Talbot effect reinterpreted

Paul Latimer and Randy F. Crouse

Pattern generation in Talbot planes has generally been interpreted in terms of image formation, the
repetitive slits are said to make repetitive images of themselves. In this context, Fourier optics
developments have correctly predicted the positions of some but not all of the Talbot planes. Now,
wave-optics methods are used to obtain general expressions for the positions of all known Talbot planes
and the lateral positions of the diffraction fringes within them. These equations predict the key features of
the Talbot effect, and they better relate multiple-slit diffraction in the Fresnel and Fraunhofer domains.
Key words: Talbot effect,diffraction, multiple-slit diffraction in the Fresnel domain, moire patterns.

I. Introduction v(2T) + (2T), where v is an integer that may be zero

Forty-five years after Talbot first demonstrated that
a grating will produce distinct patterns in certain
planes close to it,' Rayleigh2 offered the first explana-
tion of any part of this effect. Rayleigh, by the use of
equations for sound waves from a remote source in
tubes, found these Talbot planes to be a distance R2
from the grating: R = nT, where T = a2 /,, a is the
distance between adjacent slits, and n is an even
integer.
Seventy-six years later, Cowley and Moodie used
methods of Fourier optics to predict two sets of n
values [see Eq. (3.12) of Ref. 3]. Depending on the
assumptions, n was found to be either an even integer
or any integer, only. Then Winthrop and Worthing-
ton4 used another Fourier optics development to
predict that n should be an even integer only.
However, experimental
evidence developed that
showed the existence of other Talbot planes in which
n is not an integer." To date, Fourier optics has not
predicted any of these noninteger Talbot planes.
However, Rogers6 said, "It is found convenient to
make use of the work of Rayleigh (1881) and Cowley
and Moodie (1957) and establish a basic unit for D
(our R), viz. 2a2/X, and to put D in the form D =
2a2 /(nX) where n is an order number which is gener-
ally, but not invariably, an integer." Thus he sug-
gested n = 2/p where this p is an integer. Soon
thereafter, Winthrop and Worthington4 proposed R =
The authors are with the Department of Physics, Auburn
University, Auburn, Alabama 36830.
Received 18 December 1990.
0003-6935/92/010080-10$05.00/0.
© 1992 Optical Society of America.
[see Ref. 4, Eqs. (18) and (19)]. They explained, "It is
convenient to write P = N/n, if possible, where n, N
are integers, for in this case well-defined Fresnel
images are observed, especially if n is comparatively
small." Thus they suggested n = 2N/p where N andp
are integers.
These equations for noninteger Talbot planes were
only proposed as convenient guides to the observed
locations of Talbot planes; there were no claims of
physical justification. Their equations can indeed
predict all the observed Talbot planes. However the
lack of physical constraints on the integers literally
allows the equations to predict an infinite number of
Talbot planes, most of which cannot be observed.
At Rayleigh's suggestion, these patterns in Talbot
planes have been interpreted as images (i.e., Fourier
images) of the grating; a repetitive object is said to
make repetitive images of itself.2 6 However these
images are not quite like conventional ones. Despite
the lack of a general physical explanation of the
locations of all Talbot planes, the Talbot effect has
seen wide applications. .
We first used this effect to generate moir6 patterns
from two sets of Talbot fringes in studies of obscura-
tion by light scattering. However the common inter-
pretation of these moire patterns in terms of scalar
grating images neglects some effects of coherence. To
avoid this problem, Tepichin and Ojeda-Castanedall
used two incoherent laser beams instead of one to
generate moir6 patterns. Now instead of interpreting
the Talbot plane patterns as Fourier images of the
grating, we interpret them as multiple-slit diffraction
patterns produced by conventional mechanisms such
as those of Fraunhofer grating diffraction.3

