3.5 × 3.5 μm GaN blue micro-light-emitting diodes with negligible sidewall surface nonradiative recombination

Abstract

Micro-light-emitting diode displays are generating considerable interest as a promising technology for augmented-reality glasses. However, the fabrication of highly efficient and ultra-small ( <3 μm) micro-light-emitting diodes, which are required for augmented-reality applications, remains a major technical challenge due to the presence of strong sidewall nonradiative recombination. In this study, we demonstrate a 3.5 × 3.5 μm² blue GaN micro-light-emitting diode with negligible sidewall nonradiative recombination compared with bulk nonradiative recombination. We achieve this by using an ultralow-damage dry etching technique, known as neutral beam etching, to create the micro-light-emitting diode mesa. Our 3.5 × 3.5 μm² micro-light-emitting diode exhibits a low decrease in external quantum efficiency of only 26% at a current density of 0.01 A/cm², compared with the maximum external quantum efficiency that is reached at the current density of ∼3 A/cm². Our findings represent a significant step towards realizing micro-light-emitting diode displays for augmented-reality glasses.

Full text

Article https://doi.org/10.1038/s41467-023-43472-z
3.5 × 3.5 μm2GaN blue micro-light-emitting
diodes with negligible sidewall surface
nonradiative recombination
Xuelun Wang
1,2,3
, Xixi Zhao
2
, Tokio Takahashi
2
, Daisuke Ohori
4
&
Seiji Samukawa
4,5
Micro-light-emitting diode displays are generating considerable interest as a
promising technology for augmented-reality glasses. However, the fabrication
of highly efficient and ultra-small ( <3 μm) micro-light-emitting diodes, which
are required for augmented-reality applications, remains a major technical
challengeduetothepresenceofstrongsidewall nonradiative recombination.
In this study, we demonstrate a 3.5 × 3.5 μm2blue GaN micro-light-emitting
diode with negligible sidewall nonradiative recombination compared with
bulk nonradiative recombination. We achieve this by using an ultralow-
damage dry etching technique, known as neutral beam etching, to create the
micro-light-emitting diode mesa. Our 3.5 × 3.5 μm2micro-light-emitting diode
exhibits a low decrease in external quantum efficiency of only 26% at a current
density of 0.01 A/cm2, compared with the maximum external quantum effi-
ciency that is reached at the current density of ∼3A/cm
2.Ourfindings repre-
sent a significant step towards realizing micro-light-emitting diode displays for
augmented-reality glasses.
Microdisplays using semiconductor micro-light-emitting diodes
(micro-LEDs) as light emitters are considered to be ideal candidates
for next-generation virtual reality (VR) and augmented reality (AR)
smart glasses that require high-resolution, high-luminance, and
energy-efficient displays1. However, the near-eye operating environ-
ment of VR/AR microdisplays imposes strict limitations on the size
and power consumption of the micro-LEDs. For AR displays, a reso-
lution of at least 4000 pixels per inch (ppi) is required to achieve a
high-quality display, which corresponds to a pixel pitch of approxi-
mately 6 μm2–4. To achieve this resolution, the maximum size of the
micro-LEDs for each primary color should be reduced to below 3 μm.
Heat generation during the operation of AR microdisplays is not
allowed because the display is located very close to the human eye.
Unfortunately, approximately only 50% of the injected electrical
power is converted into light, even in state-of-the-art red, green, and
blue semiconductor LEDs5. The remaining 50% of electrical power is
converted into heat, which increases the temperature of the LED
package without an efficient thermal management design6. However,
it would be impractical to implement an active thermal dissipation
design in AR glasses owing to the light-weight requirement. The most
effective way to suppress heat generation is to reduce the operation
current density of micro-LEDs. A simple estimation of a 4000-ppi
microdisplay using GaN micro-LEDs suggests that the temperature
increase can be suppressed to approximately 20 °C by reducing the
operating current density to the order of 1A/cm2(Supplemen-
tary Fig. 2).
The fabrication of highly efficient sub-3-μmmicro-LEDsinthe
<1 A/cm2current density region is a significant challenge because of
Received: 13 May 2023
Accepted: 10 November 2023
Check for updates
1
GaN Advanced Device Open Innovation Laboratory, National Institute of Advanced Industrial Science and Technology (AIST), Furo-cho, Chikusa-ku, Nagoya,
Japan.
