Research Article Vol. 11, No.

4 / April 2023 / Photonics Research 549

Systematic study on size and temporal

dependence of micro-LED arrays for display
applications
JUNYANG NIE,1,2 ZHI ZHANG,2 KONGJIE CHEN,2 YIFENG LIU,2 SHUANG HE,2 ZHONGHANG HUANG,2
XIAONING ZHANG,1 JIE SUN,2,3,6 FAN ZHANG,4 YONGSHENG WU,4 AND QUN YAN1,2,5,7
1
Faculty of Electronic and Information Engineering, Xi’an Jiaotong University, Xi’an 710049, China
2
Fujian Science & Technology Innovation Laboratory for Optoelectronic Information of China, and National and Local United Engineering
Laboratory of Flat Panel Display Technology, College of Physics and Information Engineering, Fuzhou University, Fuzhou 350100, China
3
Quantum Device Physics Laboratory, Department of Microscience and Nanotechnology, Chalmers University of Technology, Göteborg 41296,
Sweden
4
Fujian Prima Optoelectronics Company Ltd., Fuzhou 350000, China
5
Rich Sense Electronics Technology Inc., Jinjiang 362200, China
6
e-mail: [email protected]
7
e-mail: [email protected]

Received 4 November 2022; accepted 11 January 2023; posted 18 January 2023 (Doc. ID 479593); published 9 March 2023

Micro-LEDs are one of the most promising candidates for next-generation displays, yet they are inconvenienced
by the efficiency reduction induced by the sidewall defects when pursuing further scaled-down device dimensions.
We have systematically investigated both the size and temporal dependence of micro-LEDs. Micro-LED arrays
with a mesa size ranging from 7 to 100 μm were prepared for display purposes. The luminance and external
quantum efficiency (EQE) were measured and discussed. Surprisingly, micro-LED arrays with a smaller mesa size
exhibit a higher EQE under 100 ns pulse duration operation when compared with longer pulse duration oper-
ations. Under certain short-pulsed excitation, a 7 × 7 μm2 micro-LED array even exhibits a >20% higher EQE as
compared to the direct current (DC) or the long duration pulse operation condition. We thus concluded that the
notorious efficiency reduction induced by sidewall defects in small-sized micro-LED arrays could be significantly
reduced by applying short-pulse voltages. © 2023 Chinese Laser Press
https://doi.org/10.1364/PRJ.479593

1. INTRODUCTION of the micro-LED array but also influences its optical and
The next-generation displays demand high pixel density, high electrical performance [23–29]. In micro-LEDs, the brightness,
resolution, and ultra-high luminance, which is difficult to external quantum efficiency (EQE), current spreading, heat
achieve with current display techniques [1–5]. As an extension management, and color shift are all found to be size dependent
of the conventional light emitting diodes (LEDs), micro-LED [30–32]. Ideally, smaller micro-LEDs should have more supe-
is considered to be a promising candidate. A GaN-based micro- rior properties than larger ones, not only because they have bet-
LED is capable of quick response, long lifetime, low consump- ter current spreading but also because they have more sidewalls,
tion, high brightness, and extraordinary stability, and it is which should lead to better light extraction and heat dissipa-
dimensionally advanced over the current techniques [6–8]. tion. In real life, unfortunately, as the sidewalls are defective,
With the combined efforts of both academia and industry, the peak EQE at smaller mesa sizes is significantly lower than
the industrialization of micro-LED display can be rightly expected, and the overall performance is often worse than with
expected [9–21]. larger micro-LEDs. Therefore, the size effect of the micro-
Ever since Lin et al. [22] published research on the first LEDs is essential. However, the size-dependent behaviors of
monolithic InGaN micro-LED array in 2000, the mesa size micro-LEDs have been so far investigated mainly at the single-
effect on the performance of micro-LEDs has received much pixel level under static conditions. Array properties of
attention. Embodied as the emitting region of an individual micro-LEDs for real-life display applications are relatively
LED fabricated by dry etching from the p-GaN layer to the underinvestigated. We note that in a real display system, nu-
n-GaN layer of the GaN epitaxial structures, the mesa size merous pixels should switch on/off based on the image load.
not only determines the pixel density and the integration level Moreover, micro-LED displays typically operate in a pulse

2327-9125/23/040549-09 Journal © 2023 Chinese Laser Press

 550 Vol. 11, No. 4 / April 2023 / Photonics Research Research Article

mode. A frequency of at least 60 Hz large is applied to avoid p-GaN / MQWs / Function layers
n-GaN
flicker for human vision. In fact, an even higher frequency is (a) u-GaN

required for a better display quality [33–36]. Ultra-high- 3D GaN

frequency driving is in favor of building a future display that sapphire

can deliver 3D moving images [37]. Yet, although the LED’s ITO
Mesa

temporal dependence is a focus issue in optical communication n-GaN

areas [38–40], its impact on micro-LED displays has rarely (b) u-GaN

3D GaN Mesa
been studied. sapphire

In this paper, the size-dependent behaviors of micro-LED ITO

arrays have been systematically studied, and the temporal Mesa
n-GaN Isolation procedure
dependence of micro-LED arrays under applied voltage pulse (c) u-GaN
has also been thoroughly researched. GaN-based monolithic 3D GaN

