DOING PHYSICS WITH MATLAB

COMPUTATIONAL OPTICS
FOUNDATIONS OF SCALAR
DIFFRACTION THEORY

Ian Cooper
School of Physics, University of Sydney
ian.cooper@sydney.edu.au

DOWNLOAD DIRECTORY FOR MATLAB SCRIPTS

View document: Numerical evaluation of the Rayleigh-Sommerfeld

diffraction integral of the first kind

SCALAR DIFFRACTION INTEGRALS

The integral theorem of Helmholtz and Kirchhoff

Kirchhoff boundary conditions

on aperture SA E (r )  EA (r ) and E (r )  EA (r )

 e jkr 
E (P) 
1
4  A  r 2 
E ( r )  jkr  1 rˆ  rEA (r ) n dS
SA  

Rayleigh-Sommerfeld diffraction integral of the second kind

1 e jkr
2  r
E (P)  E (r ) n dS
SA

Rayleigh-Sommerfeld diffraction integral of the first kind

1 e jkr
E (P) 
2  E (r ) r3
z p ( jkr  1) dS
SA

Doing Physics with Matlab op_diffraction_integrals_theory.docx 1

The Rayleigh-Sommerfeld region includes the entire space to the right of the
aperture. It is assumed that the Rayleigh-Sommerfeld diffraction integral of the first
kind is valid throughout this space, right down to the aperture. There are no
limitations on the maximum size of either the aperture or observation region, relative
to the observation distance, because no approximations have been made.

Doing Physics with Matlab op_diffraction_integrals_theory.docx 2

THE INTEGRAL THEOREM OF HELMHOLTZ AND
KIRCHHOFF

The basic idea of the Huygens-Fresnel theory is that the light disturbance at a point
P arises from the superposition of secondary waves that proceed from a surface
situated between this point P and the light source. Kirchhoff formulated a sound
mathematical basis for this principle based upon solutions of the homogeneous wave
equation, where the value at an arbitrary point P in the field can be expressed in terms
of the values of the solution and its first derivatives at all points on an arbitrary closed
surface surrounding the point P.

We consider strictly monochromatic scalar waves that are solutions of the wave
equation

1  2 E (r , t )
 2 E (r , t )  0 (1)
c 2 t 2

where E (r , t ) is electric field (or monochromatic optical disturbance) and can be

written as

E (r , t )  E (r ) e j  t  E (r ) e j k ct (2)

Solutions of the wave equation for plane waves traveling the direction + r and - r are
r  t )
+ r direction: E ( r , t )  Eo e j ( k
r  t )
- r direction: E ( r , t )  Eo e j ( k (3)

Solutions of the wave equation for spherical waves traveling the direction + r and - r
are
E
+ r direction: E (r , t )  o e j(k r   t )
r
E
- r direction: E (r , t )  o e j ( k r   t )
r

In vacuum, the space-dependent part then satisfies the Helmholtz equation

2 E (r )  k 2 E (r )  0 (4)

Doing Physics with Matlab op_diffraction_integrals_theory.docx 3

In problems involving boundaries it is often convenient to study the properties of the
difference between two solutions of an equation rather than one of the solution alone,
since the boundary conditions become simpler to handle.

Take the origin at the observation point P, at which the electric field has the value
E(0). The two wave fields, E( r ) the unknown field to be calculated and Et( r ) a trial
solution of the Helmholtz equation satisfy the equation

E (r ) 2 Et (r )  Et (r ) 2 E (r )   E (r ) k 2 Et (r )  Et (r ) k 2 E (r )  0 (5)

at all points except at the origin r = 0. Now, integrate (5) throughout a volume V
bounded by a surface S and by Green’s theorem, the volume integral can be changed
to a surface integral:

  E(r ) E (r )  E (r ) E(r )  dV    E(r )E (r )  E (r )E(r )  n dS  0

2 2
t t t t
V S

  E(r )E (r )  E (r )E(r ) 

S
t t n dS  0 (6)

n being the inward pointing normal to the surface S. We assume that both E( r ) and
Et( r ) possess continuous first- and second-order partial derivatives within and on this
surface. Because the integrand is zero in (5), the integrals in (6) are also zero.
Furthermore, the volume containing P must be free of any sources of optical radiation
(including secondary sources of reradiated or scattered light).

