Exercises and Solutions of the book

Light-Emitting Diodes
2nd edition
E. F. Schubert
(Cambridge University Press, Cambridge UK 2006)

1
 Chapter 2

Exercise: Minority carrier lifetimes. Calculate the minority carrier lifetime in p-type GaAs at doping
concentrations of 1015 and 1018 cm–3 using a bimolecular recombination coefficient of B = 10–10 cm3/s.
Assume that one could fabricate GaAs without any impurities. What is the carrier lifetime in intrinsic
GaAs with an intrinsic carrier concentration of 2 × 106 cm–3 ?
Solution: τn = 10 μs for NA = 1015 cm–3
τn = 10 ns for NA = 1018 cm–3
τ = 2500 s for undoped GaAs.
Discuss how the modulation speed of communication LEDs is affected by the radiative lifetime and the
doping concentration.
Solution: When the injection current of an LED is switched off instantaneously at t = 0 and the
radiative lifetime is τ, then light emission from the LED essentially ceases after the time t = τ. Thus an
LED cannot be switched “on” or “off” faster than the time τ. Assuming that both the switch-on and
switch-off time is τ, then the maximum modulation frequency is given by f ≈ 1 / (2τ).
We have shown that the radiative lifetime τ depends strongly on the doping concentration. Thus by
highly doping the active region of a device, the radiative lifetime τ is shortened and the maximum
modulation speed is increased.

2
Exercise: Concentration of point defects. Assume that the energy required to move a substitutional
lattice atom into an interstitial position is Ea = 1.1 eV. What is the equilibrium concentration of interstitial
defects of a simple cubic lattice with lattice constant a0 = 2.5 Å?
Solution: The concentration of lattice atoms of a simple cubic lattice is given by N = a0–3 =
6.4 × 1022 cm–3. The concentration of interstitial defects under equilibrium conditions at room temperature
is then given by

N defect = N exp (− Ea / kT ) = 2.7 × 10 4 cm −3 .

Note that the calculated concentration of defects is small when compared to the typical concentrations of
electrons and holes. If the defect discussed here forms a level in the gap, non-radiative recombination
through the defect level can occur.

3
 Chapter 3

Exercise: Radiative efficiency. Analyze the temperature dependence of the radiative lifetime based on the
van Roosbroeck–Shockley model and the non-radiative lifetime based on the Shockley–Read model and
predict the temperature dependence of the radiative efficiency in semiconductors.
Solution: The radiative recombination rate (van Roosbroeck–Shockley rate) has a weak temperature
dependence, and it depends on temperature according to R = B n p, where B ∝ T –3/2, as concluded from
Eqs. (3.24) and (3.25). The concentrations n and p mostly depend on the excitation strength (injection
current) and can be assumed to be temperature independent. Thus, it is R ∝ τradiative–1 ∝ T –3/2.
The non-radiative recombination rate (Shockley–Read) has a strong temperature dependence and it
increases very rapidly with increasing temperature. The Shockley–Read recombination rate includes the
term cosh [(ET – EFi) / (kT )]. Because cosh x = ½ (ex + e–x) and (ET – EFi) is either < 0 or > 0 (depending on
the location of the trap with respect to the intrinsic Fermi level, EFi), one of the exponential functions
(ex or e–x) dominates, so that an approximately exponential temperature dependence is obtained. Thus, it is
R ∝ τnon-radiative–1 ∝ exp [– 1 / (kT )].
The exponential dependence (Shockley–Read: R ∝ exp [– 1 / (kT )] ) is much stronger than the power-
law dependence (Shockley–Read: R ∝ T –3/2 ), so that the radiative efficiency of semiconductors strongly
decreases with increasing temperature.

4
 Chapter 4

Exercise: Critical points of diode current–voltage characteristics. The I–V characteristics of diodes are
frequently characterized in terms of four critical points, namely forward voltage one, Vf1, forward
voltage two, Vf2, forward voltage three, Vf3, and reverse saturation current, Is, specified at the operating
current (e.g. 100 mA), a small forward current (e.g. 10 μA), a very small forward current (e.g. 1 μA), and
at negative bias (e.g. – 5 V), respectively. The critical points are shown in Fig. 4.6.
(a) Explain the relevance of the critical points.
(b) Two GaInN diodes have the following data: (1) Vf1 = 3.2 V, Vf2 = 2.5 V, Vf3 = 2.3 V, Is = 0.8 μA;
(2) Vf1 = 3.4 V, Vf2 = 2.0 V, Vf3 = 1.8 V, Is = 0.8 μA. Which device has the more favorable
characteristics?
Solution: (a) For devices emitting at the same peak wavelength, Vf1 should be as low as possible, as high
values indicate a high series resistance. The forward voltage two, Vf2, should be as high as possible (as
close to Vf1 as possible), as low values of Vf2 indicate excessive sub-threshold leakage. The same
argument applies to Vf3. The reverse saturation current should be as low as possible as high values of Is
indicate excessive leakage paths (e.g. surface leakage or bulk leakage mediated by surface states, bulk
point defects and dislocations). Low values of Vf1, high values of Vf2 and Vf3, and low values of Is are
consistently correlated with high device reliability. (b) Device (1) has more favorable characteristics due
to lower series resistance and lower sub-threshold leakage.

