Practice Final, EE231

This is just for you to practice, if you wish. Do not turn your solutions: you will not be
graded for it.

Problem 1

We want to compute the oscillation frequency of a semiconductor laser, taking into

account frequency pulling. The only reason for choosing a semiconductor in this problem
is first that its gain medium can be very short, which conveniently limits the number of
possible oscillation frequencies, and second that the gain at threshold is fairly high, so
that frequency pulling is significant, as we discussed in class. However, we need not be
concerned about the differences in the laser physics between a semiconductor laser and
the lasers we studied in class. In fact, for simplicity, we assume here that as a gain
medium the semiconductor behaves like a homogeneously broadened medium. We take
the center wavelength of its gain curve to be at a free-space wavelength of λa = 806 nm,
and its full width at half maximum to be ∆λa = 0.2 nm. The laser is made of a
semiconductor waveguide (no need to consider the mode aspect of the beam) with a
length L = 500 µm and a refractive index of 3.7. The reflectors are formed by the end
faces of the semiconductor, which are polished flat and perpendicular to the waveguide
axis. One face has been coated to have a reflectivity of R1 = 100%. The other one has a
reflectivity R2 = 10%. Assume that the two mirrors are lossless, and that there is no other
loss in the cavity.

(1) Calculate the axial mode spacing for this cavity. Throughout this problem, retain five
significant digits to avoid rounding errors.

SOLUTION:
∆νax=c/(2Ln) = 8.1025*1010 Hz

(2) What is the formula for the resonant frequencies supported by the cold cavity, i.e., in
the absence of gain? Label the number of the resonant frequency q, as we did in class.

SOLUTION:
Since the reflectors are flat, g1 and g2 are 0, so the formula for the resonant frequencies of
the cavity is: νq= qc/(2Ln) or ωq= πqc/(Ln)

(3) Which value of q that will oscillate?

SOLUTION:
We just need to find the value of q for which νq is closest to c/λa = 3.7195*1010 Hz. This
turns out to be q=4591.

(4) Calculate the value of the round-trip loss of the cavity (recall that the approximation
we used in class for small loss may not be applicable here).

SOLUTION:
We use the exact formula from slide 8.8 for the mirror loss: -ln(R1R2) = 2.3026.

(5) Calculate the value of frequency pulling of the active cavity, i.e., when it is lasing.

SOLUTION:
g m (ω ) 2(ω − ω a )
We use the formula from slide 11.20: δω q = − ∆ω ax with gm(ω) =
2π ∆ω a
2.3026 (gain = loss) and ω= ω4591 (the resonant frequency of the cold cavity). The other
parameters are as given above. The result is: δωq = -1.4061*1011 rad/s, or δνq = -
2.2379*1010 Hz.

(6) What is the actual frequency of the laser oscillation? Also give your answer in terms
of wavelength. The answer should be of course very close to a free-space wavelength of
806 nm.

SOLUTION:
Using ωosc = ω4591+ δωq, we find a final frequency that corresponds to a free-space
wavelength of 805.97 nm.

Problem 2

We discussed early in the quarter the possible presence of anisotropy in the response of
an electron cloud to an electric field, and how it translates into an atomic susceptibility,
and hence an atomic cross-section, that depend on the atom's orientation in space and on
the state of polarization (SOP) of the field. We ignored these effects in the course not to
obscure the issues we investigated and to simplify the algebra, but anisotropy is very real,
albeit often small, and in many laser materials it must be accounted for. This problem
explores one manifestation of this anisotropy in solid-state amplifiers, which is
polarization-dependent gain (PDG). This effect is detrimental, and it was critical for fiber
amplifier designers to get rid of it before long-haul fiber communication links were
deployed in the mid-90s.

Consider a host doped with anisotropic laser ions, for example Nd3+. The presence of
anisotropy means that each ion has an absorption and emission cross-section that depends
on direction and polarization. In its own Cartesian coordinates (a, b, c), each ion has a
pump absorption cross-section that has an ellipsoidal distribution with axis values σa, σb,
and σc, with σa = σb ≠ σc (see Fig. 1). For example, pump light incident along c and
polarized along a will experience an absorption cross-section σa (see Fig. 1), and σb (=
σa) if it is polarized along b. Pump light incident along a and polarized along c will
experience an absorption cross-section σb, etc. The ratio of major to minor cross-sections
is the anisotropy ε = σc/σa. When ε = 1, the ellipsoid is a sphere and absorption is
isotropic (independent of SOP). When ε = 0, the ellipsoid is a line, and the ion is as
anisotropic as can be: it absorbs only light polarized along c (a rare case). Similarly, the
signal (laser) emission cross-section for this same ion has an ellipsoidal distribution with
different axis values σa', σb', and σc', but (1) the same orientation in space and (2) the
same anisotropy ε as the absorption ellipsoid. Finally, all laser ions have the same
absorption ellipsoid (and the same emission ellipsoid), except that the ellipsoids are
randomly oriented in space within the laser host, as illustrated in Fig. 2.

