y
k
t
s
k
x
k
z
kk
t
p
k
x
k
m
e
y
=
k
x
k
t
s
k
y
k
z
kk
t
p
k
y
k
m
e
z
=
k
t
k
p
k
z
k
m; (5)
which is a central result for the subsequent calculations. Note
that s and p are in fact functions of k
x
and k
y
and should in fact
be written as s(k
x
; k
y
) and p(k
x
; k
y
) when they will in a later
paragraph appear behind the integration sign.
3. HERTZ POTENTIALS AND THEIR
ANGULAR SPECTRUM REPRESENTATION
We will only consider free-space situations: the sources are
infinitely far away from our observation region. The electric
and magnetic fields can be defined in S.I. units in terms of
their scalar and vector potential [5]:
E = grad
A
t
B = rot A; (6)
where the potentials are linked by the Lorentz gauge
div A
0

t
= 0.
An equivalent and alternative set to these potentials are the
Hertz electric and magnetic vector potentials P
e
and P
m
,
which in free space are, just like E and B, also solutions of
the homogeneous vectorial dAlembert wave equation

2
P

1
c
2

2
P

t
2
= 0; = e; m.
The link between the regular potentials and the electric Hertz
vector is given by [6]:
= div P
e
A =
0

0
P
e
t
and for the magnetic Hertz vector by
= 0
A = rot P
m
.
Combining this with Eq. (6), hence eliminating the scalar and
vector potential, results for the two cases resp. in the expres-
sion of the fields as a function of the Hertz vectors. For the
electric Hertz vector:
E = grad div P
e

0

2
P
e
t
2
= rot rot P
e
B =
0
H =
0

t
(rot P
e
). (7)
And for the magnetic Hertz vector:
E =

t
(rot P
m
)
B =
0
H = grad div P
m

1
c
2

2
P
m
t
2
= rot rot P
m
. (8)
Because of the source-free space, the magnetic induction B
reduces to the magnetic field strength H. By adding Eqs. (7)
and (8), we find the general expression for the fields in terms
of the Hertz vectors:
E = grad div P
e

1
c
2

2
P
e
t
2

(rot P
m
)
t
B =
0
H =
0

0
(rot P
e
)
t
grad div P
m

1
c
2

2
P
m
t
2
. (9)
Note that the dimensions of the electric Hertz vector are V
m
and V
s
for the magnetic Hertz vector.
These expressions do not look at first sight like simplifying
life. The use of the Hertz potentials is advantageous, however,
in case the current sources are all parallel to a common direc-
tion, say the z direction [57]. In this situation, and far from
the sources, Eq. (7) for the electric Hertz potential reduces to
the set
P. Muys Vol. 29, No. 5 / May 2012 / J. Opt. Soc. Am. B 991
E
x
=

2
P
ez
xz
E
y
=

2
P
ez
yz
E
z
=

2
P
ez
z
2

0

2
P
ez
t
2
H
x
=
0

2
P
ez
yt
H
y
=
0

2
P
ez
xt
H
z
= 0; (10)
which show that the electric Hertz vector P
e
reduces to its
scalar component P
ez
:
P
e
= (0; 0; P
ez
) (11)
This fact enormously simplifies the mathematical manipula-
tions and was exploited to analyze a number of vectorial
diffraction effects in [3]. The attractive aspect of the scalar
Hertz vector representation as compared to, e.g., the mathe-
matically equivalent modal representation is that it arrives at
the same rigorous results as the modal representation, but in a
much shorter and more elegant way. To give a quick example
of its usefulness, consider the electric field of an electric di-
pole oscillating along the z axis. In every direction starting
from the origin, the electric dipole field has another orienta-
tion but the corresponding Hertz vector is nevertheless every-
where in space a scalar, oriented along the z axis, independent
of the radiation direction [3].
For laser applications, we are exclusively interested in di-
rected-energy solutions of the vectorial Maxwell equations,
i.e., in beams, rather than in fields. Of course we then take
the z direction as the propagation direction of the field energy.
So, using the Hertz potentials, vectorial beams can be comple-
tely described by a scalar function. Specifically, whereas
P
e
(R; t) in general is a solution of the vectorial wave equation,
its z component P
ez
(x; y; z; t) in Eq. (11) is now a scalar
solution of the Helmholtz wave equation with the boundary
condition P
ez
(x; y; 0; t) = P
ez
(x; y) exp(jt). The two-
dimensional (2D) Fourier transform of the Hertz potential
P
ez
(x; y) in the plane z = 0, is defined as
p
ez
(k
x
; k
y
) = k
2
_