5
Exercise: Grading of heterostructures. Assume that the conduction band discontinuity of an
AlGaAs/GaAs heterostructure is given by ΔEC = 300 meV and that the structure is uniformly doped with
donors of concentration ND = 5 × 1017 cm–3. Over what distance should the interface be graded in order to
minimize the resistance occurring in abrupt heterostructures?
Solution: Calculating the depletion layer thickness from Eq. (4.22) yields WD = 30 nm. Thus the
heterostructure should be graded over 30 nm to minimize the heterostructure resistance. The graded
region should have two parabolic regions as shown in Fig. 4.9 (b).

6
Exercise: Carrier leakage over a barrier. Electrons in the active region of a GaAs structure have a
concentration of 2 × 1018 cm–3. Calculate the current density of the carrier loss over the barrier for barrier
heights of 200 and 300 meV, assuming an electron mobility of 2000 cm2/(V s) and a minority carrier
lifetime of 5 ns. Compare the calculated leakage current to LED injection currents of 0.1–1.0 kA/cm2.
Solution: The Fermi level in GaAs with electron density of 2 × 1018 cm–3 is 77 meV above the
conduction band edge. Assuming that the effective density of states in the barrier is the same as in the
GaAs active region, the carrier concentrations at the edge of the barrier are 3.9 × 1015 cm–3 for a 200 meV
barrier and 8.3 × 1013 cm–3 for a 300 meV barrier. The diffusion constant, as inferred from the Einstein
relation, is Dn = 51.7 cm2/s. The diffusion length is then given by Ln = (Dn τn)1/2 = 5.1 µm. The leakage
current is calculated using Eq. (4.27), and one obtains 63 A/cm2 for the 200 meV barrier and 1.3 A/cm2
for the 300 meV barrier. Comparison with diode current densities of 0.1–1.0 kA/cm2 suggests that leakage
currents can be a significant loss mechanism, particularly for small barrier heights.

7
Exercise: Carrier overflow in a double heterostructure. Consider electrons in a GaAs double
heterostructure with a barrier height of ΔEC = 200 meV and an active region thickness of WDH = 500 Å.
Calculate the current level at which the electron well overflows.
Solution: Using Nc = 4.4 × 1017 cm–3 and B = 10–10 cm3/s, one obtains from Eq. (4.33) a current level
of Jmax = 3990 A/cm2.

8
Exercise: Drive voltages of LEDs. Calculate the approximate forward diode voltage of LEDs emitting in
the blue, green, and red parts of the visible spectrum. Also calculate the forward diode voltage of LEDs
emitting at 870 nm and 1.55 µm.
Solution: Emission color Wavelength Photon energy Drive voltage
Blue 470 nm 2.6 eV 2.6 V
Green 550 nm 2.2 eV 2.2 V
Red 650 nm 1.9 eV 1.9 V
IR 870 nm 1.4 eV 1.4 V
IR 1550 nm 0.8 eV 0.8 V

9
 Chapter 5

Exercise: LED efficiency. Consider an LED with a threshold voltage of Vth = Eg / e = 2.0 V with a
differential resistance of Rs = 20 Ω, so that the I–V characteristic in the forward direction is given by
V = Vth + I Rs. When the device is operated at 20 mA it emits a light power of 4.0 mW of energy h ν = Eg.
Determine the (a) external quantum efficiency, (b) internal quantum efficiency, and (c) power efficiency,
assuming that the light-extraction efficiency is 50%.
Solution: (a) External quantum efficiency: Number of emitted photons per second is given by
4.0 mW / 2.0 eV = 4.0 × 10–3 CV s–1 / (2.0 × 1.602 × 10–19 CV) = 1.25 × 1016 s–1. Number of injected
electrons per second is given by 20 mA / e = 20 × 10–3 C s–1 / (1.602 × 10–19 C) = 1.25 × 1017 s–1. Thus the
external quantum efficiency is 1.25 × 1016 s–1 / 1.25 × 1017 s–1 = 10%. (b) Using ηexternal = ηinternal ×
ηextraction and using the light-extraction efficiency value of 50%, the internal quantum efficiency is 20%.
(c) The power efficiency is given by Poptical / (I V ) = 4 mW / [20 mA × (2.0V + 20 mA × 20 Ω)] =
4 mW / 48 mW = 8.33%

