Received: 1 January 2019 Revised: 12 February 2019 Accepted: 13 February 2019

DOI: 10.1002/jsid.760

INVITED PAPER

Prospects and challenges of mini‐LED and micro‐LED

displays

Yuge Huang SID Student Member1 | Guanjun Tan SID Student Member1 |
Fangwang Gou SID Student Member1 | Ming‐Chun Li2 | Seok‐Lyul Lee SID Fellow2 |
Shin‐Tson Wu SID Fellow1

1
College of Optics and Photonics,
Abstract
University of Central Florida, Orlando,
Florida We review the emerging mini/micro–light‐emitting diode (LED) displays
2
AU Optronics Corp., Hsinchu, Taiwan featuring high dynamic range and good sunlight readability. For mini‐LED
backlit liquid crystal displays (LCDs), we quantitatively evaluate how the
Correspondence
Shin‐Tson Wu, College of Optics and device contrast ratio, local dimming zone number, and local light profile affect
Photonics, University of Central Florida, the image quality. For the emissive mini/micro‐LED displays, the challenges of
Orlando, Florida 32816, USA.
ambient contrast ratio and size‐dependent power efficiency are analyzed. Two
Email: [email protected]
figure‐of‐merits are proposed for optimizing the optical and electrical
Funding information performances of mini/micro‐LED displays.
a.u.Vista, Inc.
KEYWORDS
ambient contrast ratio, halo effect, high‐dynamic range, internal quantum efficiency, local dimming,
mini/micro‐LED, size effect, sunlight readability

1 | INTRODUCTION characteristic but lacks the latter, while OLED is oppo-

site. Both LCD and OLED camps are working hard to
Presently, liquid crystal display (LCD)1 and organic improve their own drawbacks.
light‐emitting diode (OLED) display2 are two dominat- Recently, mini‐LED and micro‐LED displays10–12 are
ing technologies for smartphones, pads, monitors, and attracting extensive attentions for their HDR,13 high
TVs. Each technology has its own pros and cons.3 For ambient CR (ACR),14 thin profile, and low power
example, LCD's major advantages are long lifetime, high consumption.12,15–17 When a mini‐LED array is employed
peak brightness, and low cost, while OLED's distinctive as LCD backlight, the local dimming technology15,16,18–24
features are true black state4 and ultrathin profile, could boost the panel's CR25 to 1 000 000:1. When
which enables flexible displays. They are comparable mini/micro‐LED chips are integrated in self‐emissive dis-
in color gamut,5 resolution, motion picture response plays, ie, without LCD panel, it presents an excellent dark
time,6 and power consumption. However, LCD has two state as well as several times higher peak luminance than
shortcomings to overcome: limited contrast ratio LCDs and OLED displays.12,26 Moreover, their simple
(CR ~ 1000‐5000:1) and flexibility. On the other hand, structure, freeform shape factor, high aperture ratio, wide
OLED's major challenges are compromised lifetime as viewing angle, and wide operation temperature range
luminance increases7,8 and relatively higher cost. In could make these displays ubiquitous for indoor and
order to faithfully reproduce nature scenes, high outdoor applications.26,27
dynamic range (HDR) is critically important.9 To In this paper, we will first introduce mini/micro‐LED
achieve HDR, a display device should have high peak displays, emphasizing on their common challenges and
luminance (Lp > 1000 cd/m2) and excellent dark state potential solutions. In Section 2, we will present mini‐
(less than 0.01 cd/m2). LCD has the former LED backlit LCD beginning from optical system
J Soc Inf Display. 2019;27:387–401. wileyonlinelibrary.com/journal/jsid © 2019 Society for Information Display 387
388 HUANG ET AL.

structure, followed by common issues of local dimming— Next, we use an exemplary candle picture as shown in
halo effect and clipping effect—and finally some pro- Figure 2 to illustrate the light modulation process of
posed solutions. A simplified simulation model is utilized mini‐LED backlit LCDs. Here, the backlight consists of
to evaluate the quantitative contribution of each design 12 × 24 local dimming zones and each zone contains
factor. In Section 3, we will discuss mini/micro‐LED as 6 × 6 mini‐LEDs in order to achieve a desired luminance.
emissive displays. In this category, two approaches can According to the image content, the mini‐LEDs in each
be considered for achieving full colors: (1) color conver- dimming zone are predetermined to show different gray
sion, such as using blue LED to pump green and red levels, as Figure 2A depicts. After passing through the dif-
phosphors or quantum dots,28–31 and (2) RGB LED fuser, the outgoing light spreads out uniformly before
chips.32,33 The former has been reviewed recently,34 while reaching the LCD panel (Figure 2B). The gray level of
for the latter, we will analyze two important issues: each LCD pixel is controlled by a thin‐film‐transistor
internal reflection and chip size dependent internal quan- (TFT), and each CF only transmits the designated color.
tum efficiency (IQE). A quantitative system evaluation Finally, a full‐color image as Figure 2C is generated.
method will be proposed, followed by exemplary optimi-
zation suggesting the best device structure and LED chip
size. Although the high‐yield mass production of small‐ 2.2 | Challenges of local dimming LCDs
chip micro‐LED (less than 50 μm) is still under active
development, the fabrication of mini‐LED with larger Mini‐LED BLU enables a new LCD with high peak lumi-
chip size (100‐500 μm) is relatively mature so that com- nance, HDR, and thin form factor,26 and in the meantime
mercial panels are stepping into market at a reasonable suppressing the undesired halo effect and clipping effect.
price. Our work would provide useful guidelines for sys- Conventional edge‐lit LCDs15,16 feature thin profile, but
tem design optimizations of mini/micro‐LED displays. the light guide plate is relatively thick if a high‐
luminance large‐area LED array is adopted. On the other
2 | M I N I‐LED B ACKLIT LCDS hand, conventional direct‐lit LCDs with fewer number of
LEDs20,22 can provide high luminance and HDR, but a
2.1 | Mini‐LED backlit LCD system relatively long travel distance is needed to ensure good
backlight uniformity. In comparison, the small chip size
Conventional LCDs suffer from a relatively low CR, and and large number of mini‐LEDs make the light to spread
some possible causes are nonuniform alignment of the out evenly so that the required optical distance between
liquid crystal (LC) layer, scattering of the color filters LED and diffuser is shorter.
(CFs), and diffraction from the pixelated electrodes.35 To Halo effect and clipping effect are common issues in
boost CR, local dimming with spatially segmented back- local dimming LCDs. Halo effect is the light leakage from
light unit (BLU) is an effective approach. Each segment, bright objects to adjacent dark areas. Clipping effect
the so‐called local dimming zone, is controlled comes from the insufficient luminance in a local
independently. With 10‐bit backlight modulation, the dimming zone when the adjacent zones are dimmed.
CR can increase from 1000 to 5000:1 to approximately Figure 3 schematically shows these two effects. The cen-
1 000 000:1. A schematic mini‐LED backlit LCD is shown ter of the local dimming zones are xzone = 0, ±1, ±2, …
in Figure 1. For discussion purpose, let us assume each with interval Δxzone = 1. In Figure 3, only the center zone
mini‐LED has a square shape. The emitted light propa- at xzone = 0 is at peak luminance while the surrounding
gates a distance (eg, adhesive layer) before reaching the zones are dimmed. Ideally, the luminance of each zone
diffuser. The distance and scattering strength of the should be uniform and independently controlled, as
diffuser need to be optimized so that the outgoing light Figure 3A shows. However, in practice, the intensity of
is spatially uniform before entering the LC layer. each local dimming zone is contributed by not only the