passive-matrix micro-LED arrays of the same configuration, sapphire

i.e., 300 μm pixel pitch and 48 × 48 resolution, were prepared ITO n-metal
with the mesa dimension ranging from 7 × 7 μm2 to
Mesa
n-GaN n-metal
100 × 100 μm2 . The luminance and EQE were obtained based (d) u-GaN

on a realistic setting of operating with a 6 × 6 pixel array. Some 3D GaN

sapphire
of our results are consistent with previous studies, i.e., as a re- PV
sult of more sidewall defects, under a direct current (DC) op- ITO
Mesa
n-GaN
n-metal

eration condition, micro-LED arrays with smaller mesa sizes (e)

n-GaN
u-GaN
are not ideal in both brightness and efficiency. Surprisingly,
3D GaN
however, when employing a high-frequency pulse excitation, sapphire
contact window

micro-LED arrays with smaller mesa sizes perform much better, p-metal PV
ITO p-metal
having a higher EQE than the large micro-LEDs. This supe- Mesa
n-GaN
n-metal

n-GaN
riority increases sharply as the excitation pulse duration de- (f) u-GaN
creases. When the pulse of excitation (2.75 V) is as short as 3D GaN
100 ns, the normalized luminance of the 7 × 7 μm2 micro- sapphire

LEDs has a >20% increase than the DC case. The mechanism Fig. 1. Process flow of the 5-step lithography for the micro-LED
of this phenomenon was discussed in the context of displays. The left side is the side view and the right side is the top view.
carrier injection/recombination and trap filling dynamics. We The figure is not drawn to scale. (a) The initial state of the epi-wafer,
concluded that the efficiency reduction induced by sidewall de- (b) the current spreading region and mesa fabrication, (c) the pixel
fects, especially for the smaller micro-LEDs, can be reduced isolation, (d) the n-contact metal deposition, (e) the p-contact window
while operating in small time duration in the order of 100 ns, formation, and (f ) the p-contact metal deposition.
which is very promising and exciting. The findings in this
research are of great significance for guiding the design of
micro-LEDs and their drivers working in practical arrays
and high-frequency conditions for future micro-LED displays. n-electrodes (2200-nm-thick Cr/Al/Ti/Pt/Au) were patterned
on the samples with a standard lift-off process using negative
photoresist, followed by 10 min annealing at 250°C in N2 , and
2. EXPERIMENT consequently, column lines of the PM structure were formed.
Four-inch commercial sapphire-based blue GaN LED epitaxial In the fourth step, contact windows for p-type injection were
wafers were adopted to fabricate our devices. On 600-μm-thick opened with ICP-RIE after the deposition of a passivation (PV)
patterned sapphire substrates (PSSs), the epi-layers were grown layer (1000 nm SiO2 ). In the fifth step, p-electrodes were pat-
by using metal organic chemical vapor deposition (MOCVD) terned with the same metal lift-off and annealing as before.
and primarily consisted of a 2-μm-thick 3D-GaN, a 2-μm- Row lines and finger pads for further flexible printed circuit
thick u-GaN, and a 4.5-μm-thick multilayer composed of (FPC) bonding were thus formed. Finally, the sapphire sub-
n-GaN, multi-function layers, multi-quantum-wells (MQWs), strate was thinned to 300 μm with chemical mechanical polish-
and p-GaN. ing (CMP), and the wafer was diced into six dice. Each die,
As shown in Fig. 1, with a 5-step lithography process, comprising a micro-LED array with a specific mesa dimension,
bottom-emitting monolithic micro-LED arrays driven by a pas- was connected to a pre-developed peripheral circuit through
sive matrix were fabricated. We first deposited a 75-nm-thick an FPC.
indium tin oxide (ITO) layer on the p-GaN. In the first step, Details of the as-fabricated micro-LED device can be seen
we defined both the ITO current spreading region and the from the optical micrographs illustrated in Fig. 2. All micro-
GaN mesa altogether (self-alignment) by a 2-step etching pro- LEDs located in the same column share a common cathode,
cess composed of ITO wet etching in a FeCl3 and HCl mix- while those in the same row share a common anode. Hence,
ture, followed by inductively coupled plasma reactive ion each micro-LED is addressable by providing a forward voltage
etching (ICP-RIE) of GaN using the same photoresist mask through its corresponding line and row [24]. To reduce the
(PM). In the second step, ICP-RIE was used again for a deeper interconnect resistance, thick multi-layer metals were em-
etching [Fig. 1(c)] to isolate all the pixels. In the third step, ployed. In that case, metal wire breaking is prone to occur
 Research Article Vol. 11, No. 4 / April 2023 / Photonics Research 551

