Diffraction theory

Scalar diffraction theory

History of diffraction theory
From vector to a scalar theory
Some mathematical Preliminaries
Hemholtz equation
Green’s theorem
The Kirchhoff formulation of diffraction by a planar screen
The integral theorem of Helmholtz and Kirchhoff
The Kirchhoff boundary conditions
Fresnel-Kirchhoff
Fresnel Kirchhoff diffraction formula
The Rayleigh-Sommerfeld (R-S) formulation of diffraction
Alternative Green’s function
The R-S diffraction Formula
Comparison between Kirchhoff and R-S theories
G
Generalization
li ti tto nonmonochromatic
h ti waves
Angular spectrum of a plane wave
Physical interpretation of the angular spectrum
Propagation of the angular spectrum
Effects of a diffraction aperture
p on the angular
g spectrum
p
Propagation phenomenon as a linear spatial filter

1
Diffraction
• Diffraction is a phenomena in the realm of physical optics.
Applicable to all waves such as acoustic and EM waves.
• Diffraction is a limiting factor on data processing and imaging
system performance. So we need to understand it to design better
systems.
• Diffraction should not be mistaken with
– Refraction: change of direction of propagation of light due to a
change in index of refraction of the environment
– Penumbra: finite extend of a source causes the light transmitted
from an aperture to spread away from it. There is no bending of
light involved in Penumbra effect.
• Diffraction (Sommerfeld): any deviation of light rays from rectilinear
that cannot be interpreted as reflection or refraction
refraction.
• Diffraction is caused by confinement of the lateral extend of a wave
(obstruction of the wavefront) and its effects are most pronounced
when size of the confinement is comparable to the wavelength of the
light.

2
 History of diffraction theory
• 1665 Grimaldi reported diffraction for the first time.
• 1678 Huygens attempted to explain the phenomenon
– Each point on the wavefront of a disturbance is considered to be a new source of a
“secondary” spherical disturbance. Then the wavefront at later instances can be
found by constructing the “envelope” of the secondary wavelet.
• 1700 Progress
1700s P on wave theory
th was suppressed d by
b theth factf t that
th t Newton
N t favored
f d the
th
corpuscular theory of light (geometrical optics).
• 1804 Thomas Young introduced the concept of interference to the wave theory of light
(production of darkness from light).
• 1818 Augustin Jean Fresnel used the wavelets from Huygens theory and Young’s Young s
interference theory letting the wavelets interfere mutually to calculate distribution of light
in diffraction patterns with excellent accuracy.
• 1860 Maxwell identified light as electromagnetic field.
• 1882 Gustav Kirchhoff put the Fresnel and Maxwell
Maxwell’s s ideas together.
together He made two
assumptions about the boundary values of the light incident on surface of an obstacle
that were not absolutely correct but an approximation and constructed a theory that
exhibited excellent agreement with experimental results. He concluded
– The amplitudes and phases ascribed to the secondary sources of Huygens wavelets
are logical
l i l consequences off th the wave nature
t off th
the lilight.
ht
• 1892 Poincare; 1894 Sommerfeld proved that the boundary values set by Kirchhoff are
inconsistent with one another. So Kirchhoff’s formulation of Huygens-Fresnel principle is
regarded as the first approximation although under most conditions it yields excellent
results
results.
• 1896 Sommerfeld modified the Kirchhoff’s theory using theory of Green’s function. The
result is Rayleigh-Sommerfeld diffraction theory.
• 1923 Kottler: first satisfactory generalization of the vectorial diffraction theory.
3
From vector to a scalar theory I
• In all of these theories light is treated as a scalar phenomenon.
• At boundaries the various components of the electric and magnetic
fi ld are coupled
fields l d th
throughhMMaxwellll equations
ti and
d cannott b
be ttreated
t d
independently.
• We stay away from those situations when using scalar theory.
• Scalar theory yields correct values under two conditions:
– The diffracting aperture must be large compared with a
wavelength.
– The diffracting fields must not be observed too close to the
aperture.
• Our treatment is not g good for some optical
p systems
y such as
diffraction from
– high-resolution gratings
– Small pits on optical recording media
• Read Goodman 3.2 From a Vector to a Scalar Theory

4
From vector to a scalar theory II
F electromagnetic
For l t ti waves propagating
ti ini media
di with
ith
the following properties, an scalar wave equation
is obeyed by all components of the field vectors Ex , E y , Ez , H x , H y , H z .
n 2 ∂ 2 u ( P, t )
∇ u ( P, t ) − 2
2
=0
c ∂t 2
Depth of
u ( P, t ) is any of the scalar field components x, y, z at time t. penetration in the
media is few
Properties: wavelengths.
linear; if u1 ( P, t ) and u2 ( P, t ) are solutions to the wave euation, then Not much effect on
total wavefront
α u1 ( P, t ) + β u2 ( P, t ) is a solution, passing through
the aperture
isotropic; properties are ind
independent
ependent of direction of polarization of the wave,
wave
homogeneous; permitivity is constant throughout the region of propagation,
nondispersive; permitivity is independent of frequency over the region of propagation,
nonmagnetic;
ti magnetic bilit iis equall tto μ0
ti permeability
At the boundaries, the above criteria are not met and coupling betwen electric and
magnetic components of the EM wave happens.
We can use the scalar theory if the boundaries are small portion of the
total area through which the wave is passing.

