You are on page 1of 4

Design and demonstration of a new kind of aperture for

getting expected diffraction patterns


Yijia Wu, Shimin Cao, Limin Hua, and Daomu Zhao*
Department of Physics, Zhejiang University, Hangzhou 310027, China
*Corresponding author: zhaodaomu@yahoo.com
Received November 11, 2013; revised December 20, 2013; accepted January 3, 2014;
posted January 6, 2014 (Doc. ID 201132); published February 5, 2014
In the regime of Fresnel diffraction, a novel algorithmis proposed for aperture design for getting expected diffraction
patterns. Experiments have verified the feasibility of this method. It may be used in beam transition, optics
communication, information encryption, and other related fields. 2014 Optical Society of America
OCIS codes: (050.1220) Apertures; (050.1970) Diffractive optics; (100.3190) Inverse problems; (350.6980) Transforms.
http://dx.doi.org/10.1364/OL.39.000801
It is well known that the diffraction pattern can be well
determined by using the famous HuygensFresnel prin-
ciple if the incident light beam and the shape of aperture
are given [1]. On the contrary, it is challenging to solve
the inverse problem, i.e., how to determine the shape of
aperture if the incident beam and the diffraction pattern
are given. Since the inverse problem is important, many
methods and algorithms have been proposed. For exam-
ple, the inverse scattering method in electromagnetic op-
tics was proposed to reconstruct the grating profile by
Roger and Maystre [2]. Phase retrieval is another useful
method, and several different retrieval algorithms were
introduced with thorough comparison by Fienup [3].
The diffraction aperture can also be reconstructed by us-
ing the algorithm based on inverse Collins formula and
fast Fourier transform [4]. In this Letter, a new method
of aperture design is introduced in the condition that
the incident beam and the diffraction pattern are given.
It is known that a generalized form (i.e., Collins for-
mula) of the diffraction integral written in terms of ray
transfer matrix elements is given by [5,6]:
E
2
(x
2
; y
2
; z) =
ik
2B
e
ikz
ZZ
s
E
1
(x
1
; y
1
) exp

ik
2B
[A(x
2
1
y
2
1
)
2(x
1
x
2
y
1
y
2
) D(x
2
2
y
2
2
)|

dx
1
dy
1
;
(1)
where the z axis is the principal optical axis, the aperture
is located at the x
1
y
1
plane where z = 0. E
1
(x
1
; y
1
) is the
electric field of an incident light in the apertures plane,
E
2
(x
2
; y
2
; z) is the diffractive electric field. A, B, D are
elements of ray transfer matrix, domain of integration
s corresponding to the transparent part of the aperture.
For free space, A = D = 1, B = z, Collins formula,
Eq. (1), is reduced to
E
2
(x
2
; y
2
; z) =
e
ikz
iz
ZZ
s
E
1
(x
1
; y
1
)
exp

ik
2z
[(x
1
x
2
)
2
(y
1
y
2
)
2
|

dx
1
dy
1
:
(2)
Consider a square aperture whose side length is 2a and
the center of the square is at the origin of the coordinate
frame. For normally incident monochromatic plane
light, in the condition of Fraunhofer diffraction, i.e.,
([(iz)x
2
1
[; [(iz)y
2
1
[) 1, Eq. (2) is reduced to:
E
2
(x
2
; y
2
; z) =
4a
2
z
e
i
z
(x
2
2
y
2
2
)
sinc

2ax
2
z

sinc

2ay
2
z

:
(3)
For simplicity, we have omitted the factor e
ikz
i (this will
not change the relative light intensity in the same x
2
y
2
plane). Besides, the wavelength is chosen as the natural
unit of length and the incident electric field E
1
is chosen
as the natural unit of electric field, i.e., = 1, E
1
= 1.
Therefore, Eq. (3) describes the Fraunhofer diffraction
generated by a square aperture whose center is at the ori-
gin of the coordinate frame, for normally incident mono-
chromatic plane light.
The aperture in our method is divided and designed in
the following way: the whole square aperture whose side
length is 200a, is divided into 100 100, 10,000 squares in
all and each squares side length is 2a. Each square is ei-
ther transparent or opaque. So all the transparent squares
form the transparent part of this aperture, in other words,
determine the actual shape of the aperture.
The diffractive electric field distribution produced by a
certain transparent square, whose center is at coordi-
nates (x
0
; y
0
), can be expressed by:
E
2
(x
2
; y
2
; x
0
; y
0
; z)
=
4a
2
z
exp