P
ez
(x; y) exp[j(k
x
x k
y
y)|dxdy
(12)
and is called its angular spectrum. The factor k
2
before the
integration signs is required to give the spectrum the same
physical dimensions as the original. The inverse Fourier trans-
formation is given by
P
ez
(x; y; z = 0) = P
ez
(x; y)
=
1
4
2
k
2
_

p
ez
(k
x
; k
y
) exp[j(k
x
x
k
y
y)|dk
x
dk
y
. (13)
The factor of 14
2
appears for mathematical reasons, the fac-
tor k
2
for physical reasons. Note that p
ez
is not the angular
spectrum of the electric field in the plane z = 0, but rather
of the electric Hertz vector. The general solution of the scalar
wave equation for P
ez
(x; y; z; t) in the halfspace z > 0 is now
given by [8]
P
ez
(x; y; z; t) = exp(jt)
1
4
2
k
2
_

p
ez
(k
x;
k
y
) exp[j(k
x
x
k
y
y k
z
z)|dk
x
dk
y
; (14)
which is called the angular spectrum representation of the
electric Hertz potential P
ez
(x; y; z; t). Equation (14) shows that
for beam-like solutions, we do not need a 3D Fourier trans-
form as solution of the wave equation, the dimensionality
of the problem is reduced by one unit.
A. Vectorial Angular Spectrum Representation
A scalar field E(R) satisfying the scalar Helmholtz equation,
can be represented as in Eq. (14) by its angular spectrum
representation, i.e., the beam is linearly polarized. What hap-
pens if the beam has a vectorial character? Just by extending
the argument from one dimensional (1D) to 3D, we look for a
representation like
E(R) =
1
4
2
k
2
_

e(k
x
; k
y
) exp[j(k
x
x k
y
y k
z
z)|dk
x
dk
y
.
But once E
x
(R) and E
y
(R) are known, then E
z
(R) is fixed,
because the electric field has no divergence. We hence need
only two scalar functions to determine the 3D electric field.
This should be reflected in the form of the vectorial angular
spectrum.
Assuming harmonic time variation, we now take the 2D
Fourier transform of E and H at z = 0 in Eq. (9) and call them
the vectorial angular spectra e and h of the field vectors, or the
spectral field vectors in short
e(k
x
; k
y
) = k
2
_

E(x; y; 0) exp[j(k
x
x k
y
y)|dxdy (15)
h(k
x
; k
y
) =k
2
_

H(x; y; 0)exp[j(k
x
xk
y
y)|dxdy. (16)
It should be noted that the vectorial character of the angular
spectrum, i.e., its polarization state, is the same as that of the
original field. Hence e and h inherit the polarization state of E
and H (and also their physical dimensions) Since the diver-
gence of E and H is zero in free-space, this means for their
spectral vectors
e k = h k = 0;
i.e., the spectral vectors are transverse to the wave vector, just
like the original field vectors.
We now take the 3D Fourier transform of the field Eq. (9)
and use the theorem that the Fourier transform of the deriva-
tive of a function is equal to jk times the Fourier transform of
the original function. In working out the expressions for
z = 0, one should take into account relations such as

x
[p
e
(x; y; 0)| =
p
e
x

z=0
;
which is easy to prove by writing out the derivatives according
to their definition as a limit of the differential quotient for
x 0. Finally, we arrive at the angular field spectra as func-
tions of the angular Hertz spectra
992 J. Opt. Soc. Am. B / Vol. 29, No. 5 / May 2012 P. Muys
e(k
x
; k
y
) = (k:p
e
)k k
2
p
e
kck p
m
(17)
h(k
x
; k
y
) =
0
kck p
e
(k:p
m
)k
0
k
2