Note: terms such as  n give the normal component of the gradient at the surface
of integration.

Consider a trial solution to be

Eot j k r
Et ( r )  e (7)
r

which is a spherical wave radiating from the origin at P. This spherical wave Et (r )
plays the role of a mathematical auxiliary function – a “probe” which we use for
investigating the optical field E (r ) . This is a virtual spherical wave and has nothing
to do with any real spherical waves that act as the source functions.

This trial function has a singularity at the point r = 0 and since Et( r ) was assumed to
be continuous and differentiable, the point P must be excluded from the domain of
integration in equation (6). This is done by surrounding P by a small sphere of radius
 and taking the surface S to be divided into two parts, an arbitrary outer surface So
and the small inner spherical surface Si of radius .

Doing Physics with Matlab op_diffraction_integrals_theory.docx 4

 dS

So n
r

V
P
 n
Si

Therefore, the gradient of the trial function is

 E e jkr    Eot e jkr  1

Et (r )    ot  ˆ   2  jkr  1 Eot e rˆ
jkr
 r  r  (8)
  r  r  r

However, after we make the substitutions into of equations (7) and (8) into (6) we can
take these waves to describe spherical waves which are radiated by elements of the
surface S and which arrive at the point P at a distance r from dS.

Therefore, (6) become

  E(r )E (r )  E (r )E (r ) 

So
t t n dS    E (r )E (r )  E (r )E (r ) 
Si
t t n dS  0

(9)

The contribution from Si can be evaluated directly, since over the small sphere of
radius  we consider E() to be constant and equal to E(0). The unit vector n is
parallel to  , so we have  n = 1 and we can substitute  d for dS where d is an
2

element of the solid angle. Therefore, the integral over the inner surface can be
written as

Doing Physics with Matlab op_diffraction_integrals_theory.docx 5

   1 ˆ 
  Eot e   E (0) ( j k   1)  2   E (0) 
jk
Ii  n  2 d
Si  
 
 

I i  Eot e j k    E (0) ( j k  ) d   E (0) d     E (0) n d 


 Si Si Si 

limit   0

I i   4 Eot E (0)
(10)

The value of the electric field at the observation point P is

 rˆ Eo e j k r 
 
 E ( r ) 2  j k r  1 Eot e jk r
 E ( r )  n dS  4 Eot E (0)  0

So 
r r


 e jk r 
E (0) 
1
4  E (r )  r 2   j k r  1 rˆ  r E (r ) n dS (10)
So  

Thus E(0) can be found if the values for E( r ) and E (r ) are known at all points on
the outer surface surrounding the point P. This is one form of the integral theorem of
Helmholtz and Kirchhoff.

This theorem expresses E(0) in terms of both E( r ) and E( r ). It may, however be
shown that from the theory of Green’s functions, that either E( r ) or E( r ) on So are
sufficient to specify E(0) at every point P with So.

In applying the integral theorem of Helmholtz and Kirchhoff to diffraction problems,

the surface So is divided into three: 1) the diffraction aperture (opening) SA, 2) the
screen blocking the radiation SB, and 3) since the surface must be closed, we must
include a mathematical surface S of radius R centre about the point P, so that when
R   the integral over the integral in (10) will approach zero or R is chosen so large
there is insufficient time for a contribution from this surface to have reached the point
P.

Doing Physics with Matlab op_diffraction_integrals_theory.docx 6

 p
rp s
SB n
Q
R
S P

SA
n

S
sources

So = SA + SB + S

It remains to integrate (10) over SA + SB. However, the actual field and its gradient
over the surface are almost impossible to determine except in the simplest cases
because for real physical apertures will partially absorb, reflect and scatter the
incident radiation.

The first approximation to make in integrating (10) is to ignore the real boundary
values and apply the so-called Kirchhoff boundary conditions and are the basis of
Kirchhoff’s diffraction theory. These replace the actual field in the presence of the
aperture by the incident field for the open aperture EA (r ) and by zero elsewhere:

Kirchhoff boundary conditions:

on SB E (r )  0 and E (r )  0 (11a)

on SA E (r )  EA (r ) and E (r )  EA (r ) (11b)

Then (10) becomes

 e jkr 
E (0) 
1
4  A  r 2
E ( r )   jkr 1 rˆ  rEA (r )  n dS (12)
SA  

The Kirchhoff boundary conditions are an approximation, since at the edge of the
opening, the amplitude of the wave does not suddenly jump from a finite value to
zero. The approximation is better when the aperture is large compared to the
wavelength.