FIGURE 1 Schematic diagram of mini–

light‐emitting diode (LED) backlit liquid
crystal display (LCD)
HUANG ET AL. 389

FIGURE 2 Light modulation of mini–light‐emitting diode (LED) backlit liquid crystal display (LCD): A, mini‐LED backlight modulation;
B, luminance distribution of the light incident on the liquid crystal (LC) layer; and C, displayed image after LCD modulation

FIGURE 3 Schematic show of halo

effect and clipping effect in local dimming
liquid crystal displays (LCDs): A, ideal and
B, practically obtainable local dimming
intensity profiles; C, target intensity
profile after LCD modulation;
D, practically obtainable intensity profile
with halo effect and clipping effect

aligned light source but also the light leakage from adja- A variety of local dimming algorithms have been
cent zones, as Figure 3B depicts. As a result, the intensity developed to suppress these two effects, from the basic
in the center zone is “clipped” to one half (purple area), “maximum,” “average” methods, to the complex point
and the light leaks to adjacent zones forming “halo” spreading function (PSF) integrations.19,21,23 In 2013,
(yellow area). Afterward, a LCD panel modulates the Burini et al compared different algorithms and conducted
light from the BLU (red lines) to get finer details (blue optimization to find the best trade‐off point between halo
lines). While the target light profile is plotted in and clipping effects with power constraint.24
Figure 3C, the displayed image quality could be degraded From the hardware aspect, an infinitely high CR or
as Figure 3D shows. pixel‐level dimming could eliminate these two effects. In
390 HUANG ET AL.

a practical HDR LCD design, increasing the number of

local dimming zones could reduce the dark area affected
by halo effect (the yellow area in Figure 3B), while a
higher LCD CR can effectively suppress the halo effect
in the bright zones (the little light leakage in the central
zone as indicated by the yellow area in Figure 3D).
Methods for reducing zone crosstalk is helpful to mitigate
both halo effect and clipping effect. In the following part,
we will demonstrate why mini‐LED BLU is promising to
function as a highly independent local dimming
controller.

2.3 | Mini‐LED BLU solutions

FIGURE 4 Simulated LabPSNR for high‐dynamic range (HDR)
The system configuration of mini‐LED backlit LCD deter- display systems with various local dimming zone numbers and
contrast ratio
mines the severity of halo effect and clipping effect and
affects the total thickness of BLU. The number of local
required for a multidomain vertical alignment (MVA)
dimming zones and LCD's CR have the dominant
LCD with CR = 5000:1. However, if an LCD's CR is lower
impacts on local dimming effect. However, between two
than 1000:1, then even 10 000 zones is still inadequate.
comparable panels, sometimes the one with fewer local
dimming zones could exhibit a better performance, which
is contradictory to the general trend. This conflict comes 2.3.2 | LED light expansion
from the different optical designs, where LED light
expansion and local light confinement also jointly From mini‐LED BLU to LC layer, the light profile of each
contribute to the final local dimming performance. In LED could expand from the original square‐shaped
the following, we will discuss the influence of each factor Lambertian distribution to a Gaussian‐like profile. The
and then suggest the corresponding optimization strate- final profile can be influenced by several factors including
gies. The following discussions are based on a 6.4‐inch the LED emission aperture, the distance between mini‐
smartphone placed at 25‐cm viewing distance, but these LEDs and diffuser, and other optical layers such as
results can be scaled up and applied to large‐size panels brightness enhancement film (BEF). Figure 5 depicts an
as well. exemplary one‐dimensional light profile. Here, we
assume there are six mini‐LEDs (NLED = 6) located at
xLED = ±0.5, ±1.5, and ±2.5, with an interval ΔxLED = 1.
2.3.1 | Number of local dimming zones In reality, there are 6 × 6 mini‐LEDs in the central dim-
and LCD CR ming zone. They are turned‐on together, while the adja-
cent zones are dimmed to the dark state. In Figure 5,
In 2018, Tan et al demonstrated that mini‐LED BLU
each black curve depicts the light profile entering the
could effectively suppress halo effect if the LCD CR and
LC layer from each individual mini‐LED, and the blue
the density of local dimming zones are properly chosen.13
curves delineate the single‐zone light profile. Because
By simulating the displayed images of a mini‐LED backlit
the light experiences propagation and diffusion before
LCD with different system configurations and conducting
entering the LC layer, here, we use Gaussian function
subjective experiments, they found the peak signal‐to‐
to fit the expanded single‐LED light profile:
noise ratio in the CIE 1976 L*a*b* color space (LabPSNR)
can be used as a metric to evaluate the halo effect. When " #
LabPSNR > 47.7 dB, only less than 5% people could ðx LED −x LED c Þ2
I ðx LED Þ ∝ exp − ; (1)
differentiate the displayed image on a mini‐LED backlit 2σ 2
LCD from the original picture.
Figure 4 shows the correlation between the LCD CR where xLED_c is the locus of the source LED, and σ is
and local dimming zone number. The black dashed lines an expansion characteristic parameter.
represent that the halo effect is unnoticeable. From In Figure 5, the vertical red dashed lines denote the local
Figure 4, we find that approximately 3000 local dimming dimming zone borders. As we can see, a small σ/xLED helps
zones is required for a fringing‐field switching (FFS) LCD confine the light in the local area (Figure 5A) while more
with CR = 2000:1, and approximately 200 zones are than one‐half of the light energy would spread outside
HUANG ET AL. 391

