Optical Source: Light Emitting Diode (LED)

Author: Dr. Dharmendra Dixit ∗

Outline
• Critical parameters of optical sources

• Light-matter interactions in semiconductors

• Recombination of e-h pairs

• Efficiency of carrier recombination

– Radiative recombination
– Non-radiative recombination
– Internal quantum efficiency

• Structures of LED

– Homo junction structure

– Hetero junction structure

• Spectral width and its model

• Optical power generated by LED

• External quantum efficiency

• Modulation bandwidth of LED

∗
MNNIT Allahabad, Prayagraj, Uttar Pradesh

1
1 Critical parameters of optical sources
Here, we talk about optical sources for the design of optical transmitters. The optical source should
have the capability of encoding the information (’0’ and ’1’) as multiple amplitude or phase levels, or a
combination of both. There are several critical parameters, which can be found need to be taken into
account for choosing a relevant source.

Optical transmitter

Electric signal Electric Optical Light wave

Transmit Source

Figure 1: A optical transmitter.

• Electrical Transmit: It derives an optical source to give modulation of the light wave carrier.

• Optical Source: It provides the electrical-optical conversion.

• Possible optical sources: LED/LASER diodes

Critical parameters:

1. Emission wavelength of the source

• The attenuation of the optical fiber channel is dependent on the wavelength λ.

2. Spectral width:

• The spectral width indicates the range of wavelengths over which the source emits light. A
narrow spectral width is desirable in coherent communication systems as it reduces dispersion
and enhances spectral efficiency, while a broader width may lead to increased pulse broadening
and reduced transmission performance.

3. Modulation capability and bandwidth:

• The source should support both amplitude and phase modulation. Equally important is the
modulation bandwidth, which defines how rapidly the source can be modulated. In the case of
direct modulation, this bandwidth directly determines the maximum achievable data rate.

4. Divergence/ability to couple into a fiber

• The divergence of the optical source determines how well the emitted light can be focused
into the small core of an optical fiber. A lower divergence angle improves coupling efficiency,
minimizing power loss at the input, whereas a larger divergence may lead to poor coupling and
reduced overall system performance.

2
 5. Power/energy consumption/efficiency:

• Efficiency is a key factor when selecting an optical source, as it influences both system perfor-
mance and operating cost. For example, LEDs typically consume less power but have lower
efficiency compared to laser diodes, which deliver higher output power and better conversion
efficiency, making them more suitable for long-distance, high-speed communication.

6. Compactness, reliability, cost-effectiveness

• Optical sources used in communication systems are expected to be physically small and ro-
bust so that they can be easily integrated into fiber-optic modules. LEDs are generally more
cost-effective and reliable for short-distance, low-cost systems, while laser diodes, though more
expensive, offer compact designs with high reliability for advanced, long-haul applications.

Possible sources: LEDs/LASER diodes

2 Light-Matter Interaction in Semiconductors

Semiconductors have:

• Conduction band: Contains electrons.

• Valence band: Contains holes.

• Bandgap energy Eg : Energy difference between the conduction band minimum and valence band
maximum.

Two types of bandgap exist:

1. Direct bandgap: Conduction band minimum and valence band maximum occur at same momen-
tum.

CB Energy band-gap
EC Eg = EC − EV

EV
VB

Figure 2: E-k diagram of direct band-gap semiconductor.

3
 2. Indirect bandgap: Momentum mismatch requires phonon involvement; inefficient for light emis-
sion.

CB momentum mismatch
EC k1 ̸= k2

EV
VB

k1 k2 k

Figure 3: E–k diagram of an indirect band-gap semiconductor.

2.1 Fundamental Processes

Three primary processes occur:
1. Absorption: Photon energy hν promotes electron from valence to conduction band.
2. Spontaneous Emission: Electron-hole recombination occurs randomly, emitting incoherent pho-
tons.
3. Stimulated Emission: An incident photon triggers coherent recombination, amplifying light.
Carrier injection via forward-biased p-n junctions creates excess electrons and holes, enabling emission.

2.2 Population Inversion

Emission requires population inversion, where conduction band electron density exceeds thermal equi-
librium. Achieved through sufficient drive current.

2.3 LED vs. LASER Operation

• LEDs: Operate via spontaneous emission, producing incoherent light.
• Laser diodes: Operate via stimulated emission, producing coherent light.

2.4 Carrier Recombination Dynamics

Let n(t) be carrier density. Carrier recombination follows:
dn n
=−
dt τ
Solution:
n(t) = n0 e−t/τ
where τ is the carrier lifetime. This lifetime depends on:

4
 • Radiative recombination

– Electrons and holes recombine to emits photons.