2
Research Institute for Advanced Electronics and Photonics, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, Japan.
3
Institute of Materials and Systems for Sustainability, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, Japan.
4
Institute of Fluid Science, Tohoku University,
Aoba-ku, Sendai, Japan.
5
Institute of Communications Engineering, College of Electrical and Computer Engineering, National Yang Ming Chiao Tung
University, Hsinchu, Taiwan. e-mail: xl.wang@aist.go.jp;seiji.samukawa.e2@nycu.edu.tw
Nature Communications | (2023) 14:7569 1
1234567890():,;
1234567890():,;
Content courtesy of Springer Nature, terms of use apply. Rights reserved

the presence of strong Shockley-Read-Hall nonradiative recombina-
tion arising from mesa sidewall defects. Typically, GaN micro-LEDs are
produced by inductively coupled plasma (ICP) etching of a planar LED
wafer, which introduces high-density defects acting as nonradiative
recombination centers to the sidewall surfaces of the micro-LED mesa
owing to ion bombardment and deep ultraviolet photon irradiation7,8.
As a result, strong sidewall surface nonradiative recombination occurs,
reducing the emission efficiency of micro-LEDs by decreasing the chip
size9,10, especially in sub-3-μm micro-LEDs where the micro-LED
dimension becomes comparable to the diffusion length of minority
carriers in GaN11,12.Thisefficiency reduction was particularly significant
in the current density region below 1 A/cm2, since the external quan-
tum efficiency (EQE) of a high-quality InGaN/GaN LED typically showsa
maximum at a current density of a few to tens of A/cm2and decreases
with decreasing current density owing to enhanced Shockley-Read-
Hall nonradiative recombination13.
Intrinsic surface defects such as dangling bonds in the outermost
surface layer also play an important role in sidewall surface non-
radiative recombination of GaN micro-LEDs. These defects act as
nonradiative recombination centers and cause pinning of the Fermi
level at the sidewall surface14,15, leading to the bending of the surface
band energy near the surface area and the accumulation of carriers at
the sidewall surface. This accumulation provides an additional channel
for nonradiative surface recombination. Jiang et al. conducted a the-
oretical investigation into the impact of intrinsic surface states on GaN
blue micro-LEDs grown on a c-plane sapphire substrate16. They found
that the presence of intrinsic surface states reduced the maximum
internal quantum efficiency (IQE) from 58% to 24% when the chip size
was decreased from 300 to 3 μm, even without plasma-etching-
induced surface damage.
Despite the considerable efforts to reduce sidewall surface
nonradiative recombination in GaN micro-LEDs smaller than 10 μm,
this problem persists. To address this issue, Wang et al. used neutral
beam etching (NBE), an ultralow-damage dry etching technique for
semiconductor materials17, to fabricate GaN blue micro-LEDs18,19.
Their 6 × 6 μm2GaN micro-LED displayed EQE vs. current density
characteristics similar to those of a 40 × 40 μm2device in the current
density region higher than 1 A/cm2, indicating a significant reduction
in dry-etching-induced nonradiative recombination. Ley et al.
reported the first 2 μm InGaN/GaN blue micro-LED fabricated by ICP
etching, where potassium hydroxide (KOH) treatment of mesa side-
walls was employed to remove plasma-induced defects, and the
sidewall surface was further passivated by a thin Al
2
O
3
layer depos-
ited by atomic layer deposition20. However, the EQE of the fabricated
micro-LED decreased to approximately 17% of its peak EQE at a cur-
rent density of 15 A/cm2when the current density was decreased to
0.1 A/cm2.
In this work, we successfully fabricate a 3.5 × 3.5 μm2GaN blue
micro-LED with negligible mesa sidewall nonradiative recombination
compared with bulk nonradiativerecombination of the epitaxial wafer,
using the NBE process. Our devices exhibit a low decrease in EQE
(approximately 26%) at a current density of 0.01A/cm2compared to
thepeakEQEobservedaround3A/cm
2. Analysis of the EQE char-
acteristics and measurement of the surface potential of the NBE-
etched sidewall surface using Kelvin force microscopy (KFM) reveals
that our devices effectively suppress nonradiative recombination
related to not only sidewall defects generated during mesa etching but
also intrinsic surface states.