0
p
m
; (18)
where we also made use of the dispersion relation = kc to
further simplify the expressions. So, in reciprocal k space, the
relations between the field spectra and the Hertz spectra are
no longer differential equations but become vector equations.
What we will do in this paragraph is work with the spectral
field vectors e and h to finally obtain an angular spectrum re-
presentation of the fields E and H, through the intermediary of
the Hertz vectors and their angular spectrum representation,
according to the Fourier transforms indicated in the following
diagram:
E; H e; h

P
e;m
p
e;m
.
Note: the spectral vectors p
e
and p
m
should not be confused
with the intrinsic vector p.
B. Transverse Spectral Hertz Vectors
Next, we proceed to decompose each angular vectorial spec-
trum of the Hertz potentials p
e
and p
m
in a part that is parallel
to k and a part that is perpendicular to k:
p

(k
x
; k
y
) =

= e; m
k:

= 0
k

= 0. (19)
On substituting Eqs. (19) in Eq. (17,18), the parallel compo-
nents of the spectral Hertz vectors all drop out and what rests
is
e(k
x
; k
y
) = k
2

(k
x
; k
y
) kck
m

(k
x
; k
y
)
h(k
x
; k
y
) =
k
2

(k
x
; k
y
) kc
0
k
e

(k
x
; k
y
). (20)
C. TE/TM Decomposition
The considerations in Subsection 3.A were valid for fields in
general. Because we consider beam propagation along the z
axis, the TE/TM decomposition theorem for the electric and
magnetic field can be invoked here [9], using the z axis as axis
along which the decomposition is stated:
{E; H] = {E
TE
; H
TE
] {E
TM
; H
TM
]
E
TE
e
z
= 0; H
TM
e
z
= 0. (21)
The transverse character of the field vectors E and H now has
to be conveyed to the spectral vectors e and h. We will pro-
ceed by taking a detour and first determine how e and h de-
pend on the spectral Hertz vectors. The reason is, as we will
see shortly, that the spectral Hertz vectors take on very simple
expressions. And once the spectral Hertz vectors are known,
it is again easy to deduce the spectral field vectors from them.
We now consider a source of oscillating dipoles, all oscil-
lating in the same direction, which we choose to be the z
direction. The Hertz vectors have the same orientation as
the currents [6]; hence the Hertz vectors have only a z com-
ponent. We assume that the dipole source is located at z =
so that the halfspace z > 0 is source free and the field differ-
ential equations become homogeneous and hence simpler. In
this halfspace, the Hertz vectors keep their z orientation.
From here on, we assume that the Hertz vectors have only
a z-dependent component:
P
e
(R) = P
ez
(R)e
z
; P
m
(R) = P
mz
(R)e
z
.
Hence, also their angular spectra p
e
and p
m
are only z depen-
dent. P
ez
and P
mz
are solutions of the homogeneous, scalar
Helmholtz equation.
We now revert to the angular spectra of the Hertz vectors.
D. TM Case
For the moment, we only know that, by Fourier transforming
the original electric Hertz vector, we obtain
p
e
= p
ez
e
z
.
Now, transforming Eq. (21), we see that h
TM
e
z
= 0. We use
Eq. (20) to calculate this scalar product. This leads after some
algebra to: p
m
= 0. Next to the coordinate transformations
Eq. (4,5) for the intrinsic vectors, this small equation is the
second central result of this paper. It shows that the spectral
magnetic Hertz vector is zero for a TM beam, leading to ex-
tremely simple equations, despite the vectorial character of
the beam. This is shown by again using the intrinsic coordi-
nate transformations, so that the angular spectra now become
e
TM
(k
x
; k
y
) = p
ez
[k
z
(k
x
e
x
k
y
e
y
) (k
2
x
k
2
y
)e
z
|
= kk
t
p
ez
(k
x
; k
y
)p(k
x
; k
y
)a
TM
p (22)
h
TM
(k
x
; k
y
) =
0
kck e
z
p
ez
=
0
kc(k
y
e
x
k
x
e
y
)p
ez
=