10
Exercise: Light escape from planar GaAs, GaN, and polymer LED structures. The refractive indices of
GaAs, GaN, and light-emitting polymers are 3.4, 2.5, and 1.5, respectively. Calculate the critical angle of
total internal reflection for GaAs, GaN, and for polymers. Also calculate the fraction of light power that
can escape from a planar GaAs and GaN semiconductor structures and a polymer LED structure.
What improvement can be attained if a planar GaAs LED is encapsulated in a transparent polymer of
refractive index 1.5, if the reflection at the polymer–air interface is neglected?
Solution:
Critical angle for total internal reflection:
GaAs φc = 17.1° GaN φc = 23.6° Polymer φc = 41.8°.
Fraction of light that can escape:
GaAs 2.21% GaN 4.18% Polymer 12.7%.
Improvement of the GaAs planar LED due to polymer encapsulation: 232%.

11
Exercise: LED-to-fiber coupling efficiency. Consider a GaAs LED with a point-like light-emitting
region located in close proximity to the planar GaAs LED surface. An optical fiber has an acceptance
angle of 12° in air. What fraction of the light emitted by the active region can be coupled into the fiber?
Assume a GaAs refractive index of 3.4. Neglect Fresnel reflection losses at the semiconductor–air and
air–fiber interfaces.
Solution: The acceptance angle in the semiconductor is obtained from Snell’s law and is 3.5°. Thus
0.093% of the power emitted by the active region can be coupled into the fiber.

12
 Chapter 6

Exercise: Temperature dependence of diode forward voltage. Experimentally determined linear

temperature coefficients (dVf / dT ) for GaAs diodes range from 1.2 to 1.4 mV/K. Calculate the linear
temperature coefficient of the forward voltage of a GaAs diode with NA = ND = 2 × 1017 cm–3 at room
temperature. What is the decrease in forward voltage if the ambient temperature is increased from 20 to
40 °C and the internal heating in the diode can be neglected?
Solution: For GaAs with α = 5.41 × 10–4 eV/K and β = 204 K, one obtains at room temperature
dVf / dT = – 1.09 mV/K. The decrease in diode voltage for the 20 °C temperature increase is ΔVf =
21.9 mV.

13
Exercise: Compensation of the temperature dependence of an LED with a drive circuit. Consider an
LED with a characteristic temperature T1 = 100 K, a turn-on voltage of 1.4 V at 20 °C, a temperature
coefficient of the turn-on voltage of –2.1 mV/K, and a linear I–V characteristic with a differential
resistance of 5 Ω for forward voltages larger than the turn-on voltage. Assume that the temperature
dependence of the emission intensity is given by I = I |300 K exp [ – (T – 300 K) / T1 ].
Design a drive circuit consisting of a constant-voltage source and a resistor, which compensates for
the temperature dependence of the emission intensity of the LED so that the LED emission intensity is the
same at the water freezing-point temperature (0 °C) and 60 °C. The LED should draw 20 mA at the
freezing-point temperature.
Solution: At 60 °C, the current needs to be 36.4 mA in order to keep the emission intensity
independent of temperature. Constructing a load line that intersects the 0 °C and 60 °C diode I–V
characteristic at 20 mA and 36.4 mA, respectively, yields the following values for the drive circuit:
Constant-voltage source with V = 1.6 V and series resistance of 2.7 Ω.

14
 Chapter 8

Exercise: Current crowding occurring at very high current levels in devices with current-spreading
layer. In device structures with vertical current flow (current flowing from top to bottom of chip), the
current-spreading layer ensures that the current spreads out over the entire p-n junction area. However, as
the current increases to very high levels, the current tends to crowd under the top contact. This is
illustrated in Fig. 8.7 (a) and (b). Explain the phenomenon of current crowding occurring at very high
current densities.
Solution 1: The equation for the current-spreading length has the dependence Ls ∝ J0–1/2. Thus, as the
current density increases, Ls decreases, and the current “bunches” under the top contact.
Solution 2: An intuitive explanation for current crowding can be obtained from the equivalent circuit
shown in Fig. 8.7 (c). At very high current densities, the resistors that represent the p-n junction decrease
(whereas the resistors representing the current-spreading layer remain constant), thereby causing the
current to flow directly downward from the top contact.