The measurements of the current-voltage-luminance

(I-V-L) characteristics under the DC operation were first con-
ducted to get a fundamental understanding of the as-fabricated
micro-LED devices. Based on both statistical and photometri-
cal measurement considerations, a cluster of 6 × 6 pixel arrays
from each type of micro-LED devices was selected for measure-
ment [shown in Fig. 4(d), where the white dotted box refers to
the entire micro-LED array]. A synchronous I-V-L test was car-
ried out with a multifunctional integration luminescent device
test system (D3000-16CH) composed of a high-voltage source-
measurement unit and a TOPCON integrated spectroradiom-
eter SR-3A.
Typical diode characteristics with mesa size dependence are
seen in Fig. 4(a). Micro-LEDs with a larger mesa size generate a
Fig. 2. Microscopic photograph of a 7 × 7 μm2 micro-LED array higher current under the same voltage, yet micro-LED arrays
with one pixel turned on and an enlarged view of an active pixel picked with smaller mesa dimensions withstand higher current den-
out from the micro-LED arrays with mesas ranging from 7 × 7 μm2 to
sities. Indeed, with a higher surface-to-volume ratio, a smaller
100 × 100 μm2 .
micro-LED is recognized to have less current crowding as well
as better heat dissipation capability [30,31].
Figure 4(b) plots the luminance characteristic curves as a
at the row and column crossing points. Thus, the semiconduc- function of the injection current over different micro-LED ar-
tor right beneath the row lines was preserved from the etching rays. As the insert shows, a measurement field with a diameter
in order to lift up the rows so that they match the height of the of 1000 μm, which is several times larger than the pixel pitch of
columns better. In Fig. 2, the row lines appear seemingly dis- the micro-LED arrays (i.e., 300 μm), was applied to collect the
crete, and the column metals look much darker, as they are of a light from the selected 6 × 6 pixel array with a 0.2° solid angle
lower altitude (deposited directly on the PSS). This intercon- emission upon the sapphire side. Consequently, spectrum in-
nect layout design reduces the altitude difference between the formation was obtained, based on which the luminance as well
row and column metal lines, effectively reducing the risk as the EQE was calculated. With the largest emitting area, the
of broken metal lines and the insulation layer (SiO2 ) at the 100 × 100 μm2 micro-LED array exhibits the highest lumi-
intersections. nance of nearly 1,000,000 cd∕m2. The 7 × 7 μm2 micro-LED
array exhibits a peak luminance of only slightly more than
20,000 cd∕m2 . This notwithstanding, a luminance of
3. RESULTS AND DISCUSSION ∼3,000 cd∕m2 at the maximum efficiency of the 7 × 7 μm2
A matrix of 48 × 48 micro-LEDs was built for display study, as device has been observed, much higher than the luminance
shown in Fig. 3. The as-fabricated micro-LED arrays are avail- (at maximal EQE) reported earlier [24–26]. Because the area
able to display both characters and pictures pixel by pixel after used to calculate luminance includes the region between the
being connected to a pre-developed peripheral circuit. A micro- pixels, the luminance can be increased just by reducing the pixel
LED array with a mesa size of 50 μm displaying the Chinese pitch. For device study, it is more convenient to have the lu-
characters “Fuzhou University” and the English characters minance calculated only from the emitting regions, which we
“Micro LED” is shown in Fig. 3. define as “brightness” in this paper, just to remove the influence
from different aperture ratios between the emitting areas and
pixel areas. The single micro-LED brightness as a function of
mesa size is shown in the insert semi-log plot in Fig. 4(b). At
low current density (phase 1), where most carriers are injected
smoothly into the active region, the brightness of a small micro-
LED is as good as that of a large device. The evolving slopes are
also almost equally large. At high current density, curves of
smaller LEDs have a lower slope than the low current density
case. (Note that large devices cannot withstand high current
densities due to current crowding, and hence they have no
data.) This is because the carrier dynamics is in the second
phase, where defects play a significant role. Smaller micro-
LEDs possess more sidewall related defects (in relative terms)
and higher non-radiative recombination rates, and thus the
slopes decline to a greater extent. As the current density is in-
Fig. 3. Images of the 50 μm micro-LED array (48 × 48) in oper- creased even further (phase 3) where most defect related traps
ation. Both Chinese characters and English characters are shown in are filled, an approximately linear increase in the amount of the
the display. output photons would occur.
 552 Vol. 11, No. 4 / April 2023 / Photonics Research Research Article

Fig. 4. I-V-L characterization of the micro-LED arrays based on the measurement of a clustered 6 × 6 pixel array. (a) The I-V characteristics of the
micro-LED arrays. The upper inset shows the photograph of the micro-LED arrays in full operation under 3 V bias. (b) The luminance as a function
of current density of the micro-LED arrays. The upper inset shows the photos taken during the measurement. The lower inset plots the relation
between the luminance of one single LED pixel (in different sizes) and the current density. (c) The EQE as a function of current density of the micro-
LED arrays. The inset is the same curves plotted in a semi-log scale. (d) The selected cluster of 6 × 6 pixel array in each micro-LED device as the data
source.

An EQE analysis helps us get a deeper insight. The EQE is luminosity function provided by the Commission International
defined as de l’Eclairage chromaticity (CIE), respectively.
np The EQE as a function of the current density is plotted in
EQE
· 100 %, (1) Fig. 4(c) based on the formula above. Consistent with the car-
ne
rier dynamics discussed earlier, the EQEs first go up and then
where np and ne are the rate of emitted photons and injected down. Maxima of the EQE can be found from the insert log-
electrons, respectively. Assuming a non-Lambert reflector of the arithmic scale plot, where larger micro-LEDs exhibit higher
light shape distribution emitted by the micro-LED array, with maxima at lower current densities. This can be explained by
luminance LV cd∕m2 , Eq. (1) can be calculated as the different weights between the non-radiative Shockley–
R 780 Read–Hall (SRH) recombination rate and the radiative recom-
π · L V · n · p2 · e λ · F λdλ
· R 780380
bination rate that resulted from the variation of surface-to-
EQE
, (2)
h · c · i · 683 lm∕W V λ · F λdλ volume ratio over different micro-LED arrays. The 100 ×
380
100 μm2 micro-LED array, with minimum sidewall defects
where n, p, e, h, c, i, λ, F λ, and V λ are the active pixels in relative term, naturally has the highest maximum EQE value
number, the pixel pitch, the charge of an electron, the Planck’s (∼15% at 31.6 A∕cm2 current density and 187;900 cd∕m2
constant, the velocity of light, the forward current, the wave- luminance). The 7 × 7 μm2 micro-LED array exhibits a maxi-
length, the relative electroluminescence (EL) intensity, and the mum EQE of 12.13% at 278.6 A∕cm2 and 3465 cd∕m2 .
 Research Article Vol. 11, No. 4 / April 2023 / Photonics Research 553