80 APPLIED OPTICS / Vol. 31, No. 1 / 1 January 1992

Approach
Problems
A complete explanation of the Talbot effect would
require information about (a) the longitudinal
positions of the various Talbot planes, (b) the
absolute lateral positions of the intensity maxima or
bright fringes within them, and (c) the details of
the intensity profile within such planes. In the con-
text of image formation, problems (a) and (b) have
been substantially interpreted as two separate one-
dimensional problems. Partial solutions to (a) were
obtained but little attention was given to (b) and (c).
However, in the wave-optics context, (a) and (b) are
two parts of a single two-dimensional problem. We
now use this approach to attack (a) and (b); problem
(c) is not considered.
Basic Relations
If a grating is illuminated with a point source of
monochromatic light, high-contrast fringes are formed
in certain Talbot planes close to it. Figure 1 shows a
point source of light, P1 , that illuminates the grating a
distance R1 away from it. A bright fringe is formed at
a
--
L _
PI -II
RI
Fig. 1. Schematic diagram of the apparatus for producing the Talbot effect.A light source P illuminates a part of a grating that generates
a bright spot at point P2 in a Talbot plane. P, and P2 are distances R, and R2 from the grating. Some of the light goes through slit d, which is
a distanceyj from a perpendicular dropped from P, to the grating, a distancey2 from the perpendicular from P2 .
P2 in a Talbot plane a distance R2 from the grating. In
practice, a Talbot plane is effectively defined as one
near the grating where fringe (image) patterns of
maximal contrast are formed.
Let H narrow slits of the grating be illuminated.
The net amplitude A at P2 will be
H
A= Ah exp(ikh), (1)
h=1
where Ah and 4hare the amplitude and phase of the
wave from one slit. A bright fringe in a Talbot plane
occurs when A and the intensity I = A2 are large
compared with those at nearby points. The incident
light travels in the x direction; the slits differ in their
lateral, y, coordinates. They extend in the z direction.
The distance from P to P2 through a given slit in
Fig. 1 is L = R, + Al + R2 + A2 . If y is the distance
from the slit center to a perpendicular to the grating
from point P and y <<R the extra distance traveled
on that side of the grating is
A = y'/(2R), (2)
wherey and R areyl and R ory2 and R2.
9
e Y2
yI
jc
1 January 1992 / Vol. 31, No. 1 / APPLIED OPTICS 81
Numerical Predictions Talbot Plane and Fringe Positions
For remote and normal illumination, a numerical Two-Slit Systems
method based on these equations was devised to
predict I patterns at arbitrary R2 values. Then the Consider the illumination of point P2 by only two slits
phase of each wavelet is 2rA 2/X. For 20 equally of the grating, which is normally illuminated by P1
spaced narrow slits, this code calculates I at 2000 when R = m. Figure 3 shows two such gratings
points opposite to the grating in a potential Talbot where the waves of the two slits interact in different
plane of arbitrary R2. Figure 2 shows the incident modes. P2 is illuminated by a fixed slit, c, and a
wave front, the grating, and the more important variable one, v, which may be d, e, etc. The slits are a
Talbot fringes (points) that were found. distance ja apart, where j is an integer. On the left,
The strongest fringes are depicted by large circles; v = d andj = 1; on the right, v = e andj = 2.
they are in Talbot planes extending into the page at Equation (2) indicates that the difference in the
R2 = nT, where n is an integer. These planes can be paths of waves going to P2, Al-, = A - A., is given by
thought of as the fundamental Talbot planes. There
are also Talbot planes with fringes of twice that u=(y D-y2)I(2R)
=y0 ujalR, (3)
frequency as indicated by x's where n = 1/2, 2/3, etc.
Still other planes contain fringes of three times that whereyav= (y, + y,)/ 2 andy, - y, =ja. The termya, is
frequency; they are denoted by triangles, where n = the distance from P 2 to a point opposite one on the
1/2, 2/3, etc. grating that is midway between v and c (see Fig. 3).
This numerical method is so simple that it effec- On the left, wherej = 1,yauis measured from a point
tively requires the reliability of its results. We use halfway between slits c and d. On the right, wherej =
Fig. 2 as a fundamental description to which the 2, Yavis measured from slit d which is midway
results of other developments are related. between c and e. Note that, while we discuss points

