Professional Documents
Culture Documents
net/publication/241196857
CITATIONS READS
2 567
1 author:
Taco D. Visser
Vrije Universiteit Amsterdam
172 PUBLICATIONS 4,141 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Taco D. Visser on 17 March 2015.
C HAPTER 15
Y OUNG’ S I NTERFERENCE
E XPERIMENT:
T HE L ONG AND S HORT OF I T
Taco D. Visser
Dedicated to Emil Wolf, in great appreciation of his friendship.
Better known today is his two-slit experiment (see Fig. 2) in which the light em-
anating from the slits projects a diffraction pattern of alternating dark and bright
fringes [3]. The analogy between this experiment and the behavior of water waves
clearly demonstrates the wavelike character of light.
Another great contribution was Young’s suggestion that light vibrations, un-
like sound waves, are transverse. This idea was confirmed in a series of beautiful
experiments by Fresnel.
Recently, a poll was held among physicists asking them what they thought was
the most beautiful physics experiment ever performed [4]. If the experiments are
Taco D. Visser 321
Figure 2 Young’s illustration of two interfering waves, taken from Ref. [3].
ranked according to the number of times that they were cited, the result is the
following:
This result shows the historical importance of Young’s experiment, both in its orig-
inal form using light, and in its more recent version using electrons [5], which
elegantly demonstrates the wave character of particles.
We assume that the field that is incident upon the pinholes is partially coherent.
A question arises naturally: is the field in the region of superposition everywhere par-
tially coherent? To answer this question, we introduce a quantitative measure of the
strength of the field correlations at the pair of observation points P1 (r1 ), P2 (r2 ) at
frequency ω. This quantity is the spectral degree of coherence [10, Sect. 4.3.2], which
is defined as
W(r1 , r2 , ω)
µ(r1 , r2 , ω) = √ , (3)
S(r1 , ω)S(r2 , ω)
In Eq. (4), U(ri, ω) represents the scalar field at position ri (i = 1, 2), the asterisk
denotes the complex conjugate, and the angular brackets denote the average taken
over an ensemble of monochromatic realizations. The spectral density of the field at
a position r at frequency ω is given by the diagonal element of the cross-spectral
density, viz.,
Let us now examine the spectral degree of coherence in the far zone of a Young’s
interference experiment. The field at the observation point P1 (r1 ) is given by the
sum of the field contributions of the two pinholes, i.e.,
ikA eikR11 eikR21
U(r1 , ω) = – U( r1 , ω) + U( r2 , ω) , (6)
2π R11 R21
where Rij denotes the distance Qi Pj (i, j = 1, 2), k = ω/c is the wavenumber as-
sociated with frequency ω, c being the speed of light, and A is the area of each
pinhole. The prefactor on the right-hand side of Eq. (6) stems from the Huygens–
Fresnel principle [11, Sect. 8.8].* In a similar manner we find that
ikA eikR12 eikR22
U(r2 , ω) = – U( r1 , ω) + U( r2 , ω) . (7)
2π R12 R22
eikRij
Kij = . (8)
Rij
We first turn our attention to pairs of points that lie in the plane x = 0, i.e., the
plane that bisects the line joining the two pinholes and is perpendicular to the
* It is to be noted that the inclusion of obliquity factors in our analysis would not affect our results;
cf. Ref. [12].
324 Young’s Interference Experiment: The Long and Short of It
screen:
r1 = (0, y1 , z1 ), (9)
r2 = (0, y2 , z2 ). (10)
On substituting from Eqs. (6) and (7) into Eq. (4), while making use of Eqs. (11)
and (12), we obtain the expression
kA 2
*
W(r1 , r2 , ω) = K11 K22
2π
1/2
where
are the spectral densities of the field at pinholes Q1 and Q2 , respectively, denotes
the real part, and
is the spectral degree of coherence of the field at the two pinholes. Also, on substi-
tuting from Eqs. (6) and (7) into Eq. (5), while again using Eqs. (11) and (12),
we obtain the formulae
1/2
Combining Eqs. (13), (17), and (18), we find for the spectral degree of coherence
the expression
* K
K11 22
µ(r1 , r2 , ω) = , (19)
|K11 ||K22 |
Taco D. Visser 325
and hence
That is, the spectral degree of coherence of the field at any two points in the plane x = 0
is unimodular, irrespective of the state of coherence of the field that is incident upon the two
pinholes.