Such a small drop in EQE is acceptable, as future optimization brightness faster than the larger ones. Table 1 lists the rise time
based on better heat management along with higher light ex- of the micro-LED arrays. Micro-LED arrays with mesa sizes
traction efficiency (LEE) in small LEDs may compensate for smaller than 50 μm exhibit a significantly faster response under
this drop [30,41,42]. After peak EQEs are reached in Fig. 4(c), the pulsed excitations.
all curves exhibit a dramatic decline and finally overlap to- The micro-LED photon intensity response time is mainly
gether. At this stage, the radiative recombination continuously affected by two factors: the carrier lifetime and the resistance-
increases with the current density, and the non-radiative SRH capacitance (RC) constant. The latter can be estimated by
recombination centers are mostly saturated, but another factor, ρ·d ε·s ε·E ε·V
i.e., the current crowding, starts to be serious and dominates R·C
·

, (3)
s d J J ·d
the performance. The curve down bending occurs earlier for
larger micro-LEDs, as they are more prone to current crowding. where ρ, d , s, ε, E, and J are the resistivity, thickness, area,
As mentioned previously, micro-LED displays typically op- dielectric constant, electric field, and current density across
erate in a pulse mode. We then investigate the transient behav- the micro-LED active region, respectively. As illustrated in
iors of micro-LED arrays under pulsed excitations. As the Fig. 4(a), the upper inset, smaller micro-LEDs can provide
schematic diagram in Fig. 5(a) shows, our 6 × 6 micro-LED higher current densities. Calculations based on Eq. (3) indicate
pixel array device connected in a series with a 10 Ω resistor that the RC constants of the arrays are in the order of ns for the
was driven by a pulse voltage from a signal generator, where a micro-LED size under 100 μm, with the applied voltage of
high-sensitivity photomultiplier tube (PMT) was employed to 2.75–3.5 V and current density in the range of a few 10s to
measure the emitting light (EL) intensity and investigate the 100s A∕cm2 . Therefore, the measured rise time is mainly de-
temporal response characteristics. A relatively long-duration- termined by the carrier lifetime. As the measured t R also in-
pulsed signal (i.e., 100 kHz frequency, 50% duty cycle) was cludes the intrinsic RC time constant of the instrument, the
applied in Fig. 5(c). The electroluminescence (EL) intensity actual response time should be shorter than the measured data
from the micro-LED array measured by the PMT and the volt- shown in Fig. 5.
age on the series resistor were both collected by a multichannel Interestingly, a closer look at Figs. 5(d)–5(f ) reveals over-
oscilloscope. shoots in the normalized micro-LED EL intensity. We found
Generally, the response of a micro-LED to a pulse voltage that the overshoots only occurred when the current density ex-
excitation can be divided into three steps: rising, stabilizing, and ceeds a specific value. In this situation, the normalized EL in-
falling [Fig. 5(b)]. The rise time (t R ) is defined by the time tensity reached its maximum, followed by a small oscillation
interval between the 10% and 90% points of the stable signal. until a stable value was reached. The difference between the
As shown in Figs. 5(d)–5(f ), where transient EL intensity overshoot peak and the value of the steady state was larger
curves over different micro-LED arrays were normalized for in the smaller micro-LEDs. Consequently, it can be concluded
a better comparison, smaller micro-LEDs reach the peak that, for micro-LEDs, the injection and recombination

Fig. 5. (a) Schematic diagram of the transient characteristic measurement system. (b) The typical response of a micro-LED to a pulse excitation.
(c) The 100 kHz pulse signal with 50% duty cycle and voltage varied from 2.75 to 3.25 V. (d)–(f ) The normalized EL intensity of the micro-LED
arrays as a function of time under 100 kHz pulse operation with applied voltage from 2.75 to 3.25 V. (The 100 × 100 μm2 micro-LED array under
3.25 V bias excitation is undermeasured due to the output limit of the pulse signal generator.)
 554 Vol. 11, No. 4 / April 2023 / Photonics Research Research Article