Results
Fabrication of the 3.5 × 3.5 μm2micro-LED
The NBE system employs a unique carbon electrode featuring a high-
aspect-ratio aperture positioned between the ICP discharge chamber
and the etching chamber. As accelerated ions pass through the aper-
ture,chargeexchangewiththecarbonelectrodeefficiently neutralizes
them, resulting in a neutral beam with controlled kinetic energy
directed towards the sample surface for etching17,19. Moreover, the
carbon aperture blocks deep ultraviolet photons, allowing for ultralow
damage etching of semiconductor materials. Overall, the NBE process
is an effective means of providing semiconductor material etching
with minimal damage.
Commercial InGaN/GaN blue LED wafers grown on patterned
(0001) c-plane sapphire substrates using metal-organic vapor phase
epitaxy (MOVPE) were utilized as epitaxial materials. The active
region comprised 16 pairs of InGaN/GaN multiple quantum wells
(MQWs), emitting light at approximately 458 nm. Figure 1a illustrates
the schematic of the micro-LEDs fabricated in this study. Square-
shaped micro-LED mesas with sizes ranging from 3 to 20 μm were
etched by NBE, using SiO
2
deposited by plasma-enhanced chemical
vapor deposition (PECVD) asa mask. In addition, a series of reference
samples were fabricated using the conventional ICP process, fol-
lowed by a KOH solution treatment to remove the plasma-induced
damage layer. Subsequently, a 150-nm-thick SiO
2
film was deposited
on the sample surface by PECVD as an electrical isolation and surface
passivation layer. Next, a self-aligned Ni/Au p-type Ohmic contact
was prepared on the mesa top (Supplementary Fig. 3), and a Cr/Au
n-type contact was formed on the n-GaN surface (Supplementary
Fig. 3). The thickness of the entire layer was adjusted such that the
surface of the n-contact was at the same height as the p-contact. The
device process was completed by the deposition of an Au/Sn multi-
layer on both the p-type and n-type Ohmic contacts, which acted as a
eutectic bonding layer. Figure 1b, c depict 45° tilted scanning elec-
tron microscopy (SEM) images of a 3-μm micro-LED fabricated by the
NBE process and the same device where the SiO
2
passivation layer
was removed by wet etching using buffered hydrofluoric acid (BHF)
to expose the details of the micro-LED mesa, respectively. The micro-
LED mesa comprised four vertical sidewall surfaces with a very
smooth surface morphology. These surfaces are close to the non-
polar m- and a-planes of the GaN crystal based on the crystalline
orientation of the epitaxial wafer, but they may not represent the
exact m-anda-planes of GaN because of the limited precision of
device processes. To avoid misleading, we refer to these surfaces as
“m-plane-like surface”or “a-plane-like surface”hereafter. The depth
of the mesa is approximately 700nm. The presence of vertical
nonpolar-like sidewall surfaces indicates that the chemical reaction is
the dominant mechanism in the NBE etching of GaN. The mesa size
was approximately 3.5 × 3.5 μm2, slightly larger than the designed
size of 3 × 3 μm2. The size of the Au/Sn multilayer and Ni/Au p-contact
(not clearly revealed in Fig. 2c) was 2 × 2 μm2. Hereafter, we refer to
the device size by the actual size measured by SEM observation, e.g.,
a 3.5 × 3.5 μm2micro-LED.
After completing the device fabrication process, the micro-LED
chip wasdiced into a 1 × 1 mm2size and bonded to an Si submount with
an electrical injection circuit using the flip-chip eutectic bonding
technique (Supplementary Fig. 4). The light emission was extracted
from the sapphire substrate side and measured using a Si photodiode
calibrated to detectemission with a limit of 0.5 nW (Thorlabs, S130VC).
The Si photodiode was placed approximately 4 mm from the micro-
LED chip, which corresponds to a collection half angle of approxi-
mately 52°. The light emission under high current densities was also
measured using a 2-inch integrating sphere (Thorlabs, S142C) with a
detection limit of 1 μW to evaluate the absolute emission effi-
ciency value.