This laser material is pumped with pump light propagating along z and linearly polarized
along y, and it is used to amplify a signal co-propagating along z. We want to calculate
the unsaturated gain (weak signal and no depletion of the ground level) for an x-polarized
signal and for a y-polarized signal as a function of ε. To do so, and for simplicity, ina first
stage we assume that this population of randomly oriented ions has the same effect as if
their individual c axes were aligned either along x, along y, or along z, with equal
probability. Take the case of an ideal four-level laser.

a) Write the expression of the population inversion for each of the three groups of ions as
a function of pump intensity and total ion population N0.

SOLUTION:
See handwritten solution below

b) Apply the usual gain coefficient expression γ0 = σe∆N to write the contribution of each
group to the total gain coefficient for a signal polarized along x, then write the total gain
coefficient γ0,x seen by the signal. Express your result in terms of a first-order polynomial
of ε (times some constants).

c) Repeat b) to derive the total gain coefficient γ0,y for a signal polarized along y. Express
your result in terms of a second-order polynomial of ε (times the same constants as in b)).

d) Sketch by hand the ratio D = γ0,x/γ0,y, which is a measure of PDG, as a function of ε,

for ε values between 0 and 2 (a few points will be sufficient to show the trends). If your
result is correct, you should find that D(ε = 1) = 1, as expected.
e) D(ε) is not very symmetric, unlike expectation. The reason is that we oversimplified
the random distribution of dipoles. To get a more accurate answer, repeat the calculation
of D(ε) assuming now that the ions can only be oriented (again with equal probability)
either along x, along y, or at 45° to x and y (we ignore the effect of the dipoles oriented
along z). Hint: for the 45° orientation, the pump sees a pump absorption cross-section (σa
+ σc)/√2 for either x or y polarization, and the signal an emission cross-section (σa' +
σc')/√2 again for either polarization. Sketch D(ε). You should find now that D(ε = 0) =
D(ε = ∞).
Problem 3

Consider an optical fiber of the kind discussed and modeled in class, with a circular core
of radius a = 4.0 µm and refractive index 1.460, surrounded by a cladding of index 1.454.
It carries light at a wavelength of 1.55 µm. Light coming out of the far end of the fiber
diffracts in free space. The object of this problem is to figure out what lens we should use
to recollimate it into a free-space Gaussian beam with a specific size. You can use any of
the (relevant) formulas presented in class.

(1) How many modes does the fiber core support, including both linear states of
polarization?

SOLUTION:
First, we find the V-number of the fiber: V = 2πaNA/λ. The NA is given by (1.462-
1.4542)1/2 = 0.132. Hence V = 2.144. This is below 2.405, the cutoff frequency for the
LP11 modes, so this is a single-mode fiber. There are two LP01 mode, with orthogonal
polarizations, so the total number of modes is two.

(2) What is the value of the fundamental mode 1/e2 power radius?

SOLUTION:
Using the approximate formula W01 = a(0.65 + 1.619V-1.5 + 2.879 V-6), we get
W01 = 4.78 µm.

(3) If we let the output of the fiber diverge in free space to infinity, and measure its
divergence angle as defined in class, what is the value of this angle?

SOLUTION:
The far-field divergence angle is that given for a Gaussian beam in lecture 2:
tan(θ) = λ/(πW01). Using W01 = 4.78 µm gives θ = 0.103 radian.

We want to recollimate this output beam with a lens such that it forms a beam with a size
that fits the requirements of a particular experiment. The experiment consists in the
following. The recollimated beam is launched into a cylindrical cell filled with a
biological species to carry out some spectroscopy measurements. The diameter of the cell
is 1 mm, and its length is L = 10 cm. Take the index of the medium in the cell to be 1,
and assume that the input and output windows of the cell, through which the beam will be
entering and exiting the cell, are appropriately anti-reflection coated so that the beam
suffers no reflection losses at is enters and exits the cell. We want the beam to be focused
at the center of the cell, and in such a way that the beam intensity on axis does not vary
by more than 10% from one end of the cell to the other. We also want the beam power
not to be cropped by the aperture of the cell by more than 10%, i.e., we want 90% of the
beam power to go through the input window.
(4) Calculate the value of the 1/e2 power radius W0 that the beam should have at its waist
to meet these two requirements.

