30.

Diffraction and the Fourier Transform

Diffraction examples
Diffraction by an edge
Arago spot

The far-field
Fraunhofer Diffraction
Some examples

Simeon Poisson Francois Arago

(1781 - 1840) (1786 - 1853)
 Reminder: Fresnel-Kirchoff diffraction
aperture
(x0, y0)
Coordinates: (x1, y1)
• the plane of the aperture: x1, y1 r01  z 2   x0  x1    y0  y1 
2 2

• the plane of observation: x0, y0

z
(a distance z downstream)
observation
region

   x12  2 x0 x1   y12  2 y0 y1   
E  x0 , y0  
 exp  jk 
  2 z

2 z
  Aperture  x1 , y1  E  x1 , y1  dx1dy1
 

Quadratics in the exponent: a messy integral

A plane wave incident on a sharp edge

Fresnel diffraction integral, in one dimension:




1    x  x 2  
E ( x0 )  exp  jkz exp  jk  0 1   dx1
j z   2 z  
x1  0

geometrical exact diffraction result

shadow
Intensity

The irradiance
exactly at the edge is
25% of the value far
from the edge.

position x0
Diffraction by an Edge

Light passing
by an edge

Electrons passing
by an edge
An interesting manifestation of
diffraction effects: The Spot of Arago
If a beam encounters a circular “stop”, it develops a hole, which fills in as
it propagates and diffracts:
r r
Eincident E(r,0) E(r,z)

beam stop Interestingly, the

hole fills in from
the center first!
The Spot of Arago
Why does it happen?
According to Huygen’s principle,
every point on the perimeter of the
disc radiates a spherical wave.
incoming
plane
wave

Obviously, the distance from each

circular spherical wave’s point of origin to a
beam stop point on the central axis of the disc
is the same.

Constructive interference on the center line!

In 1818, Poisson used Fresnel’s theory to predict this phenomenon. He

regarded this as proof that Fresnel’s wave theory was nonsense, and
that light must be a particle and not a wave. But almost immediately,
Arago experimentally verified Poisson’s prediction.
 Simplification of Fresnel diffraction
Recall the Fresnel diffraction result:


  (2 x0 x1  2 y0 y1 ) ( x12  y12 )  
E  x0 , y0   exp  jk     Aperture(x1 , y1 ) dx1dy1
  2z 2 z  

Let D be the largest dimension of the aperture: D2 = max(x12 + y12).

Our first step, which allowed us to obtain the Fresnel result, was the
paraxial approximation:
D
z  D or  1
z
Note that this approximation does not contain the wavelength.

A more severe approximation is suggested by noticing that the integral

simplifies A LOT if only we could neglect the quadratic terms x12 and y12.

If kD2/2z << 1, then we could do that…

Approximation #2: involving the wavelength
This new approximation, which does contain , is equivalent to:
D 2 
 
z kD  D
So D/z is not merely required to be less than 1, but is required to be
less than D, which is generally smaller than one for visible light.

Note: if the aperture is a slit of width D = 2b, this condition becomes:

2b 1 4 b 2
 or  1
z kb z

b2
Recall our definition of the Fresnel number for a slit of width 2b: N
z

we see that this new approximation is equivalent to: 4N << 1.

Approximation #2: Fraunhofer diffraction
kD 2
Apply this approximation:  1 to the Fresnel diffraction result:
2z


  (2 x0 x1  2 y0 y1 ) ( x12  y12 )  
E  x0 , y0   exp  jk     Aperture(x1 , y1 ) dx1dy1
  2z 2 z  

In this case, the quadratic terms are tiny, so we can ignore them.

As in Fresnel diffraction, we’ll typically assume a plane wave incident

field, we’ll neglect the phase factors, and we’ll explicitly write the
aperture function in the integral:


 jk 
E  x0 , y0   exp   x0 x1  y0 y1   Aperture(x1, y1) dx1dy1
 z 

Fraunhofer Diffraction: when the quadratic terms can be ignored.

Joseph von Fraunhofer
Fraunhofer achieved fame by developing recipes for
the world’s finest optical glass. He also invented
precise methods for measuring dispersion of glass,
and discovered more than 500 different absorption
lines in sunlight, most due to specific atomic or
molecular species at the sun’s surface.

Joseph von Fraunhofer

1787-1826
These are still known
as Fraunhofer lines.

He had almost nothing to do with Fraunhofer diffraction that we’re

discussing today. He did, however, invent the diffraction grating,
which we will discuss next lecture.
He was almost an exact contemporary of Augustin Fresnel
(1788 – 1827). But it is unlikely that they ever met.
 The Fraunhofer regime
How far away is far enough? We must have both z >> D and z >> D2/.

Example #1: green light ( = 0.5 m) Example #2: microwaves ( = 3 cm)

If D = 1 millimeter, then: If D = 10 centimeters, then:

 D 2  1000   10 
2 2

z    6.3 meters
D 2
z    1 meter
 0.5  3

If D = 10 microns, then: If D = 1 millimeter, then:

 10   1
2
D 2
D 2 2
z    630 microns z    0.1 millimeters
 0.5  30
But notice that z >> 1 mm is required
in this case for the paraxial
approximation to also be true.
Fraunhofer diffraction is a Fourier transform


 jk 
E  x0 , y 0   exp    x0 x1  y0 y1   Aperture( x1 , y1 ) dx1dy1
 z 

This is just a Fourier Transform! (actually, two of them, in two variables)

Interestingly, it’s a Fourier Transform from

position, x1, to another position variable, x0 (in
another plane, i.e., a different z position).