Doing Physics with Matlab op_diffraction_integrals_theory.docx 7

Strictly speaking the boundary conditions (11) are mathematically inadmissible. A
theorem in Riemann’s theory of functions implies that if the value of a function and
its first derivative vanish along a segment then the function vanishes everywhere, in
fact the assumed boundary conditions even contradict each other and if we would
calculate the value at P on the surface S, the boundary conditions would not be
reproduced.

In principle, a scalar approach can be performed for each component of the vector
wave, but in practice this is rarely necessary. We can assume that the radiation from
the opening is much larger than from the blocking screen. Also, well into the opening,
all possible directions of polarization are possible for the radiation. Often, edge effects
are only predominant in a region within less than a wavelength from the edges.
Therefore, in most practical applications the use of a scalar approach is justified.
However, in the scalar approach, all information about the polarization of the wave is
lost.

Doing Physics with Matlab op_diffraction_integrals_theory.docx 8

GREEN’S FUNCTION
A SIMPLIFIED FORMULATION OF HUYGENS’S PRINCIPLE

Rayleigh-Sommerfeld diffraction integral of the first kind

Rayleigh-Sommerfeld diffraction integral of the second kind

The mathematical contradiction posed by the Kirchhoff boundary conditions can be

avoided by substituting for the trial spherical wave function Et ( r ) given by (7) by the
Green’s function G belonging to the surface surrounding the point P. This function is
defined by the following conditions:
within the volume V 2G  k 2G  0 (13a)

on SB G=0 (13b)

as r  0 G  Et (r ) (13c)

as r   r (G n  j k G)  0 (13d)

as before, r is the distance from the point P and (13d) is called the radiation condition.
G has a singularity only at the point P (r = 0) and is continuous everywhere else
within the volume V. G differs from Et ( r ) because of the additional condition (13b).
As a result of this condition the term containing the term E (r ) in equation (9)
vanishes and therefore (9) simplifies to

1
E (P) 
4  E (r ) G(r ) n dS (14)
So

Now, we need to specify only the boundary values of E (r ) and not its gradient. This
approach has the practical advantage of leading to the simpler form of the integral as
given in (14) as compared to (12).

This approach again is only an approximation and is valid for sufficiently small
wavelengths. The real field does not vanish completely behind the blocking screen,
nor is the field entirely unaffected by the presence of the screen, at least not within
distances of the order of magnitude of a wavelength from the edge of the screen.

However, the applicability of the Green’s function method is restricted to the special
case of a plane screen, for this is the only case for which the Green’s function can be
conveniently expressed.

Doing Physics with Matlab op_diffraction_integrals_theory.docx 9

Note: The wavelength  is really only defined for plane monochromatic waves and
can lose its significance in wave types encountered in diffraction. However, we can
always interpret  as the length 2 c / which is defined for all monochromatic
radiation processes and which for plane waves is identical with the actual
wavelength.

The method of images is used to determine the appropriate Green’s function.

S(xd, ys, zs)

rsq
screen
z=0
O(0, 0, 0)
Q(xq, yq, zq) 
+Z
n
rpq
cos = zp / rpq
P(xp, yp, zp)

Construction of the Green’s function for a plane screen

We construct the mirror image S of the point P with respect to the plane of the screen
z = 0. For an arbitrary point Q(xq, yq, zq) where zq > 0, we form the Green’s function

j k rpq j k rsq
e e
G  (15)
rpq rsq
where

rpq 2  ( x p  xq )2  ( y p  yq )2  ( z p  zq )2

rsq 2  ( xs  xq )2  ( ys  yq )2  ( zs  zq )2

x p  xs y p  ys z p   zs (16)

We have to determine G to evaluate (14). Firstly,

d  e pq  rpq d  e sq  rsq
jk r jk r
G
      (17)
zq drpq  rpq  zq drsq  rsq  zq
   

Doing Physics with Matlab op_diffraction_integrals_theory.docx 10

If we now place Q on the screen, we have

rpq  rsp  r

rpq rsq zp zp
  cos   (18)
zq zq rpq r

G G   e jk r 
G   2   cos  (19)
n zq r  r 

  e jkr  1
  ( j k r  1)e
jk r
 (20)
r  r  r 2

Substituting (19) and (20) into (14) we obtain and we can integrate over the area SA.