i
z
[(x
2
x
0
)
2
(y
2
y
0
)
2
|

sinc

2a(x
2
x
0
)
z

sinc

2a(y
2
y
0
)
z

: (4)
The total diffractive electric field is the superposition of
every transparent squares contribution. So far the di-
rect problem has been solved by means of Collins
formula.
In the following we will describe the algorithm of
aperture design in order to deal with the inverse problem,
which can be summarized as:
February 15, 2014 / Vol. 39, No. 4 / OPTICS LETTERS 801
0146-9592/14/040801-04$15.00/0 2014 Optical Society of America
(a) Draw the expected diffraction pattern on computer.
(b) All the aperture squares are endowed with an initial
state, either transparent or opaque state.
(c) Estimate the state of every aperture square. Every
square is estimated once in one round of estimation.
When a certain square is estimated, change its state. If
this change enables the diffraction pattern to be closer
to the expected pattern, execute this change; otherwise,
remain the former state unchanged.
(d) After several rounds estimation, the diffraction pat-
tern can be close to the expected pattern. The corre-
sponding aperture satisfies the requirement for getting
expected diffraction pattern.
The parameters of the optical system and the reason
for the parameters chosen are stated below. The diffrac-
tion can be classified by Fresnel number N
F
[6]. In the
case of N
F
1, it is almost geometrical optics. So the
shape of the aperture is nearly the same as the expected
diffraction pattern, which is meaningless. When N
F
1,
it is in the regime of Fraunhofer diffraction. In this case,
the size of the diffraction pattern is much larger than the
size of the aperture, it is difficult to get a complicated
diffraction pattern. So we choose the parameters as
N
F
1, i.e., Fresnel diffraction. Specifically, we choose
a = 0.3 and z = 100 (as previously mentioned, the wave-
length is chosen as the natural unit, i.e., = 1), in this
condition

i
z
x
2
1

i
z
y
2
1

i
z
a
2

=
9
10; 000
1;
which satisfies the condition used in the derivation of
Eq. (3). So the diffraction generated by a single aperture
square belongs to Fraunhofer diffraction, while the entire
aperture produces Fresnel diffraction, which is the key
point of parameters chosen in this aperture design.
In Fresnel diffraction, the size of the diffraction pattern
and the size of the aperture are in the same order of mag-
nitude. So we restrict the expected diffraction pattern in
a large square whose side length is also 200a and whose
center is at the z axis (the same as the aperture). Ex-
pected diffraction pattern is also divided into
100 100, 10,000 squares in all, each squares side length
is 2a. Figure 1 is the schematic diagram of the aperture
and the expected diffraction pattern.
For example, a 100 100 pixels picture (English word
optics) is drawn on computer as the expected diffraction
pattern [see Fig. 2(a)]. White pixels in the figure indicate
that the corresponding squares are expected to be illumi-
nated, while black pixels indicate the corresponding
squares are expected to be unilluminated.
In order to evaluate the degree of closeness between
the expected pattern and the actual diffraction pattern, in
the following we will introduce a criterion. In fact, the
diffractive field intensity and the expected field intensity
should be normalized at first. The normalization is indis-
pensable, since the diffraction pattern is determined by
the relative intensity instead of the absolute intensity. So
the comparison between the absolute intensity is mean-
ingless.
In the expected pattern, illuminated square (white
pixels) expected intensity is I
exp
(x
2
; y
2
) = 1 and unillu-
minated square (black pixels) expected intensity is
I
exp
(x
2
; y
2
) = 0, where (x
2
; y
2
) are the coordinates of
the center of each square on the pattern. We could define
the normalized expected intensity as
I

exp
(x
2
; y
2
) =
I
exp
(x
2
; y
2
)
P
x
2
;y
2
I
exp
(x
2
; y
2
)N
; (5)
where N is the number of squares in the diffraction pat-
tern (N = 10; 000 in this algorithm).
The diffraction pattern generated by a certain shape of
aperture can be described by
I(x
2
; y
2
) =