Usually, the Fourier “conjugate variables” have

reciprocal units (e.g., t and , or x and k).

The conjugate variables here are really x1 and kx0/z,

which do have reciprocal units.
Fraunhofer diffraction is a Fourier transform


  kx0  
In one dimension: E  x0   exp  j   x1  Aperture( x1 )dx1
  z  
aperture
(x0, y0)

(x1, y1)
r01  z 2   x0  x1    y0  y1 
2 2

z observation
region

So, the light in the Fraunhofer regime (the “far field”) is simply the
Fourier Transform of the apertured field!

Knowing this makes the calculations a lot easier…

Fraunhofer diffraction pattern for a slit
In this case, the problem is a single
Fourier transform (in x), rather than two
of them (in x and y):


 jk
E  x0   exp    x0 x1   Aperture  x1 dx1
 z 

The aperture function is simple:

1 b  x1  b
Aperture  x1   
2b
0 otherwise
But we know that the Fourier transform of a rectangle written here
in terms of the
function (of width 2b) is a sinc function: Fresnel number

sin  kx0b z  sin  2 Nx0 b 

2
N = b /z

FT  Aperture  x1    
kx b z
0 2 Nx b 0

How very satisfying! This is exactly the answer we saw last lecture,
for the Fresnel diffraction result in the limit of very large z.
 Fraunhofer Diffraction
from a Square Aperture

A square aperture (edge length = 2b)

just gives the product of two sinc
functions in x and in y. Just as if it were
two slits, orthogonal to each other.

Diffracted
FT  Square Aperture  x1 , y1    irradiance

sin  kx0b z  sin  ky0b z 


kx0b z ky0b z
Diffracted
field
 Fraunhofer diffraction from
a circular aperture
A plot of J1(r)/r

The 2D Fourier transform of a circular

aperture, radius = b, is given by a first zero at
r = 3.83
Bessel function of the first kind:

J1  k  b z 
FT Circular aperture  x1 , y1   
k b z
0 7.5 15
where   x  y is the radial
2 2
1 1
coordinate in the x1 - y1 plane.

Most of the energy falls in the central

region, for values kb/z < 3.83

Diffracted E-field plotted in 2D

The Airy pattern
A circular aperture yields a diffracted pattern
known as an “Airy pattern” or an “Airy disc”.

Sir George Biddell Airy

1801-1892

3.83


J1  r 
2

dr
J1  k  b z 
2
r
Diffracted Irradiance 
k b z
0

 0.838


J1  r 
2

The central spot contains about 84% of the total energy: dr

r
0
The size of the Airy pattern
How big is the central spot, where most of the energy is found?

Define the size of the spot as spot,

the radial distance from the center
spot to the first zero:

k  spot b z  3.83
3.83 z z
 spot   3.83
kb 2 b

The diameter of the aperture is D = 2b, so we can write:

3.83  z z
 spot   1.22
 D D
Diffraction from small and large
circular apertures

2spot
Small aperture

This is a good illustration

of the Scale Theorem!

1
 spot 
D
Large aperture
Fraunhofer Diffraction: an interesting example
Randomly placed identical holes yield a diffraction pattern whose
gross features reveal the shape of the holes.

Hole Diffraction
pattern pattern

Square holes

Round holes
The Fourier Transform of a random
array of identical tiny objects
Define a random array of two-dimensional delta-functions:
n
Rand ( x, y )    ( x  xi , y  yi ) Sum of rapidly
i 1
varying sinusoids
n (looks like noise)
F  Rand ( x, y )   exp[ j (k x xi  k y yi )]
i 1

If OneHole(x,y) is the shape of an individual tiny hole, then a random

array of identically shaped tiny holes is: convolution

Holes ( x, y )  Rand ( x, y ) * OneHole( x, y )

The Fourier Transform of a random array of identically shaped tiny holes is:

F {Holes ( x, y )}  F {Rand ( x, y )} F {OneHole( x, y )}

 
rapidly slowly
varying varying
Diffraction from Slit Diffraction
Pattern Pattern
multiple slits

Infinitely many equally spaced

slits (a Shah function!) yields a
far-field pattern which is the
Fourier transform; that is, the
Shah function.
Laser speckle is a diffraction pattern.
When a laser illuminates a
rough surface or passes
through a region where it
can scatter a little bit, the
result is a “speckle”
pattern. It is a diffraction
pattern from the very
complex surface.

Don’t try to do this Fourier

Transform at home.

But people do. Computing

the inverse FT of a speckle
pattern can give information
about the degree of
roughness of a surface.
There are situations where this Fourier
transform idea is not so useful
Example: light passing by an edge

In this case, the effective “width” of the

slit, D, is infinite. It is impossible to reach
the Fraunhofer regime of z >> D2/.

geometrical exact diffraction result

shadow
Intensity

position x0