A X 0D X A A X (
.C1 A A a

ap A

x -&-- …* … -A- --
A

-4 - - - - -
A

- -X --- -

A A A

A IL A

A X X A A X (
IL A A

IL IL A

x- -A-- - ---- - - -A-----

I
A A A

40 IL AL IL

A X X A A X (
IL A AL

A IL IL

a e --- A
- ------- X IL
----
-- - -

A
- - -X - -A -------

40 IL A A
-)
C
A X
0 X A A X 0D
IL A IL

A A a
- -
d A
- -- … -a- ----

A
-- -

A
-

IL A A

A X X A AL X
A A A

A IL A

- - - - - -X- - -A --… -… -a- - - -- -- X - - - - - - - - - - - - - - -- -.- -A -------

-
A A A

A I A

I I

1/2 1 4/3 5/3 2 5/2 3

Fig. 2. Some of the predictions by a numerical method [Eqs. (1) and (2)] of the brightest fringes or images (dots) beyond a 20-slit grating
that is normally illuminated by a remote source. The fringes occur in Talbot planes; the n values of some of these planes are indicated.

82 APPLIED OPTICS / Vol. 31, No. 1 / 1 January 1992

 C
0 L
C
I ov . 0
j=I j=2 0
_ >

I
caD,
a
C en
f
TP2.o 0
> 0 Ye

e e T.:
C
-o-
U)
cv ~0
d d V

CC
_ R2|

Fig. 3. Fringe generation by parts of two gratings as illuminated normally by remote sources. In each case, the illumination of P2 in a
Talbot plane is only by slits c and v. On the left, v = d (j = 1); on the right v = e (j = 2). Constructive interference lines, slope = Y_ IR 2 , are
shown as dotted lines. On the left, this line intersects the grating at a point midwaybetween the slits. On the right, the line intersects a slit
d, which is midway between the generating slits (j = 2) c and e. The slits containing small circles generate the dotted lines.

and lines in the x-y plane, in three dimensions these
terms actually refer to lines and planes that extend in
the z direction.
At an interference maximum, the phase difference
+ of the waves from the two slits is
= 27rm = 2irA/X,
Waves from the slits will constructively interfere at
all points on all lines. Physically, the m value of a line
is the phase difference, in cycles, of the two waves; it
is the amount by which the phase of the wave from
the upper slit leads to that from the lower slit. A
double-slit pattern may be observed in any plane of
(4)

where A is A,,of Fig. 3 and m is an integer. Combining

Eqs. (3) and (4),

ya = mX R2 ja = ma R2 IjT = maxljT, (5)

where x has been substituted for R2.

If d = ja is the distance between the slits and 0
sin 0 tan 0, Eq. (5) may be written as a)
d sin 0 = mX. (6) ~0
C0

This is a well-known equation for the angular loca-

tion of diffraction maxima from a double slit. Until
U)
now, it had been developed only from Fraunhofer
postulates [see Eq. (16g) of Ref. 13]. -. a
Equation (5) can be interpreted as defining a 0

straight line in terms of its slope, ma/(jT), and -0

intercept, a point midway between the two generating
slits. Waves from the slits will constructively inter-
fere at all points on such a line when R >>a. We call it
a constructive interference line.
Equation (5) defines many such lines with different
values of m. Figure 4 shows some of these lines.
Fig. 4. Apair of slits, normally and remotely illuminated. Emanat-
ing from a point midway between them are constructive interfer-
ence lines of the labeled slope indices m = 0 to +4. Bright
double-slit fringes could be observed in any plane of observation at
an arbitrary distance from the grating where these lines intersect
this plane.

