The document summarizes key aspects of microLED device physics. It discusses:
1) Efficiency considerations from both a device and system perspective, including series resistance and voltage drops.
2) Factors influencing external quantum efficiency such as internal quantum efficiency, light extraction efficiency, and their governing parameters.
3) Challenges of achieving low contact resistance on p-type GaN materials and the role of work function and doping.
4) Dependence of efficiency and electrical characteristics on microLED size due to increasing surface-to-volume ratios and sidewall damage from dry etching.
3. Brian Kim
Efficiency of device and system
3
From a system point of view, power efficiency is important.
Requires lower series resistance to reduce IR drop: ITO, contacts for n-side and p-side.
A local cathode design in every pixel is better solution than a common cathode design.
In addition, the Vth variation in the LED must be suppressed. (cause the higher Vdd to operate in the saturation region)
𝜼𝒑 =
𝒍𝒎
𝑾
=
𝒄𝒅 𝝅
𝑨 𝑽
=
(𝒄𝒅 𝒎𝟐
)
⁄ 𝝅
(𝑨 𝒎𝟐
⁄ ) 𝑽
= 𝜼𝒍
𝝅
𝑽
Current efficiency
(device level efficiency)
Power efficiency
(system level efficiency)
n-GaN
p-GaN
hν
Id
Id
hν
~1V
2~2.5V
2~4V
Sidewall damage:
Light leaky &
carrier losses
P-contact:
Resistance &
reflectance
Backplane TRs
𝜼𝒍 =
𝐜𝐝
𝑨
=
𝒄𝒅 𝒎𝟐
⁄
𝑨 𝒎𝟐
⁄
=
𝑳
𝑱 Including IR
drop terms
4. Brian Kim
Power efficiency: system efficiency
4
Wherein,
• WTR= ID x VTR: power consumption in transistor
• WLED= ID x VD: power consumption in LED
• WR = ID x VR: power consumption in contact and line
• W: WR+WLED (power consumption of system)
VDD
VR
VLED
R
Id
VTR
𝒕𝒓 𝑳𝑬𝑫 𝑹
LED
current
(I
ds
)
Vdd
Voltage
Vss
Vth
VDS
(consumed by Tr)
Q-point
VLED
(consumed by LED)
5. Brian Kim
EQE: formula
5
Wherein:
• 𝑒 𝑐ℎ𝑎𝑟𝑔𝑒 𝑜𝑓 𝑎𝑛 𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑛 = 1.602 × 10 𝐴 𝑠 [𝑛𝑜𝑡𝑒 𝑡ℎ𝑎𝑡 1𝐶𝑜𝑙𝑢𝑚𝑏 = 1𝐴 1𝑠𝑒𝑐]
• h ( Plank’s constant)= 6.626 ×10−34 J s [note that 1Joule = 1W.s]
• c (speed of light)= 2.998 ×108 m/s
• ∴ =
. ×
. ×
= 1.23995 × 10
• 𝐸𝑄𝐸 𝑢𝑛𝑖𝑡 = 𝑢𝑛𝑖𝑡𝑙𝑒𝑠𝑠,
𝒕𝒐𝒕𝒂𝒍 𝒓𝒂𝒅𝒊𝒐𝒎𝒆𝒕𝒓𝒊𝒄 𝒑𝒐𝒘𝒆𝒓
𝒂𝒗𝒈.𝒐𝒇 𝒓𝒂𝒅𝒊𝒐𝒎𝒆𝒕𝒓𝒊𝒄 𝒑𝒐𝒘𝒆𝒓
𝒕𝒐𝒕𝒂𝒍 𝒄𝒖𝒓𝒓𝒆𝒏𝒕
𝒄𝒉𝒂𝒓𝒈𝒆 𝒐𝒇 𝒆𝒍𝒆𝒄𝒕𝒓𝒐𝒏
𝒓𝒂𝒅𝒊𝒐𝒎𝒆𝒕𝒓𝒊𝒄 𝒑𝒐𝒘𝒆𝒓
𝒉𝒄
𝒂𝒗𝒈.