Table 1. Response Speed of the Micro-LED Arrays to those at 100 kHz in Fig. 6, and the EL intensity curves with
longer pulse durations are the natural extension of the short
t R (ns)
duration ones with no additional features. In Fig. 6, after
Mesa Size 2.75 V 3.00 V 3.25 V the initial overshoot, the EL intensity eventually reaches a stable
7 × 7 μm 2
35 32.8 32 value, i.e., the DC baseline, and finally gets down to zero when
10 × 10 μm2 40.6 39.5 37.8 the input pulse ends. For 10% and 20% duty ratios, there is
20 × 20 μm2 41.8 44 43.9 simply no time for the voltage to stabilize at the baseline plateau
30 × 30 μm2 65 51.8 48 before the signal disappears. For the three voltages tested in the
50 × 50 μm2 136.5 65 52.5
100 × 100 μm2 173.5 125
7 μm devices, the smaller the voltage, the lower the DC output
baseline. However, the main message of Fig. 6 is that, as long as
the duty cycle is short enough, the baseline does not matter
processes are both size- and time-dependent. In the initial stage anymore, since the overshoot dominates. Thus, smaller
of the injection, carriers access the active area smoothly. micro-LEDs possess a high potential for operating at applied
The photons generated by the MQW reach a maximum voltage pulses with very short time durations.
number after the rising period, seen as the overshoot peak On the above basis, a luminance loss mechanism under
in Figs. 5(d)–5(f ). Soon after, the non-radiative recombination high-frequency-pulsed excitation is deduced, i.e., the longer
that is mainly due to the defective sidewalls also occurs, and the the rise time, the greater the luminance loss compared with
space-charge field created by those carriers trapped in the de- the DC case. Moreover, it can be well expected that, when fur-
fects will hinder the subsequent carrier injection flow by virtue ther reducing the pulse width, where the overshoot is large
of the Coulombic repulsive force. As a result, the light outputs enough to overcompensate the luminance loss in the rising
of the micro-LED array exhibit a decline. Eventually, in the and falling period, an enhanced performance of the micro-
third phase, the forward field and the defect-induced space- LED arrays would be observed. Figure 7 plots the normalized
charge field reach an equilibrium state. A steady current flows luminance as a function of the current density, which was cal-
across the device, and the EL intensity reaches a stable value. culated by dividing the pulse luminance (LV pulse ) by the DC
This mechanism is related to the sidewall defects and is more luminance (LV DC ),
pronounced when the micro-LED size gets smaller. With a bet- LV pulse
ter current spreading and a higher surface-to-volume ratio, car- Normalized Luminance
: (4)
LV DC
riers in the smaller micro-LEDs are more likely to reach the
mesa edge and finally be captured by the sidewall defects.
Such a process does not reach equilibrium instantaneously, Theoretically, the normalized luminance defined in this way
meaning it requires time (∼100 ns based on our results) for should equal the duty cycle. Experiment results agree well with
the electrons to be captured by the defect traps. the analyses. For pulsed frequency ≤100 kHz, where lumi-
Therefore, the 7 μm micro-LEDs, with the maximal surface- nance loss induced by the rising and falling period is negligible,
to-volume ratio, exhibit the maximal difference between the all micro-LED arrays exhibit a normalized luminance close to
overshoot peak and the stable value. In addition, the overshoots the theoretical value. Nevertheless, at high frequency, the nor-
occur at an initial 100 ns after applying voltage on the 7 μm malized luminance falls below the theoretical values, a trend
micro-LED arrays. As a result, an applied voltage pulse with an that is more pronounced with the mesas increasing from
order of 100 ns can operate mostly in the overshoot part of the 10 × 10 μm2 to 100 × 100 μm2 . This gap widens further at
transient response of the micro-LED arrays, which can signifi- the decreasing duty cycle and forward voltage. Exceptionally,
cantly affect the performance of the micro-LED arrays in a pos- the 7 × 7 μm2 micro-LED array does not have this gap, which
itive manner. The EL intensity of the 7 × 7 μm2 device array, as even exhibits a normalized luminance higher than the theoreti-
a function of time under the 1 MHz pulse operation, is plotted cal value under 2.75 V. In these cases, overshoots dominate the
in Fig. 6. The EL intensity under the DC operation was also performance of the 7 × 7 μm2 micro-LED array, as illustrated
measured and used as a baseline value (see arrows in Fig. 6). We in Fig. 6. Efficiency drop induced by the sidewall defects has
find that the transient behavior at 1 MHz is qualitatively similar not fully taken the effect.

Fig. 6. EL intensity of the 7 × 7 μm2 device array as a function of time under 1 MHz pulse operation. (a) 2.75 V, (b) 3.00 V, and (c) 3.25 V.
 Research Article Vol. 11, No. 4 / April 2023 / Photonics Research 555

Normalized Luminance 1.0

Theoretical Value
7 μm 2.75 V 100 kHz
1.0
Theoretical Value
7 μm 2.75 V 1 MHz the short-pulse operation mode, such brightness boosting effect

Normalized Luminance
0.8 10 μm
20 μm
0.8 10 μm
20 μm
can be applied in a real display. This can be achieved by using
30 μm 30 μm
0.6
50 μm
0.6
50 μm scrambled pulse width modulation (S-PWM) or dispersed
100 μm 100 μm
0.4 0.4 PWM techniques [43], where a single original pulse in each
0.2 0.2 subfield is split into several sub-pulses with the same total time
0.0
10 20 50 75
0.0
10 20 50 75
of applied voltage [as shown in Fig. 9(a)]. For a conventional
1.0
duty cycle (%)
1.0
duty cycle (%) LED display, the luminance is determined by the time span of
Theoretical Value Theoretical Value
7 μm 3.00 V 100 kHz 7 μm 3.00 V 1 MHz the applied voltage. For the short-pulse operation mode, how-
Normalized Luminance

Normalized Luminance
0.8 10 μm 0.8 10 μm

0.6
20 μm
30 μm 0.6
20 μm
30 μm
ever, the luminance is determined by the number of short
50 μm 50 μm