FIGURE 5 Simulated spatial profiles of local dimming backlight units (BLUs) with different σ/xLED values

the zone when σ/xLED is large (Figure 5C). Such a crosstalk 2.3.3 | Local light confinement
could impair the local dimming function and give rise to
the unwanted halo effect and clipping effect. To reduce crosstalk between adjacent local dimming
Figure 6 shows that for a given number of LEDs in a zones without compromising uniformity, optical struc-
local dimming zone (NLED), better image fidelity (higher tures such as bank isolation36 or lens collimation37 can
LabPSNR) can be obtained by a smaller σ/xLED, corre- be employed in a period of zone pitch (pzone). Ideally, a
sponding to a smaller LED emission aperture and shorter rectangular light profile can generate uniform local dim-
optical distance. The latter leads to a thinner panel ming backlight without crosstalk. Whereas in practical
profile. However, the associated challenges are thermal designs, only flattop profile can be realized, which can
management, manufacturing yield, and especially the be described by a super‐Gaussian function as
compromised luminous uniformity. Figure 5A shows that
   
if the LED light does not spread wide enough, the resul- x zone −x zone c β
I ðx zone Þ ∝ exp −   : (2)
tant backlight intensity could be very sensitive to the σ
spatial location. Therefore, a proper σ/xLED should be
selected. For instance, σ/xLED = 0.5 could provide greater Similar to above discussion, here, we assume the cen-
than 97% backlight uniformity, which enables unnotice- ter of the local dimming zones (xzone_c) are xzone = 0,
able halo effect on a local dimming LCD with 2 × 2 LEDs ±1, ±2, … with interval Δxzone = 1. In Figure 7, each black
per local dimming zone and CR = 2000:1 (Figure 6B). In curve depicts a spatial profile of light generated by the
Figure 6A to 6C, if we compare the LabPSNR values at zone under its curve center, while the red dashed lines
σ/xLED = 0.5 and an identical CR, we find that a smaller delineate the borders of the zone at xzone_c = 0. We set
NLED leads to a higher LabPSNR. The reason is that, here, σ/xzone ~ 0.5 in order to obtain good overall uniformity,
we use the same LED dimension parameters and panel as the blue curves indicate. Figure 7A to 7C shows that
size for simulation. In other words, the smaller NLED, as β increases from 2 to 25, the crosstalk is reduced so
the larger number of local dimming zones, therefore the that the clipping effect is lessened accordingly. Although
higher LabPSNR. In a mini‐LED backlit LCD system, the uniformity is improved noticeably from Figure 7A to
σ/xLED can be obtained by Gaussian fitting the expanded 7C, the abrupt luminance change at zone borders is still
spatial luminous profile of a single mini‐LED. observable (Figure 7C) at a large β. If the compensation

FIGURE 6 Simulated LabPSNR for high‐dynamic range (HDR) display systems with different NLED. The blue, red, and yellow lines stand
for contrast ratio (CR) = 1000:1, 2000:1, and 5000:1, respectively
392 HUANG ET AL.

FIGURE 7 Simulated spatial profiles of different local dimming backlight unit (BLU) with different β

at borders is not performed carefully, the incongruous without any LCD panel. Presently, the major technical
lines may be noticeable in the actual display panel. In challenges are in three aspects: fabrication yield and cost
practical manufacturing, uneven distribution of local due to mass transfer, ACR due to strong internal reflec-
dimming zone and misalignment between dimming zone tion, and decreased IQE as the chip size decreases. The
and compensation may aggravate this issue. high cost is associated with the relatively low fabrication
Figure 8 demonstrates that good light confinement yield.38 Defects could be generated by LED chips, parti-
(high β) helps improve image quality. As β increases, cles, and the complex massive transfer procedure.27,39
LabPSNR increases initially but saturates as β exceeds To ensure display quality, color uniformity should be
4.5. This implies local light confinement is helpful to strictly controlled over the whole panel through multiple
certain degree. In contrast, high CR and short pzone help transfers.32 Taking a 4K full‐color display as an example,
enhance the LabPSNR value more obviously. When if the process yield is 99.99%, then there are approxi-
β > 2, an unnoticeable halo effect can be achieved for mately 2200 bad subpixels to be repaired. A yield as high
the LCDs with CR > 1000:1 (blue lines), 2000:1 (red as 99.9999% is required in order to reduce the number of
lines), and 5000:1 (yellow lines) with pzone = 1 mm bad subpixels to approximately 22 counts. Ideally, a good
(Figure 8A), 2 mm (Figure 8B), and 6 mm (Figure 8C), display should be defect‐free. In order to improve yield
respectively. In practice, β can be extracted from a mini‐ and accelerate production speed, a two‐step mass transfer
LED enhanced LCD by super‐Gaussian fitting the spatial approach has been developed.33 In the first step, “good”
luminous profile of single‐lit local dimming zone. mini/micro‐LEDs are transferred from epitaxial wafers
to an interposer substrate or cartridge array. After that,
the patterned LEDs are transferred to display substrate.38
3 | M I N I / M I C R O‐L E D EM I S S I V E From the cost management viewpoint, small LED chip
DISPLAYS size is preferred. The estimated die cost of Samsung's
146‐inch 4K micro‐LED TV “The Wall” by Yole Develop-
In Section 2, we discussed strategies to achieve HDR dis- ment is approximately $30 000, making the price unaf-
play with mini‐LED enhanced local dimming LCDs. fordable for average consumers. Similar to other
From here on, we will introduce mini/micro‐LED as technologies, the initial high cost could be reduced dra-
emissive displays: each LED chip serves as a color pixel matically as the manufacturing technique becomes