5
The Helmholtz equation
For a monochromatic wave u ( P, t ) the scalar field can be written as
u ( P, t ) = A( P ) cos[2πν t − φ ( P)] = Re{U ( P)e − j 2πν t } where
U ( p) = A( p )e jφ ( P ) space dependent part of the field
e − j 2πν t time dependent part of the field
P can be any of the space coordinates x, y, or z
Substituting the scalar field in scalar wave equation
n 2 ∂ 2u ( P, t )
∇ u ( P, t ) − 2
2
=0
c ∂t 2

− j 2πν t n 2 ∂ 2U ( P)e − j 2πν t

∇ U ( P )e
2
− 2 =0
c ∂t 2

⎛ 2 n 2 (− j 2πν ) 2 ⎞ − j 2πν t n 2 (2πν ) 2 2π

⎜ ∇ U ( P) − U ( P) ⎟ e = 0 with wavenumber k = =
⎝ c 2
⎠ c 2
λ
(∇ 2 + k 2 )U ( P ) = 0 time-independent Helmholtz equation.
Complex amplitude of any monochromatic wave propagating in vacuum
or in homogeneous dielectric media has to obey the Helmholtz equation.

6
Gauss’s Theorem

Gauss's theorem:
connecting surface integral and volume integral of a vector
total outflow of flux
from the volume V

∫V
∇iU dV =
net outflow
∫ S
Uids
of flux per total outflow of flux
unit volume from the surface S

Green's
Green s theorem is a corollary of Gauss's
Gauss s theorem

7
Green’s Theorem: a mathematical tool
If U and G are two scalar functions
∇i(U ∇G ) = U ∇i∇G + ( ∇U )i( ∇G ) -
∇i(G∇U ) = G∇i∇U + ( ∇G )i( ∇U )
∇i(U ∇G − G∇U ) = U ∇ 2G − G∇ 2U
Assuming U , G , their first and second derivatives are continuous
over the volume V and on the surface S enclosing the V

∫∫∫ V
∇i(U ∇G − G∇U )dV = ∫∫∫ (U ∇ 2G − G∇ 2U )dV
V

Using the Gauss's theorem convert the volume integral on LHS

∫∫
S
(U ∇G − G∇U )ids = ∫∫∫ (U ∇ 2G − G∇ 2U )dV
V

If we take the gradiant in the outward normal direction n the LHS

can be written in a scalar form since n || s at everyy point.
p
∂G ∂U
∫∫S ∂n ∂n − = ∫∫∫V ∇ − ∇ U )dV ← Green's theorem
2 2
(U G ) ds (U G G
8
Physical meaning of the Green’s function I

Imagine an inhomogeneous linear differential euation

d 2U ( x) dU ( x)
a2 + a1 + a0U ( x) = V ( x)
dx 2 dx
V ( x) is a driving force
U ( x) is the solution for a known set of boundary conditions (BC)
G ( x) is a solution to the equation with V ( x) → δ ( x − x ') the impulse driving force
and the same BCsBCs.
G ( x) is an impulse response and we can expand U ( x) in terms of Gs
U ( x) = ∫ V ( x '))G ( x - x '))dx '
G ( x) known as the Green's function of the problem.
G ( x) may be regarded as an auxiliary function chosen cleverly to
solve our problem.

9
Physical meaning of the Green’s function II
Imagine an oscillator
d 2U ( x)
a2 2
+ a0U ( x) = V ( x)
dx
V ( x) is a driving force
U ( x) is the solution for a known set of boundary conditions (BC)
G ( x) is impulse response
U ( x) = ∫ V ( x '))G ( x - x '))dx '
x'

The solution is convolution of the driving force with the impulse respons
of the system.
In case of diffraction application of Green's theorem will yeild different
variations of the diffraction theory based on the choice of Green's function.

10
Application of Green’s Theorem in scalar
diffraction theory
Goal: calculation of the complex disterbance U at an observation point
i space, P0 , using
in i G Green's
' Th
Theorem.
∂G ∂G
∫∫S ∂n ∂n− = ∫∫∫V ∇ − ∇
2 2
(U G ) ds (U G G U )dV

Green's theorem is the prime foundation of the scalar diffraction theory.

To apply it to the diffraction problem we need to have a proper choice of
1) an auxiliary
ili ffunction
ti G (G
(Green's
' function)
f ti )
P1
2) a close surface S V n
P0 is an arbitraryy point
p of observation
P1 is an arbitrary point on the surface S
We want solution of the wave equation U ( P0 )
.
P0
ro1
S

at P0 in terms of the value of the solution

and its derivatives on the surface S.
11
The integral theorem of Helmholtz & Kirchhoff
1) Choice of Green
Green's
s function: a unit-amplitude spherical wave expanding
e jkr01
about point P0 (impulse). G has to be a solution of the wave equation. At P1: G ( P1 ) =
r01
2)) Treating
g the discontinuity
y at P0 byy isolating
g it with Sε P1
4
3) New surface & volume: S ' = S + Sε ; V ' = V − πε 3
3 V’
4) Use Green's theorm with Green's function of r01
e jkr01
G ( P1 ) =
r01
and Helmholtz equation (∇ 2 + k 2 )U = 0

Note both G and U are the solutions of the same

.
ε
P0 n
S

wave euation. After all G is the impulse response and Sε n

U is the disturbance. We want to find U ( P0 ) or field after the aperture.
At the linit of ε → 0 we get (follow from Goodman page 41):
1 ⎧ ∂U e jkr01 ∂ e jkr01 ⎫
U ( P0 ) = ∫∫S ⎨⎩ ∂n r01 - U ∂n r01 ⎬⎭ ds
4π
This result is known as the integral thorem of Helmholtz and Kirchhoff. It has important role in
development of the scaler theory of diffraction. U ( P0 ) , the field at point P0 is expressed in terms
of the "boundary values" of the wave on any closed surface surrounding that point.
12
 Fresnel-Kirchhoff diffraction formula I
Problem: diffraction of light by an aperture in an infinite opaque screen.
The field U at P0 behind the aperture is to be calculated.
r01 is the distance from aperture to observation point.
point
Assumptions: r01 >> λ and k >> 1/ r01 Wave
S2