We next consider pairs of points that are each other’s mirror image with respect
to the plane y = 0, i.e.,
r1 = (x1 , y1 , z1 ), (21)
r2 = (x1 , –y1 , z1 ). (22)
On substituting from Eqs. (6) and (7) into Eq. (4), while using Eqs. (23) and (24),
we obtain the expression
kA 2
W(r1 , r2 , ω) = |K11 |2 S1 (ω) + |K22 |2 S2 (ω)
2π
1/2 *
On substituting from Eqs. (6) and (7) into Eq. (5), while again using Eqs. (23)
and (24), we find that
Substituting from Eqs. (25), (26) and (27) into definition (3) yields the result
µ(r1 , r2 , ω) = 1. (28)
Stated in words, the spectral degree of coherence of the field at two points that are each
other’s mirror image with respect to the plane y = 0, is always unity, i.e., the field at P1
326 Young’s Interference Experiment: The Long and Short of It
and P2 is fully coherent and cophaseal, irrespective of the state of coherence of the field at
the two pinholes.
This result has a clear physical meaning: according to the spectral interference
law [10, Sect. 4.3.2], if the light at P1 and P2 is combined (for example, by cou-
pling the light into two fiber tips) and used in a second Young’s experiment, the
resulting interference pattern will have fringes with perfect visibility.
We emphasize that the surprising results expressed by Eqs. (20) and (28) also
hold in the limiting case when each pinhole is illuminated by a separate laser. It is to
be noted that taking obliquity factors into account does not alter the outcome of our
analysis (for a proof of this, see [12]). Moreover, these results have recently been
generalized to the case of partially coherent, partially polarized electromagnetic
beams [13].
where we have used Eqs. (6) and (7). In the far zone the factors Kij have the
approximate form
exp[ik(Rj – r̂j · ri )]
Kij ≈ , (31)
Rj
where Rj = |rj | is the distance from the origin to the point of observation Pj , and
r̂j is a unit vector pointing in the direction OPj . On substituting from Eq. (31)
Taco D. Visser 327
respectively. The phase, φµ (r1 , r2 , ω), of the spectral degree of coherence was
calculated for the case in which x2 is varied while y2 , z2 , and r1 are kept fixed.
An example of its discontinuous behavior is depicted in Fig. 6. The change by an
amount ±π of the phase across a phase singularity is clearly seen.
Figure 6 Illustrating the discontinuous behavior of the phase φµ of the spectral de-
gree of coherence across a phase singularity. In this example r1 , y2 , and z2 are kept fixed
while x2 is varied. Here k = 0.333 × 107 m–1 , d = 0.1 cm, µ12 (ω) = 0.8 + 0.3 i,
r1 = (0, 0, 1.5) m, y2 = 0.9 mm, and z2 = 1.5 m.
Taco D. Visser 329
We have also studied the phase of the spectral degree of coherence in a plane
parallel to the screen that contains the apertures. An example is shown in Fig. 7.
The vertical line indicates the location of a phase singularity, i.e., a set of points
P2 (r2 ) for which µ(r1 , r2 , ω) = 0, and hence the phase of the spectral degree
of coherence is singular. Four pairs of contour lines show the discontinuity of the
phase φµ across the phase singularity. In all four cases the phase undergoes a jump
equal to ±π, in agreement with the asymptotic behavior predicted by Eq. (32).
We note that the phase singularities of the spectral degree of coherence can
easily be observed. Detecting the phase change requires interfering the light from
the vicinity of the pair of points P1 and P2 . This can be done by coupling the
light from these points into another Young’s interference experiment (for example,
by using optical fibers) and observing the behavior of the spectral interference
fringes produced in this secondary experiment as the point P2 is moved across
the phase singularity (see Fig. 8). In Fig. 9 the fringe patterns shown would be
observed in this secondary experiment for a selection of points P1 , P2 . The point
P1 (r1 ) was chosen as in Fig. 7, and the point P2 was taken along a line of constant
phase at several points in the vicinity of the phase singularity; the choices of P2
(A, B, C, D, and E) are illustrated in Fig. 7. It can be seen in Fig. 9 that the π phase
change results in the minima of the secondary fringe pattern becoming maxima,
and vice versa, in accordance with the spectral interference law [10, Sect. 4.3.2].