0.4
100 μm
0.4
100 μm pulses. Each pulse in the order of 100 ns delivers a constant
pulse of light. The number of pulses in a time frame,
0.2 0.2
e.g., 16.7 ms for 60 Hz display, will determine the luminance
0.0 0.0
10 20 50 75 10 20 50 75 level (or gray scale) of a display.
duty cycle (%) duty cycle (%)
1.0
Theoretical Value
1.0
Theoretical Value Based on the experiment results of our micro-LEDs, the
3.25 V 100 kHz 3.25 V 1 MHz
peaked brightness at the initial ∼100 ns shows a higher value
7 μm 7 μm
Normalized Luminance

Normalized Luminance

0.8 10 μm 0.8 10 μm
20 μm 20 μm
0.6 30 μm
50 μm
0.6 30 μm
50 μm
compared to the steady state. Therefore, it can be well expected
0.4 0.4 that, by taking advantage of the overshoot mechanism, the total
0.2 0.2 turn-on voltage time for achieving the same level of brightness
0.0 0.0 at short pulse operation mode is less than the conventional op-
10 20 50 75 10 20 50 75
duty cycle (%) duty cycle (%) eration mode by using continuous voltage. Consequently, lower
Fig. 7. Normalized luminance as a function of the duty cycle with power consumption can be expected since a smaller total
various forward voltages and frequencies. duty ratio of the voltage pulse would be applied. As shown in
Fig. 9(b), the relative luminance of the 7 μm micro-LED gen-
erated within a 100 ns duration pulse can be estimated by in-
A similar scenario can be obtained in the EQE analysis at a tegrating the corresponding EL intensity curve. Thereby, under
high frequency. Assuming the same spectral distribution as the 2.75 V pulse excitation, a 25.5% luminous enhancement of the
DC case, the EQE under the pulse operation can then be esti- 100 ns short pulse is acquired, compared with that upon the
mated by Eq. (2). Figure 8 plots the EQE at 1 MHz and various steady state. Based on this argument, for a frame time T [as
duty cycles. Except in the 7 μm case, the EQE increases with illustrated in Fig. 9(c)], the short-pulse PWM mode can deliver
the duty cycle. Again, three input voltages are chosen (2.75 V, a 25.5% higher luminance, compared with the conventional
3.00 V, and 3.25 V), which are roughly on the left, middle, and PWM mode.
right of the maximum EQE point, according to the DC oper-
ation [Fig. 4(c)]. For all the three voltages, the 7 μm micro- 4. CONCLUSION
LED array is seen to outperform other devices for small duty We have studied both the size and temporal dependence of
cycles, especially at 10%. For instance, Fig. 8(a) shows an GaN micro-LED arrays. The luminance and EQE of micro-
EQE > 11% for all the duty ratios, whereas other micro- LED arrays are found to be varied by the mesa size ranging
LED arrays exhibit variable degrees of EQE decline upon from 7 to 100 μm. Under the DC operation, the luminance
the reduction of the duty cycle. At the 10% duty cycle, the and EQE decrease with the reduction of the mesa size. In con-
7 × 7 μm2 micro-LED array shows the highest EQE of trast, micro-LED arrays with smaller mesa dimensions are ob-
12.05%. On the other hand, the 100 μm devices are really served with higher luminance and EQE under a 1 MHz pulse
not suitable under this operation condition. At duty cycles operation and a short duty cycle. Reducing the pulse duration
<75%, it is far less efficient compared to the others. to 100 ns, the 7 μm devices demonstrate a 20% increase in
It is evident that the fast response speed along with the over- EQE compared to the DC case. A hypothesis based on carrier
shoot mechanism enables the 7 × 7 μm2 micro-LED array injection/recombination and trap filling dynamics is used to
(small-sized micro-LEDs) to operate at a high frequency explain the effect. The efficiency drop in small micro-LEDs
with a pulse width in the order of 100 ns. By employing may be drastically reduced by applying voltages of very short

Fig. 8. Estimated EQE under different pulse operations with 1 MHz frequency and various duty cycles. (a) 2.75 V, (b) 3.00 V, and (c) 3.25 V.
 556 Vol. 11, No. 4 / April 2023 / Photonics Research Research Article

Fig. 9. (a) Conventional PWM mode compared with short-pulse PWM mode. (b) The relative luminance of a 7 μm micro-LED generated within
a 100 ns duration. (c) The short-pulse mode combined with overshootings.

pulses, in the order of 100 ns. Small sized micro-LEDs thus REFERENCES
possess a very high potential for next-generation displays. 1. Q. Yan, “Emissive displays rise and fall–and rise again,” Inf. Disp. 32,
4–35 (2016).
Funding. Fujian Provincial Projects (2021HZ0114, 2. Z. Wang, X. Shan, X. Cui, and P. Tian, “Characteristics and tech-
2021J01583, 2021L3004); Fujian Science Technology niques of GaN-based micro-LEDs for application in next-generation
display,” J. Semicond. 41, 041606 (2020).
Innovation Laboratory for Optoelectronic Information of
3. Y. Huang, E.-L. Hsiang, M.-Y. Deng, and S.-T. Wu, “Mini-LED, micro-
China (2020ZZ110, 2021ZZ122); National Key Research LED and OLED displays: present status and future perspectives,”
and Development Program of China (2021YFB3600104). Light Sci. Appl. 9, 105 (2020).
4. T. Wu, C.-W. Sher, Y. Lin, C.-F. Lee, S. Liang, Y. Lu, S.-W. Huang
Acknowledgment. We also thank local government Chen, W. Guo, H.-C. Kuo, and Z. Chen, “Mini-LED and micro-LED:
funding for Rich Sense Electronics Technology Inc. promising candidates for the next generation display technology,”
Appl. Sci. 8, 1557 (2018).
5. W. Henry and C. Percival, “55-2: Invited Paper: ILED displays: Next
Disclosures. The authors declare no competing interests. generation display technology,” SID Int. Symp. Dig. Tech. Papers 47,
747–750 (2016).
Data Availability. All data supporting the findings of this 6. J. Y. Lin and H. X. Jiang, “Development of microLED,” Appl. Phys.
study are available within the paper. Lett. 116, 100502 (2020).
 Research Article Vol. 11, No. 4 / April 2023 / Photonics Research 557