Choice of S : a plane surface plus a impinging S1

spherical cap S = S1 + S 2
e jkr01 n
∑ . P0
R

Choice of Green's function: G = P1

r01
r01
We apply the Helmholtz-Kirchhoff integral theorem
1 ⎛ ∂U ∂G ⎞
U ( P0 ) = ∫∫ ⎜ G − U ⎟ ds to find U ( P0 )
4π S1 + S2 ⎝
∂n ∂n ⎠
Somefeld radiation condition: if the disturbance U vanishes
at least as fast as a diverging spherical wave then:
as R → ∞ contribution of S 2 to U ( P0 ) vanishes.
13
Kirchhoff’s Boundary conditions
The screen is opaque and the aperture is shown by Σ
1) Across the surface Σ, the field and its derivatives are exactly
the same as they would be in the absence off the screen.
2) Over the portion of S1 that lies in the geometrical shadow
of the screen the filed and its derivative are identically zero
zero.
1 ⎛ ∂U ∂G ⎞
U ( P0 ) = ∫∫Σ ⎝⎜ ∂n ∂n ⎠⎟ ds
G − U
4π S2
Wave
Condition 1 is not exactly true since close impinging S1
to the boundaries the field is disturbed.
Condition 2 is not true since the n
∑ .
r01
P0
R

shadow is never perfect and some P1

field will extend behind the walls.
walls
On these portions of S1 :
For Σ >> λ this is an OK approximation. ∂ U ∂ 2U
U = = =0
∂n ∂n 2 14
Fresnel-Kirchhoff diffraction formula II S2
S1
Withe above assumptions and k >> 1/ r01 we arrive at

U ( P0 ) =
1 ⎧e ⎡ ∂U
jkr01
⎤⎫
∫∫∑ ⎨⎩ r01 ⎢⎣ ∂n − jjkU cos((n, r01 ) ⎥⎦ ⎬⎭ ds n
∑ . P0
R

4π r01
P1
If the aperture is illuminated by a single spherical wave
Ae jkr21
U ( P1 ) = located at point P2 , at a distance r21 from
f P1.
r21
If r21 >> λ we can show that (problem 3.3)
A ⎧ e jk ( r21+ r01 ) ⎡ cos(n , r01 ) − cos(n , r21 ) ⎤ ⎫
U ( P0 ) = ∫∫∑ ⎨⎩ r21r01 ⎢ ⎥ ⎬ ds
jλ ⎣ 2 ⎦⎭
The Fresnel-Kirchhoff diffraction formula that
holds
h ld only
l ffor a single
i l point

The Fresnel-Kirchhoff diffraction formula

i t source ill
illumination
i ti .P2
r21
P0
r01
n P1
is symmetrical with respect to P0 and P2 .
A point source at P0 will produce the same effect at P2 .
This result is known as: reciprocity theorem of Helmholtz
15
Huygens’ wavelets
Huygens-Fresnel
Huygens Fresnel theory: the light disturbance at a point P0 arises from
the superposition of secondary waves that produced from a surface
situated between this point and the light source.
If we rewrite
A ⎧ e jk ( r21+ r01 ) ⎡ cos(n , r01 ) − cos(n , r21 ) ⎤ ⎫
U ( P0 ) = ∫∫ ⎨ ⎢ ⎥ ⎬ ds
jλ ∑
⎩ r21r01 ⎣ 2 ⎦⎭
e jkr01
U ( P0 ) = ∫∫ U '( P1 ) ds where
∑ r01
illuminating wavefront
⎡ Observation angle Illumination angle ⎤
1 ⎡ Ae jkr21 ⎤ ⎢ cos(n , r01 ) − cos(n , r21 ) ⎥
U '( P1 ) = ⎢ ⎥ ⎢ ⎥
jλ ⎣ r21 ⎦ ⎢ 2 ⎥
⎣ ⎦
Seems like U ( P0 ) arises from sum of infinite fictitious sources with
amplitudes and phases expressed by U '( P1 ).
We used point source to get this result but it is possible to generalize
this result for any illlumination by using Rayleigh-Sommerfeld theory.
16
The Rayleigh-Sommerfeld Formulation of
diffraction
Potential theory: If a two-dimensional potential function and its normal
derivaive vanish together along any finite curve segment,
segment then the
potential function must vanish over the entire plane.
This is also true for solution of a three-dimensional wave equation.

1) The Kirchhoff boundary conitions suggests that the diffracted field

mustt be
b zero everywhere
h behind
b hi d th
the aperture.
t N t ttrue.
Not
2) Also close to aperture the theory fails to produce the observed
diffraction field
field.
Inconsistencies of the Kirchhoff theory were removed by Sommerfeld.
He elliminated the need of imposing boundary values on the
disturbance and its derivative simultaneously.

17
An alternative Green’s function I
Ob
Observed
d fi
field
ld strength
t th iin tterms off th
the incident
i id t fifield
ld and
d itits normall d
derivatives:
i ti
1 ⎛ ∂U ∂G ⎞
U ( P0 ) =
2π ∫∫S1 ⎜⎝ ∂n G − U ⎟ ds
∂n ⎠
(Fresnel-Kirchhoff diffraction formula)

Conditions for validity:

1) The scalar theory holds
2) Both U and G are solutions of the homogeneous scalar wave equation
3) The sommerfeld radiation condition holds i.e. if the disturbance U
vanishes at least as fast as a diverging spherical wave then:
as R → ∞ contribution of S 2 to U ( P0 ) vanishes.