1 January 1992 / Vol. 31, No. 1 / APPLIED OPTICS 83

observation where the lines hit the screen; R is
arbitrary.
For any pair of narrow slits in Fig. 1, the total path
of the light on going from P1 to P2 is R, + R2 + A1 + A2 .
Constructive interference will occur at P2 when
A, + A2 = mX.
Then, substituting Eq. (3) into Eq. (7), we obtain
(jayla)IR, + (jay2_,)IR2 = mA-
If i is the angle of incidence, YI21 /R, = tan(i) sin(i);
if 0 is the angle of diffraction, y2a,/R2 sin(0). Then
Eq. (8) can be written as
d[sin(i) + sin(O)] = m.
This equation for diffraction close to the slits is the
counterpart of Eq. (6) for oblique illumination. It is
identical to Eq. (17f) of Ref. 13 as developed from
Fraunhofer postulates.
(7) Multiple-Slit Systems
Any two slits of a grating can be interpreted as a pair,
the members of which are a distanceja apart. In Fig.
5 we show four adjacent slits of a grating; three
(8) overlapping pairs c and d, d and e, and e and f.
Constructive interference lines [Eq. (5)] are drawn
from each of thesej = 1 pairs. The pairs c and d and d
and e overlap in that they share slit d. At a junction of
lines from these pairs, the waves from both c and e
will be in phase with that from d; all three waves are
(9) in phase. If a constructive interference line from pair

I I I

a
Cr

e
0
a
3r.
-I.-
C:
0
-0 d
.5
C:

n=j/q 1/10 1/5 1/3 1/2

Fig. 5. Four slits of a normally and remotely illuminated grating. These slits are interpreted as three overlapping pairsj = 1. Each pair, c
and d, etc. generates a set of constructive interference lines: m = 0, + 1, +2, etc. When two or more of these lines intersect at a plane
indicated by a vertical dashed line, the waves from all the parent slits that are involved constructively interfere. Bright fringes occur at these
junctions. When such lines of slope indices m and m' intersect, R2 = nT, where n ='j2 Iq = 1/q and q = m = m'. There is also an unlabeled
Talbotplane at n =j2/q = 4/6 = 2/3. It is omitted here since fringes in this plane are actually generated by wave interactions in thej = 2
mode.