7. Y. Huang, G. Tan, F. Gou, M.-C. Li, S.-L. Lee, and S.-T. Wu, 26. Y. Liu, K. Zhang, B.-R. Hyun, H. S. Kwok, and Z. Liu, “High-brightness
“Prospects and challenges of mini-LED and micro-LED displays,” InGaN/GaN micro-LEDs with secondary peak effect for displays,”
J. Soc. Inf. Disp. 27, 387–401 (2019). IEEE Electron Device Lett. 41, 1380–1383 (2020).
8. V. W. Lee, N. Twu, and I. Kymissis, “Micro-LED technologies and 27. E.-L. Hsiang, Z. He, Y. Huang, F. Gou, Y.-F. Lan, and S.-T. Wu,
applications,” Inf. Disp. 32, 16–23 (2016). “Improving the power efficiency of micro-LED displays with optimized
9. K. Ding, V. Avrutin, N. Izyumskaya, Ü. Özgür, and H. Morkoç, “Micro- LED chip sizes,” Crystals 10, 494 (2020).
LEDs, a manufacturability perspective,” Appl. Sci. 9, 1206 (2019). 28. J. Zhu, T. Takahashi, D. Ohori, K. Endo, S. Samukawa, M. Shimizu,
10. X. Zhou, P. Tian, C.-W. Sher, J. Wu, H. Liu, R. Liu, and H.-C. Kuo, and X.-L. Wang, “Near-complete elimination of size-dependent effi-
“Growth, transfer printing and colour conversion techniques towards ciency decrease in GaN micro-light-emitting diodes,” Phys. Status
full-colour micro-LED display,” Prog. Quantum Electron. 71, 100263 Solidi 216, 1900380 (2019).
(2020). 29. A. Daami, F. Olivier, L. Dupré, F. Henry, and F. Templier, “59-4:
11. W. Meng, F. Xu, and Z. Yu et al., “Three-dimensional monolithic Invited Paper: Electro-optical size-dependence investigation in GaN
micro-LED display driven by atomically thin transistor matrix,” Nat. micro-LED devices,” SID Int. Symp. Dig. Tech. Papers 49, 790–
Nanotechnol. 16, 1231–1236 (2021). 793 (2018).
12. X. Shi, Y. Zuo, P. Zhai, J. Shen, Y. Yang, Z. Gao, M. Liao, J. Wu, J. 30. Z. Gong, S. Jin, Y. Chen, J. McKendry, D. Massoubre, I. M. Watson, E.
Wang, X. Xu, Q. Tong, B. Zhang, B. Wang, X. Sun, L. Zhang, Q. Pei, Gu, and M. D. Dawson, “Size-dependent light output, spectral shift,
D. Jin, P. Chen, and H. Peng, “Large-area display textiles integrated and self-heating of 400 nm InGaN light-emitting diodes,” J. Appl.
with functional systems,” Nature 591, 240–245 (2021). Phys. 107, 013103 (2010).
13. K. Kishino, N. Sakakibara, K. Narita, and T. Oto, “Two-dimensional 31. F. Olivier, S. Tirano, L. Dupré, B. Aventurier, C. Largeron, and F.
multicolor (RGBY) integrated nanocolumn micro-LEDs as a funda- Templier, “Influence of size-reduction on the performances of GaN-
mental technology of micro-LED display,” Appl. Phys. Express 13, based micro-LEDs for display application,” J. Lumin. 191, 112–116
14003 (2019). (2017).
14. Z. Liu, C.-H. Lin, B.-R. Hyun, C.-W. Sher, Z. Lv, B. Luo, F. Jiang, T. 32. R.-H. Horng, C.-X. Ye, P.-W. Chen, D. Iida, K. Ohkawa, Y.-R. Wu, and
Wu, C.-H. Ho, H.-C. Kuo, and J.-H. He, “Micro-light-emitting diodes D.-S. Wuu, “Study on the effect of size on InGaN red micro-LEDs,”
with quantum dots in display technology,” Light Sci. Appl. 9, 83 (2020). Sci. Rep. 12, 1324 (2022).
15. Y. Weng, G. Chen, J. Nie, S. Que, S.-H. Song, Y. Yu, F. Zhang, H. Liu, 33. J.-H. Kim, S. Shin, K. Kang, C. Jung, Y. Jung, T. Shigeta, S.-Y. Park,
X. Zhou, Y. Zhang, J. Sun, J.-K. Song, C. Wu, T. Guo, and Q. Yan, H. S. Lee, J. Min, J. Oh, and Y.-S. Kim, “15-1: PWM pixel circuit with
“Hybrid device of blue GaN light-emitting diodes and organic light- LTPS TFTs for micro-LED displays,” SID Int. Symp. Dig. Tech. Papers
emitting diodes with color tunability for smart lighting sources,” 50, 192–195 (2019).
ACS Omega 7, 5502–5509 (2022). 34. C.-H. Chu, F. Wu, and S. Sun, “31.2: High PPI micro-LED display
16. Y. Wu, J. Ma, P. Su, L. Zhang, and B. Xia, “Full-color realization of based on PWM technology,” SID Int. Symp. Dig. Tech. Papers 49,