FIGURE 8 Simulated LabPSNR for high‐dynamic range (HDR) display systems with various pzone. The blue, red, and yellow lines stand
for liquid crystal display's (LCD's) contrast ratio (CR) = 1000:1, 2000:1, and 5000:1, respectively
HUANG ET AL. 393

mature. However, device structure should be optimized 3.1.1 | ACR calculation and metric of
beforehand. In the following sections, we will discuss optical performance
some design strategies by analyzing the optical
(eg, ACR) and electrical (eg, IQE) performances in detail. Figure 10 depicts the device structure of an RGB
mini/micro‐LED emissive display, in which the LED
array is encapsulated by bonding layers and a protection
3.1 | Ambient CR
glass. For this device structure, the luminous reflectance
(RL) can be described by
The CR of a display device is usually measured at dark
ambient. In the presence of ambient light, the CR could
be deteriorated dramatically because of the surface and RL ¼ Rex þ Rin ¼ Rs þ AP⋅ð1 − Rs Þ⋅RL LED ⋅T: (4)
interface reflections. Under such a circumstance, ACR is
a more meaningful metric to compare because it is what In Figure 10 and Equation 4, Rs, AP, and RL_LED stand
the viewer actually experiences.14 The ACR can be for surface reflectance, aperture ratio, LED luminous
expressed as follows: reflectance, respectively, and T represents the transmit-
tance of the reflected ambient light from LED through
I am additional optical components, such as CF and circular
Lon þ ⋅RL polarizer (CP). In each pixel, the RGB LED chips only
ACR ¼ π : (3)
I am occupy a portion of the pixel, and the rest area is covered
Loff þ ⋅RL
π by black matrix (BM); as a result, the AP is usually small.
In Equation 4, RL consists of two terms: external reflec-
tion (Rex) at the air‐glass interface and internal reflection
Assuming luminous reflectance RL = 4% and CR ≈ 106:1
(Rin) by LEDs, as illustrated by the red arrows and cyan
(off‐state luminance Loff ≈ 0), we simulate the images
arrows in Figure 10, respectively. To reduce RL, BM, CF,
with different display on‐state luminance (Lon) and ambi-
and CP are helpful to reduce Rin, while antireflection
ent light illuminance (Iam). Results are summarized in
(AR) surface treatment helps reduce Rex.
Figure 9. At a given peak luminance, as the environment
From the layout of two pixels depicted in Figure 11,
light gets stronger (Iam increases), ACR decreases and the
the aperture ratio and characteristic LED chip size (s)
displayed image is gradually washed out. To improve the
are defined as
ACR of an LCD, a straightforward way is to boost the dis-
play luminance, say from 1000 to 2500 cd/m2. However,
the light leakage in dark state also increases, resulting emission area 3s2
AP ¼ ¼ 2; (5)
in a limited ACR. whole area p

FIGURE 9 Simulated displayed images with different peak luminance and ambient illuminance. The ambient contrast ratio (ACR) is
marked on the right bottom corner of each picture
394 HUANG ET AL.

FIGURE 10 Scheme demonstration of

ambient light reflection on mini/micro–
light‐emitting diode (LED) emissive
display panels. Red arrows and cyan
arrows correspond to external reflection
and internal reflection, respectively

leakage. Here, we assume the display has a high intrinsic

CR = Lon/Loff > 1 000 000:1 and Loff < < (Iam/π) · RL. To
ensure a reasonably good sunlight readability, we also
assume ACR > > 1. Under such conditions, Equation 3
can be approximated as

I am
α⋅L0 þ ⋅RL π⋅L0 α π⋅L0 α
ACR ≈ π ¼1þ ⋅ ≈ ⋅ ; (7)
I am I am RL I am RL
⋅RL
π

FIGURE 11 Pixel layout and dimensions of a mini/micro–light‐

where
emitting diode (LED) display. Each color pixel consists of three
RL ¼ Rs þ AP⋅ð1 − Rs Þ⋅β⋅R0 ⋅T: (8)
R/G/B subpixels
Equation 7 suggests a quantitative metric to evaluate the
pffiffiffiffiffiffiffi optical performance of a mini/micro‐LED emissive
s ¼ w⋅l: (6)
display system. The first term π · L0/Iam represents the
ratio of intrinsic display luminance to ambient lumi-
Here, w and l denote the width and length of a single
nance, which depends on the applied LED current and
LED chip, respectively, and p stands for the pixel pitch
the ambient condition. Differently, the second term
length. Because BM absorbs the incident ambient light,
α/RL originates from display optics. Therefore, we call it
only the light falls on the aperture would be internally
as the figure‐of‐merit of optical design (FoMo):
reflected. This explains why the Rin term is related to
AP in Equation 4. α
FoM o ¼ : (9)
The optical structure influences RL and the on‐state RL
display luminance Lon. For easier comparison, we define
The influence of the optics part is governed by the
a display luminance coefficient α with Lon = α · L0 and
numerical coefficients in Equations 8 and 9, such as α,
a LED reflectance coefficient β by RL_LED = β · R0. Here,
β, T, Rs, and AP. In the following sections, we will ana-
L0 and R0, respectively, stand for the on‐state luminance
lyze how each optical component affects the ACR.
and LED luminous reflectance of the benchmark struc-
ture: mini/micro‐LED with indium tin oxide (ITO)
electrode, well‐aligned BM, and without any additional 3.1.2 | LED electrode
optical element, such as CF or CP. When we replace
the bottom electrode from ITO to another material, the The bottom electrode of LED affects the light emission
display luminance could be boosted by α times, while efficiency. Besides ITO, multilayer metal electrode can
the LED reflectance is changed by β times. These effec- also be used for LEDs to achieve good ohmic contact. A
tive coefficients are the properties of electrode materials typical structure for forming multilayer metal contact
and should be obtained through simulations or experi- contains three parts40: (1) a thin layer physically attached
ments. For an emissive mini/micro‐LED display, the to the semiconductor to form good ohmic contact, eg, a
ideal off‐state display luminance Loff should be zero. thin ITO41; (2) intermediate layers serving as a diffusing
But in reality, it may have a small crosstalk‐induced light barrier (eg, noble metals Pt, Pd, and Re as well as
HUANG ET AL. 395

