6.

013 Recitation 24:

Dispersion and Interferometry
Pulse spreading due to dispersion

As mentioned previously, one of the key properties of an optical fiber is its dispersion.
The phase velocity of a single frequency propagating along the fiber is simply given by
vp = ω/k = (εµ)-1/2 . But, since ε can vary with frequency, the phase velocity is generally
a function of frequency. Then, as we have seen before, the envelope of a signal
(necessarily comprised of a band of frequencies) travels with the group velocity vg =
dω/dk. If, in addition, the group velocity also varies with frequency, different parts of the
signal will travel at different group velocities. The signal will spread and distort. This
kind of dispersion is referred to as group velocity dispersion, or GVD.

We can make a simple estimate of how rapidly a pulse spreads in the presence of GVD.
Consider a pulse of duration τp. It must necessarily (by Fourier analysis) be
comprised of a frequency spectrum with a width of at least ∆ω = 1/τp. As the pulse
propagates it spreads, as illustrated in Figure 1, to a width of ∆t at a distance of L.

∆t
τp

Figure 1: Pulse spreading with distance caused by group velocity dispersion (GVD)

This amount of spreading occurs because the frequency components ω1 and ω2 separated
by ∆ω become separated by

L L d 1 d2k
∆t = − =L⋅ ( ) ⋅ ∆ω = L ⋅ ⋅ ∆ω (1)
v g1 v g2 dω v g dω 2
We have used (1/vg) = dk/dω and assumed a small change in vg between ω1 and ω2. So,
the spreading is directly proportional to d2k/dω2 and to the distance L. It is also apparent,
since ∆ω = 1/τp, that the shorter the initial pulse duration τp is , the faster it spreads.

Unlike in metallic waveguides or other strongly guiding structures, the dispersion of

optical fibers is rather strongly influenced by the dispersion of the material itself. Figure
2 shows how the GVD of optical fiber varies with wavelength. The units of d2k/dω2 are,
as per convention, ps2/km. (A picosecond is 10-12 sec.)

1
 100 SiO2

50 Modified by
waveguide design

β2 (ps2/km)
0.5 1.0 1.5 2.0 λ (µm)

-50

-100

Figure 2: The group velocity dispersion of optical fiber vs. wavelength

The solid line is the GVD for the fused silica (SiO2) glass material by itself. It has a
value of zero at a wavelength of about 1.3 µm, which was therefore initially thought to be
the optimum wavelength for optical communications. With the advent of erbium-doped
fiber amplifiers (EDFA’s) that operate at wavelengths near 1.55 µm, attention moved to
those longer wavelengths. Fiber designers found too that they could shift the zero GVD
point into the 1.55 µm regime by proper design of the waveguide (generally by making
the core smaller and more tightly guiding. It is even possible to make a length of fiber
with one sign of GVD to cancel out the GVD of another fiber with the opposite sign. (In
fiber with positive d2k/dω2, lower frequencies travel faster; and in fiber with negative
d2k/dω2, higher frequencies travel faster.)

Example: Standard single mode optical fiber has a dispersion of 20 ps2/km at 1.55 µm.
If 10ps pulses are used to transmit data at a rate of 40Gb/s, the pulses will broaden to
25ps and begin to interfere strongly with each other after a distance of L =
(25ps)(10ps)/20ps2/km = 12.5 km. This is considerably less than the usual 50 km
between amplifiers and/or the total span in main trunk lines of 500 km. So, dispersion
compensation is necessary.

More generally: The reason for describing GVD in terms of d2k/dω2 is best understood
by considering the expansion of the propagation constant k in a Taylor series around a
particular center frequency ωo.

dk d2k 1
k(ω) = k o + ( ω − ωo ) + ⋅ (ω − ωo ) 2 + ....... (2)
dω dω 2 2

Since propagation of the wave is determined by e − jk ( ω) z , it becomes clear that the first
derivative dk = 1 is the coefficient of the jkz phase term that varies linearly with
dω vg
2
frequency, and that the second derivative d k2 = d ( 1 ) , the GVD term, is related to the
dω dω v g
jkz phase that varies quadratically with frequency.