X
x
0
;y
0
E
2
(x
2
; y
2
; x
0
; y
0
; z)

2
; (6)
where the notation of summation
P
x
0
;y
0
indicates all the
transparent squares (x
0
; y
0
) on the aperture. Since the
diffractive electric field at the center of a certain square
can be a good approximation of the average electric field
of this square, (x
2
; y
2
) in Eq. (6) are the coordinates of the
center of each square on the pattern [the same as Eq. (5)].
The diffractive intensity is normalized by:
Fig. 1. Schematic diagram of the aperture and the expected
diffraction pattern.
Fig. 2. Design and experimental demonstration for expected
diffraction pattern (English word optics). (a) Expected diffrac-
tion pattern, (b) designed aperture, (c) numeric simulation pat-
tern, and (d) experimental pattern.
802 OPTICS LETTERS / Vol. 39, No. 4 / February 15, 2014
I

(x
2
; y
2
) =
I(x
2
; y
2
)
P
x
2
;y
2
I(x
2
; y
2
)N
: (7)
After normalization, the average light intensity of
both the expected pattern and the diffraction pattern
are equal to 1, which makes the comparison more reason-
able. The criterion value is defined by normalized
intensity:

X
x
2
;y
2
([I

(x
2
; y
2
) I

exp
(x
2
; y
2
)[
2
): (8)
Apparently, smaller means that the diffraction pattern is
closer to the expected pattern and vice versa. So is used
as the criterion of the degree of closeness.
Figure 1 shows that x
2
x
0
and y
2
y
0
in Eq. (4) can
be: 198a; 196a; 2a; 0 (notice that x
2
, x
0
, y
2
, y
0
are
the coordinates of the center of a certain square). So we
calculate the diffractive field data when x
2
x
0
and
y
2
y
0
take the value of 198a; 196a; 2a; 0, re-
spectively. So the number of the data is 199
2
, all the data
are stored in a data table. When the program is running,
the corresponding data can be called according to the
relative coordinates of the aperture square and the dif-
fraction pattern square. A certain number of squares
(400 in our computer program) are randomly chosen
to be endowed with transparent initial state, while
others are in the initial state of opaque.
In one round of estimation, every of the 10,000 aper-
ture square is estimated once. The order of the estimation
is random. When a certain square is estimated, change its
state. If this change enables the diffraction pattern to be
closer to the expected pattern [the degree of closeness is
described by criterion , see Eq. (8)], execute this
change; otherwise, leave the former state unchanged.
After several rounds estimation (20 rounds in our pro-
gram, the program will terminate earlier if in a certain
round, the number of the squares changed is <10),
the diffraction pattern can be close to the expected
pattern and the corresponding aperture satisfies the
requirement.
Criterion decreases rapidly in the first few rounds
estimation and the decrease rate is getting slower and
slower with the increasing of estimation rounds. In prac-
tice, 20 rounds are enough for to reach a plateau.
Since the initial states of the aperture squares are ran-
domly set and the order of estimation is also random, it is
possible to get different apertures that correspond to the
same expected diffraction pattern. Usually, these aper-
tures are similar but not identical. This is reasonable be-
cause the phase information of the diffraction pattern is
not provided in this problem.
Figure 2(a) (word optics) is chosen as the expected
pattern. We execute the program described as above.
The result of the execution by the algorithm is shown
in Fig. 2(b). White pixels indicate the transparent square
of the aperture and black pixels indicate the opaque
square.
The diffraction pattern generated by the aperture in
Fig. 2(b) can be simulated on computer. The numeric
simulation result of diffraction is shown in Fig. 2(c).
Comparison between Figs. 2(a) and 2(c) shows that
though the diffraction pattern deviates, the expected pat-
tern in some sense, one can identify the word optics
easily.
Equation (4) shows that if all the physical quantities
along x axis and y axis (x
0
; y
0
; x
2
; y
2
; a) are magnified
n times and meanwhile z is magnified n
2
times, the
diffraction result will remain unchanged. This scale
invariance offers great convenience for experimental
demonstration.