0
_
kk
t
p
ez
(k
x
; k
y
)s(k
x
; k
y
)b
TM
s. (23)
The angular field spectra are shown to be simple functions of
the intrinsic vectors and the scalar angular electric Hertz spec-
trum. In Eq. (23), the magnetic vector only depends on s,
which has no z component, according to Eq. (3); so therefore
the fields given by Eq. (22,23) are indeed TM. This is in com-
plete agreement with Eqs. (10) which show that the magnetic
field has no z component. [Eq. (10) is not written in the Four-
ier domain, but for the polarizations this does not matter.]
E. TE Case
In a similar way as in the former case, this corresponds to:
p
e
= 0
p
m
= p
mz
e
z
.
With this substitution, Eqs. (14) and (15) become, making use
of the transformations (5):
e
TE
(k
x
; k
y
) = kck e
z
p
mz
= kc(k
y
e
x
k
x
e
y
)p
mz
= ckk
t
p
mz
(k
x
; k
y
)s(k
x
; k
y
)a
TE
s (24)
P. Muys Vol. 29, No. 5 / May 2012 / J. Opt. Soc. Am. B 993
h
TE
(k
x
; k
y
) =
1

0
p
mz
[k
z
(k
x
e
x
k
y
e
y
) (k
2
x
k
2
y
)e
z
|
=
1

0
kk
t
p
mz
(k
x
; k
y
)p(k
x
; k
y
)b
TE
p. (25)
Hence we see again that the spectral vectors have become
very simple functions of the intrinsic vectors and the scalar
spectral magnetic Hertz potential.
Note that the magnitudes of the magnetic spectral vectors
are just proportional copies of the magnitudes of the electric
spectral vectors.
F. Spectra of the Transverse Hertz Vectors
The fact that the spectral vectors only have a z dependence
further simplifies their general polarization expressions (18)
and (19). Comparing Eq. (22) with Eq. (17) gives
e
TM
= k
2

= kk
t
p
ez
p;
so we deduce that

(k
x
; k
y
) =
k
t
k
p
ez
(k
x
; k
y
)p. (26)
In the same way, we find:

(k
x
; k
y
) =
k
t
k
p
mz
(k
x
; k
y
)p. (27)
So finally in the spectral domain, the relations between the
fields and their Hertz vectors is
e
TM
= k
2

h
TM
= kc
0
k
e

e
TE
= kck
m

h
TE
=
k
2

. (28)
G. Angular Spectral Representation
The electric field of the beam propagating along the z axis has
been decomposed in its TE and TM components:
E(R) = E
TM
(R) E
TE
(R)
and its angular spectrum, given by Eq. (22) and (24) in
e(k
x
; k
y
) = e
TM
(k
x
; k
y
) e
TE
(k
x
; k
y
).
This means that the angular spectral representations of the
fields look like
E
TM
(R) =
1
(2)
2
_