refractory metals Ti, W, Ta, and Mo); and (3) highly con-
ductive metal (eg, Au) for bonding. The optical property
of mini/micro‐LED electrode depends on the exact elec-
trode structure employed. Here, we take two typical elec-
trode materials, transparent ITO and reflective Ag
electrode,42 as examples to show how display perfor-
mance can be influenced by the optical properties of
LED electrode. The LED structures used in our simula-
tion are drawn in Figure 12. When an electric current is
applied, the multiple quantum well (MQW) layer could
emit light in upward and downward directions. While
one‐half of the light transmits the transparent ITO and
gets lost in the structure of Figure 12A, the Ag electrode
in Figure 12B works as a bottom reflector to recycle the FIGURE 13 Simulated intensity of incident D65 light source and
downward light, indicating αAg = 2. However, one draw- reflected light from a mini/micro–light‐emitting diode (LED) with
back is that it increases RL_LED from RL_ITO = 5.4% to ITO and Ag electrode
RL_Ag = 92.3%. For displays, we need to consider human
perception when calculating the luminous reflectance: transmitted spectra after CFs (dashed lines). In contrast,
the incident ambient light passes through the CFs twice
λ2
∫ V ðλÞSðλÞRðλÞdλ due to the reflection of bottom electrode. Thus, in
RL LED ¼ λ1 λ2 ; (10) Figure 14B, we plot the D65 incident ambient light (solid
∫λ1 V ðλÞSðλÞdλ line) and the outgoing light after passing through the CFs
twice (dashed lines). The obtained effective coefficients
where V(λ) is the photopic human eye sensitivity func-
are αCF = 0.75 and TCF = 0.184.
tion, R(λ) is the spectral reflectance, and S(λ) is the spec-
A broadband CP consists of a linear polarizer, a half‐
trum of the ambient light (CIE Standard Illuminant D65).
wave plate, and a quarter‐wave plate. The linear polarizer
Figure 13 depicts the D65 incident light and the
blocks half of the LED light, corresponding to αCP = 0.5.
simulated reflected light of ITO‐ and Ag‐embedded LEDs.
The merit of using a CP is to suppress the internal reflec-
The index matched incident medium is used in our
tion from the bottom electrode. Because TCP < < Rs, we
simulations. From the data shown in Figure 13, and using
set TCP ~ 0. However, a serious drawback is that the
Equation 10, we find βAg = RL_Ag/RL_ITO = 17.
added CP reduces the panel's flexibility. This is
undesirable for flexible displays.
3.1.3 | CFs and CP
3.1.4 | Surface reflection
Each CF transmits about 80% of the corresponding emitted
RGB LED light but absorbs about two‐thirds of the inci- As shown in Equation 4, surface reflection plays an
dent ambient (white) light. Figure 14A shows typical important role in external ambient reflection. A normal
RGB LED emission spectra (solid lines) and the glass‐air surface reflectance is Rs ≈ 4.0%, while

FIGURE 12 Mini/micro–light‐emitting
diode (LED) structures with A,
transparent ITO electrode and B, Ag
reflective electrode
396 HUANG ET AL.

the performance of different designs, we summarize the

above‐mentioned parameters in Tables 1 and 2.
Figure 15A depicts the calculated FoMo as a function
of AP, which is enlarged in Figure 15B. The highest FoMo
for each AP is obtained by optical structures with Ag elec-
trode (Ag, Ag + CF, and Ag + CP). Ag + CF design is
favored for 0.24% < AP < 1.5%, while Ag design and the
Ag + CP are preferred for the smaller and the larger
AP, respectively. Overall speaking, a high FoMo (greater
than 200) could be maintained with the optimal struc-
tures for each AP. That means, if the display luminance
in the default ITO design is 500 cd/m2, the display would
well qualify for the requirements of both indoor
(ACR = 1000:1 under 314 lux lighting) and outdoor
(ACR = 10:1 under 31416 lux strong sunlight)
applications.
The influence of Rs on FoMo can be seen in Figure 15B
and 15C. Compared with bare glass surface (Rs = 4.0% in
Figure 15C), AR‐coated surface shows an eight times
smaller Rs (0.5% in Figure 15B). By plotting FoMo with
eight times scale difference, we find that AR coating helps
boost FoMo by eight times, and the corresponding AP is
reduced to one‐eighth. This rule can be applied to predict
the influence of other surface treatment. If Rs is reduced
by n times from 4.0%, similar profile as Figure 15C can
be obtained, except for an n‐time smaller AP scale and
an n‐time larger FoMo scale. After that, the optimal struc-
ture for each AP can be obtained.

FIGURE 14 A, The RGB emission spectra (solid lines) and the

TABLE 1 Optical parameters in various designs
transmittance after passing through a color filter (CF) (dashed
lines). B, The D65 white light source (solid line) and the Structure α β T Rs
transmittance after passing through a CF twice (dashed lines). The
ITO 1 1 1 Rs_AR
RGB line colors stand for RGB LED/CF colors, respectively
ITO + CF αCF 1 TCF Rs_AR
manufacturers have been devoting extensive efforts to ITO + CP αCP 1 TCP Rs_AR
lower Rs. DisplayMate has found that Macbook Pro has Ag αAg βAg 1 Rs_AR
an impressive Rs = 0.5% and Rs = 2% for iPad mini 4
Ag + CF αAg · αCF βAg TCF Rs_AR
touch panel. While developing antiglare AR solutions
for touch panels, several groups, such as TruVue, Ag + CP αAg · αCP βAg TCP Rs_AR

Dexerials, and NSG, have achieved Rs < 1%. Here, we Abbreviations: CF, color filter; CP, circular polarizer.
use the state‐of‐the‐art Rs_AR = 0.5% for the following
analyses.
TABLE 2 Parameters used for simulation

3.1.5 | Optimal structure Ag αAg 2

βAg 17
Under a given LED emittance and viewing environment,
CF αCF 0.75
the ACR is proportional to FoMo, as Equations 7 and 9
TCF 0.184
and indicate. For example, if the display luminance in
CP αCP 0.5
the default design (ITO without CF nor CP) is
TCP Approximately 0
1000 cd/m2, then FoMo = 100 indicates an ACR = 100:1
under 3142 lux overcast daylight. In other words, the opti- AR Rs_AR 0.50%

cal structure with a higher FoMo is favorable. To compare Abbreviations: AR, antireflection; CF, color filter; CP, circular polarizer.
HUANG ET AL. 397

FIGURE 16 A, Current density‐dependent internal quantum

FIGURE 15 FoMo of different optical structures as a function of efficiency (IQE) and B, light‐emitting diode (LED) current‐
AP: A, full‐scale plot for Rs = 0.5%; B, enlarged plot for Rs = 0.5%; C, dependent IQE for different LED dimensions. (c) IQEp and the
full‐scale plot for Rs = 4.0% corresponding current for different LED dimensions

3.2 | Electrical driving IQE As the current density (or current as Figure 16B
depicts) increases, IQE increases to a peak value (IQEp)
Due to the surface defect‐generated sidewall effect,43 and then declines. Details depend on the chip
impaired quantum efficiency on small‐chip mini/micro‐ size. As the blue line shows in Figure 16C, IQEp
LEDs has been observed. Several groups have reported increases drastically as s (defined in Equation 5)
current density‐dependent IQE with the trend shown in increases from 4 to 50 μm and then saturates in the 50
Figure 16A.44–46 to 500 μm region.
398 HUANG ET AL.