micro-LED displays,” Nanomaterials 10, 2482 (2020). 337–338 (2018).
17. F. Gou, E.-L. Hsiang, G. Tan, Y.-F. Lan, C.-Y. Tsai, and S.-T. Wu, 35. L. Svilainis, “LED PWM dimming linearity investigation,” Displays 29,
“High performance color-converted micro-LED displays,” J. Soc. 243–249 (2008).
Inf. Disp. 27, 199–206 (2019). 36. D. Gacio, J. M. Alonso, J. Garcia, L. Campa, M. J. Crespo, and M.
18. C.-M. Kang, J.-Y. Lee, D.-J. Kong, J.-P. Shim, S. Kim, S.-H. Mun, S.-Y. Rico-Secades, “PWM series dimming for slow-dynamics HPF LED
Choi, M.-D. Park, J. Kim, and D.-S. Lee, “Hybrid full-color inorganic drivers: the high-frequency approach,” IEEE Trans. Ind. Electron.
light-emitting diodes integrated on a single wafer using selective area 59, 1717–1727 (2012).
growth and adhesive bonding,” ACS Photon. 5, 4413–4422 (2018). 37. Q. Yan, “Micro-LED driving the third wave of technology revolution of
19. Z. Pan, C. Guo, X. Wang, J. Liu, R. Cao, Y. Gong, J. Wang, N. Liu, Z. information display,” in Proceedings of the International Display
Chen, L. Wang, M. Ishikawa, and Z. Gong, “Wafer-scale micro-LEDs Workshops (2018), Vol. 25, pp. 2–3.
transferred onto an adhesive film for planar and flexible displays,” Adv. 38. Y.-H. Chang, T.-C. Hsu, F.-J. Liou, C.-W. Chow, Y. Liu, H.-C. Kuo,
Mater. Technol. 5, 2000549 (2020). C.-H. Yeh, and P.-H. Yang, “High-bandwidth InGaN/GaN semipolar
20. J. Xiong, E.-L. Hsiang, Z. He, T. Zhan, and S.-T. Wu, “Augmented micro-LED acting as a fast photodetector for visible light communi-
reality and virtual reality displays: emerging technologies and future cations,” Opt. Express 29, 37245–37252 (2021).
perspectives,” Light Sci. Appl. 10, 216 (2021). 39. Y. Cai, J. I. H. Haggar, C. Zhu, P. Feng, J. Bai, and T. Wang, “Direct
21. F. Gou, E.-L. Hsiang, G. Tan, P.-T. Chou, Y.-L. Li, Y.-F. Lan, and S.-T. epitaxial approach to achieve a monolithic on-chip integration of a
Wu, “Angular color shift of micro-LED displays,” Opt. Express 27, HEMT and a single micro-LED with a high-modulation bandwidth,”
A746–A757 (2019). ACS Appl. Electron. Mater. 3, 445–450 (2021).
22. S. X. Jin, J. Li, J. Z. Li, J. Y. Lin, and H. X. Jiang, “GaN microdisk light 40. Y. Huang, Z. Guo, X. Wang, H. Li, and D. Xiang, “GaN-based high-
emitting diodes,” Appl. Phys. Lett. 76, 631–633 (2000). response frequency and high-optical power matrix micro-LED for vis-
23. K. Behrman and I. Kymissis, “Enhanced microLED efficiency via stra- ible light communication,” IEEE Electron Device Lett. 41, 1536–1539
tegic pGaN contact geometries,” Opt. Express 29, 14841–14852 (2020).
(2021). 41. S. Lan, H. Wan, J. Zhao, and S. Zhou, “Light extraction analysis of
24. S. Bornemann, J. Gülink, V. Moro, J. C. Gil, S. Wolter, G. Schöttler, D. AlGaInP based red and GaN based blue/green flip-chip micro-
Bezshlyakh, J. Daniel Prades, A. Dieguez, and A. Waag, “Processing LEDs using the monte carlo ray tracing method,” Micromachines
and characterization of monolithic passive-matrix GaN-based 10, 860 (2019).
microLED arrays with pixel sizes from 5 to 50 μm,” IEEE Photon. 42. H.-Y. Ryu, J. Pyo, and H. Y. Ryu, “Light extraction efficiency of GaN-
J. 13, 8200209 (2021). based micro-scale light-emitting diodes investigated using finite-
25. Y. Liu, K. Zhang, F. Feng, K.-W. Chan, S.-Y. Yeung, H.-S. Kwok, and difference time-domain simulation,” IEEE Photon. J. 12, 1600110
Z. Liu, “The size and temperature effect of ideality factor in GaN/ (2020).
InGaN multiple quantum wells micro-light-emitting diodes,” J. Soc. 43. http://www.micros.com.pl/mediaserver/info-uimbi5031gts.pdf.
Inf. Disp. 29, 948–960 (2021).