ECE 604, Lecture 33

Fri, April 12, 2019

Contents
1 High Frequency Solutions 2
1.1 Tangent Plane Approximations . . . . . . . . . . . . . . . . . . . 2
1.2 Fermat’s Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Generalized Snell’s Law . . . . . . . . . . . . . . . . . . . . . . . 4

2 Gaussian Beam 5
2.1 Derivation of the Paraxial/Parabolic Wave Equation . . . . . . . 5
2.2 Finding a Closed Form Solution . . . . . . . . . . . . . . . . . . . 6
2.3 Other solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Printed on April 12, 2019 at 15 : 28: W.C. Chew and D. Jiao.

1
ECE 604, Lecture 33 Fri, April 12, 2019

1 High Frequency Solutions

1.1 Tangent Plane Approximations
We have learnt that reflection and transmission of waves at a flat surface can
be solved in closed form. The important point here is the physics of phase
matching. Due to phase matching, we have the law of reflection, transmission
and Snell’s law.
When a surface is not flat anymore, there is no closed form solution. But
when a surface is curved, an approximate solution can be found. This is obtained
by using a local tangent-plane approximation when the radius of curvature is
much larger than the wavelength. Hence, this is a good approximation when
the frequency is high or the wavelength is short. This is similar in spirit that we
can approximate a spherical wave by a local plane wave at the spherical wave
front when the wavelength is short.
When the wavelength is short, phase matching happens locally, and the law
of reflection, transmission, and Snell’s law are satisfied approximately as shown
in Figure 1. The tangent plane approximation is the basis for the geometrical
optics (GO) approximation. In GO, light waves are replaced by light rays.
The reflection and transmission of these rays at an interface is estimated with
the tangent plane approximation. This is also the basis for lens or ray optics
from which lens technology is derived. It is also the basis for ray tracing for
high-frequency solutions.

Figure 1:

2
ECE 604, Lecture 33 Fri, April 12, 2019

1.2 Fermat’s Principle

Fermat’s principle (1600s) says that a light ray follows the path that takes the
shortest time between two points.1 Since time delay is related to the phase
delay, and that a light ray can be locally approximated by a plane wave, this
can be stated that a plane wave follows the path that has a minimal phase
delay. This principle can be used to derive law of reflection, transmission, and
refraction for light rays. It can be used as the guiding principle for ray tracing.

Figure 2:

Given two points A and C in two different half spaces as shown in Figure 2.
Then the phase delay between the two points, per Figure 2, can be written as

P = ki · ri + kt · rt (1.1)

As this is the shortest path according to Fermat’s principle, another other path
will be longer. In other words, if B were to move to another point, a longer
path will ensue, or that B is the stationary point of the path length or phase
delay. Specializing (1.1) to a 2D picture, then the phase delay as a function of
xi is stationary. In this Figure 2, we have xi + xt = const. Therefore, taking
1 This eventually give rise to the principle of least action.

3
ECE 604, Lecture 33 Fri, April 12, 2019

the derivative of (1.1) with respect to xi , one gets

∂P
= 0 = ki − kt (1.2)
∂xi
The above yields the law of refraction that ki = kt , which is just Snell’s law. It
can also be obtained by phase matching.

1.3 Generalized Snell’s Law

Figure 3: Courtesy of Capasso’s group.

Metasurfaces are prevalent these days due to our ability for nano-fabrication and
numerical simulation. One of them is shown in Figure 3. Such a metasurface
can be thought of as a phase screen, providing additional phase shift for the light
as it passes through it. Moreover, the added phase shift can be controlled to be
a function of position due to advent in fabrication technology and commercial
software for numerical simulation.
To model this phase screen, we can add an additional function Φ(x, y) to
(1.1), namely that
P = ki · ri + kt · rt − Φ(xi , yi ) (1.3)
Now applying Fermat’s principle that there should be minimal phase delay, and
taking the derivative of the above with respect to xi , one gets
∂P ∂Φ(xi , yi )
= ki − kt − =0 (1.4)
∂xi ∂xi
The above yields that the generalized Snell’s law (Capasso) that
∂Φ(xi , yi )
ki − kt = (1.5)
∂xi

4
ECE 604, Lecture 33 Fri, April 12, 2019

It yields the fact that the transmitted light can be directed to other angles
due to the additional phase screen.

2 Gaussian Beam
We have seen previously that in a source free space

∇2 A + ω 2 µεA = 0 (2.1)

∇2 Φ + ω 2 µεΦ = 0 (2.2)

The above are four scalar equations with the Lorenz gauge

∇ · A = −jωµεΦ (2.3)

connecting A and Φ. We can examine the solution of A such that

A(r) = A0 (r)e−jβz (2.4)

where A0 (r) is a slowly varying function while e−jβz is rapidly varying in the
z direction. This is primarily a quasi-plane wave propagating predominantly in
the z-direction. We know to be the case in the far field of a source, but let us
assume that this form persists less than the far field. Taking the x component
of (2.4), we have2

Ax (r) = Ψ(r)e−jβz (2.5)

where Ψ(r) = Ψ(x, y, z) is a slowly varying envelope function of x, y, and z.

2.1 Derivation of the Paraxial/Parabolic Wave Equation

Substituting (2.5) into (2.1), and taking the double z derivative first, we arrive
at
∂2 
 2 
−jβz
 ∂ ∂ 2
Ψ(x, y, z)e = Ψ(x, y, z) − 2jβ Ψ(x, y, z) − β Ψ(x, y, z)
∂z 2 ∂z 2 ∂z
(2.6)

Consequently, after substituting the above into the x component of (2.1), we

obtain an equation for Ψ(r), the slowly varying envelope as

∂2 ∂2 ∂ ∂2
Ψ + Ψ − 2jβ Ψ + Ψ=0 (2.7)
∂x2 ∂y 2 ∂z ∂z 2
2 Also, the wave becomes a transverse wave in the far field, and keeping the transverse

component suffices.