15
 Chapter 10

Exercise: Lambertian reflectors in LEDs. Assume a lambertian reflector with reflectivity 1.0 that is
incorporated in a lossless GaAs LED structure with refractive index of 3.5. Assume that the outside
medium is air. Calculate the critical angle of the escape cone, the probability that a reflected light ray falls
within the escape cone and the average number of reflection events before a photon escapes from the
high-index GaAs layer.
Solution: Critical angle Θc = 16.6°; Probability of escape p = 8.2%; Average number of reflection
events before a photon escapes N = 11.7.
Would a hypothetical planar reflector that reflected light coming from any incoming direction towards the
surface normal be useful? Is there a physical principle that prevents a reflector from reflecting light in
such a way?
Solution: Although such a reflector would be very useful, such a reflector would unfortunately violate
the conservation of radiance theorem (previously called the conservation of brightness theorem), which
states that it is impossible to increase the radiance of light by a passive optical system beyond a value of
L/n2 where L is the radiance in vacuum and n is the refractive index of the medium in which the light
propagates.

16
 13.5 Dislocations in III–V nitrides

Chapter 13

Exercise: Activation of Mg acceptors in GaN. Mg acceptors in GaN have an activation energy of

Ea = 200 meV. (a) Calculate the fraction of acceptors that are ionized at 300 K for an acceptor
concentration of NMg = 1018 cm–3 using the formula p = (g–1 NMg Nv)1/2 exp (– Ea / 2kT ) where g is the
acceptor ground state degeneracy (g = 4) and Nv is the effective density of states at the valence band edge
of GaN. (b) What would be the activation of acceptors if a hydrogen atom were bonded to each acceptor?
Solution: (a) Using the formula given above, one obtains that only about 6% of the acceptors are
ionized. (b) If acceptors are passivated, p-type conductivity cannot be established.

17
Exercise: Cracking. Why does cracking occur in epitaxial layers that are under biaxial tensile strain but
not in epilayers that are under biaxial compressive strain?
Solution: Wafer bowing and ultimately cracking of an epitaxial film that is under biaxial tensile strain
releases the strain energy stored in the film. For epitaxial layers that are under compressive strain, the
strain energy can be released by wafer bowing, film buckling, and film delamination. Due to the
compressive strain, there is “no room” for fissures or cracks, so that cracks generally do not form in
compressively strained films.
The strain energy stored in a homo-epitaxial film, that is lattice mismatched to the substrate, is
proportional to the thickness of the film. As the thickness of a strained film increases, it will at some point
become energetically more favorable to reduce the strain energy by creating misfit dislocations and
cracks. Thus, at a certain thickness, the film will form misfit dislocations to release the strain energy. The
critical thickness, at which a homo-epitaxial film starts forming misfit dislocations, is given by the
Matthews−Blakeslee law (Matthews and Blakeslee, 1976). As the film thickness increases further, misfit
dislocations do not suffice to release the strain energy, so that at some point the film will start to crack.
A formula for the critical thickness at which a film under biaxial tensile strain starts to crack was given by
Hearne et al. (2000).

18
 Chapter 14

Exercise: Transmission through a Fabry–Perot cavity. Derive Eq. (14.1) by calculating the transmitted
wave intensity in terms of a geometric series as illustrated in Fig. 14.1 (a).
Solution: As illustrated in Fig. 14.1 (a), the amplitude of the electric field of the transmitted wave,
ET, is obtained by the following sum
ET = E0 t1 t2 + E0 t1 t2 r1 r2 ei 2φ + E0 t1 t2 r12 r22 ei 4φ + E0 t1 t2 r13 r23 ei6φ + K
(
= E0 t1 t2 1 + r1 r2 ei 2φ + r12 r22 ei 4φ + r13 r23 ei6φ + K )
where t and r are the electric field Fresnel transmittance and reflectance coefficients, respectively, and
φ = 2π ( n d / λ) is the phase change incurred by the wave when traveling the distance between the two
reflectors. Using the formula for the geometric series, i.e. 1 + x + x2 + x3 + … = 1 / (1 – x), we obtain
ET t1 t 2 t1 t 2
t = = =
i 2φ 1 − r r cos ( 2φ) − i r1 r2 sin (2φ)
E0 1 − r1 r2 e 1 2
Making the transition from electric-field amplitude to electric-field intensity yields
2
2 A A2 t12 t 22
T = t = = =
B − iC B2 + C 2 1 − 2 r1 r2 cos (2φ) + r12 r22 (cos 2 2φ + sin 2 2φ)
T1 T2
=
1 + R1 R2 − 2 R1 R2 cos (2φ)
what was to be shown.