84 APPLIED OPTICS / Vol. 31, No. 1 / 1 January 1992

e and f also passes through the same point, the waves
from all four of the slits will constructively interfere.
This argument can be extended to more slits.
Bright fringes generated by the j = 1 mode of
interaction could be observed at all of the intersec-
tions of constructive interference lines that are
marked with dashed lines in Fig. 5. Note that the
fringes all fall in straight lines; thus the existence of
Talbot planes is predicted in Figs. 2 and 5. Other
studies of the Talbot effect have simply assumed that
the planes exist.
To determine the x values of these planes, we
rewrite Eq. (8) for slits c and d:
ja(ylaVCd)IRl +ja(yaCd)1R 2 = MCdX,
and for slits d and e:
ja(Yla-d,)Rl +ja(y2avd,)1R
2 = indeX-
Lines of these pairs will cross only if
ylau -cd =ja, Y2av-de- Y2av-cd = ja, and mde
Mde - q. Physically, q is the difference, in cycles,
mCd =
in the phase differences of the two constructive
interference lines from overlapping slit pairs that
cross. Then subtracting (10) from (11), we obtain
1/R, + 1/R 2 = 1/(nT),
where T = a 2 / X and n =j2/ q. The definitions ofj and
q place practical limits on the predicted n values. This
solves problem (a).
Equation (12) was previously derived on the basis
of various assumptions by Cowley and Moodie3 and
for R1 = ooby using other assumptions by Rayleigh2
and by Winthrop and Worthington.4 In each case, the
solution to the problem required that n be an integer,
usually an even one. Then Rogers6 and Winthrop and
Worthington4 suggested ad hoc equations that pre-
dict nonintegral values of n. On the other hand, Eq.
(12), as it is based on a single set of assumptions, in
general predicts all known Talbot planes.
To determine the y-coordinates of fringes, Eq. (8)
may be rewritten iny~,awhere R1 and R2 are subject to
the constraints of Eq. (12). There is a special case of
interest. Here, the light is normally incident, in which
caseYla = 0. Theny2a, can be obtained from Eq. (8) or
Eq. (5) with Eq. (12). If illumination is normal and
R= , this becomes
Y = mna/j = majlq.
If j > 1, there will bej overlapping series of fringes,
each with its own sets of origins. Equation (13a) gives
the coordinates of fringes in one series only. In
practice, we may want Y2avfor all fringes, regardless of
their series. For this,
Y2av= malq.
To determine the fringe spacing b for arbitrary R1,
we let two adjacent fringes from Eq. (8) be generated
by slit pair c and d. One fringe appears aty2av, the next
of the same series at y,,, + b'. The slope indices of
these lines are, respectively, m and m'. For the
adjacent fringes, (m - m') = 1. Then from Eqs. (10)
and (12) and these relations we obtain
b'= a RI + R2\ (14a)
q~ R, _
Again, this b' is the distance between adjacent fringes
of a given series. If b is the distance between adjacent
fringes of all series, then
(10) a (R, + R 2 )
(14b)
q R=
Equations (8), (13), and (14) solve problem (b).
(11)
Can Constructive Interference Lines Explain All Talbot
Ylau-de - Fringes?
•4 Md. Let To gain better insight into Talbot plane formation, we
examine the generation of fringes in Talbot planes by
constructive interference lines for wave interactions
in the = 1 and = 2 modes in more detail. While
there will be simultaneous wave interactions in many
(12) j modes, these two are the most important for planes
close to the grating.
Figure 5 showsj = 1 lines. Talbot planes generated
by these lines are indicated with dashed vertical lines.
These constructive interference lines are produced by
continuously overlapping pairs of slits. At the planes
labeled with vertical lines the waves from all slits are
exactly in phase at the junctions. Note that all
junctions satisfy q = m - m'. Sincej = 1, these lines
(the edges of Talbot planes) satisfy n =j 2 /q = 1/q.
The lateral positions of fringes in Fig. 5 are of
interest. The j = 1 constructive interference lines
emanate from points midway between the slits. Note
that some line junctions in all labeled planes are
opposite to these midpoints, m = 0 in Eq. (13).
Figure 6 shows fringe formation when the waves
interact in thej = 2 mode. Constructive interference
lines of two types are generated by the two alternat-
ing series of slits: c, e, g, etc., and b, d, f, h etc.,
respectively. Solid constructive interference lines are
drawn from slits c, e, g, etc.; they are generated by
waves from b, d, e, and h, respectively. Dotted lines
are generated by waves from c, e, etc. Some of the
Talbot planes are labeled with solid vertical lines.
(13a) Some of the points in these planes near slit h have
been labeled with a square, diamonds, triangles, and
circles. Some, but not all, of these junctions ofj = 2
constructive interference lines are at the locations of
bright fringes.
In the n =j 2
/m = 4/3 plane of Fig. 6, the junctions
are pure, i.e., lines from overlapping pairs of only one
type intersect there and all the waves are exactly in
(13b) phase. In the other labeled Talbot planes in Fig. 6, the
junctions involve constructive interference lines of
both types in about equal numbers. Then the result-
1 January 1992 / Vol. 31, No. 1 / APPLIED OPTICS 85
 L