EQE analysis as a function of current density
Figure 2depicts the EQE measured as a function of current density
for four micro-LEDs with sizes ranging from 3.5 × 3.5 μm2to
20.5 × 20.5 μm2, which were fabricated using the NBE process. For
comparison, a 3 × 3 μm2sample etched using ICP and treated with a
KOH solution is also shown. The EQE was calculated using the
Article https://doi.org/10.1038/s41467-023-43472-z
Nature Communications | (2023) 14:7569 2
Content courtesy of Springer Nature, terms of use apply. Rights reserved

following equation:
EQE = Pλe
Ihc ð1Þ
where Pis the measured light output power, λis the peak light emis-
sion wavelength, Iis the injected current, cis the speed of light in a
vacuum, his the Planck constant, and eis the elementary charge. All
four EQE curves of the NBE samples exhibited a maximum at a current
density of approximately 3 A/cm2. The current density at which EQE
peaks is indicative of the nonradiative recombination velocity and can
be used as a figure of merit to evaluate LED efficiency20,21.Alower
current density at the peak EQE implies a lower nonradiative recom-
bination rate and thus, a higher EQE for devices with similar radiative
and Auger recombination rates. Therefore, the observation of a similar
current density at the peak EQE for all the four micro-LEDs fabricated
bytheNBEprocesssuggeststhatallthefourdeviceshaveasimilar
nonradiative recombination rate. Here, we assumed that allthe devices
have a similar Auger recombination rate. This is a reasonable
assumption because Auger recombination rates of GaN blue LEDs
have been shown to be nearly size-independent for LEDs smaller than
100 µm22. Moreover, the peak EQE was found to increase as the chip
size decreased. The peak EQE of the 3.5× 3.5 μm2chip was calculated
to be as high as 37.5% from the total emission power measured using
the integrating sphere. Although the differences in light-extraction
efficiency need to be considered, the aforementioned EQE value is
substantially higher than those reported for GaN blue micro-LEDs
with comparable dimensions20. However, the most notable finding
revealed in Fig. 2is that all NBE-etched devices exhibited very
slow decreases in EQEs with decreasing current density from the
current density at peak EQEs. We defined an efficiency droop as
ðEQEpeak EQE0:01A=cm2Þ=EQEpeak to quantitatively evaluate the effi-
ciency decrease in the low current density region, where EQE
peak
and
EQE0:01A=cm2are the peak EQE and the EQE at a current density of
0.01 A/cm2, respectively. The efficiency droop for the 3.5 × 3.5 μm2,
Fig. 2 | Current density dependence of EQE. EQE of micro-LEDs fabricated by the
NBE process with different sizes ranging from 3.5 × 3.5 μm2to 20.5 × 20.5 μm2as a
function of current density is shown. Here, a uniform current density over the
whole mesa area was assumed since the spacing between the edge of the metal
contact and the edge of the chip (approximately 0.75 μm, see Fig. 1c) is small
enough than the current spreading length (see Fig. 3c), and the chip size measured
from SEM observation was used to calculate thecurrent density. A 3 × 3 μm2device
fabricated by the ICP process and treated with a KOH solution was also given for
comparison. The solid up-pointing triangle (open up-pointing triangle), solid circle
(open circle), solid right-pointing triangle (open right-pointing triangle), and solid
diamond (open diamond) represent the EQEs calculated from the light emission
power measured using a photodiode (an integrating sphere) for the 3.5 ×3.5 μm2,
6.5 × 6.5 μm2,10.5×10.5μm2, and 20.5 × 20.5μm2micro-LEDsfabricated by theNBE
process, respectively. The solid and open down-pointing triangles represent the
EQEs calculated from the light emission power measured using a photodiode and
integrating sphere, respectively, for the 3 ×3 μm2micro-LED fabricated by the ICP
process.
a
b
c
3.7 Pm
Au/Sn
SiO2
Patterned sapphire substrate
n-GaN
SiO2
p-GaN
Ni/Au
InGaN/GaN
Multiple
quantum well
Au/Sn Au/Sn
Cr/Au
3.5 Pm
a-plane-like surface
m-plane-like surface
m-plane-like surface
2 Pm
Ni/Au
Au/Sn
[1100]
[1120]
Fig. 1 | Micro-LED structure. a Cross-sectional schematic illustration of the micro-
LEDs fabricated in this work. Thesapphire substrate was thinned from thebackside
to 20 0 μm and polished to a mirror surface. bA 45° tilted SEM image of a
3.5 × 3.5 μm2micro-LED fabricated by the NBE process is shown. Scalebar, 3 μm. cA
45° tilted SEM image of the same micro-LED after the SiO
2
passivation layer was
etched off by BHF solution. Scale bar, 3 μm.