SOLUTION:
First, let’s deal with the peak intensity constraint: If the beam is focused at the center of
the cell, we know that the beam waist is there. So we need to find a beam waist w0 such
that at z = ±5 cm, the on-axis intensity is at least 90% of the intensity at z=0. We know
1
that the on-axis intensity goes like and that
w( z ) 2
zλ 2
w( z ) = w0 1 + ( z / z R ) 2 = w0 1 + () . In order for the intensity on axis at the ends of
πw02
the cell (z = ±5 cm) be at least 90% of the intensity on axis at the center of the cell, we
must impose that w2(z = ±5 cm) ≤ w02/0.90. Inserting the above expression for w(z) as a
function of w0 and solving for w0 gives w0 ≥ 3zλ / π , or w0 ≥ 272 µm.

Now, let’s deal with the power that passes through the aperture. The fractional power
carried by a Gaussian beam with radius w0 that passes through an aperture of radius R is
P(R)
= 1 − exp[−2(R / w)2 ] . In this expression, w is the beam radius at the aperture, i.e.,
Ptot
at z = 5 cm. We want the aperture radius R to be such that P(R) ≥ 0.90Ptot. Using our
expression for w(z) in terms of w0 above, we can solve this inequality and get
w(z) ≥ 2R 2 / | ln(0.1) | . Since the aperture radius is R = 0.5 mm, we get w = 466 µm.
Inserting this value in our expression for w(z) versus w0, we find that the only possible
values of w0 that satisfy this last inequality are 57 µm ≤ w0 ≤ 430 µm.
Looking at the intersection of the two conditions, we find that the range of w0 that
satisfies our two criteria is 272 µm ≤ w0 ≤ 430 µm.
Alternatively, you could solve the two inequalities graphically, as shown in the graph
below.
(5) We recollimate the free-space beam coming out of the fiber with a lens to produce the
beam that goes through the cell (i.e., a beam of 1/e2 power radius W0). Assume the lens is
placed right against the input window of the cell (distance = 0). What should be the focal
length of the lens?

SOLUTION:
To find the necessary focal length, we can simply use the formula wf=fλπwin and find that
we need f=3mm. (Here, we have chosen wf = 350 µm. As shown above, any value of wf
between 272 µm and 430 µm works.)

Problem 4.

Consider a Nd-doped silica fiber amplifier operating at a signal wavelengh λs = 1060 nm

and pumped with 20 mW of absorbed pump power at a wavelength λp = 808 nm. Take
the pump and signal mode areas to be 20 µm2, the emission cross-section σe = 2.5 10-24
m2, and the lifetime of the upper laser level τ = 500 µs.

(1) What is the value of the unsaturated (i.e., small-signal) power gain G0?

SOLUTION:
στ Pabs
We have g 0 = = 5.0844, and G0 = exp(g0) = 161.493.
hω p A
(2) What is the value of the power gain when a signal of power equal to 10 mW is
coupled into the amplifier? You will need to solve a transcendental equation to get the
answer; do it by trial and error on a calculator, which only takes a few tries. Only push
the accuracy to the second significant decimal (e.g., a gain of 6.54).

SOLUTION:
We use the equation G=G0exp(-(G-1)I1/Isat) and solve numerically to find G= 2.28.

(3) What is the signal power at the output of the amplifier?

SOLUTION:
I2 = G I1, so Pout = G Pin = 22.8 mW

(4) What is the extraction efficiency of this amplifier for this input power, i.e., the ratio of
the extracted intensity to the maximum possible extracted intensity?

SOLUTION:
The maximum possible extracted intensity is Iextr,max=g0Isat=Iavail, and the extracted
intensity is I2-I1. Putting in the numbers we get an extraction efficiency of 84%.

This problem is simple. It just requires plugging numbers in the right formulas.

Problem 5.

Like the Newtonian telescope we discussed in class, a Galilean telescope consists of two
lens of focal lengths f1 and f2, but the focal length f1 is negative, meaning that this lens is
a diverging lens. To avoid sign confusion, write it as f1 = -|f1|. Also, the lens spacing is no
longer f1 + f2 but f2 -|f1|.

(1) Write the ABCD matrix for this telescope, from the input surface of the first lens to
the output surface of the second lens (i.e., not accounting for any propagation distance
before and after the telescope).

SOLUTION:
We can use the derivation on slide 2.30, replacing f1 with -|f1|. The ABCD matrix is:
 f 2 / f1 f 2 − f 1 
M = 
 0 f / f 
 1 2 

(2) By applying the ABCD law to a collimated input beam with a Rayleigh range much
larger than |f1|, show that the beam coming out of this telescope is not focused (as it
should for a telescope), and that it is magnified transversely in the ratio of M = f2/|f1|, i.e.,
its radius is M times larger than the radius of the input beam.
SOLUTION:
Aqin +B
The ABCD law tells us q out = . Using the ABCD matrix above, and solving for
Cqin + D
the real and imaginary parts of qout to obtain Rout and wout, we find that Rout is very large if
zR >> f1, and wout/win = f2/|f1|.