>
3 h
.I)
a

ai
_
f

I i n In i .A 17 __

n = Ii/ ;icS
Io 1I
2
j /q = 4/8 4/6 4/4 4/3 4/2
Fig. 6. Fringe formation by waves interacting in thej = 2 mode. Two different sets of constructive interference lines are generated by slit
combinations involving every other slit. Slits d, f, h, etc., generate the solid constructive interference lines that emanate, respectively, from
slits e, g, etc. Similarly dotted constructive interference lines from slits d, f, h, etc., are generated by waves from slits c, e, g, i. Some of the
Talbot planes, which contain fringes generated by these lines, are marked by vertical solid lines. Some of the line junctions at the level of slit
h have also been labeled with a square, diamonds, triangles, and circles. At most junctions, both solid and dotted lines intersect. Then the
resultant disturbances depend on the relative phases of the waves from the two series of slits.

ing disturbances depend on the relative phases of the
two sets of waves.
The relative phases of the two sets of waves at the
large square opposite slit h were determined. With
the dotted lines, it is found that the waves at the
square from slits f and d are, respectively, 1 and 4
cycles behind those from h. From Eq. (2), it is found
that waves from slits g and e are respectively,. 1/4 and
9/4 cycles behind those from h. Evidently, at the
square the amplitude phasors of the two series of slits
are about equal in magnitude and 1/4 cycle ('r/2 rad)
out in phase with each other. Such resultant ampli-
tudes fromj = 2 interactions will produce fringes but
they will not be as bright as those in the n = 1 plane
due to j = 1 interactions where all waves are exactly
in phase.
In the n = 2/3 plane in Fig. 6, three of the line
junctions have been labeled with triangles. At each,
the waves from the two series of slits are also found to
be 1/4 cycle out of phase.
Some of the planes near the grating contain junc-
tions corresponding to wave interactions in both the
j = 1 andj = 2 modes, and the n = 1/2, and n = 1
planes. To relate interactions better in these two
modes, Fig. 7 shows j = 1 constructive interference
lines, as in Fig. 5, redrawn on the scale of Fig. 6
(j = 2). The special points of Fig. 6 are included.
Figure 7 reveals that all junctions of j = 1 lines for

86 APPLIED OPTICS / Vol. 31, No. 1 / 1 January 1992

 C

fl = 1/5 1/2 1
Fig. 7. Agratingwithj = 1 constructive interference lines in the format of Fig. 6. The labels of the specialline junctions of Fig. 6 have been
copied. Fringe generation is constrained by requirements that n =j 2 q and that the junctions of constructive interference lines of the same
series should cross and these lines should be generated by continuously overlapping pairs of slits. In general, wave interactions of the lowest
possible mode,j value, are the most important (see text).

n > 1 involve lines that are not of overlapping pairs of
slits. Thus fringes in planes beyond n = 1 are
generated only byj > 1 interactions.
We relate events that occur at these junctions in
the n = 1 planes of Figs. 6 and 7. There are two
fringes per slit spacing in Fig. 6 and one fringe per slit
in Fig. 7. In Fig. 6 the point in the n = 1 plane
opposite slit h is labeled with a large diamond, the
point below it with a medium diamond. It is found
that the two types of waves that meet at the large
diamond are 1/2 cycle out of phase with each other;
they destructively interfere. However, at the medium
diamond, the waves of the two series of slits are in
phase; a bright fringe is generated. Of the junctions in
the n = 1 plane of = 2 lines in Fig. 6, only those that
are junctions of the = 1 lines are a real fringes.
Finally, four junctions in the n = 1/2 plane in Fig. 6
are marked with circles. At two of them, the waves
from the different series of slits prove to be 1/2 cycle
out of phase so they cancel out. At the other two
junctions the waves from the different series are
found to be in phase-they constructively interfere.
These examples for n = 1/2 and n = 1 suggest, and it
can indeed be proven, that when junctions of lines of
different modes of interaction intersect in the same
plane, only those of the lowest j value lead to bright
fringes.
Discussion
Equations (6) and (9) above predict the angular
positions at which constructive interference occurs
near two slits. These equations are identical to well-
known equations for the same slits as derived from
Fraunhofer postulates. Both developments fix an
angle, the orientation of a straight line. The develop-
ment for near the slits also fixes an intercept of such a