– Radiative recombination lifetime: τr
– Radiative recombination rate: Rr = n
τr

• Nonradiative recombination

– Electrons and holes recombine to emits phonons

– Defect states, Auger recombination
– Nonradiative recombination lifetime: τnr
– Nonradiative recombination rate: Rnr = n
τnr

• The non-radiative recombination is temperature, because larger the temperature, larger is the prob-
ability of phonon emission.

• Another the useful parameter is the carrier lifetime (τc ), which is a combination of radiative and
non-radiative recombination lifetimes, defined as follows
1 1 1
= +
τc τr τnr

2.5 Internal Quantum Efficiency

Defined as:
Radiative rate Rr n/τr τnr
ηint = = = =
Total recombination rate Rr + Rnr n/τr + n/τnr τr + τnr
Non-radiative processes reduce ηint and are often temperature dependent.

2.6 Material Considerations

• Direct bandgap materials (e.g., GaAs) are preferred for emission devices.

• Indirect bandgap materials (e.g., Si) are unsuitable for efficient light emission due to momentum
mismatch.

5
3 Structures of LED
• Forward biased p-n junction

• Active layer sandwiched between p-type and n-type

Current

ets
Metal contact

fac
d
ve
ea
Cl
p-type

n-type

Figure 4: LED

Structures of
LED

Homo Junction Hetero Junction

Structure Structure

Figure 5: Structures of LED.

3.1 Homo Junction Structure:

Materials/substrate with same band-gap for either sides (p-type/n-type ) of junction.

• Recombination typically happens in the depletion region

• Typical width of depletion region : 1µm-10µm

• The recombinations in this structure can actually happen in a widely distributed fashion over this
entire depletion region.

6
 • No possibility of confinement of the carriers
• No possibility of confinement of the light
• Inefficient way of generating light, it results in a very low internal quantum efficiency

3.2 Hetero Junction Structure:

A heterojunction typically employs a double heterostructure, which consists of p-type and n-type regions
separated by a thin active layer. Unlike a homo-junction, this active layer has a smaller bandgap (Eg )
compared to the surrounding materials (Eg1 ). The thickness of the active region is about 0.1 µm, which is
much thinner than the depletion region.
Eg < Eg1
This design provides two major advantages:
• Carrier Confinement
– To achieve high radiative recombination
• Optical Confinement
– Allows light to be confined, control of emission area
Since materials with larger bandgaps generally exhibit smaller refractive indices, the lower-bandgap active
region has a higher refractive index, thereby supporting strong optical confinement.

Metal n-type n-type n-type p-type p-type Metal

Contact GaAs Ga1−x Alx As Ga1−y Aly As Ga1−x Alx As GaAs Contact
Eg1 Eg Eg1

Injected electrons

electron-hole recombination
Energy

Injected holes

Refr. Index

Figure 6: Double hetero structure LED: layer stack and energy-band diagram with carrier injection and
recombination.

• GaAlAs: λ = 0.81 − 0.87 µm

• InGaAsP: λ = 1 − 1.65 µm
• Si is not used
• Band gap tuned with composition
– Eg:
x
In1−x Gax Asy P1−y , 0 ≤ y ≤ 1, = 0.45 for lattice matching
y

Eg (y) = 1.35 − 0.72y + 0.12y 2

7
4 Spectral Width and its Model
Now, we will find out the largest or the shortest wavelength/frequency that can be emitted from the
structure.
• Longest wavelength: decided by bandgap Eg

Eg = EC − EV = h νmin

λmax (µm) = 1.24

Eg (eV )

• Shortest wavelength: ?
– The shortest wavelength will get decided by the highest energy level difference between which,
the transitions can occur. This highest occupied energy level separation is decided by the
1) joint density of state of these energy states
2) joint occupation probability of these energy states

4.1 Model for Spectral Width

• Spectral width is decided by rate of spontaneous emission
– Optical joint density of states
– Joint occupation probability
• Without derivation, we can say rate R of spontaneous emission:
 
h ν−Eg
q −
R = R0 h ν − Eg e kT

where R0 is a constant, k is Boltzmann constant and T is temperature (in Kelvin).