𝒊
𝒆
𝒓𝒂𝒅𝒊𝒐𝒎𝒆𝒕𝒓𝒊𝒄 𝒑𝒐𝒘𝒆𝒓 ×𝒂𝒗𝒈.
𝒊
𝒆
𝒉𝒄
𝒓𝒂𝒅𝒊𝒐𝒎𝒆𝒕𝒓𝒊𝒄 𝒑𝒐𝒘𝒆𝒓×𝒂𝒗𝒈.
𝟏𝟐𝟑𝟗.𝟗𝟓 × 𝒊
n-GaN
p-GaN
hν
Id
Id
hν
6. Brian Kim 6
Internal quantum efficiency (IQE)
Defines ‘the number of photons generated relative to
the number of electrons’ inside the LED.
The key is how efficiently electrons and holes
generate photons. The controlling factors are defects
in the epi structure, band structure (e.g., band
bending due to piezoelectric polarization caused by
strain), and sidewall damage etc.
Light extraction efficiency (LEE)
Defines how efficiently photons created inside a LED
can be emitted into the air.
The governing parameters are the geometry of the die
and the reflectivity of the p-side reflector etc.
EQE: IQE x LEE
Internal
Quantum
Efficiency
(IQE)
Light
Extraction
Efficiency
(LEE)
External
Quantum
Efficiency
(EQE)
x =
n-GaN
p-GaN
hν
Id
Id
hν
7. Brian Kim
Light extraction parameters
7
Chip
shaping
Effect of sidewall
treatment
Effect of
encapsulation
Die size effect
The role of sidewall
could be different in
both cases.
Effect of
reflectance
in p-side
Sidewall
angles
Surface
morphology
such as PSS for a
better light
extraction
P-contact
size
Circle vs.
Square etc.
• Refractive indices of
dielectric
• Reflectance of metal
(e.g. RI of metal)
Microlens
Beam
shaping
LED
Die
height
8. Brian Kim 8
The die surrounded by 3-surfaces having distinct
functions:
p-side in the bottom, sidewall, and light exit plane.
The photons emanating from the quantum well are
traveling and hit the walls with the following
interactions:
reflection, absorption, and transmission etc.
The density of photon in the space of GaN decrease
exponentially.
Based on a modeling experiment, the higher
extraction efficiency from the light exit plane and the
higher reflectance on the p-side show rapid decay of
photon density.
LEE: motion of photons
Governing interaction
Surface
Extraction/beam shaping, bouncing back
Top surface
Leakage, bouncing
Sidewall
Absorption, bounding
p-side reflector
Extraction
Leakage
Absorption
Nearly isotropic
source
Light emission
surface
Sidewall
P-contact
/reflector
9. Brian Kim
p-contact: ohmic contact & reflectance
9
As a p-contact metal, the work function must be large to
lower the natural potential between the semiconductor and
the metal.
Work function: p-GaN: (~7.6 eV), n-GaN: about 4.1-4.2 eV,
Band gap of GaN: 3.4eV
Owing to the wide energy band gap and absence of large
work function ɸM, it is not easy to form low-resistivity
Ohmic contacts to p-GaN.
Thus, low resistance tunneling contact by heavy doping and
surface treatment to reducing actual bandgap lower than
Schottky-Mott barrier height are being used.
In terms of electrical model, contact resistance is a
parameter that hinders device efficiency, so it should be
lowered as much as possible.
1V
~2V
3~4V
Sidewall damage:
Leaky & carrier loss
P-contact:
Resistance & reflectance
BackplaneTRs
~IMΩ
(20m)
The role of p-side contact should be considered from an electrical and optical perspective, and it requires a
reasonable Ohmic contact resistance and high reflectance.