1 January 1992 / Vol. 31, No. 1 / APPLIED OPTICS 87

line. However, this point is not normally encountered
in connection with diffraction as observed at remote
locations. In those situations, the finite dimensions of
the gratings are neglected. Nevertheless, the concept
of a constructive interference line as introduced here
for considerations of diffracted light observed close to
the slits is still applicable to conventional Fraunhofer
double-slit diffraction.
To test Eq. (6) for a double-slit system, we used a
He-Ne laser and a double-slit system of an elemen-
tary physics laboratory. Generally, Fraunhofer double-
slit diffraction is demonstrated by observing the
pattern at R2 = co, here approximated as 1 m away.
We found that the pattern could indeed be observed
continuously with little qualitative change on reduc-
ingR2 from 1 m to 0.1 m, the value of T = a2 /,. Hence
the Fraunhofer domain for double-slit effects actually
begins close to the slits.
Fraunhofer and Fresnel diffraction are generally
distinguished on the basis of how the phase of a
wave that takes an indirect path varies with its
maximum lateral displacement. Fraunhofer diffrac-
tion occurs when this phase is retarded in direct
proportion to the displacement. Presently close to the
slit, this phase shift is proportional to displacement to
the second power [Eq. (2)]. However, for two slits
only, the phase difference can always be expressed as
a first-order function of the displacement. This re-
lates to the fact that a straight line can always be
drawn through two points. Hence the Fraunhofer
domain should indeed begin close to the slits. On the
other hand, if three or more slits are involved in
fringe formation, the Fresnel domain will extend
much further out. Then Fraunhofer diffraction oc-
curs only in the limiting case where q ->0 and R2 -> co.
In the laboratory we used Rochi gratings, i.e., 150
lines in., with 20 lines illuminated with a He-Ne laser
beam to observe the Talbot effect. It is easy to produce
the patterns when n is a small integer; indeed, the
patterns do look like grating images. In addition to
integral planes, we have seen fringes in the n = 1/2
and 1/3 planes, and others. However, most applica-
tions of the Talbot effect have used only the integral
Talbot planes.
It is much easier to predict fractional Talbot planes
with constructive interference lines than to produce
them in the laboratory. To determine why, we used
our numerical method to predict intensity curves at
various R2 values. This method indicates that at the
appropriate R2 values, sharp fringes should indeed be
observable in many fractional Talbot planes. How-
ever, small errors in R2, which might occur with
simple apparatus, lead to quite different patterns.
There may also be some problems relating to non-
negligible slit width that are not considered here.
While Talbot first reported the effect in 1836, only
now do we have general expressions for the positions
of the Talbot planes and of fringes within them. Why
has this taken so long? We speculate that part of the
problem has been in the approach. The Talbot effect
has generally been interpreted as an image generator;
88 APPLIED OPTICS / Vol. 31, No. 1 / 1 January 1992
then the evaluation of R2 is a one-dimensional prob-
lem. Investigations thereof have sought mainly to
predict the distance from the grating to the various
Talbot planes [problem (a)]. The lateral positions of
fringes [problem (b)] have received less interest. On
the other hand, the present wave-optics approach
treats the Talbot effect as a two-dimensional problem
and achieves general solutions to both problems (a)
and (b).
Rayleigh offered little explanation of how his sound
equations predicted that n must be an even integer.
However, we have formulated an optical explanation
assuming a laterally central wave source and a double
Lloyd's mirror system. This model predicts that n can
be an even integer but not an odd one, as Rayleigh
found.
The earlier investigations of the Talbot effect
suggested that there are various types of Talbot