• Plot a curve for R vs. ν

R
Eg
ν0 = h

ν0 νmax ν

Figure 7: R versus ν plot

• Peak at h νmax = Eg + kT
2

• Full width at half maximum (FWHM): ≈ 1.8 k T

h
Hz
• FWHM at room temperature: 11 THz

1.8×1.38×10−23 ×300
FWHM= 1.8 k T
h
= 6.63×10−34
= 11.23982 × 1012 Hz≈ 11 THz

8
5 Optical Power Generated by LED

I LED Poptical
Input ηint Output

Figure 8: LED as a system

Notations

I = Injected current
q = Charge of an electron
Popt = Optical power generated by LED (in Watts)
h = Planck’s constant = 6.63 × 10−34 J.s
ν = frequency of emitted photon
Rr = Rate of radiative recombination
Rnr = Rate of non-radiative recombination
ηint = Internal quantum efficiency

Rate of injection of carriers = Number of carriers injected in active region

I
=
q
Each injected carrier will either radiatively recombine or non-radiatively recombine, hence, we can also
write
I
Rr + Rnr =
q
We have
Rr
ηint =
Rr + Rnr
Now, the optical power generated by LED can be calculated as

Poptical = Energy of one photon × Rate of gneration of photons

= hν × Rate of radiative recombinations
= hν × Rr
= hν × ηint × (Rr + Rnr )
I
= hν × ηint ×
q
Finally, we can write
!
h ν ηint
Popt = I
q

9
5.1 L-I or P -I Curve
• L-I or P -I curve characterizes optical sources (LED and Laser diodes).

• A typical L-I or P -I curves is shown in Figures 9 and 10.

• Popt denotes optical/light power.

• It is possible to differentiate or identify between LED and LASER based on their P -I curves

Popt

slope= h ν qηint

0
I

Figure 9: P -I curve of LED

Popt

0
I

Figure 10: P -I curve of LASER

6 External Quantum Efficiency

External loss due to

• Fresnel reflection

• Total internal reflection (TIR)

• Absorption

10
6.1 Loss due to Fresnel Reflection
• Fresnel reflection occurs due to change of medium.
• Consider the two medium shown in Figure 11. The reflection coefficient r is defined as
n1 − n2
r=
n1 + n2
where n1 and n2 denote the refractive indices of medium-1 and medium-2, respectively, Er is reflected
electric field amplitude, and Ein is incident electric field amplitude. This formula is valid for only
normal incidence. If the wave is incident at angle, this formula will get modified and it will modified
for two polarization: s-polarization and p-polarization.

Medium-1 Medium-2
n1 n2

Figure 11: Two different meduim

• Consider the LED shown in Figure 12. Assume that light is coming out from n-side. Also both
regions are transparent.
• For n1 = n and n2 = 1 (air), we can write
n−1
r=
n+1

p-region n-region
n1 = n
air n2 = 1

Figure 12: LED surface emitting structure

• Reflectance is the ratio of powers. It can be given by

n−1
2
Pr E2

R= = 2r = r2 =
Pin Ein n+1
where Pr is reflected power and Pin is incidence/input power.
• The power Ptr available at the output is given as
n−1
 2
Ptr = Pin − Pr = Pin − Pin
n+1

11
 • The efficiency η1 due to Fresnel reflection can be calculated as
n−1
2
Ptr

η1 = =1−
Pin n+1
• After simplification, we can write
4n
η1 =
(n + 1)2

6.2 Loss due to Total Internal Reflection (TIR)

p-region n-region
n1 = n
θi
θi
air n2 = 1

Figure 13: TIR in LED

   
• θi > θc = sin−1 n2
n1
= sin−1 1
n
, If TIR occurs,-there is no output
• θi < θc
 
• For n = 3.5, angle of incidence θc = sin−1 1
3.5
≃ 16.60 . It implies that normal incidence approxi-
mation is ok.
• Consider, a point source is emitting in all directions, into a solid angle, 4π. If it is emitting only in
a specific direction of a cone.
• Available output: Fraction of emission within a cone of angle θc

θi

• To find the solid angle of a cone, one has to integrate it 0 to θc in a spherical coordinate system.
Z θc
Solid angle of the cone = 2π sin θ dθ = 2π (1 − cos θc )
0

• The efficiency η2 due to TIR can be calculated as

Solid angle of cone 2π (1 − cos θc ) 1
η2 = = = (1 − cos θc )
4π 4π 2
 s 
1 1 1 1
 q 
η2 = (1 − cos θc ) = 1− 1 − sin2 θc = 1− 1−
2 2 2 n2
Taylor expansion with one term
1 1 1
  
η2 ≃ 1− 1− 2 =
2 2n 4n2

12
 • For n = 3.5
1 1
η2 = = = 0.0204 ≃ 2%
4n2 4 × (3.5)2
We are getting only 2% of that light out.