19
Exercise: Optical mode density. Derive Eq. (14.8), i.e. optical mode density in a 1D space. Also derive
the 3D and 2D optical mode density.
Solution: To derive the 3D optical mode density, we consider a cubic volume with length L, and
volume V = L3, as shown in Fig. 14.4. The possible k vectors of an optical wave inside the volume are
given by
2π 2π π for m = 1, 2, 3 …
k = m = m = m
λ0 2L L
where m is the mode index and λ0 is the wavelength of the fundamental optical mode. Note that the
modes are equidistant in k space. Thus the “volume element” of one mode in k space is given by
dk x dk y dk z = ( π / L) 3 .
Furthermore, the spherical volume in k-space defined by the wave vector k is given by V = (4/3) π k3.
Since k is restricted to positive values, the volume is reduced to only the positive quadrant and we use
only one eighth of the space, i.e.
1 4
V = πk3 .
8 3
Thus the number of modes is obtained by dividing the volume in k space used by one optical mode
through the volume element in k space, i.e.
1 4 πk3
1 1
Nk = 8 3 ×2 = k3 V .
3 3 π2
(π / L)
where the factor of 2 (“× 2”) is due to the two possible polarizations of the optical mode. Using the free-
space dispersion relationship k = ω (ε μ)1/2 = 2πν n /c, one obtains
3
1 1 ⎛ 2πνn ⎞
Nν = ⎜ ⎟ V .
3 π2 ⎝ c ⎠
Thus the density of optical modes per unit volume per unit frequency is given by
1 dN ν n3
ρ( ν ) = = 8π ν 2 3D optical mode density
V dν c3
For two degrees of freedom (2D case), we derive the 2D optical mode density as follows: The area in
k space of one optical mode is given by
dk x dk y = ( π / L) 2 .
The area in k-space defined by the wave vector k is given by (1/4) π k2. Thus
2 2
(1 / 4) π k 2 1 2 and
1 ⎛ 2πνn ⎞ ⎛ νn ⎞
Nk = ×2 = k A Nν = ⎜ ⎟ A = 2π ⎜ ⎟ A
(π / L) 2 2π 2π ⎝ c ⎠ ⎝ c ⎠
where the area A = L2. We thus obtain the density of optical modes per unit area per unit frequency
1 dN ν n2
ρ( ν ) = = 4π ν 2D optical mode density
A dν c2
For one degree of freedom (1D case), we derive the 1D optical mode density as follows: The length in
k space of one optical mode is given by
dk x = π / L .
The length in k-space defined by the wave vector k is given by (½) k. Thus

20
 (1 / 2) k 1 1 2πνn 2ν n
Nk = ×2 = kL and Nν = L = L .
π/ L π π c c
We thus obtain the density of optical modes per unit length per unit frequency
1 dN ν 2n
ρ( ν ) = = 1D optical mode density
L dν c
what was to be shown.

21
 Chapter 16

Exercise: Photometric units. A 60 W incandescent light bulb has a luminous flux of 1000 lm. Assume
that light is emitted isotropically from the bulb.
(a) What is the luminous efficiency (i.e. the number of lumens emitted per watt of electrical input power)
of the light bulb?
(b) What number of standardized candles emit the same luminous intensity?
(c) What is the illuminance, Elum, in units of lux, on a desk located 1.5 m below the bulb?
(d) Is the illuminance level obtained under (c) sufficiently high for reading?
(e) What is the luminous intensity, Ilum, in units of candela, of the light bulb?
(f) Derive the relationship between the illuminance at a distance r from the light bulb, measured in lux,
and the luminous intensity, measured in candela.
(g) Derive the relationship between the illuminance at a distance r from the light bulb, measured in lux,
and the luminous flux, measured in lumen.
(h) The definition of the cd involves the optical power of (1/683) W. What, do you suppose, is the origin
of this particular power level?
Solution: (a) 16.7 lm/W. (b) 80 candles. (c) Elum = 35.4 lm/m2 = 35.4 lux. (d) Yes.
2 2
(e) 79.6 lm/sr = 79.6 cd. (f) Elum r = Ilum. (g) Elum 4πr = Φlum.
(h) Originally, the unit of luminous intensity had been defined as the intensity emitted by a
real candle. Subsequently the unit was defined as the intensity of a light source with specified
wavelength and optical power. When the power of that light source is (1/683) W, it has the
same intensity as the candle. Thus this particular power level has a historical origin and
results from the effort to maintain continuity.