Article https://doi.org/10.1038/s41467-023-43472-z
Nature Communications | (2023) 14:7569 3
Content courtesy of Springer Nature, terms of use apply. Rights reserved

6.5 × 6.5 μm2,10.5×10.5μm2,and20.5×20.5μm2micro-LEDs fabri-
cated by the NBE process were calculated from Fig. 2to be
approximately 26%,34.7%, 36.7%, and 37.2%, respectively. The findings
of this study reveal a surprising result that the efficiency droop
decreases with decreasing chip size. The 3.5 × 3.5 μm2micro-LED
exhibited the lowest efficiency droop of approximately 26%. This
observation contrasts with the behavior of conventional ICP-etched
GaN micro-LEDs, where the efficiency droop typically increases rapidly
with chip size reduction below 10 μm because of the enhanced
sidewall nonradiative recombination, even with KOH treatment9,10,23.
These results indicatethat sidewall nonradiative recombination, which
typically increases when the chip size decreases, was effectively
suppressed in the micro-LEDs fabricated using the NBE process to a
level that can be neglected relative to the bulk nonradiative
recombination of the epitaxial wafer. In other words, the emission
efficiency is only limited by the bulk nonradiative recombination
which should be size-independent. Considering a size-independent
nonradiative recombination rate (a similar IQE) in NBE-etched
micro-LEDs, the increase in peak EQEs when the chip size decreases,
can be explained by an increase in light-extraction efficiencies in
smaller micro-LEDs, which resulted from enhanced sidewall light
extraction20,23. On the other hand, the 3 × 3 μm2micro-LED fabricated
by the ICP process (Supplementary Fig. 5) exhibited a lower peak
EQE of approximately 28.5%, a higher current density at peak EQE
of approximately 9 A/cm2, and a much larger efficiency droop of
approximately 60% at a current density of 0.01 A/cm2compared to the
NBE-etched devices. These differences clearly indicate the existenceof
strong sidewall nonradiative recombination in the ICP-etched device,
and that KOH treatment cannot entirely remove sidewall nonradiative
recombination, consistent with previous reports12,20,23–25.
To further confirm the influence of sidewall nonradiative recom-
bination, we examined the intrinsic EQE of the epitaxial wafer by fab-
ricating a 200 × 200 μm2LED with a small (3 μm in diameter) p-contact
at the center, using the same processes as those employed for the
micro-LEDs in Fig. 2(Supplementary Fig. 6). The microscopic photo-
graph of the 200 × 200 μm2LED and the 30° tilted high-magnification
SEM image of the 3-μm diameter p-contact are shown in Fig. 3a, b,
respectively. In this case, the current is injected into the device
through the p-contact and spreads out from the p-contact with a dif-
fusion length L
s
,whichisdefined by the following equation:
t=ρLsrc+Ls
2

J0
e
nidealitykT
!
ln 1 + Ls
rc
 ð2Þ
where trepresents the thickness of the p-type layer, ρdenotes the
resistivity of the p-type layer, r
c
denotes the radius of the p-contact
metal, J
0
represents the current density at the edge of the p-contact,
which can be calculated from the injection current Iby I=πðrc+LsÞ2
(see inset of Fig. 3c), n
ideality
is the ideality factor of the LED, eis the
elementary charge, kis the Boltzmann constant, and Tis the
temperature26. Additionally, ρwas measured by Hall measurement
to be approximately 18 Ωcm. n
ideality
was determined to be approxi-
mately 2.18 from the I–V curve (Supplementary Figs. 7 and 8). The
radius of the Ni/Au p-contact was 1.5 μm(r
c
). We assumed that sidewall
nonradiative recombination was completely avoided if the current
spreading length was much shorter than the distance between the
p-contact and the mesa edge (approximately 100 μm). Consequently,
the measured EQE solely reflected the intrinsic emission efficiency of
the epitaxial wafer.