Unfortunately, the LEDs cannot be always driven at LEDs can be always driven at the desired IQEp, corre-
the sweet spot if analog driving with 100% duty cycle is sponding to the ηL in digital driving scheme. Therefore,
adopted. Taking a 65‐inch 4K2K TV as an example, the it is reasonable to use LED size‐dependent IQEp as a
panel area is Apanel = 1.16 m2. Assuming display metric to evaluate power efficiency. Here, we define an
luminance Lon = 1000 cd/m2 and luminous efficiency electrical figure‐of‐merit (FoMe) as
ηL = 200 lm/W,47 the electric power of the panel is
Ppanel = 18.2 W, as calculated from FoM e ¼ IQE p ; (15)

π⋅Lon ⋅Apanel ¼ ηL ⋅Ppanel : (11) which is plotted as the blue line in Figure 16C.

Applying a typical forward voltage V f = 3 V, the current

flowing through the LEDs in the whole panel is 3.3 | Optimization strategy

I panel ¼ Ppanel =V f ¼ 6:1A: (12) Until now, we have discussed the AP‐dependent FoMo
and the LED chip size‐dependent FoMe. Let us return to
For the 4K2K resolution, the number of RGB LEDs the basic questions of designing a mini/micro‐LED
(NLED) and the single‐LED current (ILED) are emissive display: what are the optimal LED chip size
and optical structure? In order to answer these questions,
N LED ¼ 3840 × 2160 × 3; (13) we need to consider both optical and electrical perfor-
mances together. Thus, let us define an optical‐electrical
I LED ¼ I panel =N LED ¼ 0:24μA: (14) figure‐of‐merit (FoMoe) as

Figure 16B shows the relationship between IQE and ILED. FoM oe ¼ FoM o ⋅FoM e : (16)
To be noticed, the IQE is relatively low at such a small
current (as the blue dashed lines mark), resulting in low The FoMoe indicates how high an ACR can be obtained
ηL and inadequate luminance. Therefore, a higher by the user with a given power consumption.
ILED ~ 1 μA (red dashed lines) may be required to produce Figure 17 shows the FoMoe as a function of s for differ-
Lon = 1000 cd/m2. Although the above‐mentioned param- ent pixel pitch length. We only plot the structures with
eters may vary in different panels, the calculated ILED Ag electrode because Figure 15 shows that it has better
should be in the same order, which is about 100× lower optical performance. As demonstrated in Figure 17, a
than the ILED for optimal IQE (50‐1000 μA for s ≤ 50 μm). peak FoMoe exists in the small s region for the Ag (blue
As a result, the energy efficiency is low in analog driving lines) and Ag + CF (red lines) designs. This is because a
scheme. In order to boost ηL, pulse width modulation small AP helps reduce internal reflection. In contrast,
(PWM) with low duty cycle can be utilized so that the for the Ag + CP design, CP helps suppress internal

FIGURE 17 Simulated FoMoe as a

function of s for different p: A, p = 73 μm
for d = 25 cm (smartphone); B,
p = 145 μm for d = 50 cm (gaming
monitors); C, p = 375 μm for 65‐inch 4K
TV; D, p = 3 mm for 12.3‐m‐wide 4K
scope display for movie theaters. The pitch
length p is calculated by viewing distance
d and 1‐arcminute human eye acuity. The
blue, red, and yellow lines stand for the
Ag, Ag + CF, and Ag + CP structure,
respectively
HUANG ET AL. 399