If the Green's function of Kirchhoff theory was modified so that either G

or ∂G / ∂n vanished over entire surface S1 , then there is no need to im pose
boundaryy condtions on both U and ∂U / ∂n.

18
An alternative Green’s function II
Somerfeld
S f ld argued
d th
thatt one G
Green's
' function
f ti that
th t meets
t th
these criteria
it i isi
composed of two identical point sourcers at two sides of the aperture,
mirror image of each other, oscillating with a 1800 phase difference:
e jkr01 e jkr01
G− ( P1 ) = −
r01 r01
Now G = 0 on the plane of aperture
aperture.
Kirchhoff's BC may be applied only on U
Σ1
1 ⎛ ∂U ∂G ⎞
U ( P0 ) = ∫∫ ⎜
2π 1 ⎝ ∂n
G − U ⎟ ds
S

S1 = Σ1 + Σ + Σ 2
∂n ⎠
. P0
r01
P1 Σ
.
P0
1 ⎛ ∂U ∂G ⎞ r01
2π ∫∫Σ1 +Σ2 ⎝ ∂n
we need U ( P0 ) = ⎜ G − U ⎟ ds = 0 n
∂n ⎠
G |S1 = 0 so if we require only U = 0 on Σ1 and Σ2 Σ2
∂U
then does not need to be zero to make U ( P0 ) |Σ1 +Σ2
∂n
With the new BC on the U only there is no conflict with
the potential theorem. 19
The Rayleigh-Sommerfeld diffraction
Formula
With the Green's function G−
e jkr01 e jkr01
G− ( P1 ) = −
r01 r01
the U ( P0 ) takes the form
1 e jkr01
U I ( P0 ) = ∫∫ U ( P1 ) cos(n , r01 )ds or
jλ 1S r01
Assuming r01 >> λ
Now applying the Kirchhoff BC only on U and not on ∂U / ∂n we get
1 e jkr01
U I ( P0 ) = ∫∫ U ( P1 ) cos(n , r01 )ds
jλ Σ r01
And U will not vanish on the other side of the aperture.
aperture

20
Rayleigh-Sommerfeld Diffraction formula
With the Green
Green's
s function G+ two sources oscillating in phase with each other
e jkr01 e jkr01
G+ ( P1 ) = + the U ( P0 ) takes the form
r01 r01
1 ∂U ( P1 ) e jkr
jk 01
U II ( P0 ) = ∫∫Σ ∂n r01 ds
2π
Now for the spacial case illumination: . r01 .
P0
e jkr21 P0 r01
a diverging spherical wave from point P2 : U ( P1 ) = A n P1
r21
We apply the Kirchhoff BC only on U and not on ∂U / ∂n and
using G− we get
A e jk ( r21 + r01 )
U I ( P0 ) = ∫∫Σ r21r01 cos(n, r01 )ds
jλ
and and G+ gives
A e jk ( r21 + r01 )
U II ( P0 ) = ∫∫Σ r21r01 cos(n, r21 )ds
jλ
Where the angle between n and r21 is greater than 900.
This is Rayleigh-Sommerfeld Diffraction formula where we assumed r21 >> λ
21
Comparison of the Kirchhoff and Rayleigh-
Sommerfeld (R-S)
(R S) theorem
e jkr01 e jkr01 e jkr01 e jkr01 e jkr01
G− ( P1 ) = − ; G+ ( P1 ) = + , GK ( P1 ) =
r01 r01 r01 r01 r01
G
Green functions
f off the Sommerfeld
S f formulation
f Green function of the
Kirchhoff formulation

∂G−
On the surface Σ we can show that G+ = 2GK and = 2GK
∂n
1 ⎛ ∂U ∂GK ⎞
For the Kirchhoff theory: U ( P0 ) = ∫∫ ⎜ G − U ⎟ ds
4π Σ ⎝ ∂n ∂n ⎠
K

−1 ∂GK
For the R-S theory: U I ( P0 ) =
2π ∫∫Σ ∂n
U ds

1 ∂U
U II ( P0 ) = ∫∫ GK ds
2π Σ ∂n
We can see that
U ( P ) + U II ( P0 )
U ( P0 ) = I 0
2
Summary: the Kirchhoff solution is the arithmatic average of the two
Rayleigh-Sommerfeld solutions. 22
Comparison of the Kirchhoff and R-S theory
Kirchhoff theory:
Obliquity factor ψ

A e jk ( r21+ r01 ) ⎡ cos( n , r01 ) − cos(n , r21 ) ⎤ A e jk ( r21+ r01 )

U ( P0 ) = ∫∫∑ r21r01 ⎣⎢ ⎥ ds = ∫∫ ψ ds
jλ 2 ⎦ jλ ∑ r21r01
Obliquity factor ψ
jk ( r21 + r01 )
A e A e jk ( r21 + r01 )
R-S theory: U I ( P0 ) = ∫∫ cos(n , r01 ) ds = ∫∫Σ r21r01 ψ ds
jλ Σ r21r01 jλ
Obliquity factor ψ
jk ( r21 + r01 )
A e A e jk ( r21 + r01 )
U II ( P0 ) =
jλ ∫∫Σ r21r01 ∫∫
cos(n , r21 ) ds = ψ ds
jλ Σ r21r01