Figure 3c shows the current density dependence of the calculated
current spreading length and measured EQE. The EQE reached a
maximum at a current density of approximately 40A/cm², which was
one order of magnitude larger than that observed in Fig. 2.Thiswas
due to thefact that the current densitydecreases outside the p-contact
with increasing distance from the p-contact edge, as depicted in the
inset of Fig. 3c. When the current density under the p-contact is
increased to the value at peak EQE in Fig. 2, the local EQE underneath
the p-contact will reach a maximum. However, further current
p-GaN
Ni/Au
Au/Sn
SiO2
a
b
c
Au/Sn
p-contact
n-contact
LED mesa
Ls
rc
J0
10
-3
10
-2
10
-1
10
0
10
1
10
2
0
2
4
6
8
10
12
14
16
18
Current density (A/cm2)
External Quantum Efficiency (%)
0
5
10
15
20
25
Current spreading length Ls ( m)
Fig. 3 | 200 × 200 μm2LED with a smallp-contact. a A microscopic photograph of
the 200 × 200 μm2LED with a small Ni/Au p-contact (3 μm in diameter) to inves-
tigate the intrinsic EQE of the epitaxial wafer is shown. A stripe-shaped n-contact
surrounding the 200 μm mesa was employed. Scale bar, 100 μm. bA high-
magnification 30° tilted SEM image showing the details of the p-contact. Scale bar,
1μm. cMeasured EQE (the blackcurve) and calculated current spreading length L
s
(the red curve) as a function of current density. The inset illustrates schematically
the variation of current density along the radial direction, where J
0
,r
c
,andL
s
represent the current density at the edge of the p-contact, the radius of the p-
contact, and the current spreading length, respectively.
Article https://doi.org/10.1038/s41467-023-43472-z
Nature Communications | (2023) 14:7569 4
Content courtesy of Springer Nature, terms of use apply. Rights reserved

injection was needed for the position-averaged EQE to reach a max-
imum, as the EQE outside the p-contact decreased with distance from
the p-contact edge. As a rough approximation, we assumed that the
EQE vs. current density curve shifted by one order of magnitude
towards a higher current density with respect to those given in Fig. 2,
where the current density was spatially uniform. Therefore, the effi-
ciency droop at the current density of 0.1 A/cm2in Fig. 3cwhichwas
estimated to be approximately 55% are considered corresponding to
that at the current density of 0.01 A/cm2in Fig. 2. The current
spreading length at the current density of 0.1 A/cm2(approximately
6.3 μm) was much shorter than the distance between the p-contact and
the mesa edge (100 μm), so we assumed that the influence of sidewall
nonradiative recombination could be neglected, and that the above
efficiency droop represented the intrinsic EQE characteristics of the
epitaxial wafer.
Considering the uncertainties related to the non-uniform current
density distribution in the 200 × 200 μm2device, the above efficiency
droop is in reasonable agreement with that of the 20.5 × 20.5 μm2
micro-LED in Fig. 2. This suggests that the efficiency droops of the NBE-
etched micro-LEDs shown in Fig. 2were solely determined by non-
radiative recombination in the InGaN/GaN bulk layers, and that
sidewall-related nonradiative recombination did not give rise to any
further reduction in the emission efficiency. This was a surprising
conclusion, as nonradiative recombination related to intrinsic sidewall
surface states is generally believed to inevitably exist. Nonetheless, our
results suggest that both nonradiative recombination induced by
sidewall damage generated during mesa etching and that induced by
intrinsic sidewall surface states resulting from surface dangling bonds
in micro-LEDs fabricated by the NBE process were suppressed to a level
that made the influence of sidewall nonradiative recombination on
micro-LED efficiencies negligible compared to bulk nonradiative
recombination.
Surface potential measurement of NBE-etched sidewall surfaces
Intrinsic surface states induced by dangling bonds can cause surface
band bending, which can be characterized using KFM27,28.Inthisstudy,
we investigated the surface band bendingof the NBE-etchedsurface by
KFM, using the sample shown in Fig. 4a. The wafer was grown on a
(0001) c-plane n-type freestanding GaN substrate with a resistivity of
approximately 0.025 Ωcm, using MOVPE. The substrate was mis-
oriented 0.4° towards the [1-100] direction. The layer structure con-
sisted of 0.5-μm-thick Si-doped GaN layer (electron concentration of
approximately 3 × 1018 cm−3), 0.5-μm-thick unintentionally doped GaN
layer (n-type conductivity), a 5-period InGaN/GaN blue-emitting MQW
layer (total thickness of approximately 80 nm), and 0.5-μm-thick
unintentionally doped GaN layer. The sample was subsequently pro-
cessed into a 1-μm-pitched grating with equal line and space widths,
using photolithography and NBE etching (Fig. 4a). The etching depth is
approximately 0.9 μm. The sample was then cleaved along the line-
stripe direction, and KFM observations were performed on the
exposed m-plane-like surface etched by NBE. An as-cleaved sample
without a grating pattern was measured as a reference. Figure 4b
shows the measured surface potentials of the NBE-etched samples.