22
Exercise: Luminous efficacy of radiation and luminous efficiency of LEDs. Consider a red and an
amber LED emitting at 625 and 590 nm, respectively. For simplicity, assume that the emission spectra are
monochromatic (Δλ → 0). What is the luminous efficacy of radiation of the two light sources? Calculate
the luminous efficiency of the LEDs, assuming that the red and amber LEDs have an external quantum
efficiency of 50%. Assume that the LED voltage is given by V = Eg / e = hν / e.
Assume next that the LED spectra are thermally broadened and have a gaussian lineshape with a
linewidth of 1.8 kT. Again calculate the luminous efficacy of radiation and luminous efficiency of the two
light sources. How accurate are the results obtained with the approximation of monochromaticity?
Solution: The LED emitting at 625 nm has a luminous efficacy of radiation of 219.2 lm/W.
The LED emitting at 590 nm has a luminous efficacy of radiation of 517.0 lm/W.
The LED emitting at 625 nm has a power efficiency of 50% and a luminous efficiency of 109.6 lm/W.
The LED emitting at 590 nm has a power efficiency of 50% and a luminous efficiency of 258.5 lm/W.
We use a gaussian lineshape with a full-width at half-maximum (FWHM) of 1.8 kT = 1.8 × 25 meV =
45 meV. The FWHM and standard deviation of a gaussian function, σ, are related by: FWHM =
2 (2 ln2)1/2 σ = 2.355 σ. Thus the gaussian function’s standard deviation is σ = 19.11 meV or 6.02 nm (for
the 625 nm LED) and 5.37 nm (for the 590 nm LED). A numerical calculation reveals the following:
625 nm LED: Luminous efficacy of radiation = 221.4 lm/W Luminous efficiency = 110.7 lm/W
590 nm LED: Luminous efficacy of radiation = 515.7 lm/W Luminous efficiency = 257.8 lm/W
Comparison of the results obtained with the approximation of monochromaticity with the results obtained
by the accurate numerical calculation reveals a difference of only a few percent.

23
 Chapter 19

Exercise: Color rendering. The color of a physical object, as seen by a human being, is not just a
function of the object but also a function of the light source illuminating the object! In fact, the color of
an object can depend very strongly on the light source illuminating the object. Some light sources do
render the natural colors of an object (true color rendering) while some light sources do not (false color
rendering).
(a) What is the color of a yellow banana when illuminated with a red LED?
(b) What is the color of a green banana when illuminated with a yellow LED?
(c) Could it be advantageous for a grocer to illuminate meat with red LEDs, bananas with yellow LEDs,
and oranges with orange LEDs?
(d) Is it possible for two physical objects of different colors to appear to have the same color under
certain illumination conditions?
(e) Why are low-pressure Na vapor lights used despite their low color-rendering index?
(f) What would be the advantage and disadvantage of using green LEDs for illumination?
Solution:
(a) Red. (b) Yellow. (c) Yes – but his truthfulness in displaying fruit could be questioned. (d) Yes.
(e) Because of their high luminous efficiency (and thus low electricity consumption). (f) High
luminous efficacy would be an advantage but low color-rendering properties would be a
disadvantage.

24
 Chapter 22

Exercise: Modal dispersion in waveguides. Calculate the time delay between the slowest and the fastest
modes, and the maximum possible bit rate for a 1 km long multimode fiber waveguide with core
refractive index n 1 = 1.45 and cladding refractive index n 2 = 1.4.
Solution: Using Snell’s law (Eq. 22.2), one obtains θc ≈ 15º. The time delay calculated from
Eq. (22.3) for a 1 km long fiber amounts to Δτ = 170 ns. The minimum time required to transmit one bit
of information is given by Δτ. This yields an approximate maximum bit rate of
fmax = 1/170 ns = 5.8 Mbit/s. The calculation shows that modal dispersion can be a significant limitation in
optical communication. Graded-index multimode fibers or single-mode fibers are therefore required for
high-speed communication systems.