(k
x
; k
y
) exp(jk:R)dk
x
dk
y
H
TM
(R) =
c
0
(2)
2
k
_

k
e

(k
x
; k
y
) exp(jk:R)dk
x
dk
y
E
TE
(R) =
c
(2)
2
k
_

k
m

(k
x
; k
y
) exp(jk:R)dk
x
dk
y
H
TE
(R) =
1
(2)
2

0
_

(k
x
; k
y
) exp(jk:R)dk
x
dk
y
. (29)
Or, alternatively,
E
TM
(R)=
1
(2)
2
k
_

k
t
p
ez
(k
x
;k
y
)pexp(jk:R)dk
x
dk
y
H
TM
(R)=
1
(2)
2

0
_ _

k
t
p
ez
(k
x
;k
y
)sexp(jk:R)dk
x
dk
y
E
TE
(R)=
c
(2)
2
_

k
t
p
mz
(k
x
;k
y
)sexp(jk:R)dk
x
dk
y
H
TE
(R)=
1
(2)
2
k

0
_

k
t
p
mz
(k
x
;k
y
)pexp(jk:R)dk
x
dk
y
. (30)
For clarity and brevity, we have suppressed in the above in-
tegrals the dependency of the intrinsic vectors on k
x
and k
y
.
Equation (30) show the electric and magnetic fields as func-
tions of the angular spectra of the two Hertz vectors and of the
intrinsic polarization vectors s and p. In the next paragraph,
we will express the fields directly using their angular spectra.
This will lead to a relation between the angular spectra of the
field vectors and of the Hertz vectors.
4. ANGULAR SPECTRUM OF THE VECTOR
FIELDS
The angular spectrum representation of the vector electric
field has been given in [9], expressed in Cartesian coordinates
as:
E
x
(x; y; z) =
1
4
2
k
2
_

A
x
(k
x
; k
y
) exp(jk:R)dk
x
dk
y
E
y
(x; y; z) =
1
4
2
k
2
_

A
y
(k
x
; k
y
) exp(jk:R)dk
x
dk
y
E
z
(x; y; z) =
1
4
2
k
2
_

_
k
x
k
z
A
x
(k
x
; k
y
)

k
y
k
z
A
y
(k
x
; k
y
)
_
exp(jk:R)dk
x
dk
y
. (31)
Here again, we see that A
z
in fact is a function of A
x
and A
y
.
Following another strategy to solve the same vectorial
wave equation, [1] started from the modal representation,
as commonly used in guided wave problems, combining it
with the angular spectrum representation. The advantage of
this approach is that it clearly identifies and separates the con-
tributions of the transverse electric fields (TE and TM) right
from the start. In laboratory Cartesian coordinates, this finally
leads to
E(R) = E
TM
(R) E
TE
(R)
= E
x
(R)e
x
E
y
(R)e
y
E
z
(R)e
z
E
TM
(R) =
1
4
2
k
2
_

_
k
z
kk
t
(k
x
e
x
k
y
e
y
)
k
t
k
e
z
_
A
TM
(k
x
; k
y
) exp(jk:R)dk
x
dk
y
E
TE
(R) =
1
4
2
k
2
_

_
k
y
k
t
e
x

k
x
k
t
e
y
_
A
TE
(k
x
; k
y
) exp(jk:R)dk
x
dk
y
. (32)
The angular spectra A
TM
and A
TE
are in fact 2D Fourier trans-
forms defined by the total fields E
x
and E
y
in z = 0 [1]
994 J. Opt. Soc. Am. B / Vol. 29, No. 5 / May 2012 P. Muys
A
TM
(k
x
; k
y
) = k
2
_

k
k
t
k
z
(k
x
E
x
k
y
E
y
) exp[j(k
x
x k
y
y)|dxdy
A
TE
(k
x
; k
y
) = k
2
_

1
k
t
(k
y
E
x
k
x
E
y
) exp[j(k
x
x k
y
y)|dxdy; (33)
but due to the vector character of the fields, the mathematical
expressions of the integrands are more complex than in the
scalar case. Note again that we only need to know E
x
and
E
y
. So [1] was able to separate the propagating field into
two transverse contributions, which [2,4] did not indicate in
this way. We will see, however, that in [4], essentially the same
result was obtained. Expressions for the magnetic field are
not given by [1]. It is nevertheless clear that E
TE
is lacking
a z component, as it should for the electric field of a TE mode.
Again using the vector transformations (4), these expres-
sions (32) now take on a very simple form:
E
TM
(R) =
1
4
2
k
2
_

pA
TM
(k
x
; k
y
) exp(jk:R)dk
x
dk
y
E
TE
(R) =
1
4
2
k
2
_

sA
TE
(k
x
; k
y
) exp(jk:R)dk
x
dk
y
(34)
so that the vectorial angular spectrum of the total vectorial
electric field E(R) = E
TM
(R) E
TE
(R) is given by
A(k
x
; k
y
) = A
TM
p A
TE
s; (35)
which is exactly the form mentioned in [4], although in a
slightly different format, denoted there as A = A
p
p A
s
s
without explicitly pointing to the transverse character of
the contributions A
s
and A
p
.
The same reasoning can be repeated for the magnetic
angular spectra:
B(k
x
; k
y
) = B
TM
s B
TE
p. (36)
The vectorial angular spectrum representation of the total
fields now become compactly
E(R) =
1
4
2
k
2
_