Distorted images appearing just beneath the etching bottom position
are attributed to dimples observed around the corner between the
vertical m-plane-like surface and the bottom c-plane in the SEM image
shown in Fig. 4a.
Although a uniform potential image was observed along the
grating stripe direction, the surface potential tended to increase
from the MQW position toward the sample edge. In Fig. 4c, we pre-
sent the surface potentials of the NBE-etched sample and the as-
cleaved reference sample as a function of the distance from the
sample edge. KFM measures the difference in contact potential
between the atomic force microscopy (AFM) probe tip and the GaN
surface. This difference can be expressed by (χ
tip
-χ
GaN
)−SBB −(E
C
-
E
F
), where χ
tip
is the electron affinity (or Fermi level) of the AFM tip,
χ
GaN
is the electron affinity of GaN, SBB is the surface band bending of
GaN, E
C
is the conduction band bottom of GaN, and E
F
is the Fermi
level of GaN27. Since χ
tip
,χ
GaN
,E
C
, and E
F
are the same for the NBE-
etched and reference samples, the measured potential difference in
Fig. 4creflects only the difference in the surface band bending
between the two samples. The m-plane GaN surface is typically
a
c
b
KFM observation
(m-plane-like surface)
Surface potential (mV) +1000
-1000
Si-doped GaN
Etching bottom
MQW
Sample edge
0.0 0.5 1.0 1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
Surface potential (V)
Distance from the edge (μm)
measurement
artifacts
measurement
artifacts
NBE etched sample
As-cleaved
sample
MQW
etching bottom
Fig. 4 | Surface potentialmeasurement by KFM.a A tilted SEM image of the 1-μm-
pitched grating for KFM observation is shown. The dashed line indicates the cleave
direction. Scale bar, 1 μm. bA typical measured surface potential image is shown.
The scan area was 1.5 μm×1.5μm. The MQW and Si-doped GaN positions were
markedbased on growth thickness. The etching bottom position was markedusing
the NBE etching depth obtained from SEM observation. Scale bar, 200 nm.
cSurface potential as a function of distance from the edge of the NBE-etched (the
red curve) and as-cleaved (the blue curve) reference samples. The small fluctua-
tions marked by dashed circles are due to measurement artifacts.
Article https://doi.org/10.1038/s41467-023-43472-z
Nature Communications | (2023) 14:7569 5
Content courtesy of Springer Nature, terms of use apply. Rights reserved

reported to show upward band bending (towards the vacuum level)
of the order of a few hundred meV14,28. In this study, we observed that
the NBE-etched sample had a slightly higher (0.3–0.4 V) surface
potential and upward surface band bending compared to the refer-
ence sample in the region between the etching bottom and MQW,
while the two samples exhibited similar surface potentialsaround the
MQW position. This suggests that the NBE-etched sample has a
similar or slightly larger number of surface states compared to the
reference sample. However, as the measurement position moved
towards the sample edge from the MQW position, the surface
potential started to increase, and we speculate that this could be due
to the adsorption of Cl on the etched surface, because this area was
exposed to the Cl neutral beam for a longer time29. The KFM results,
together with the EQE data presented in Fig. 2, indicate that the
surface states on the NBE-etched sidewall surface are not efficient
nonradiative recombination centers. Further investigation is neces-
sary to comprehend the unusual behavior of surface states in NBE-
etched GaN and InGaN. Nonetheless, we have demonstrated that it is
possible to fabricate InGaN/GaN micro-LEDs with negligible sidewall
surface nonradiative recombination.
Mechanism for the low efficiency droop in the 3.5 × 3.5 μm2
micro-LED
Finally, we would like to briefly discuss the mechanism behind the
observed decrease in efficiency droop with decreasing chip size in
micro-LEDs fabricated by the NBE process, as shown in Fig. 2. It is well
established that patterning InGaN/GaN MQWs into sub-micrometer-
sized nanostructures can lead to strong strain relaxation, which in
turn weakens the quantum-confined Stark effect and enhances
emission efficiency30–32. Strain relaxation is also present in InGaN/
GaN mesa structures with diameters of a few to tens of micrometers,
with relaxation mainly occurring near the mesa surface over a width
of a few hundred nanometers33,34. Therefore, we presumably attrib-
uted the low-efficiency droop observed in the 3.5 × 3.5 μm2micro-
LED to the strain relaxation effect occurring near the mesa surface.