.P2
r21 .
P0
. r01
P0
.
r01 P0 r01
n P1 n P1

23
 Comparison of the Kirchhoff and R-S theory
Obliquity factor of both Kirchhoff and R-S theory
⎧ 1 ⎫
⎪ 2 [cos(n , r01 ) − cos(n , r21 )] Kirchhof theory ⎪
⎪ ⎪
ψ = ⎨ cos(n , r01 )
⎪− cos(n , r )
⎪
⎩
21
First R-S solution
Second R-S solution
⎬
⎪
⎪
⎭
.
P2
r21
P1 r01
θ
. P0

When a point source is at a very far distance

n

⎧1 ⎫
⎪2 [1 + cos θ ] Kirchhof theory ⎪
⎪ ⎪
ψ = ⎨cos(θ ) First R-S solution⎬
⎪1 R S solution ⎪
Second R-S
⎪ ⎪
⎩ ⎭
In summary: for small angles all three solutions are identical
When observation point or illumination source are
far away, the angles are small. P0
. r01
P1
r01
θ
.
P0
R-S solution requires the diffracting screens be planar. n
Kichhoff solution is not limitted to planar surfaces
surfaces.
For most applications both are OK.
We will use the first R-S solution for simplicity.
24
Huygens-Fresnel Principle

25
Generalization to Non-monochromatic waves I
We generalize the R-S's first solution to nonmonochromatic waves (chromatic?)
Monochromatic time function of the disturbance:
{
R U ( P )e − j 2πν t
u ( P, t ) = Re }
Time dependent chromatic functions: u ( P1 , t ) at the aperture, u ( P0 , t ) observation point
u ( P1 , t ), u ( P0 , t ) in terms of their Fourier transforms. Let's change the variable ν ' = -ν
∞ ∞

∫ U ( P ,ν )e ∫ U ( P , −ν ')e
j 2πν t − j 2πν ' t
u ( P1 , t ) = 1 dν = 1 dν '
−∞ −∞
∞ ∞

∫ U ( P ,ν )e ∫
j 2πν t
u ( P0 , t ) = 0 dν = U ( P0 , −ν ') e − j 2πν 't dν '
−∞ −∞ complex amplitudes monochromatic
of the disturbance elementary
at frequency ν ' function of
freuency ν '

Wh
Where U ( P1 ,ν ) = F {u ( P1 , t )} and
d U ( P0 ,ν ) = F {u ( P0 , t )}

We see that the cromatic function is sum

.
P0 r01
u ( P1 , t )
P1
r01
θ
.
u ( P0 , t )
P0
of the monochromatic functions over n
different frequencies.

26
Generalization to Non-monochromatic waves II
∞ ∞

∫ U ( P1 ,ν )e ∫
j 2πν t − j 2πν ' t
u ( P1 , t ) = dν = U ( P1 , −ν ')
) e dν '
−∞ −∞

u ( P0 , t ) =
∞

∫ U ( P0 ,ν )e
−∞
j 2πν t
dν =
∞

∫ U ( P0 , −ν ') e − j 2πν 't dν '

−∞ complex amplitude elementary
l
function
t
of
. P0 r01
u ( P1 , t )
P1
r01
θ P0
.
u ( P0 , t )

of the disturbance
at frequency ν ' freuency ν' n
Now we introduce the R-S first solution for the diffracted field
U ( P1 )
Obliquity factor ψ
jkr01
1 Ae jkr21
e 1 e jkr01
U I ( P0 ) = ∫∫ cos(n , r01 ) ds = ∫∫Σ U ( P1 ) r01 cos(n, r01 )ds
jλ Σ r21 r01 jλ
ν' e j 2πν ' r01 //VV
U ( P0 , −ν ') = − j ∫∫ U ( P1 , −ν ') cos(n , r01 )ds where V = c / n
V Σ r01
This is one frequency component. Summing over all of them we get:
Elementary
Complex amplitude at each frequency function at
that frequency
⎡ ν'
∞ e j 2πν ' r01 / V ⎤
u ( P0 , t ) = ∫ ⎢ − j ∫∫ U ( P1 , −ν ') cos(n , r01 )ds ⎥ e − j 2πν 't dν '
−∞ Σ
⎣ V r01 ⎦
r
cos(n , r01 ) ∞ − j 2πν '( t − 01 )
u ( P0 , t ) = ∫∫ ∫ − j 2πν 'U ( P1 , −ν ')e V
dν ' ds
Σ 2π Vr01 −∞
27
Generalization to Non-monochromatic waves III
Next we want to relate the disturbance at the observation point
r
cos(n , r01 ) ∞ − j 2πν '( t − 01 )
u ( P0 , t ) = ∫∫ ∫ − j 2πν 'U ( P1 , −ν ')e V
dν ' ds
Σ 2π Vr −∞
01

t the
to th disturbance
di t b
∞
att the
th aperture
t location
l ti
. P0 r01
u ( P1 , t )
P1
r01 .
u ( P0 , t )
θ P0
∫ U ( P , −ν ')e
− j 2πν ' t
u ( P1 , t ) = 1 dν ' we use the identity n
−∞
∞
∞
d d
∫
− j 2πν ' t
u ( P1 , t ) = U ( P1 , −ν ') e dν '
dt dt −∞
∞
d
u ( P1 , t ) = ∫ − j 2πν 'U ( P1 , −ν ')e − j 2πν 't dν '
dt −∞

cos(n , r01 ) d r
u ( P0 , t ) = ∫∫ u ( P1 , t − 01 ) ds
Σ 2π Vr dt V
01
The wave disturbance
at P0 is linearly proportional Over all angles Incident wave at
to the time derivative of the the "retarded" time
disturbance at eah point P1 or the time that the
on the aperture wave was generated