A(k
x
; k
y
) exp[j(k
x
x k
y
y k
z
z)|dk
x
dk
y
H(R) =
1
4
2
k
2
_

B(k
x
; k
y
) exp[j(k
x
x k
y
y k
z
z)|dk
x
dk
y
.
Now, going back to the spectral Hertz vectors, Eqs. (2225),
and comparing these with Eq. (30,34,35), we see that
a
TM
= A
TM
b
TM
= B
TM
a
TE
= A
TE
b
TE
= B
TE
.
This means that the angular spectra of the field vectors and of
the Hertz vectors are related as
A
TM
= kk
t
p
ez
B
TM
=

0
_
kk
t
p
ez
=

0
_
A
TM
(37)
A
TE
= ckk
t
p
mz
=

0
_
B
TE
B
TE
=
1

0
kk
t
p
mz
.(38)
Alternatively, in function of the spectra of the transverse Hertz
potentials:
A
TM
= k
2

B
TM
=

0
_
k
2

A
TE
=

0
_
k
2

B
TE
= k
2

.
5. AXIALLY SYMMETRIC FIELDS
We now work out the general Eq. (34) for the special case of
axial symmetry of the fields around the z axis.
E(R) = E
r
(r; z)e
r
E

(r; z)e

E
z
(r; z)e
z
= E
TM
(r; z) E
TE
(r; z).
Both [1] and [2] have considered this case. The angular spec-
trum is independent of the azimuthal angle in reciprocal k
space and only depends on k
t
. Equation. (34) becomes [1]
E
TM
(r; z) =
1
2k
3
_
k
0
[jk
z
J
1
(k
t
r)e
r
k
t
J
0
(k
t
r)e
z
|k
t
A
TM
(k
t
) exp[jk
z
z|dk
t
E
TE
(r; z) =
je

2k
2
_
k
0
A
TE
(k
t
)J
1
(k
t
r)k
t
exp[jk
z
z|dk
t
; (39)
where the 2D Fourier transforms for the angular spectra now
reduce to 1D Hankel transforms
A
TM
(k
t
) =
2k
3
j

k
2
k
2
t
_
_

0
E
r
(r; z = 0)J
1
(k
t
r)rdr
A
TE
(k
t
) = 2jk
2
_

0
E

(r; z = 0)J
1
(k
t
r)rdr. (40)
J
0
(z) and J
1
(z) are the well known Bessel functions. If we re-
write Eq. (39) by just keeping the vectorial part and the dif-
ferentials in place, and absorbing the other contributions in
the symbol (.), we can more clearly see how the unit vectors
are transformed by changing from Cartesian to cylindrical co-
ordinates. Note that e

as integration variable in k space is

transformed into e

in R space, which is independent of k

x
and k
y
and hence can be brought in front of the integration
sign for the TE-field:
E
TE
=
__
s(:)dk
x
dk
y
=
__
e

(:)dk
x
dk
y
= e

_
(:)dk
t
.
The electric TE-field is purely azimuthally polarized. For the
TM-field, this simplification does not occur:
P. Muys Vol. 29, No. 5 / May 2012 / J. Opt. Soc. Am. B 995
E
TM
=
__
p(:)dk
x
dk
y
=
__ _
k
z
k
e
r