2
From this general description of dispersion in terms of the different derivatives of k(ω) it
follows that there is an easy prescription for describing the propagation and spreading of
an arbitrary temporal signal given at z=0 by E(t,0). One simply finds its Fourier
transform E(ω), multiplies E(ω) by expj[ko+(dk/dω)(ω) + (1/2)(d2k/dω2)(ω 2)]L, and then
takes the inverse Fourier transform to determine E(t,L).

Interferometry
A large variety of measurement, signal analysis, modulation and switching applications
rely upon the use of wave interference. This is especially true at optical frequencies
because, on the one hand, the short wavelengths provide high resolution and, on the other
hand, optical detection systems cannot directly observe small changes in the electric field.
Perhaps the simplest, and certainly the most famous, device for utilizing wave
interference is the Michelson interferometer illustrated in Figure 3.

l1 S

So
So
l2

S
π 2π 3π 2π
(l 2 − l 1 )
Detector λ

Figure 3: Michelson Interferometer Detected output intensity

The interferometer operation can be idealized as follows. A uniform plane wave Eo e-jkz
is incident from the left onto a beam splitter that has a reflection coefficient r and a
2 2
transmission coefficient t . Power conservation requires that 1 − r = t . Both reflected
and transmitted beams are subsequently reflected from perfectly reflecting mirrors and
return to the beam splitter after traveling distances of 2 l 1 and 2 l 2 respectively. Then
after being again transmitted and reflected respectively by the beamsplitter they combine
(in the downward direction in the figure) to produce a field

E( x ) = E o ⋅ r ⋅ t ⋅ (e − jk ( x + 2l 1 ) + e − jk ( x + 2 l 2 ) )
= E o ⋅ r ⋅ t ⋅ e − jkx ⋅ e − jk ( l 2 + l 1 ) ⋅ (e jk ( l 2 −l 1 ) + e − jk ( l 2 −l 1 ) ) (3)

The intensity S traveling in the x (downward) direction to the detector is then

1 2 1 2 2 2
S= E( x ) = E o ⋅ r t ⋅ 4 cos 2 k(l 2 − l 1 ) (4)
2η 2η

3
 2 2 1
If the beam splitter divides the power 50:50, r = t = and
2
1 2 1 2
S= E( x ) = E o ⋅ cos 2 k(l 2 − l 1 ) = S o cos 2 k(l 2 − l 1 )
2η 2η
1
= S o {1 + cos 2k(l 2 − l 1 )} (5)
2
λ
So, periodically when l 2 − l 1 = m ( for m = 0,1,2...) all of the incident intensity So
2
reaches the detector. (Where does the power go when the output is zero, i.e.
at cos 2 k(l 2 − l 1 ) = 0 ?)

Some of the most accurate distance measurements possible are made with Michelson
interferometers by monitoring the output intensity S as one of the mirrors is moved. An
optical wavelength is only about 0.5 µm, and it is easy to resolve the change in intensity
due to a length change equal to a small fraction of a wavelength. The amplitudes of
periodic motions in particular (because they are easier to distinguish from noise) can be
resolved with accuracies on the order of the size of an atom. Interferometers are used to
determine shifts in the earth’s crust, and a major research effort is currently underway to
use a interferometer to detect motion due to gravitational waves from outer space.

The Michelson interferometer is also quite often used as a “wavemeter,” a tool for
determining optical wavelengths and frequencies. A known wavelength is passed
through the interferometer for calibration and other wavelengths are identified from their
different periodicities in the output. More generally - since the electric field at the output
of the interferometer is the coherent sum of the cosinusoidal patterns produced by all the
individual incident frequencies, a plot of the output obtained by moving one of the
mirrors is in fact the Fourier Transform of the input frequency spectrum. The technique
for studying spectra this way (because it was first used for infrared wavelengths where
other forms of spectroscopy are less efficient) is called FTIR (Fourier Transform
Infrared) spectroscopy.