In summary the results of the diffration theory for the monochromatic

waves is applicable to the more general case of the chromatic waves.
28
The angular spectrum of plane a wave I
Next we want to formulate the diffraction theory in a framework of
linear, invariant systems.
Assume a transverse monochomatic wave traveleing in + z direction
incident on a transverse ( x, y ) plane
Across the z = 0 plane U = U ( x, y, 0)
Across the z = z plane U = U ( x, y, z )
Objective: to calculate the resulting field U = U ( x, y, z ) down the road z > 0
as a function of U ( x, y, 0))
∞
FT of the U at z = 0 plane: A( f X , fY ;0) = ∫ ∫ U ( x, y, 0)e − j 2π ( f X x + fY y ) dxdy
−∞
∞
And U ( x, y, 0)) = ∫ ∫ A( f X , fY ;0)
; ) e j 2π ( f X x + fY y ) dff X dffY
−∞
What is the physical meaning of these components?

So far we have looked at A( f X , fY ;0) as the spatial frequency spectrum

off the disturbance
o disturbance.
What is the direction of propagation of each these components?

29
Physical interpretation of angular spectrum
C
Considerid a simple i l plane
l wave propagatingti in
i di
direction
ti off kk:
2π
P ( x, y, z ) = e jk .r where r = xx + yy + zz and k = (α x + β y + γ z );
λ
2π
α , β , and γ are the direction cosines of k. Also |k |=
λ
Using α 2 + β 2 + γ 2 = 1, between the direction cosines we rewrite
k 2π / λx λ
α= x = = f X λ ; β = fY λ ; γ = f Z λ = 1 − ( f X λ ) − ( fY λ )
2 2
=
| k | 2π / λ λx
2π 2π
j (α x + β y ) j γz j 2π ( f X x + f y y ) j 2π 1− f X2 − fY2
P ( x, y , z ) = e = e
jk .r
e λ
=e e λ x
Now with f X = α / λ , fY = β / λ f z = 0 we can write
cos-1α
∞
A( f X , fY ;0) = ∫ ∫ U ( x, y, 0)e − j 2π ( f X x + fY y ) dxdy as
−∞
⎛α β ⎞
cos-11γ
⎛α β ⎞ ∞ − j 2π ⎜ x + y ⎟
A ⎜ , ;0 ⎟ = ∫ ∫ U ( x, y, 0)e dxdy ⎝λ λ ⎠

⎝λ λ ⎠ −∞ z
is the angular
g spectrum
p of the disturbance U ( x, y, 0).
) cos-1β
y
In summary this results shows that:
each spatial frequency component is propagating at a different angle.
30
Propagation of the angular spectrum I
Consider
C id th
the angular
l spectrum
t off th
the U across a plane
l parallel
ll l tto ( x, y )
at a distance z from it:
⎛α β ⎞
⎛α β ⎞ ∞ − j 2π ⎜ x + y ⎟
A ⎜ , ; z ⎟ = ∫ ∫ U ( x, y, z )e ⎝ λ λ ⎠ dxdy
⎝λ λ ⎠ −∞

Our goal is to find the effects of the wave propagation on the angular spectrum
⎛α β ⎞ ⎛α β ⎞
of the disurbance or the relationship of A ⎜ , ;0 ⎟ and A ⎜ , ; z ⎟
⎝λ λ ⎠ ⎝λ λ ⎠
⎛α β ⎞
⎛ α β ⎞ j 2π ⎜ x + y ⎟ α β
∞
We start from U ( x, y, z ) = ∫ ∫ A ⎜ , ; z ⎟ e ⎝ λ λ ⎠ d d
−∞
⎝λ λ ⎠ λ λ
U must satisfy the Helmholtz equation ∇ 2U + k 2U = 0 wherever there is no source.
The result is that A must satisfy the following differential equation.
d 2 ⎛ α β ⎞ ⎛ 2π ⎞ ⎛α β ⎞
2

A ⎜ , ; z +
⎟ ⎜ ⎡
⎟ ⎣1 − α 2
− β 2
⎤
⎦ A ⎜ , ;z⎟ = 0
dz ⎝ λ λ ⎠ ⎝ λ ⎠
2
⎝λ λ ⎠
coefficient

The solution can have the form:

⎛ 2π ⎞
⎛α β ⎞ ⎛α β ⎞ ⎜ j 1−α 2 − β 2 z ⎟
A ⎜ , ; z ⎟ = A ⎜ , ;0 ⎟ e⎝ λ ⎠

⎝λ λ ⎠ ⎝λ λ ⎠
31
Propagation of the angular spectrum II
⎛ 2π ⎞
⎛α β ⎞ ⎛ α β ⎞ ⎜⎝ j λ 1−α − β z ⎟⎠
2 2