6.3 Loss due to Absorption

There is one more source of loss due to absorption by the material. As we said before that a double
hetero structures has a larger band gap energy than the photon energy hν. The condition of absorption
is that the energy of incident photon of hν must be greater than the band gap of energy Eg . In a double
heterostructure, when the emitted photon crosses the junction, it experiences a band gap larger than its
energy, and hence the loss due to absorption is minimal. However, there could be defect states due to the
impurities while processing the material.
• Let us assume that the absorption coefficient per unit length is α and the length is L

p-region n-region
n1 = n

Figure 14: Absorption in LED

• The fraction of light that is absorbed is Pabs = e−αL Pin .

• The fraction of light that is available at the output is

Ptr = Pin − Pab = Pin − e−αL Pin

• The efficiency η3 due to absorption is given by

Ptr
η3 = = 1 − e−αL
Pin

• This absorption loss will be small when compared to the other two contributions:- Fresnel reflection
loss and the total internal reflection loss.

• All (Fresnel reflection, TIR and absorption) are independent process. Every photon can experience
either of that or all of that.

• Thus, the total external quantum efficiency has now three contributions-η1 , η2 , η3 and they are inde-
pendent processes. Hence, the external quantum efficiency ηext would be the product of these three
efficiencies.

ηext = η1 η2 η3

13
 • Now, the optical power available at output of LED can be given by
!
h ν ηint ηext
Pout = I
q

7 Some LED Parameters

7.1 Wall Plug Efficiency
Wall plug efficiency is a measure of how effectively an LED converts electrical power into visible light. It
is essentially a metric of how efficiently the LED utilizes the electricity it receives. When electricity passes
through an LED, it excites electrons within the semiconductor material of the LED, causing them to emit
photons of light. However, not all of the electrical energy is converted into light; some of it is lost as heat
due to various factors such as resistance in the circuit and non-radiative recombination processes in the
semiconductor material. It is defined as
Optical power availabe at output of LED
Wall Plug Efficiency =
Input Electrical power
Pout
=
I ×V
where I is foward biased current and V is voltage drop across the p-n junction.

7.2 Responsivity
The responsivity R of LED system is defined as
Optical power availabe at output of LED
R=
Input electrical current
Pout
=
I
hν
= ηint ηext
q

The unit of responsivity is Watts/Ampere.

8 Modulation Bandwidth of LED

The modulation bandwidth of an LED refers to the range of frequencies over which the LED can accurately
modulate its optical output. In other words, it determines the speed at which the LED can switch its
emitted light on and off or vary its intensity. How fast LED can turn-on and turn-off is directly related to
how fast we can change carrier density. In other words, the speed at which an LED can be turned on and
off is directly correlated with the rate at which we can change the carrier density.

8.1 Derivation of Optical Bandwidth of LED

• Here, we consider the harmonic analysis by taking the sinusoidal wave but not square wave. We will
check the response of the system by changing the current to the system. The current is given to the
system is a varying current whose frequency is ωm . ωm is the modulating frequency and our objective
is to find out how high can this modulation be for the oscillatory current. Our final end goal is to
change the intensity by the same response and check the immediate response of the system

14
• Let I be the input current to the LED. It can be given as

I = Ib + Im ejωm t

where Ib is DC biased current, Im is modulating current and ωm is modulating frequency.

• If the current is sinusoidal, the carrier density will also response in the same way.

• Let n be the carrier density in the active region of the LED. It can be given as

n = nb + nm ejωm t

• Carrier density in the active region in the presence of injection can be given by the following rate
equation:
dn I n
= −
dt q τc

• Steady state case: I = Ib , n = nb and dn

dt
=0

Ib nb
0= −
q τc
After simplification, we can get
Ib τc
nb =
q

• Now, substituting I and n in rate equation, we can get

Ib Im jωm t nb nm jωm t
jωm nm ejωm t = + e − − e
q q τc τc

Using nb = Tb τc
q
into above equation, we can get

Ib Im jωm t Ib nm jωm t
jωm nm ejωm t = + e − − e
q q q τc

!
Im nm jωm t
jωm nm e jωm t
= − e
q τc

!
Im nm
jωm nm = −
q τc

1 Im
 
nm jωm + =
τc q

Im
q
nm =  
jωm + 1
τc

15
• We can write
Im
q
nm (ω) =
jω + 1
τc

• We want to find out the response of the LED system. Basically we want Hopt (jω) of the system. It
is defined as
Response when system is excited by frequency ωm
Hopt (jω) =
Response when system is not excited by frequency (ω = 0)
carrier density at frequency ω
=
carrier density at frequency ω = 0
nm (ω)
=
nm (0)

where nm (0) is also called steady state carrier density.