Figure 7 Contours of equal phase of the spectral degree of coherence µ(r1 , r2 , ω) near
a phase singularity (the vertical line) in a plane parallel to the screen. In this exam-
ple k = 0.333 × 107 m–1 , d = 0.1 cm, µ12 (ω) = 0.8 + 0.3 i, r1 = (0, 0, 1.5) m, and
z2 = 1.5 m (after [14]).
330 Young’s Interference Experiment: The Long and Short of It
Figure 8 Combining the light from the two observation points P1 and P2 in a secondary
Young’s interference experiment using fibers. The end points of the fibers act as point
sources situated in a second screen B. The visibility of the resulting interference fringes on
a third screen C changes as the point P1 is kept fixed while P2 is scanned across a phase
singularity of µ(r1 , r2 , ω).
Figure 9 Illustrating the spectral interference pattern formed along the x-direction by
combining the light from the two observation points P1 and P2 in a secondary Young’s
interference experiment. The observation plane was taken to be at z = 1.5 m, and the
spacing of the pinholes taken to be d = 0.1 cm. The positions A, B, C, D, and E of the
points P2 are illustrated in Fig. 7. S0 is the spectral intensity normalized to the value of the
spectral intensity on the curve C.
ity of subwavelength apertures in a realistic screen, i.e., a screen with a finite thick-
ness and a finite conductivity. Not only does our analysis reveal new effects such
as the creation and annihilation of phase singularities of the Poynting vector, but
it also shows that in certain cases the two apertures can strongly influence each
other through surface plasmons that are generated on the screen. The latter will be
discussed in Sect. 15.6.
We first study a single, infinitely long slit in a thin metal plate. The slit runs
in the y-direction. The incident field, taken to have time dependence exp(–iωt),
propagates in the positive z-direction, perpendicular to the plate. In this case, a
scalar approach, as was used in the previous sections, does not suffice, so rigorous
electromagnetic theory must be used to obtain the field. Specifically, the following
integral equation for the electric field [15] has to be solved:
Êi (x, z) = Êi (x, z) – iωε ĜEij (x, z; x , z )Êj (x , z ) dx dz ,
(inc)
(34)
D
where ε = ε0 – εplate is the difference between the vacuum permittivity and the
E
permittivity of the metal plate, Ĝ is the electric Green’s tensor pertaining to the
(inc)
plate without the slit, and Êi is the ith component (i = x, y, z) of the incident
field, i.e., the field that is present when there is no slit in the plate. The integration
is over the domain D of the slit. For points which lie within the slit, Eq. (34) is a
Fredholm equation of the second kind for Ê, which can be solved numerically by
the collocation method with piecewise-constant basis functions. The electric field at
observation points outside the domain of the slit is then calculated by substituting
this solution back into Eq. (34). With the electric field determined everywhere, the
magnetic field Ĥ follows directly from Maxwell’s equations.
We are interested in the singular optics behavior of the real-valued, two-
dimensional time-averaged Poynting vector field,
1 ∗
Sz (x, z)
sinφS (x, z) ≡ ,
|S(x, z)|
Sx (x, z)
cosφS (x, z) ≡ . (36)
|S(x, z)|
It follows from these equations that φS (x, z) is singular at those points where
S(x, z) = 0.
332 Young’s Interference Experiment: The Long and Short of It
An example of the power flow field (i.e., the time-averaged Poynting vector)
near a narrow slit in a thin silver plate is shown in Fig. 10. In this example the in-
cident field is taken to be TE polarized (i.e., the Ê field is parallel to the slit). It is
seen that the field exhibits several phase singularities, namely vortices and saddles.
In addition, the aperture is seen to have a funnel-like effect on the field, corre-
Figure 11 Behavior of the time-averaged Poynting vector when the slit width is in-
creased to 250 nm. The two vortices a and b remain, whereas the four phase singularities
c, d, e, and f that were visible in Fig. 10 have annihilated each other.