An alternative device for interferometry is the Mach-Zehnder interferometer illustrated

in Figure 4.
So
t
r l1

S1
l2 r
t

S2

Figure 4: Mach-Zehnder interferometer

4
 2 2
In this case the output beam that has intensity proportional to r ⋅ t (like the output of
the Michelson) emerges to the right. Since the path lengths in this case are l 1 and
k
l 2 rather that 2l 1 and 2l 2 the output is proportional to cos 2 (l 2 − l 1 ) . As (l 2 − l 1 )
2
changes, however, intensity that doesn’t emerge to the right emerges in the downward
direction (rather than being reflected by the interferometer as is the case with the
Michelson). By power conservation the intensity of the second output is proportional to

k k
1 − cos 2 (l 2 − l 1 ) = sin 2 (l 2 − l 1 ) . Thus,
2 2
1 1
S1 = S o (1 + cos k(l 2 − l 1 ) and S2 = So (1 − cos k(l 2 − l 1 ) (6)
2 2

The Mach-Zehnder interferometer has achieved greatly increased importance with the
development of integrated optical waveguide technology. It can be implemented simply
in the form shown Figure 5. Incoming waves are divided equally into the two waveguide
arms of the interferometer. If they combine in-phase again at the second waveguide
merger, all of the input power is transmitted. If they combine out-of-phase at the second
waveguide merger, there is no output S1. (In a closed waveguide system the power
would then be reflected back out the input port, but in an integrated low-index-contrast
dielectric waveguide device the out-of-phase combination produces power in a forward
traveling mode that is not trapped by the output waveguide and that therefore radiates
away.)
V
+
So S1
−

Figure 5: The waveguide Mach-Zehnder (with modulation electrodes)

There are two important applications of the waveguide Mach-Zehnder: (1) as a

wavelength filter, and (2) as a modulator, used for putting data on optical beams.
Consider first the filter application. It follows from the wavelength dependence of the
output when l 2 ≠ l 1 . Assume that the output (S1) is a maximum at λo when
l 2 − l 1 = ∆L , i.e. π∆L/λo= mπ. Then transmission is zero for λ = λo + ∆λ when

π∆L π
= mπ +
λ o + ∆λ 2

Now, assuming that ∆L is large enough so that ∆λ << λο, we obtain

5
π∆L ∆λ π λo
(1 − ) = mπ + ⇒ ∆λ = ⋅ λo (7)
λo λo 2 2∆L

Thus, the fractional bandwidth of the filter (∆λ/λo) is inversely proportional to ∆L.

The second, modulator, application is achieved by forming the waveguides in an electro-

optic material - a material whose ε changes linearly with applied electric field. Then, a
voltage applied between electrodes around one of the interferometer arms (as illustrated
in Fig. 5) produces a change in dielectric constant that creates a different effective path
length (different phase velocity) in that arm – proportional to the voltage V. A DC bias
voltage Vo can be applied in addition to the signal voltage V(t) so that the optical output
can be made to either increase or decrease with signal voltage V(t) depending upon the
bias. In a typical device made of LiNbO3 v p = 1/ εµ o the change in index of refraction
∆n = ∆ ε / ε o ≅ 3 ⋅ 10 −10 ⋅ E where E (v/m) is the electric field in the waveguide arm.
Assuming that the voltage v is applied across a dimension of 20 µm, then the induced
optical phase delay in the arm is

2π
∆φ = (1.5 ⋅ 10 −5 ) ⋅ l ⋅ V
λ

For a λ = 1.5 µm, the desired ∆φ = π for maximum optical output change requires a
voltage length product 2.6 x 10-2 volt-m. To achieve switching with a voltage of 5 volts
then requires the length of the waveguide arm to be about 1 cm.

Example: One might also ask how high a modulation speed is possible. At the speed of
light in LiNbO3 ( vg ≅ c/2.2 ) it takes about ∆t = 7.3 x 10-11sec for light to travel beneath
the electrode. This transit time limits the modulation bandwidth to frequencies for which
∆t < T/2 = 1/2f , i.e. to f < 6.8 GHz. Higher modulation speeds can still be achieved by
using traveling wave electrodes so that the modulating voltage can move along with the
optical wave.