25
Exercise: Material dispersion in waveguides. Derive Eqs. (22.6) and (22.7). Why does material
dispersion have a much smaller significance for semiconductor lasers than for LEDs?
Solution:
Derivation of Eq. (22.6): The group refractive index is defined as
c dk d ωn .
ngr = = c = c
v gr dω dω c
In this equation we have used
2π 2π 2π ω .
k = = n k0 = n = n = n
λ λ0 c/ν c
Performing the derivative yields
⎛n ω dn ⎞ dn
ngr = c⎜ + ⎟ = n +ω .
⎝c c dω ⎠ dω
Using ω = 2πν = 2π c / λ0, and using d (1/λ0) = –λ02 dλ0, one obtains
c dn dn ,
ngr = n + 2π = n − λ0
λ 0 2π c d 1 dλ 0
λ0
what was to be shown.
Derivation of Eq. (22.7): It is vgr = c / n gr. Forming the derivative with respect to λ yields dvgr / dλ =
c (d / dλ) n gr–1 = c (–1 / n gr2) (d n gr / dλ). Thus Δvgr = c (–1 / n gr2) (d n gr / dλ) Δλ. Taking the absolute
value of both sides of the equation yields the equation that was to be shown.
Why is material dispersion much less relevant for lasers than for LEDs? According to Eq. (22.8),
material dispersion is proportional to the spectral linewidth of the source, Δλ0. Because lasers
(longitudinal multi-mode as well as single-mode lasers) have a much narrower linewidth than LEDs,
material dispersion is generally much less relevant for laser-based communication systems than it is for
LED-based communication systems.

26
Exercise: Comparison of material and modal dispersion. Consider a 62.5 µm core diameter multimode
step-index fiber of 3 km length with a core index of n1 = 1.45 and a cladding index of n2 = 1.4. Assume
that the fiber inputs come from either an LED or a laser emitting at 850 nm. Assume that the LED and the
laser have a linewidth of 50 and 5 nm, respectively. Calculate the modal and the material dispersion for
each case and explain the result.
Solution: Modal dispersion: The calculation of the modal dispersion depends only on the fiber and is
independent of the source. Using Snell’s law, one obtains θc ≈ 15º. The time delay due to modal
dispersion calculated from Eq. (22.3) for a 3 km long optical fiber amounts to Δτmodal dispersion = 510 ns.
Material dispersion: Figure 22.5 shows that at 850 nm, the material dispersion in silica is Δτmaterial dispersion /
(Δλ0 L) ≈ – 65 ps / (nm km). For a 3 km long optical fiber that is fed by a source with a 50 nm line width
(LED) and 5 nm line width (laser), the dispersion amounts to 9.75 ns and 0.975 ns, respectively. This
exercise shows that modal dispersion typically dominates over material dispersion when multimode fibers
are used.

27
Exercise: Coupling efficiency of a fiber butt-coupled to an LED. Consider an LED with a point-like
emission region that emits an optical power of 1 mW into the hemisphere. For simplicity, assume that the
intensity emitted by the LED is independent of the emission angle. What is the maximum acceptance
angle of a single-mode fiber with NA = 0.1 and multimode fiber with NA = 0.25? What is the power that
can be coupled into the two fibers?
Solution: The maximum acceptance angles of the single-mode and multimode fibers in air are
θair = 5.7° and 14.5°, respectively. The solid angle defined by an acceptance angle θair is given by
Ω = 0.031 and 0.20 for the single-mode and multimode fiber, respectively. Since the entire hemisphere
has a solid angle of 2π, the power coupled into the single-mode and multimode fibers is given by
0.0049 mW and 0.032 mW, respectively.

28
Exercise: Coupling efficiency of a fiber coupled to an LED with a lens. Consider an LED circular
emission region with diameter 20 μm coupled to a silica multimode fiber with NA = 0.2 and a core
diameter of 62.5 μm. The LED emits a power of 1 mW into the hemisphere lying above the planar LED
surface. For simplicity, assume that the LED emission intensity is independent of the emission angle.
What is the maximum power that can be coupled into the multimode fiber?
Solution: Improved coupling can be obtained by imaging the LED emission region on to the core of
the optical fiber. For maximum coupled power, a convex lens with magnification M = 62.5 μm / 20 μm =
3.125 can be used. Using the lens, the acceptance angle of the fiber is increased from θair = 11.5° to θLED =
35.9°. The solid angle defined by the LED acceptance angle θLED is given by Ω = 1.19. Since the LED
emits 1 mW into the entire hemisphere (with solid angle Ω = 2π), the power coupled into the fiber is
given by 0.189 mW.