10. Brian Kim 10
P-contact: bend bending before and after metal-semiconductor contact
https://link.springer.com/chapter/10.1007/978-3-319-10756-1_8
Metal / n-type semiconductor Metal / p-type semiconductor
m: work function of metal, s: work function of semiconductor
m > s
m < s
Rectifying (p-GaN)
Ohmic
Ohmic
Rectifying
11. Brian Kim 11
The Schottky–Mott barrier height, as defined by
For most applications, ɸSM is very large.
Owing to the wide energy band gap and absence
of large work function ɸM, it is not easy to form
low-resistivity Ohmic contacts to p-GaN.
Reflectance and Work function of various metals
P-contact: Schottky Barrier Height (SBH)
where q is the electronic charge, EG is the energy band gap and
χS is the electron affinity of p-GaN.
Electrically isolated Schottky Contact
∅
χ
∅
𝐸
𝐸
𝐸
Metal p-GaN
Metal
p-GaN
𝐸
𝐸
𝐸
∅
χ
12. Brian Kim
Achieving low Ohmic contact resistance
Use metal that has a large metal work function so that the natural potential barrier between the metal
and semiconductor is minimized.
metals with sufficiently large work functions especially in the case of p-type GaN are lacking.
Dope the semiconductor heavily so as to form a low resistance tunnel contact.
Decrease the bandgap of the semiconductor at the contact. This can be done in two ways:
1) heterostructure band discontinuities can be completely eliminated by grading
2) a superlattice can be placed between the metal contact and semiconductor bulk to allow for tunneling.
(induce polarization field near the surface by P-AlGaN/GaN
12
13. Brian Kim
P-contact: Effect of reflectance
13
The output flux decreases exponentially along with reflectance decrease.
The absorptions are being mainly happening on the reflector surface associated with an extinction
coefficient of metal (conductor), and it will convert to thermal energy (as a phonon).
A surface feature like a PSS extract more photons from the die to the free-space than a non-PSS.
R=90%, reflectance
0.90
n=2 n=3
0.81
0.73
1
n=1
Light extraction
element
Reflector
GaN
(n: number of internal reflection)
height
bottom
top
4.5
1
4
4.5
2
5
4.5
7
10
4.5
12
15
4.5
22
28
[unit: m]
14. Brian Kim 14
ABC model
Explains efficiency of microLED
Shows competition between radiative and non-radiative
recombination.
Internal quantum efficiency (IQE)
𝑰𝑸𝑬 =
𝑩𝒏𝟐
𝑨𝒏 + 𝑩𝒏𝟐 + 𝑪𝒏𝟑
A: Shockly-Read-Hall (SRH) non-radiative
strongly depend on LED size, sidewall defect with dry etch, sidewall effect defined by
the ratio perimeter/surface (P/S) of LED
B: Radiative recombination independent on LED size
C: Auger nonradiative
Independent of LED size, root cause of efficiency droop by current crowding at high
current injection
n: carrier concentration
Contributions of various recombination channels to the total
current density. https://doi.org/10.1117/12.912305
S. Y. Karpov "Simulation of light-emitting diodes for new
physics understanding and device design", ( 2012)
Experimental and fitted IQE as a function of
current density for a 100 x 100 m2 sized LED.
http://dx.doi.org/10.1063/1.4993741
Extracted coefficients A (a) and C (b) plotted
versus LED size. Coefficient A shows a large
dependence on LED size in contrary to
coefficient C that almost shows a constant
value.
(A)
(B)
(C)
(A)
(C)
(C)
(A)
(B)
B: radiative
A, C: non-radiative (loss terms)
15. Brian Kim
IQE: sidewall damage of microLED
15
There are no practical wet etchants for selective anisotropic etching of GaN, and like other wide bandgap
semiconductors, it is difficult to plasma etch (ICP) without creating rough and damaged surfaces.
Due to the relatively high bond energy (8.92 eV/atom) of GaN, the threshold ion energy for the onset of dry
etching is typically on the order of 25 eV.
Plasma etching of p-type GaN creates n-type nitrogen vacancy (VN) defects which act as shallow donor states, at
the etched surface: leading to type conversion to n-type at the etched surface, while n-type GaN becomes n+.