reflection so that RL and FoMo do not change with s. designs. Widespread application of mini/micro‐LED for
Therefore, FoMoe increases with s and then saturates, HDR displays is around the corner.
showing the same trend as FoMe (the yellow lines in
Figure 17 and the blue line in Figure 16C). Please note
ACKNOWLEDGMENT
that these results are calculated based on the parameters
listed in Table 2. Other electrode materials, CFs, surface The UCF group is indebted to a.u.Vista, Inc. for the
treatment, and electrical properties may change the line financial support.
shape in Figure 17. Nevertheless, our analyses remain
valid, and the trend for each design should be similar.
ORCID
From the comparison in Figure 17, we can see that the
optimal optical structure depends on the pixel pitch. For Shin‐Tson Wu https://orcid.org/0000-0002-0943-0440
personal flat panel displays (Figure 17A and 17B), Ag + CP
is preferred since it provides comparable or better perfor-
RE FER EN CES
mance. In the meantime, the highest FoMoe appears at a
much larger s than that of Ag and Ag + CF designs, 1. Schadt M. Milestone in the history of field‐effect liquid crystal
and the higher operating IQE implying less nonradiative displays and materials. Jpn J Appl Phys. 2009;48:03B001.
heating issue. Differently, for public displays with large 2. Tang CW, Vanslyke SA. Organic electroluminescent diodes.
pitch length (Figure 17C and 17D), Ag and Ag + CF Appl Phys Lett. 1987;51:913–915.
present a higher peak FoMoe, and the corresponding s is 3. Chen H‐W, Lee J‐H, Lin B‐Y, Chen S, Wu S‐T. Liquid crystal
within the fabrication range. For example, for a 3‐mm display and organic light‐emitting diode display: present status
pitch cinema display, its peak FoMoe occurs at and future perspectives. Light: Sci & Appl. 2018;7:17168.
s = 39 μm with the Ag design, as the blue curve in 4. Tan G, Zhu R, Tsai Y, et al. High ambient contrast ratio OLED
Figure 17C shows. The small die size yet without the and QLED without a circular polarizer. J Phys D Appl Phys.
need of a large‐area CP is not only cost‐effective but also 2016;49:315101.
advantageous for flexible displays. 5. Zhu R, Luo Z, Chen H, Dong Y, Wu S‐T. Realizing Rec. 2020
color gamut with quantum dot displays. Opt Express.
2015;23(18):23680–23693.
6. Peng F, Chen H, Gou F, et al. Analytical equation for the motion
4 | CONCLUSION picture response time of display devices. J Appl Phys.
2017;121(2):023108.
We have described two types of mini/micro‐LED dis- 7. Féry C, Racine B, Vaufrey D, Doyeux H, Cinà S. Physical mech-
plays: (1) As a LCD backlight, the zoned mini‐LED anism responsible for the stretched exponential decay behavior
enables local dimming, which helps suppress the halo of aging organic light‐emitting diodes. Appl Phys Lett.
effect and the clipping effect. Through numerical simula- 2005;87(21):213502.
tion and subjective experiments, we find that halo effect 8. Murawski C, Leo K, Gather MC. Efficiency roll‐off in organic
can be suppressed to an unnoticeable level, depending light‐emitting diodes. Adv Mater. 2013;25:6801–6827.
on the LCD's CR and local dimming zone numbers. For 9. Daly S, Kunkel T, Sun X, Farrell S, Crum P. Viewer preferences
example, around 3000 and 200 local dimming zones are, for shadow, diffuse, specular, and emissive luminance limits of
respectively, required for an FFS LCD with CR ≈ 2000:1 high dynamic range displays. SID Intl Symp Dig Tech Pap.
and an MVA LCD panel with CR = 5000:1 for a 6.4‐inch 2013;44(1):563–566.
smartphone placed at 25 cm. These results can also be 10. Park S, Xiong Y, Kim R, et al. Printed assemblies of inorganic
extended to large‐size panels according to the viewing light‐emitting diodes for deformable and semitransparent dis-
distance. A well‐confined light of each LED or each local plays. Science. 2009;325(5943):977–981.
dimming zone could reduce the interzone crosstalk and 11. Jiang HX, Lin JY. Nitride micro‐LEDs and beyond—a decade
clipping effect but may result in severe uniformity issue. progress review. Opt Express. 2013;21(S3):A475–A484.
(2) Similar to OLED displays, RGB mini/micro‐LED 12. Templier F. GaN‐based emissive microdisplays: a very promis-
emissive displays exhibit an excellent dark state but the ing technology for compact, ultra‐high brightness display
strong internal reflection may give rise to compromised systems. J Soc Inf Display. 2017;24(11):669–675.
ACR. Besides, micro‐LEDs faces low IQE problem on 13. Tan G, Huang Y, Li M‐C, Lee S‐L, Wu S‐T. High dynamic range
small‐size chips. We have also analyzed the influence of liquid crystal displays with a mini‐LED backlight. Opt Express.
each optical component and suggested the optimal LED 2018;26(13):16572–16584.
chip‐size considering both ACR and power efficiency. 14. Chen H, Tan G, Wu S‐T. Ambient contrast ratio of LCDs and
Our work would help optimize device and system OLED displays. Opt Express. 2017;25(26):33643–33656.
400 HUANG ET AL.

15. Shirai T, Shimizukawa S, Shiga T, Mikoshiba S, Käläntär K. 33. Bower CA, Meitl MA, Raymond B, et al. Emissive displays with
RGB‐LED backlights for LCD‐TVs with 0D, 1D, and 2D adaptive transfer‐printed assemblies of 8 μm × 15 μm inorganic light‐
dimming. SID Intl Symp Dig Tech Pap. 2006;37(1):1520–1523. emitting diodes. Photonics Res. 2017;5(2):A23–A29.
16. Hulze HG, de Greef P. Power savings by local dimming on a 34. Wu T, Sher C, Lin Y, et al. Mini‐LED and micro‐LED: Promising
LCD panel with side lit backlight. SID Intl Symp Dig Tech candidates for the next generation display technology. Appl Sci.
Pap. 2009;40(1):749–752. 2018;8(9):1557.

