ECE 475 - Lecture 13

February 25, 2025

1 LED device structure for ”surface emission”

• Recall that in LEDs, e− and h+ are injected into the pn junction

– Some fraction (ηIQE ) will recombine radiatively and produce a photon with energy hν ≃ Eg

• Of the photons produced internally, only a fraction of those will escape as useful light output (ηEE )

• In a conventional surface-emitting LED, the pn junction is formed vertically by epitaxial growth

of the desired semiconductor onto a compatible substrate

Figure 1: Schematic illustration of planar surface emitting LED device

Notes

• To minimize defects that can result in non-radiative recombination, which lowers ηIQE , the epitaxial

layers and the substrate ideally have the same lattice constant, a

– This is called ”lattice matching”

1
 – The lack of available/suitable materials can delay or limit the use of a particular semiconductor

for LED applications, e.g. the development of a blue LED

– The excellent lattice match between ALx Ga1−x As (for all x) and GaAs, which can easily be

produced as a large crystalline substrate, made this system the earliest successful LED system

(red light)

• Typically, light is extracted from the top surface through the p layer, as shown above

– The positive contact is kept as unobtrusive as possible (small, ring-shaped, etc) and/or semi-

transparent, e.g. using a conductor such as indium tin oxide

– Some of the light emitted towards the substrate can be reflected back

– Typically, the junction is doped pn+ , so that most of the recombination (light emission) occurs

on the p-side, near the surface

2 Output flux and external efficiency

• Consider a forward biased pn junction with bias V and current I

– The total number of charge carriers, i.e. electrons, passing through the junction per second

is I/e, where e is the electron charge

– Since ηIQE is the fraction that produces photons, we can write:

Φp = ηIQE (I/e)

where Φp is the internal photon flux [photons/sec] and

τnr
ηIQE =
τnr + τr

is the internal quantum efficiency

• It follows that the output photon flux is determined by the internal photon flux and the extraction

efficiency of the LED

Φout = ηEE Φp = ηEE ηIQE (I/e)

= ηEQE (I/e)

2
 where Φout is the number of photons per unit time emitted externally, summed over all directions

[photon/sec]

• Since each photon has energy hν ≈ Eg , the optical output power is

P0 = hνΦout

• The input electrical power is simply

Pe = IV

• We can define the ”wall-plug” efficiency, or power conversion efficiency (PCE)

P0 ηEQE (I/e)(hν)
ηPCE ≡ =
Pe IV
ηEQE (hν)
ηPCE =
eV

where ν is the frequency of the light [Hz], e is the electron charge [C], and V is the applied bias

voltage [V]

• Some representative numbers for wall-plug efficiency

Semiconductor Substrate λ (nm) ηPCE Comments

GaAs GaAs 870 − 900 10% Infrared LEDs

Alx Ga1−x As GaAs 640 − 870 5-20% Red to IR LEDs

In1−x Gax Asy P1−p InP 1000 − 1600 > 10% LEDs in communications

InGaN alloys GaN or SiC 430 − 460 2% Blue LED

InGaN alloys Sapphire 500 − 530 3% Green LED

SiC Si; SiC 460 − 470 0.02% Blue LED, low efficiency

GaP (Zn-O, doped) GaP 700 2-3% Red LED

GaP (N, doped) GaP 565 < 1% Green LED

• GaAs and InP based LEDs are the oldest and most technologically mature, developed in large part

for optical communications

• GaP and GaN based LEDs are primarily important for lighting applications

3
 Notes

• Responsivity is an alternative parameter that is sometimes used to characterize LEDs:

P0 ηIQE (hν)(I/e))
R≡ =
I I
R = ηPCE V [W/A]

where R is the LED responsivity

– Typically, R ≃ 0.01 W/A

• R is approximately constant up to some device-dependent max current

– At higher currents, the junction temperature increases and ηIQE is reduced

– Reduction in ηIQE is mainly due to a decrease in τnr , due in turn to a presence of more

high-energy phonons at elevated temperatures

– The ”cut-in” or ”turn-on” voltage of a diode is the voltage at which the forward bias current

begins to start increasing rapidly. Cut-in voltage scales with Eg in a diode

• LEDs with higher peak frequency (hν ≃ Eg ) have higher operating bias requirements, in general

3 Increasing LED extraction efficiency

• A basic surface emission/square-sided LED has low extraction effficiency due to total internal reflection

4
 Figure 2: Kasap Fig. 3.40

• However, many strategies can be employed to address this problem

– Shape of the semiconductor chip (Fig. 3.40b)

∗ Well engineered shapes can allow a greater fraction of light to escape, e.g. dome/hemisphere

∗ This typically requires complicated processing, and is therefore expensive

∗ Encapsulating the LED in epoxy is a cheap way of reshaping to improve efficiency

– Textured surfaces (Fig. 3.40c)

∗ Texturing or nanostructuring can roughen the surface to produce random scattering of

light, rather than Fresnel refraction/specular reflection

∗ Light can undergo one to two reflections before escaping

∗ Chemical etching or epitaxial growth is a relatively cheap way of achieving this

∗ Nano-scale lithographic patterning is also possible, but more expensive

– Distributed Bragg reflector (DBR) is a dielectric layer that can be added under active layers

to increase extraction ratio (Fig. 3.40d)

∗ DBRs, also known as dielectric mirrors, are manufactured to optimize the reflection co-

efficient of LED emission

∗ DBRs consist of an alternating sequence of layers of two different optical materials. Fre-

quently the thickness of these layers is one quarter of the wavelength for the design

wavelength

5
 ∗ The optical path length between each surface is one half of a wavelength, and produces

constructive interference of the reflected field and destructive interference of the trans-

mitted field through the Fresnel equations

– Modification of the emission pattern

∗ Structuring of the local environment by the active layer can change the pattern of spon-

taneous emission

∗ For example, by adding a ”resonant cavity” (Fig. 3.40e) formed by two dielectric mirrors

(DBRs) can narrow the emission spectrum by wavelength-specific mode selection (only

light of specific/desired) resonant modes can escape

∗ Modes can be formed to preferentially emit in directions that minimize TIR, however the

emission spectrum is also modified

– Encapsulation in plastic

∗ Encapsulation in transparent plastic (typically epoxy) with refractive index of n ≃ 1.5−1.6

is common in most discrete LEDs

∗ The epoxy is modified into a lens shape to improve extraction efficiency and/or modify

the emission pattern

∗ In white-light LEDs, the plastic is often a silicone loaded with inorganic phosphors

∗ These phosphors convert the monochromatic LED into longer-wavelength colors

6
Figure 3: Source: Fan, Jiajie, et al., (2021). A Gamma process-based degradation testing of silicone
encapsulant used in LED packaging. Polymer Testing. 96. 107090.

4 The spatial pattern of emitted LED light

• Surface-emitting LEDs can emit directional or non-directional beams, depending on their structure

– Directionality is favorable for optical communications, but not for lighting

• A non-encapsulated surface emitting LED has an approximately Lambertian radiation pattern

I(θ) = I0 cos θ

where I0 is the peak irradiance emitted in the surface-normal direction and θ is the angle relative

to normal

I0
– This expression predicts I(θ = 60◦ ) = 2 , i.e. the FWHMθ = 120◦

Figure 4: Typical Lambertian radiation pattern

7
• Epoxy lenses reduce (or increase) the beam divergence

• For edge emitting LEDs, waveguiding (more later) results in a more collimated (less divergent)

beam:

I(θ) = I0 coss θ

where s = 10 is typically, and FWHMθ = 30◦ − 40◦