A ⎜ , ; Z ⎟ = A ⎜ , ;0 ⎟ e
⎝λ λ ⎠ ⎝λ λ ⎠
For this solution two cases are recognized:
1) When α 2 + β 2 < 1 (true for all direction cosines) the effect of propagation
on the angular spectrum is simply a change in phase of each component.
When α 2 + β 2 > 1 (α and
2) Wh d β are no llonger th
the di
direction
ti cosines
i effect
ff t
of aperture is present here) the angular spectrum has the form:
Fourier transform of
a field distributiion
on which BCs of the ⎛ A real number ⎞
aperture is imposed ⎜ 2π ⎟
⎜j 1−α 2 − β 2 z ⎟
⎜ λ
⎛α β ⎞ ⎛α β ⎞ ⎛α β ⎞ 2π
⎟
⎜ ⎟
A ⎜ , ; z ⎟ = A ⎜ , ;0 ⎟ e = A ⎜ , ;0 ⎟ e − μ z ; μ =
⎝ ⎠
α 2 + β 2 −1
⎝λ λ ⎠ ⎝λ λ ⎠ ⎝λ λ ⎠ λ
Since z is a positive real number, the wave components are attenuating
as they propagate. They are also called evanescent waves.
Si
Simmilar
il tto th
the case off microwave
i waveguides
id there
th iis a cutoff
t ff frequency.
f
Below cutoff frequency, these evanescent waves carry no energy away
from the aperture. 32
Propagation of the angular spectrum III
⎛ 2π ⎞
⎛α β ⎞ ⎛ α β ⎞ ⎜⎝ j λ 1−α 2 − β 2 z ⎟
Substituting A ⎜ , ; Z ⎟ = A ⎜ , ;0 ⎟ e ⎠
in the U ( x, y, z ) we get:
⎝λ λ ⎠ ⎝λ λ ⎠
⎛α β ⎞
∞⎛ α β ⎞ j 2π ⎜⎝ λ x + λ y ⎟⎠ α β
U ( x, y , z ) = ∫ ∫ A ⎜ , ; Z ⎟ e d d
−∞
⎝λ λ ⎠ λ λ
⎛ 2π ⎞ ⎛α β ⎞
⎛ α β ⎞ ⎝⎜ j λ
∞ 1−α 2 − β 2 z ⎟ j 2π ⎜ x + y ⎟ α β
U ( x, y, z ) = ∫ ∫ A ⎜ , ;00⎟e ⎠
e ⎝λ λ ⎠
d d
−∞
⎝λ λ ⎠ λ λ
⎧1 x2 + y 2 < 1
⎪
circ ( x2 + y 2 ) ⎪
= ⎨1/ 2
⎪0
x2 + y 2 = 1
otherwise
⎪⎩
⎛ 2π ⎞ ⎛α β ⎞
⎛ α β ⎞ ⎜⎝ j λ 1−α − β z ⎟⎠ j 2π ⎜ x + y ⎟ α β
2 2
∞
U ( x, y, z ) = ∫ ∫ A ⎜ , ;0 ⎟ e circ α + β e ⎝ λ λ ⎠ d d
2 2
−∞
⎝λ λ ⎠ λ λ
No angular spectrum contribute to U ( x, y, z ) beyond the evanescent

wave cutoff α 2 + β 2 < 1.

33
Physical meaning of cutoff frequency
No angular spectrum contribute to U ( x, y, z ) beyond the evanescent

wave cutoff α 2 + β 2 < 1.

No imaging system can resolve a periodic structure that its period is
less than the wavelength of the light used.
2
⎛ kx ⎞ ⎛ k y ⎞
2

⎜ k ⎟ + ⎜ k ⎟ < 1 → k x + k y <| k | or x + y >| λ | .

2 2 2 2

⎝ ⎠ ⎝ ⎠
Near-Field imaging couples to the evanescent waves of a very fine
structure and recovers the phase information that would be lost otherwise.

34
Effects of a diffracting aperture on the
angular spectrum I
Now an infinite opaque screen containing a diffracting structure is placesd
in the z = 0 plane.
G l fi
Goal: find
d effects
ff t off the
th screen on the
th angular
l spectrum
t off the
th disturbance.
di t b
Amplitude transmitance function:
U t ( x, y;0)
t A ( x, y ) =
U i ( x, y;0)
U t ( x, y;0) = t A ( x, y )U i ( x, y;0) take the Fourier transform of the both sides
d use th
and the ffrequency convolution
l ti ththeorem:
⎛α β ⎞ ⎛α β ⎞ ⎛α β ⎞
At ⎜ , ⎟ = Ai ⎜ , ⎟ ⊗ T⎜ , ⎟
⎝λ λ⎠ ⎝λ λ ⎠ Convolved ⎝λ λ ⎠
with
Angular spectrum Angular spectrum A second angular
of the transmitted of the incident spectrum that is a
disturbance disturbance result of the diffracting
structure

⎛α β ⎞
⎛α β ⎞ ∞ − j 2π ⎜ x + y ⎟
where T ⎜ , ⎟ = ∫ ∫ t A ( x, y )e ⎝ λ λ ⎠ dxdy
⎝λ λ⎠ −∞

is the Fourier transform of the t A ( x, y )

35
Effects of a diffracting aperture on the
angular spectrum II
Example: for a unit amplitude plane wave illuminating the diffracting
structure the angular spectrum of the input is a delta function:
structure,
⎛α β ⎞ ⎛α β ⎞ ⎛α β ⎞ ⎛α β ⎞ ⎛α β ⎞
Ai ⎜ , ⎟ =δ ⎜ , ⎟ At ⎜ , ⎟ = Ai ⎜ , ⎟ ⊗T ⎜ , ⎟
⎝λ λ ⎠ ⎝λ λ ⎠ ⎝λ λ ⎠ ⎝λ λ ⎠ ⎝λ λ ⎠
Fourier transform of the
Transmitted angular amplitude transmitance
spectrum function of the aperture

⎛α β ⎞ ⎛α β ⎞ ⎛α β ⎞ ⎛α β ⎞
At ⎜ , ⎟ = δ ⎜ , ⎟ ⊗ T ⎜ , ⎟ = T ⎜ , ⎟
⎝λ λ⎠ ⎝λ λ⎠ ⎝λ λ⎠ ⎝λ λ⎠
If the diffracting structure is an aperture that limits the extend of the
field distribution,
⎛α β ⎞
⎛α β ⎞ ∞ − j 2π ⎜ x + y ⎟
T ⎜ , ⎟ = ∫ ∫ t A ( x, y )e ⎝ λ λ ⎠ dxdyy
⎝λ λ ⎠ −∞

the angular spectrum of the disturbance will be broadened.