The parameters chosen in the computer program is
a = 0.3 and z = 100, as stated previously. We choose a =
47.39 in experiments, so the magnification along x axis
and y axis is n = 47.390.3 158.0. Accordingly, z is
magnified into z = 100 158.0
2
2.496 10
6
. An He
Ne laser is used as the beam source in the experiment,
whose wavelength is about 633 nm. So the practical
parameters are: a = 633 nm 47.39, about 30.0 m and
meanwhile, the distance between aperture and diffrac-
tion pattern is z = 633 nm 2.496 10
6
, about 158 cm.
So the side length of each square is 2a = 60.0 m, and
the side length of the whole aperture is 200a = 6.00 mm.
The size of the diffraction pattern is also restricted in the
region of 6.00 mm 6.00 mm.
Since the HeNe laser produces a Gaussian beam in-
stead of a plane light, the latter is what we used in theo-
retical derivation. So a telescopic system formed by two
convex lenses is adopted to expand the Gaussian beam
and the aperture is illuminated by the center part of the
expanded beam, whose light intensity is relatively uni-
formly distributed. The diffractive pattern is detected
by CCD. Figure 3 is the schematic diagram of the whole
experimental apparatus.
The experimental result of diffraction pattern pro-
duced by the aperture in Fig. 2(b) is shown in Fig. 2(d).
The nonuniformity of the beam, the deviation of the aper-
ture and other reasons cause the experimental error. But
it is enough to identify the word optics and verify the fea-
sibility of this design method.
Another two typical results of design are shown as
follows. Figure 4(a) is a geometrical pattern (star),
which serves as the expected diffraction pattern here.
Figure 4(b) is the shape of the aperture designed by
the program. Figure 4(c) is the numeric simulation result
of the diffraction pattern produced by the aperture of
Fig. 4(b). Figure 4(d) is the experimental result which
verifies the feasibility of this method.
Figure 5(a) is formed by four Chinese characters,
which mean Zhejiang University, and serve as the ex-
pected diffraction pattern here. Figure 5(b) is the shape
of the aperture designed by the program. Figure 5(c) is
the numeric simulation result of the diffraction pattern
Fig. 3. Schematic diagram of the experimental apparatus.
February 15, 2014 / Vol. 39, No. 4 / OPTICS LETTERS 803
produced by the aperture of Fig. 5(b). Figure 5(d) is the
experimental result, which demonstrates the feasibility
of this design method.
In conclusion, an algorithm is proposed in this Letter
for dealing with the inverse problem of diffraction (i.e.,
determine the shape of the aperture in the condition that
the incident light beam and the diffraction pattern are
given). Apertures for getting patterns like English words,
geometrical pattern, and Chinese characters can all be
well designed by this method. Furthermore, we have ex-
perimentally verified the feasibility and validity of this
method. This new method can also provide us possible
shape of the aperture in the condition that the diffraction
pattern has been already known in experiments. This
method may be useful in a number of fields, like beam
transition, optics communication, information process-
ing, information encryption, and so on.
This work was supported by the National Natural Sci-
ence Foundation of China (NSFC) (11274273, 11074219,
and J1210046).
References
1. M. Born and E. Wolf, Principles of Optics, 7th edition
(Cambridge University, 1999).
2. A. Roger and D. Maystre, J. Opt. Soc. Am. 70, 1483 (1980).
3. J. Fienup, Appl. Opt. 21, 2758 (1982).
4. J. Li and C. Li, Appl. Opt. 47, A97 (2008).
5. S. A. Collins, J. Opt. Soc. Am. 60, 1168 (1970).
6. S. Wang and D. Zhao, Matrix Optics (CHEP-Springer,
2000).
Fig. 4. Design and experimental demonstration for geometri-
cal pattern (star). (a) Expected diffraction pattern, (b) designed
aperture, (c) numeric simulation pattern, and (d) experimental
pattern.
Fig. 5. Design and experimental demonstration for Chinese
characters (Zhejiang University). (a) Expected diffraction
pattern, (b) designed aperture, (c) numeric simulation pattern,
and (d) experimental pattern.
804 OPTICS LETTERS / Vol. 39, No. 4 / February 15, 2014

You might also like