Taco D. Visser 333
sponding to an enhanced light transmission [16]. When the slit width is increased
in a continuous manner, the four singularities below the slit (c, d, e, and f) move
together and eventually annihilate each other. In this process topological charge
is conserved. As can be seen from Fig. 11, the annihilation results in a smoother
power flow field, corresponding to a greater power transmission. Further examples
of such creation and annihilation events are given in [17]. The relation between
such events and the onset of guided modes within the slit is discussed in [18].
When the incident field is TM polarized (i.e., when the Ê field is perpendicular to
the slit), surface plasmons may be generated [19]. These are electromagnetic fields
that travel along the interface between a metal and a dielectric. The field component
normal to the interface decays exponentially. Because the surface plasmon decay
length on the interface (i.e., the propagation distance over which the amplitude of
the fields decreases by a factor 1/e) is much larger than the skin depth of the metal,
it is possible for plasmons that are generated at one slit in the screen to travel to
another slit. When a plasmon exchanges momentum with the screen, for example
at a slit, it can be converted into a propagating light field. There are conflicting
reports in the literature on whether the generation of plasmons actually helps or
frustrates the light transmission process [20,21]. In this section we demonstrate
some unexpected consequences of plasmon excitation for the light transmission
through two narrow slits.
The two-slit configuration that we analyzed is depicted in Fig. 12. A metal plate
of thickness d with two parallel slits is illuminated by a plane, monochromatic, TM
polarized wave that is normally incident upon it. The slits each have a width w and
are separated by a distance b. The plate is surrounded by vacuum. Let us write the
complex-valued relative permittivity of the metal plate as
ω m 1/2
kx = , (39)
c m + 1
334 Young’s Interference Experiment: The Long and Short of It
with c the speed of light in vacuum. Using the integral equation formalism de-
scribed in Sect. 15.5 we have analyzed the light transmission process for two par-
allel narrow slits. First, for TM polarized incident light the amplitude of the field
along a cross section was calculated some distance away from the slits. In Fig. 13
the exponential fall off of Ez , the electric field component normal to the two in-
Figure 13 The transverse field profile |Ez |2 (in arbitrary units), showing the exponential
fall off of the field amplitudes normal to the interfaces that is typical of a surface plasmon.
In this example the cross section is taken at a distance of two wavelengths from the nearest
slit. The two slits are both 30-nm wide. The silver plate has a thickness of 100 nm, and its
permittivity m = (0.05 + i2.87)2 . The slit spacing is 450 nm. The incident TM field,
traveling in the positive z-direction, has a wavelength λ = 500 nm.
Taco D. Visser 335
terfaces, that is characteristic for surface plasmons can be seen. We conclude that
plasmons are indeed generated. (Note that for TE polarized light the amplitude of
Ez is zero everywhere.)
In Fig. 14 the light transmission of a two-slit system is shown as a function of
the separation distance b between the two slits (cf. [22]). For TE polarized light,
no plasmons are generated and the spacing hardly affects the field transmission. For
TM polarized light, however, a very strong modulation of the light transmission
can be seen. The modulation period coincides exactly with the plasmon wavelength
λsp = 2π/kx , with kx given by Eq. (39). Clearly, the plasmons that are generated
at each of the two slits and travel toward the other can interfere with each other
either constructively or destructively. In the former case they give rise to enhanced
transmission (i.e., a transmission greater than one); in the latter case they cause
frustrated transmission (i.e., a transmission smaller than one).
We conclude that for TM polarized light, the light transmission through two
narrow parallel slits is dominated by the effect of surface plasmons. In contrast
to claims in the literature (Refs. [21] and [22]), they can either give rise to an
enhanced transmission or to a frustrated transmission. The effect of this surface
plasmon-induced transmission depends on the spacing between the two slits. Experi-
mental verification of this prediction has already taken place.
Figure 14 The light transmission for a two-slit system as a function of the separation
distance b between the two slits. For TE polarized light no plasmons are generated, in
contrast to the TM case. The transmission is normalized to the intensity that is incident
onto the two slits according to geometrical optics.
336 Young’s Interference Experiment: The Long and Short of It
15.7 Conclusions
References
View publication st