36
The propagation phenomenon as a linear
filter I
Consider propagation from plane z = 0 to the plane z = z
Across the z = 0 plane U = U ( x, y, 0)
Across the z = z plane U = U ( x, y, z )
Goals: a) to show that the propagation phenomenon acts like a
linear space-invariant suystem
suystem.
b) find the system's transfer function
The system is linear since it is governed by a linear wave equation or
considering the superposition integral (R-S first solution).
1 e jkr01 1 e jkr01
U I ( P0 ) =
jλ ∫∫Σ U ( P1 ) r01 cos(n, r 01 )ds = jλ ∫∫Σ U ( P1 ) r01 cos θ ds
1 e jkr01
U ( P0 ) = ∫∫ h( P0 , P1 )U ( P1 )ds with h( P0 , P1 ) = cos θ
Σ
jλ r01
To establish the space-invariance we need to derive the transfer function
of the system and show that the mapping is space-invariant.
37
Transfer function of the linear invariant systems
g 2 ( x2 , y2 ) = S { g1 ( x1 , y1} action of a linear operator on a inp
inputt
∞
g 2 ( x2 , y2 ) = ∫ ∫ g1 (ξ ,η )h( x2 , y2 ; ξ ,η )dξ dη h is the impulse response
−∞

h( x2 , y2 ; ξ ,η ) = S{δ ( x1 − ξ ; y1 − η )}
h( x2 , y2 ; ξ ,η ) = h( x2 − ξ , y2 − η ) for linear invariant systems
∞
g 2 ( x2 , y2 ) = ∫ ∫ g1 (ξ ,η ) h( x2 − ξ , y2 − η ) dξ dη = g1 * h = g1 ⊗ h
−∞
Object function Impulse response
of the system

Take Fourier transform

G2 ( f X , fY ) = H ( f X , fY )G1 ( f X , fY )
Where H is the Fourier transform
f off the impulse response.
∞
H ( f X , fY ) = ∫ ∫ h(ξ ,η )e − j 2π ( f X ξ + fYη ) dξ dη
−∞

And it is called transfer function of the system that indicates

effects of the system in the freuency domain.
38
Propagation phenomenon as a linear filter II
To establish the space-invariance we need to derive the transfer function
of the system and show that the mapping is space-invariant.
Let A( f X , fY ;0) be spatial spectrum (Fourier transform) of U ( x, y, 0) at z = 0
Let A( f X , fY ; z ) be spatial spectrum (Fourier transform) of U ( x, y, z ) at z = z
So following G2 ( f X , fY ) = H ( f X , fY )G1 ( f X , fY ) and
A( f X , fY ; z ) =H ( f X , fY ) A( f X , fY ;0) we can see H connects the two frequency
spectrums before propagation and after the propagation. We got
⎛ 2π ⎞ ⎛α β ⎞
⎛ α β ⎞ ⎜⎝ j λ
∞ 1−α 2 − β 2 z ⎟ j 2π ⎜ x + y⎟ α β
U ( x, y, z ) = ∫ ∫ A ⎜ , ;0 ⎟ e ⎠
circ α 2 + β 2 e ⎝λ λ ⎠
d d
−∞
⎝λ λ ⎠ λ λ
α β
with = f X , = fY
λ λ
∞ ⎛ 2π ⎞
1−( λ f X ) − ( λ fY ) 2 z ⎟
2
⎜j
U ( x, y , z ) = ∫ ∫ A ( f X , fY ;0 ) circ (λ f X ) + (λ fY ) 2 e ⎝ λ
e j 2π ( f X x + fY y ) df X dfY
2 ⎠

−∞
Imposes the bandwidth limitation
associated with evanescent waves
∞
U ( x, y, z ) = ∫ ∫ A( f X , fY ; z )e j 2π ( f X x + fY y ) df X dfY comparing two equations shows:
−∞
⎛ 2π ⎞
1−( λ f X ) − ( λ fY )2 z ⎟
2
2 ⎜⎝ λ
j
A( f X , fY ; z ) = A ( f X , fY ;0 ) circ (λ f X ) + (λ fY ) e
2 ⎠
39
Propagation phenomenon as linear filter III
⎛ 2π z ⎞
1−( λ f X ) − ( λ fY ) 2 ⎟
2
⎜ j
A( f X , fY ; z ) = A ( f X , fY ;00 ) circ (λ f X ) + (λ fY ) e
2 ⎝ λ
2 ⎠
i
And the transfer function is then
⎧ ⎛⎜ j 2πλ z 1−( λ f X )2 −( λ fY )2 ⎞⎟
⎪e⎝ ( λ ) λ < 1/ λ
2
H ( f X , fY ) = ⎨
⎠
f X + ( f Y ) 2

⎪⎩0 otherwise
This shows that the propagation phenonmenon can be considered as
a linear, dispersive spatial filter with a finite bandwidth.
Transmission is zero outside the circular region of radius λ -1(in spatial
frequency space)
Its transfer function is exponential.
Within the circular frequency bandwidth the modulus (or amplitude) of
the function is 1, but frequency dependent phase shifts are introduced.
Phase dispersion is largest at high spatial frequencies (below the cutoff)
A f X → 0 and
As d fY → 0 th
then phase
h di
dispersion
i vanishes.
i h
For a fixed spatial frequency pair, f X and fY :
the phase dispersion increases as z increases 40