- serve as vertical current conduction pathways in a way similar to that of threading dislocations.
Consequently, the µLED element can be thought of as being surrounded by a network of defects that are n-type
in nature.
The external quantum efficiency (EQE) of microLEDs decreases as lateral dimensions are reduced. This
originates from an increased surface-area-to-volume ratio, which increases nonradiative Shockley–Read–
Hall (SRH) recombination at the edge of the mesa.
Ion bombardment
n-GaN
p-GaN
MQW
n+
n
n-GaN
p-GaN
16. Brian Kim
Leaky on sidewall
The smaller the die, the larger the surface to
volume ratio.
- Therefore, surface defects occur more easily in small
dies during the mesa process than in large dies..
The larger dies are less sensitive owing to small
ratio (surface /volume).
16
w/ and w/o sidewall treatment
Two main improvements
1. changed the etching process during vias formation between P-
contact and hybridization pads into a softer plasma.
2. changed P-contact in order to obtain higher reflectivity and lower
contact resistance.
Size dependent characteristic of microLED
Current density–voltage characteristics of a set of microLED
device sizes. Current density is based on the area of the active
region. https://doi.org/10.1063/5.0011651
Electrical characteristic of a 7*7μm² μLED with previously reported fabrication process and optimized
process. https://doi.org/10.1002/sdtp.11615
premature turn-on
caused by leakage
current or defective
regions
Remove defective
regions; suppression
of leakage current
Before After
18. Brian Kim
Series resistance of microLED
18
Dominant contribution on series resistance
Since the mobile carrier concentration in p-GaN is one
order lower than that of n-GaN, the dominant
contribution to series resistance is p-GaN.
Effect of microLED size on series resistance
Series resistance of microLED: circuit model
p-contact
p-GaN
MQW
n-GaN
n-contact
Rs: series resistance of microLED
ρ: electrical resistivity of p-GaN
d: thickness of p-GaN
A: cross-sectional area of p-GaN
Rc: resistance of n-GaN, p- & n-contacts
Wherein:
p-GaN n-GaN + contact
19. Brian Kim
Series resistance: impact on efficiency
Origin of series resistance
Increased device impedance with die size reduction
Increased in series resistance after hybridization.
Impact on Efficiency
It does not affect current efficiency (cd/A), but has a
significant impact on power efficiency (lm/W).
(Current efficiency=cd/A, Power efficiency=lm/W)
19
Wherein,
• WTR= ID x VTR: power consumption in transistor
• WLED= ID x VD: power consumption in LED
• WR = ID x VR: power consumption in contact and line
• W= WR+WLED (power consumption of system)
VDD
VR
VLED
R
Id
Voltage
Current
On-panel
Commercialized
LED
MESA
VTR
𝒕𝒓 𝑳𝑬𝑫 𝑹
21. Brian Kim
Real diode characteristics
At low current, the measured current is larger than
the ideal current.
When the bias voltage increase to close to VBi, the
diode current, the current increase is slowed down.
Eventually, the current saturates at some value with
further increase in applied voltage.
Ideal diode characteristics
slope in the log plot = (q/kT)ln10
21
Diode characteristic curve
Source: presentation document from Hong Kong University of Science & Technology, Department of Electronic & Computer Engineering
22. Brian Kim
Measured I-V of microLED
22
Voltage
Current
LED model
Ideal LED model & w/ series, parallel
resistance
Real data
Measured microLED characteristic
24. Brian Kim
Experimental modeling of LED
24
n-MOSFET w/ resistor
(Drain side)
n-MOSFET w/ resistor
(Source side)
n-MOSFET w/
resistor & resistor
(Source side)
25. Brian Kim
How to drive for display?
What we learn from the measured I-V curves.
Very steep I-V curve.
Non-homogeneity between dies.
- forward voltage non-uniformity caused by the epi growth and fabrication process,
degradation after long time operation
Shunt: observed current flow but no emission of light.