17. Bonar JR, Valentine GJ, Gong Z, Small J, Gorton S. High‐bright- 35. Chen H, Tan G, Li M‐C, Lee S‐L, Wu S‐T. Depolarization effect
ness low‐power consumption microLED arrays. Proc SPIE. in liquid crystal displays. Opt Express. 2017;25(10):11315–11328.
2016;9768:97680Y. 36. Sun C, Chien W, Moreno I, et al. Calculating model of light
transmission efficiency of diffusers attached to a lighting cavity.
18. Seetzen H, Whitehead LA, Ward G. A high dynamic range dis-
Opt Express. 2010;18(6):6137–6148.
play using low and high resolution modulators. SID Intl Symp
Dig Tech Pap. 2003;34(1):1450–1453. 37. Woodgate GJ, Harrold J, Limited O, Park H, Heyford U, Ox O.
Micro‐optical systems for micro‐LED displays. SID Intl Symp
19. Seetzen H, Heidrich W, Stuerzlinger W, et al. High dynamic Dig Tech Pap. 2018;49(1):1559–1562.
range display systems. ACM Trans Graph. 2004;23(3):760–768.
38. Paranjpe A, Montgomery J, Lee SM, Morath C. Micro‐LED
20. de Greef P, Hulze HG. Adaptive dimming and boosting displays: key manufacturing challenges and solutions. SID Intl
backlight for LCD‐TV systems. SID Intl Symp Dig Tech Pap. Symp Dig Tech Pap. 2018;49(1):597–600.
2007;38(1):1332–1335. 39. Chaji R, Fathi E, Zamani A. Low‐cost micro‐LED displays for all
21. Lin F, Huang Y, Liao L, et al. Dynamic backlight gamma on applications. SID Intl Symp Dig Tech Pap. 2017;48(1):264–267.
high dynamic range LCD TVs. J Disp Technol. 2008;4(2): 40. Schubert EF. Light‐emitting diodesTroy: E. Fred Schubert; 2018.
139–146. 41. Wong MS, Hwang D, Alhassan AI, et al. High efficiency of III‐
22. Chen H, Ha TH, Sung JH, Kim HR, Han BH. Evaluation of LCD nitride micro‐light‐emitting diodes by sidewall passivation using
local‐dimming‐backlight system. J Soc Inf Display. 2010;18(1): atomic layer deposition. Opt Express. 2018;26(16):21324–21331.
57–65. 42. Liu Z, Wang K, Luo X, Liu S. Precise optical modeling of blue
23. Kim S, An J‐Y, Hong J‐J, Lee TW, Kim CG, Song W‐J. How to light‐emitting diodes by Monte Carlo ray‐tracing. Opt Express.
reduce light leakage and clipping in local‐dimming liquid‐ 2010;18(9):9398–9412.
crystal displays. J Opt Soc Am A. 2009;17(12):1051–1057. 43. Tian P, McKendry JJD, Gong Z, et al. Size‐dependent efficiency
and efficiency droop of blue InGaN micro‐light emitting diodes.
24. Burini N, Nadernejad E, Korhonen J, Forchhammer S, Wu X.
Appl Phys Lett. 2012;101(23):231110.
Modeling power‐constrained optimal backlight dimming for
color displays. J Disp Technol. 2013;9(8):656–665. 44. Olivier F, Daami A, Licitra C, Templier F. Shockley‐Read‐Hall
and Auger non‐radiative recombination in GaN based LEDs: a
25. Deng Z, Zheng B, Zheng J, et al. High dynamic range incell LCD
size effect study. Appl Phys Lett. 2017;111(2):022104.
with excellent performance. SID Intl Symp Dig Tech Pap.
45. Bourim EM, Han JI. Size effect on negative capacitance at
2018;49(1):996–998.
forward bias in InGaN/GaN multiple quantum well‐based blue
26. Day J, Li J, Lie DYC, Bradford C, Lin JY, Jiang HX. III‐Nitride LED. Electron Mater Lett. 2016;12(1):67–75.
full‐scale high‐resolution microdisplays. Appl Phys Lett.
46. Konoplev SS, Bulashevich KA, Karpov SY. From large‐size to
2011;99:031116.
micro‐LEDs: scaling trends revealed by modeling. Phys Status
27. Henry W, Percival C. ILED displays: next generation display Solidi Appl Mater Sci. 2018;215(10):1700508.
technology. SID Intl Symp Dig Tech Pap. 2016;47:747–750. 47. Sadeghi S, Kumar BG, Melikov R, Aria MM, Jalali HB,
28. Chen H, He J, Wu S‐T. Recent advances on quantum‐dot‐ Nizamoglu S. Quantum dot white LEDs with high luminous
enhanced liquid‐crystal displays. IEEE J Sel Top Quantum Elec- efficiency. Optica. 2018;5(7):793–802.
tron. 2017;23(5):1–11. 1900611
29. He J, Chen H, Chen H, Wang Y, Wu S‐T, Dong Y. Hybrid AUTHOR BIOGRAPHIES
downconverters with green perovskite‐polymer composite films
for wide color gamut displays. Opt Express. 2017;25(11):
Yuge Huang received her BS
12915–12925. degree in physics from Nanjing
University, China, in 2015 and is
30. Lin H‐Y, Sher C‐W, Hsieh D‐H, et al. Optical cross‐talk reduc-
tion in a quantum‐ display by a lithographic‐fabricated
currently working toward the
photoresist mold. Photonics Res. 2017;5(5):411–416. PhD degree at College of Optics
and Photonics, University of Cen-
31. Chen G‐S, Wei B‐Y, Lee C‐T, Lee H‐Y. Monolithic red/green/
blue micro‐LEDs with HBR and DBR structures. IEEE Photon- tral Florida, Orlando, FL, USA.
ics Technol Lett. 2018;30(3):262–265. Her research focuses on mini‐
32. Beckers A, Fahle D, Mauder C, Kruecken T, Boyd AR, Heuken
LED backlit LCDs and fast‐
M. Enabling the next era of display technologies by micro LED response liquid crystal devices for augmented reality
MOCVD processing. SID Intl Symp Dig Tech Pap. and virtual reality displays. She received SID Metro
2018;49(4):601–603. Detroit Academic Award in 2018.
HUANG ET AL. 401

University, Korea, in 2010. Currently, he is a principal

Guanjun Tan received a BS
engineer and fellow at AU Optronics. He is one of the
degree in Physics from University
key inventors of fringe‐field switching (FFS) liquid
of Science and Technology of
crystal display. He contributed to development and
China in 2014 and is currently
commercialization of mobile, tablet, and monitor
working toward the PhD degree
TFT‐LCD products using the FFS mode. He received
at the College of Optics and Pho-
SID Special Recognition Award in 2012 and SID Fellow
tonics, University of Central Flor-
Award in 2018.
ida. His current research interests
include head‐mounted displays,
organic LED displays, and novel liquid crystal Shin‐Tson Wu is Pegasus profes-
displays. sor at College of Optics and Pho-
tonics, University of Central
Fangwang Gou received her BS Florida. He is among the first six
degree from University of Elec- inductees of the Florida Inventors
tronic and Science and Technol- Hall of Fame (2014) and a Charter
ogy of China in 2012 and MS Fellow of the National Academy
degree from Peking University, of Inventors (2012). He is a fellow
China, in 2015. Currently, she is of the IEEE, OSA, SID, and SPIE,
working toward the PhD degree and an honorary professor of National Chiao Tung
at College of Optics and Photon- University (2018) and of Nanjing University (2013).
ics, University of Central Florida, He is the recipient of 2014 OSA Esther Hoffman Beller
USA. Her current research interests include optical Medal, 2011 SID Slottow‐Owaki Prize, 2010 OSA
devices for liquid crystal displays, augmented reality Joseph Fraunhofer Award, 2008 SPIE G. G. Stokes
and virtual reality, and micro‐LEDs. Award, and 2008 SID Jan Rajchman Prize. Presently,
he is serving as SID honors and awards committee
Ming‐Chun Li received his BS degree in Electronic chair.
Engineering, Feng Chia University, Taiwan, in 2000
and MS degree from National Sun Yat‐Sen University,
Taiwan, in 2005. Currently, he is a manager at AU
Optronics, in charge of LCD cell material and process How to cite this article: Huang Y, Tan G, Gou F,
technology development. Li M‐C, Lee S‐L, Wu S‐T. Prospects and challenges
of mini‐LED and micro‐LED displays. J Soc Inf
Seok‐Lyul Lee received his BS degree in Electronic Display. 2019;27:387–401. https://doi.org/10.1002/
Communication Engineering, Kwangwoon University, jsid.760
Korea, in 1992 and MS degree from Chonbuk National