1 e jkr
E (P) 
2  E (r ) r3
z p ( jkr  1) dS (21)
SA

(21) is known as the Rayleigh-Sommerfeld diffraction integral of the first kind.

1 2
If we make the approximation that k r >> 1 then 2 ( j k r  1)  j
r r

then (21) can be written as

j e j kr
E (P) 
  E (r ) r
cos  dS (22)
SA

This expression (22) is equivalent to Huygens’s principle – a light wave falling on an

aperture propagates as if every element dS emitted a spherical wave, the amplitude
and phase of which are given by that of the incident wave. The factor cos relates to
Lambert’s law of surface brightness.

Instead of the Green’s function (15) which satisfies the boundary conditions (13) in
which G = 0 at z = 0, we now form

j k rpq j k rsq
e e
G  (23)
rpq rsq

Doing Physics with Matlab op_diffraction_integrals_theory.docx 11

This is a function which satisfies the boundary condition G/z = 0 at z = 0.

As a result of this condition the term containing the term E (r ) in equation (9)
vanishes and therefore (9) simplifies to

1
E (P) 
4  G(r ) E (r ) n dS (24)
So

Now, we need to specify only the boundary values of E (r ) and not the value of
E (r ) .
If we substitute the values of the Green’s function on the plane z = 0 we obtain

1 e j k r
2  r
E (P)  E ( r ) n dS (25)
SA

(25) is known as the Rayleigh-Sommerfeld diffraction integral of the second kind.

The Rayleigh-Sommerfeld region – the entire space to the right of the aperture. It is
assumed that the Rayleigh-Sommerfeld diffraction integrals are valid throughout this
space, right down to the aperture. There are no limitations on the maximum size of
either the aperture or observation region, relative to the observation distance, because
no approximations have been made.

The Fresnel region (near field) – that portion of the Rayleigh-Sommerfeld region
within which the Fresnel conditions are satisfied. Both the input and output signals are
restricted to regions lying near the z axis, i.e., the lateral dimensions of which are
much smaller than the separation between the input and output planes.

z pq  L1  L2 (26)

where L1 and L2 are the maximum lateral extents of the aperture and observation
planes respective. Also it is assumed that

  L1  L2 
4
3
z pq  (27)
4

Conditions (26) and (27) are sometimes known as the Fresnel conditions.

Doing Physics with Matlab op_diffraction_integrals_theory.docx 12

The Fraunhofer region (far field) – that portion of the Fresnel region within which
the Fraunhofer condition is satisfied

 L12
z pq  (30)


In the Fraunhofer region the size of the diffraction pattern increases with increasing
distance but its shape is invariant.

The Rayleigh distance dRL

The Rayleigh distance in optics is the axial distance from a radiating aperture to a
point an observation point P at which the path difference between the axial ray and an
edge ray is λ / 4. A good approximation of the Rayleigh Distance d RL is
4 a2
d RL 

where a is the radius of the aperture. Rayleigh distance is also a distance beyond
which the distribution of the diffracted light energy no longer changes according to
the distance zP from the aperture.

zP < dRL Fresnel diffraction

zP > dRL Fraunhofer diffraction.

If we consider a circular aperture of radius a, then much of the energy passing through
the aperture is diffracted through an angle of the order   / a from its original
propagation direction. When we have travelled a distance d RL from the aperture,
about half of the energy passing through the opening will have left the cylinder made
by the geometric shadow if a / d RL  . Putting these formulae together, we see that
the majority of the propagating energy in the "far field region" at a distance greater
than the Rayleigh distance d RL  4 a 2 /  will be diffracted energy. In this region then,
the polar radiation pattern consists of diffracted energy only, and the angular
distribution of propagating energy will then no longer depend on the distance from the
aperture.

References

Born 417 Hect 501,648 Klein 360 Lipson 156 Slater 168 Sommerfeld 197
Gaskill p385

Doing Physics with Matlab op_diffraction_integrals_theory.docx 13