25
Shockley diode equation
Voltage driven Current driven
Voltage driven with load
Constant
voltage
Constant current
26. Brian Kim
Transition of microLED characteristics
26
Transition of microLED characteristics of microLED over the process.
EQE DWL
I - V
28. Brian Kim
How to implement luminance?: AM and PM driving
28
• 1 Frame=16.67msec @60Hz
• 1 line scan time=55.6us (300 lines in 1 frame=16.67ms/300)
• Equivalent luminance with PM driving=300A cd/m2
Luminance
PM
AM
A cd/m2
300A cd/m2
16.67 msec 55.6us
Time
29. Brian Kim
Gray scale modulation: passive, active, binary coding
29
1st row
2nd row
3rd row
Nth row
Adjustable
along with
luminance
16.67 msec
16.67/N
Adjustable
along with
luminance
line at a time with
different amplitude
Utilize full frame with
different amplitude
Binary coding with sub-fields
having a same amplitude
20 21 22 23 24 25
Passive Matrix LTPS CMOS
30. Brian Kim
CMOS (PWM) and LTPS (PAM)
30
A fixed Q-point vs. variable Q-points
LED
current
(I
ds
)
Vdd
Voltage
Vss Vth
VDS
(consumed by Tr)
Q-point
VLED
(consumed by LED)
LED
current
(I
ds
)
Vdd
Voltage
Vss Vth
Q-point
CMOS (PWM) LTPS (PAM)
31. Brian Kim
LED
current
(I
ds
)
Vdd
Voltage
Vss
Vth
VDS
(consumed by Tr)
Q-point
VLED
(consumed by LED)
LED
current
(I
ds
)
Vdd
Voltage
Vss
Vth
VDS
(consumed by Tr)
Q-point
VLED
(consumed by LED)
Subpixel variations
31
The LED with a high series resistance does not operate in the saturated region but operates in the linear
region.
Higher Vdd is required for all LEDs to operate in saturation areas, which in turn increases power
consumption.
Clean LED w/ variations
Variations
Vdd’
32. Brian Kim
Common cathode vs. Local cathode
Common cathode model Local cathode model
Vdd
Vss
Vdd
Vss Vss Vss Vss
Current density (A/cm2)
IR drop with CC scheme
50nA 100nA 150nA 200nA 250nA
33. Brian Kim 33
IR drop model with the common cathode
Modeling condition
- Display: 480 x 270 pixel (0.7”),
- Current: 1µA per LED,
- ITO: 10 ohm/sq and 20 ohm/sq
Result
- 20 ohm/sq: 0.63V @ center
- 10 ohm/sq: 0.315V @ center
IR drop compensation
Pulling down with negative potential: increases power
consumption.
Low resistance Ag nanowire for better conductivity:
need to be confirmed long term stability.
The ideal solution is a local cathode configuration.
IR drop modeling
12 strings of current sources are
connected by resistors
For simplicity, 1-D modeling was performed. Take consideration of
half of horizontal pixels (240 pixels) and scale down 1/20 so that
one LED model corresponds to 20x20 LEDs.
34. Brian Kim 34
Four combinations are possible as shown in the
figure.
The image quality is affected by how the LED and
driving transistor are connected.
(A) and (D) modes are not acceptable
Combination(A): Vsg is affected by the LED Vf, thus
the voltage applied to the driving transistor (Vsg) is
not uniformly controlled.
(B) and (C) modes are preferable.
Combination (C): Vsg is not affected by the LED, so
that's fine. (Vsg is free from LED Vf.)
Image quality influence by combination of LED and driving transistor
GND
S
G
D
Vdd
GND
S
G
D
Vdd
GND
D
G
S
Vdd
GND
D
G
S
Vdd
Possible connection modes of driving transistor and LED
NMOS driver PMOS driver
Common N Common P Common N Common P
(A) (B) (C) (D)
(D)
(C)
(B)
(A)
PMOS
NMOS
Driver
P side
N side
P side
N side
Common
Id
Id Id Id
[NG] [Good] [Good] [NG]