k
t
k
e
z
_
(:)dk
x
dk
y
=
_ _
J
1
k
z
k
e
r
J
0
k
t
k
e
z
_
(:)dk
t
.
The TM-vector keeps its transverse and longitudinal contribu-
tions in place, although now with other weighting coefficients,
which are Bessel functions. Also here, the unit vectors can be
brought in front of the integration symbol, since they are in-
dependent of k
t
.
6. NONDIFFRACTING VECTORIAL BEAMS
The best known nondiffracting scalar beam in the laser litera-
ture is the Bessel beam J
0
(k
t
r). This beam is linearly polar-
ized. As a short application of Eqs. (39,40), we will
consider the diffraction-free propagation of an azimuthally
polarized vectorial beam and extend in this way the scope
of [10], which analyzed scalar beams.
The general solution of the scalar wave equation for a
diffraction-free beam is given [10] as
u(x; y; z) = exp(jk
z
z)
_
2
0
a() exp[jk
t
(x cos y sin )|d
and is known under the name Whittaker integral in the math-
ematical literature. Physically, it represents a superposition of
plane waves with amplitude coefficient a() and with their
wave vector situated on a cone with base radius k
t
and height
k
z
. In other words, their angular spectrum in reciprocal k
space is given by A = (k
t
k

t
). In the vectorial case, we
use this same angular spectrum to describe nondiffracting
vector beams and substitute it in the TE-field expression of
Eq. (14) to arrive immediately at
E
TE
(r; z) =
j
2
J
1
(k
t
r) exp(jk
z
z)e

. (41)
The vectorial Bessel beam (41) is hence nondiffracting since
the transverse part is not depending on the propagation
coordinate z. This result was also obtained in [11], but by
the method of separation of the variables of the paraxial wave
equation. A mathematically equivalent statement is that the J
1
Bessel function can be represented as a Whittaker integral. Its
explicit form can be found in [3].
7. CONCLUSIONS
In summary, we have pointed out the relevance of using an
intrinsic coordinate system to unify the existing angular
spectrum representations. We have given the transformation
formulas from intrinsic to Cartesian coordinates. The intro-
duction of intrinsic coordinates much simplifies the angular
spectral representation of vectorial beams. We have intro-
duced the Hertz vector potential oriented along the z axis
and have combined it with the intrinsic coordinates to repre-
sent TE and TM beams in a particularly simple and elegant
form. This leads also to the mathematical link between the
angular spectra of the fields and of their Hertz vector. Next,
we derived expressions for a diffraction-free vectorial beam,
without invoking the paraxial approximation.
REFERENCES
1. H. Guo, J. Chen, and S. Zhuang, Vector plane wave spectrum of
an arbitrary polarized electromagnetic wave, Opt. Express 14,
20952100 (2006).
2. C.-F. Li, Integral transformation solution of free-space cylind-
rical vector beams and prediction of modified Bessel-Gauss
vector beams, Opt. Lett. 32, 35433545 (2007).
3. A. Nesterov and V. Niziev, Vector solution of the diffrac-
tion task using the Hertz vector, Phys. Rev. E 71, 046608 (2005).
4. C.-F. Li, Unified theory for Goos-Hnchen and Imbert-Fedorov
effects, Phys. Rev. A 76, 013811 (2007).
5. E. Essex, Hertz vector potentials of electromagnetic theory,
Am. J. Phys. 45, 10991101 (1977).
6. J. Van Bladel, Electromagnetic Fields (McGraw-Hill, 1964).
7. P. Varga and P. Trk, Exact and approximate solutions of
Maxwells equations for a confocal cavity, Opt. Lett. 21,
15231525 (1996).
8. C. Someda, Electromagnetic Waves, 2nd ed. (CRC, 2006).
9. D. Rhodes, On the stored energy of planar apertures, IEEE
Trans. Antennas Propag. 14, 676683 (1966).
10. J. Durnin, Exact solutions for nondiffracting beams.I. The
scalar theory, J. Opt. Soc. Am. A 4, 651654 (1987).
11. A. Nesterov and V. Niziev, Propagation features of beams with
axially symmetric polarization, J. Opt. B 3, S215S219 (2001).
996 J. Opt. Soc. Am. B / Vol. 29, No. 5 / May 2012 P. Muys