29
 Chapter 24

Exercise: Derivation of equations. Derive Eqs. (24.3), (24.4), (24.5) and (24.8).
Solution:
Derivation of Eq. (24.3): Consider an exponential decay with time constant τ, i.e. exp (– t/τ). Assume
that the time at which the amplitude has decreased to 90% and 10% of its maximum value are t90% and
t10%, respectively. Accordingly
e −t 90% / τ = 90% and e −t10% / τ = 10% .
Then
t90% t
− 10% = − ln 0.1 + ln 0.9 = ln 9 or t90% − t10% = τ ln 9
τ τ
what was to be shown. An analogous consideration applies to the rise time.
Derivation of Eq. (24.4): The voltage transfer function of an RC voltage-divider circuit is given by
V I (i ωC ) −1 1 1
H (ω) = out = = =
Vin I R + I (i ωC ) −1 1 + i ωRC 1 + i ωτ
where τ = RC; this is what was to be shown.
Derivation of Eq. (24.5): Using | H (ω3dB) |2 = | (1 + i ωτ)–1 |2 = ½, and
2 2
2 A A A2 1
H (ω) = = = =
A + iB A + iB
2 2
A +B 2
1 + ω2 τ 2
we obtain ω3dB = τ–1 = (RC) –1 or f3dB = (2π τ)–1, what is what was to be shown.
Derivation of Eq. (24.8): Using | H (ω3dB) |2 = | (1 + i ω3dB τ)–1 | = ½, and
2 A A 1
H (ω) = = =
A + iB 2
A +B 2
1 + ω2 τ 2
we obtain
1 1 .
=
2 1 + ω32dB τ 2
Solving for ω3dB yields ω3dB = 31/2 / τ or f3dB = 31/2 / (2πτ), what is what was to be shown.

30
Exercise: Rise and fall time and 3 dB frequency. Consider an LED with a rise time of 1.75 ns. Assume
that the fall time of the LED is identical to the rise time. What is the 3 dB frequency of the device? Give
the physical reasons as to why Eq. (24.8) gives only an approximate value of the 3 dB frequency.
Solution: A 3 dB frequency of 343 MHz is expected on the basis of Eq. (24.8). In practice the 3 dB
frequency can be lower or higher than the calculated value since the rise and fall are frequently not
exponential. As a practical rule, the numerical factor 1.2 in the numerator of Eq. (24.8) can vary between
1.0 and 1.5.

31
Exercise: Calculation of carrier sweep-out time. Calculate the carrier sweep-out time for typical values
of the electric field in the p-n junction depletion region, typical carrier velocity, and an active region
thickness of 0.1–1 μm.
Solution: The carrier sweep-out time can be very short. For typical diode parameters, the carrier
sweep-out time is about 1–100 ps, i.e. much shorter than the spontaneous recombination time. As an
example, let us assume that a carrier drifts with the drift-saturation velocity, which is about 107 cm/s,
across a reverse-biased active region. The time needed to drift across a 1.0 μm thick active region is given
by 1.0 μm / 107 cm/s = 10 ps.

32
Exercise: Photometric units. A 60 W incandescent light bulb has a luminous flux of 1000 lm. Assume
that light is emitted isotropically from the bulb.
(a) What is the luminous efficiency (i.e. the number of lumens emitted per watt of electrical input power)
of the light bulb?
(b) What number of standardized candles emit the same luminous intensity?
(c) What is the illuminance, Elum, in units of lux, on a desk located 1.5 m below the bulb?
(d) Is the illuminance level obtained under (c) sufficiently high for reading?
(e) What is the luminous intensity, Ilum, in units of candela, of the light bulb?
(f) Derive the relationship between the illuminance at a distance r from the light bulb, measured in lux,
and the luminous intensity, measured in candela.
(g) Derive the relationship between the illuminance at a distance r from the light bulb, measured in lux,
and the luminous flux, measured in lumen.
(h) The definition of the cd involves the optical power of (1/683) W. What, do you suppose, is the origin
of this particular power level?
Solution: (a) 16.7 lm/W. (b) 80 candles. (c) Elum = 35.4 lm/m2 = 35.4 lux. (d) Yes.
(e) 79.6 lm/sr = 79.6 cd. (f) Elum r2 = Ilum. (g) Elum 4πr2 = Φlum.
(h) Originally, the unit of luminous intensity had been defined as the intensity emitted by a
real candle. Subsequently the unit was defined as the intensity of a light source with specified
wavelength and optical power. When the power of that light source is (1/683) W, it has the
same intensity as the candle. Thus this particular power level has a historical origin and
results from the effort to maintain continuity.

33