5
ECE 604, Lecture 33 Fri, April 12, 2019

When β → ∞ , or in the high frequency limit,

∂ ∂2
2jβ Ψ  Ψ (2.8)
∂z ∂z 2

In the above, we assume the envelope to be slowly varying and β large, so that
|βΨ|  |∂/∂zΨ|. And then (2.7) can be approximated by

∂2Ψ ∂2Ψ ∂Ψ
2
+ 2
− 2jβ ≈0 (2.9)
∂x ∂y ∂z
The above is called the paraxial wave equation. It is also called the parabolic
wave equation.3 It implies that the β vector of the wave is approximately
parallel to the z axis, and hence, the name.

2.2 Finding a Closed Form Solution

A closed form solution to the paraxial wave equation can be obtained by a
simple trick4 . It is known that
0
e−jβ|r−r |
Ax (r) = (2.10)
4π|r − r0 |
is the solution to

∇ 2 Ax + β 2 Ax = 0 (2.11)

if r 6= r0 . If we make r0 = −ẑjb, a complex number, then (2.10) is always a

solution to (2.10) for all r, because |r − r0 | =6 0 always. Then
p
|r − r0 | = x2 + y 2 + (z + jb)2
1/2
x2 + y 2

≈ (z + jb) 1 + + ...
(z + jb)2
x2 + y 2
≈ (z + jb) + + ..., |z + jb| → ∞ (2.12)
2(z + jb)
And then
e−jβ(z+jb) −jβ 2(z+jb)
x2 +y 2
Ax (r) ≈ e (2.13)
4π(z + jb)

By comparing the above with (2.5), we can identify

jb −jβ 2(z+jb)
x2 +y 2
Ψ(x, y, z) = A0 e (2.14)
z + jb
3 The paraxial wave equation, the diffusion equation and the Schrodinger equation are all

classified as parabolic equations in mathematical parlance.

4 Introduced by Georges [Link] of UIUC.

6
ECE 604, Lecture 33 Fri, April 12, 2019

By separating the exponential part into the real part and the imaginary part,
and writing the prefactor in terms of amplitude and phase, we have
2 2 2 2
A0 −1 z x +y
−jβ 2(z x +y
2 +b2 ) z −bβ 2(z 2 +b2 )
Ψ(x, y, z) = p ej tan (b)
e e (2.15)
1 + z 2 /b2

The above can be rewritten as

A0 x2 +y 2 x2 +y 2
Ψ(x, y, z) = p e−jβ 2R e− w2 ejψ (2.16)
1 + z 2 /b2

where
z2 z 2 + b2
 
2b z 
w2 = 1+ , R= , ψ = tan−1 (2.17)
β b2 z b

For a fixed z, the parameters w, R, and ψ are constants. Here, w is the beam
q
2b
waist which varies with z, and it is smallest when z = 0, or w = w0 = β .
And R is the radius of curvature of the constant phase front. This can be
appreciated by studying a spherical wave front e−jβR , and make a paraxial
wave approximation, namely, x2 + y 2  z 2 to get
 2 2 1/2
−jβR −jβ(x2 +y 2 +z 2 )1/2 −jβz 1+ x z+y
2
e =e =e
x2 +y 2 x2 +y 2
≈ e−jβz−jβ 2z ≈ e−jβz−jβ 2R (2.18)

In the last approximation, we assume that z ≈ R in the paraxial approximation.

The phase ψ changes rapidly with z.
A cross section of the electric field due to a Gaussian beam is shown in
Figure 4.

Figure 4: Electric field of a Gaussian beam in the x − z plane frozen in time.

The wave moves to the right as time increases; b/λ = 10/6 (Courtesy of Haus,
Electromagnetic Noise and Quantum Optical Measurements).

7
ECE 604, Lecture 33 Fri, April 12, 2019

2.3 Other solutions

In general, the paraxial wave equation has solution of the form5
 1/2  
2 −N/2 1 2 2 2 β 2 2
Ψnm (x, y, z) = 2 e−(x +y )/w e−j 2R (x +y ) ej(m+n+1)Ψ
πn!m! w
(2.19)
 √   √ 
·Hn x 2/w Hm y 2/w
(2.20)

where Hn (ξ) is a Hermite polynomial of order n. The solution can also be

express in terms of Laguere polynomials, namely,
 1/2
2 1 β 2 2 2
Ψnm (x, y, z) = min(n, m)! e−j 2R ρ − e−ρ /w e+j(n+m+1)Ψ ejlφ
πn!m! w
√ !  2
min(n,m) 2ρ n−m 2ρ
(−1) Lmin(n,m)
w w2
(2.21)

where Lkn (ξ) is the associated Laguerre polynomial.

These gaussian beams have rekindled recent excitement in the community
because, in addition to carrying spin angular momentum as in a plane wave,
they can carry orbital angular momentum due to the complex transverse field
distribution of the beams.6 They harbor potential for optical communications
as well as optical tweezers to manipulate trapped nano-particles. Figure 5 shows
some examples of the cross section (xy plane) field plots for some of these beams.
5 See F. Pampaloni and J. Enderlein.
6 See D.L. Andrew, Structured Light and Its Applications and articles therein.

8
ECE 604, Lecture 33 Fri, April 12, 2019

Figure 5: Courtesy of L. Allen and M. Padgett’s chapter in J.L. Andrew’s book

on structured light.