plane: those where n is an integer and those where
it is a noninteger or the ratio of two integers. These
noninteger planes are indeed now predicted above.
The earlier studies collectively also suggested that the
integral planes are of two types: those where n is
even and those where it is odd. However our Eq. (12)
does not distinguish between such n values.
The present development does predict two types of
plane, depending on whetherj is even or odd. When n
is evenj is always even, and when n is odd,j is always
odd. When j is even, the origins of the constructive
interference lines fall on slits and there is always a
fringe [m = 0 in Eq. (13)] directly opposite each slit
(see Fig. 6). Whenj is odd, the origins of the construc-
tive interference lines fall midway between slits and
there is always a fringe directly opposite to these
points (see Figs. 5 and 7).
We propose two types of Talbot plane that are
distinguished by the lateral positions of the fringes.
The conditions on constructive interference lines
cause the corresponding fringes in the two types of
plane to be laterally displaced a distance a/2 apart,
the distance between the origins of the different types
of constructive interference line. The effect is seen in
Fig. 2 in both the original integral Talbot planes
treated by previous authors and in the newly ex-
plained nonintegral Talbot planes. We speculate that
the reason why the Fourier optics developments
distinguished between even and odd n is related to
these differences in the lateral positions of the fringes.
VI. Concluding Remarks
By conventional wave-optics methods, we have ob-
tained general expressions for the longitudinal posi-
tions of all known Talbot planes and also the lateral
positions of fringes within them. Detailed examina-
tions of fringe formation in the various Talbot planes
indicate that the present development can indeed
explain all known Talbot fringes and that it is un-
likely that any others beyond the scope of these
equations will be found. The present results should
be useful in applications. Furthermore, they better
relate multiple-slit diffraction in the Fresnel and
Fraunhofer domains.
References
1. F. Talbot, "Facts relating to optical science No. IV," Philos.
Mag. 9, 401-407 (1836).
2. Lord Rayleigh, "On copying diffraction-gratings, and on some
phenomenon connected therewith," Philos. Mag. 11, 196-205
(1881).
3. J. M. Cowley and A. F. Moodie, "Fourier images: I-the point
source," Proc. R. Soc. London Ser. B70, 486-496 (1957).
4. J. T. Winthrop and C. R. Worthington, "Theory of Fresnel
images. I. Plane periodic objects in monochromatic light," J.
Opt. Soc. Am. 55, 373-381 (1965).
5. E. A. Hiedemann and M. A. Breazeale, "Secondary interfer-
ence in the Fresnel zone of gratings," J. Opt. Soc. Am. 49,
372-375 (1959).
6. G. L. Rogers, "Calculations of intermediate Fourier images of
a finite line grating on a digital computer, with an application
to an unusual case," Brit. J. Appl. Phys. 14, 657-661 (1963).
7. D. E. Silva, "Talbot interferometer for radial and lateral
derivatives," Appl. Opt. 11, 2613-2624 (1972).
8. Y.Nukano and K.Murata, "Talbot interferometer for measur-
ing the focal length of a lens," Appl. Opt. 24, 3162-3166
(1985).
9. E. Keren, I. Glatt, and 0. Kafri, "Propagator for the modulat-
ing transfer function of a wide-anglescatterer," Opt. Lett. 11,
554-556 (1986).
10. K. Patorski, "Talbot interferometry with increased shear: part
3," Appl. Opt. 27, 3875-3878 (1988).
11. E. Tepichin and J. Ojeda-Castaneda, "Talbot interferometer
with simultaneous dark and bright fields," Appl. Opt. 28,
1517-1520 (1989).
12. 0. Kafri, E. Keren, K. Kreske, and Y. Zag., "Moire deflectom-
etry with a focused beam: radius of curvature, microscopy,
and thickness analysis," Appl. Opt. 29, 133-136 (1990).
13. F. A. Jenkins and H. E. White, Fundamentals of Optics,3rd ed.
(McGraw-Hill, New York, 1957).
14. M. Born and E. Wolf, Principles of Optics (Macmillan, New
York, 1959), p. 382.
1 January 1992 / Vol. 31, No. 1 / APPLIED OPTICS 89