• Hence, we can get

nm (ω) 1
Hopt (jω) = =
nm (0) 1 + jωτc

• Magnitude response
1
|Hopt (jω)| = q
1 + ω 2 τc2

|Hopt (jω)|

1
√1
2

0
ω 3dB ω

Figure 15: Magnitude response plot

• The FWHM frequency of an LED refers to the range of frequencies (or wavelengths) over which the
LED emits light at half of its maximum intensity. In the context of an LED, this term is often used
interchangeably with ”optical bandwidth. It is also called 3dB optical frequency. It is defined as
maximum frequency ω for which |H(jω)| = 21

1
|Hopt (jω)| =
2
1
|Hopt (jω)|2 =
4

16
 1 1
=
1 + ω τc
2 2 4

1 + ω 2 τc2 = 4

ω 2 τc2 = 3

√
3
ω=
τc

√
3
f=
2πτc

√
3
fFWHM =
2πτc

√
3
3dB
fopt =
2πτc

• τc is effective carrier life time. This has contribution form both radiative and non-radiative life times.
τn τnr
τc =
τn + τnr

• Question: What finally decides the modulation optical bandwidth?

• Answer: The modulation optical bandwidth, which represents the range of frequencies over which an
LED can accurately modulate its optical output, is primarily determined by the recombination time
(τc ). When the recombination time increases, the FWHM frequency decreases, consequently affecting
the modulation optical bandwidth. This relationship can be understood as follows: the longer
recombination time results in a slower response of carrier density to changes in input modulation
speed. This delay acts as a bottleneck, limiting the LED’s ability to respond rapidly to modulation
signals and thereby reducing the modulation optical bandwidth. In essence, the modulation optical
bandwidth is constrained by the speed at which carriers recombine within the LED. This concept
is crucial for designing and optimizing LED-based communication systems and other applications
requiring high-speed modulation.

• Question: Why is it called optical bandwidth?

• Answer: It is called optical bandwidth because it represents the range of optical frequencies over
which an LED emits light effectively. In semiconductor physics, carrier density, which refers to
the concentration of charge carriers (electrons and holes) within the LED, plays a crucial role in
determining the optical characteristics of the device. As the carrier density changes, so does the
intensity and spectral distribution of the emitted light. Therefore, by characterizing the response
function for carrier density, we gain insights into how the LED behaves across different optical
frequencies, akin to understanding the bandwidth of an optical signal. In essence, the term ’optical
bandwidth’ encapsulates the range of frequencies within the optical spectrum that contribute to the
LED’s emission, making it a vital parameter for evaluating and designing optical devices and systems.

17
 Ib

L
C Driver electronics

AC
signal

Figure 16: A driver electronic circuit for LED.

8.2 Electrical Bandwidth of LED

Given a certain fopt
3dB
(decided by τc ). What should be the bandwidth of the driver electronics that can
support this fopt ?
3dB

•
Popt (ω) 1
|Hopt (jω)| = =q
Popt (0) 1 + ω 2 τc2

• To find electrical 3dB bandwidth, we have to find electrical transfer function

• Recall how optical power and electrical power depend on electric current.

Pele ∝ I 2

Popt ∝ I

Based on above two relations we can also write

2
Pele ∝ Popt

• Hence, the electrical transfer function is given by

Pele (ω) P 2 (ω) 1

|Hele (jω)| = = opt = |Hopt (jω)|2 =
Pele (0) Popt (0)
2
1 + ω 2 τc2

• The 3dB electrical bandwidth is defined as maximum frequency ω for which |Hele (jω)| = 1
2

1
|Hele (jω)| =
2

1 1
=
1 + ω τc
2 2 2

18
 1 + ω 2 τc2 = 2

ω 2 τc2 = 1

1
ω=
τc

1
f=
2πτc

1
3dB
fele =
2πτc

• Relationship between fopt

3dB
and fele
3dB
is given by
√
3dB
fopt = 3 fele
3dB

References
[1] Prof. Deepa Venkitesh NPTEL video lecture : Fiber Optic Communication Technology .

[2] John M. Senior, Optical Fiber Communications: Principles and Practice, 3rd edition, Pearson, 2013.

[3] Gerd Keiser, Optical Fiber Communications, 4th edition Tata McGraw-Hill.

19