To confirm the existence of strain relaxation, we performed cathode
luminescence (CL) mapping of the emission wavelength of the
3.5 × 3.5 μm2mesa fabricated by the NBE process. As shown in Fig. 5,a
blue shift of 1–2 nm in the emission wavelength can be clearly
observed near the m-plane-like sidewall surface over a width of a few
hundred nanometers. This experiment provides direct evidence for
the existence of strain relaxation near the mesa sidewall surface of
micro-LEDs studied in this work. However, observation of efficiency
increase induced by strain relaxation in the low current density
region is challenging in micro-LEDs withstrong sidewall nonradiative
recombination because the decrease in efficiency caused by sidewall
nonradiative recombination can easily offset the efficiency increase
induced by strain relaxation. The above discussion indicates again
that the sidewall surface nonradiative recombination in micro-LEDs
fabricated by the NBE process is negligible compared with the bulk
nonradiative recombination.
Discussion
In summary, we successfully demonstrated a 3.5 × 3.5 μm2InGaN/GaN
blue micro-LED with negligible sidewall surface nonradiative recom-
bination. This was achieved through an ultralow-damage NBE etching
process. The micro-LED exhibited an EQE droop as low as 26% at a
current density of 0.01 A/cm2, compared with the peak EQE observed
around 3 A/cm2.Thisefficiency droop is approximately 11% lower than
thatofa20.5×20.5μm2device. With these results, we anticipate that it
will be possible to reduce the chip size even further to meet the
requirements of AR smart glasses, because the mesa sidewall does not
give rise to any reduction in micro-LED efficiencies.
Methods
NBE experiments
For the Cl
2
-based etching process, the ICP source power was set to
800 W and modulated at a frequency of 10 kHz with a 50% duty ratio.
ACl
2
gas flow rate of 40 sccm was used as the etching gas, and the
pressure in the etching chamber was 0.1 Pa. To enhance the eva-
poration of the In-containing etching products, the sample stage
temperature was set to 130 °C. Additionally, a bias power of 6 W was
applied to the carbon aperture in order to control the kinetic energy
of the neutral beam. The etching rate achieved under these condi-
tions for the LED wafer used in this study was approximately 5 nm/
min, though it is worth noting that the InGaN well layer may exhibit
slower etching rates due to the low volatility of In-containing etching
products.
ICP experiments and KOH etching
In the ICP experiments, an ICP machine (RIE-400iPS, Samco Inc.) was
utilized. A mixture of Cl
2
and BCl
3
gases was used as the etching gas
with flow rates of 50 sccm and 6 sccm, respectively. The ICP and RF
bias powers were set to 150 W and 5 W, respectively. The etching
pressure and stage temperature were 1 Pa and 20 °C, respectively,
resulting in an etching rate of approximately 20nm/min for the LED
wafer used in this study. After ICP etching of the micro-LED mesa, the
samplewas treated with a 48% KOH solution at approximately 25 °C for
35 min to remove ICP-induced surface damage.
KFM measurements
KFM measurements were conducted using a Bruker NanoscopeV/
Dimension Icon Glovebox AFM in a high-purity argon gas atmosphere
at room temperature, where the residual concentrations of both water
and oxygen were approximately 0.1 ppm. An Si cantilever covered with
Pt/Ir was used as the probe. After cleaving the NBE-etched sample
along the grating stripe direction, SEM observation was performed to
find a surface area with minimal height differences around the etching
bottom. A bias was applied to adjust the surface potential in the
starting area (approximately 1.5 μm from the sample edge) to zero,
enabling a comparison between the two samples. As the a-plane GaN
surface is more difficult to cleave, KFM measurements were performed
only on the m-plane.
CL mapping
A CL mapping experiment was conducted using a Schottky-type field-
emission SEM machine (JEOL, JSM-7100F). An acceleration voltage of
444
445
446
447
448
[1120]
[1100]
m-plane-like surface
a-plane-like surface
Wavelength (nm)
Fig. 5 | CLmapping of the emission wavelength. Mapping the CL peakwavelength
of a 3.5 × 3 .5μm2micro-LED mesa fabricated by the NBE-process. Scale bar, 1 μm. A
purple-to-red color scale was used to represent the wavelength.
Article https://doi.org/10.1038/s41467-023-43472-z
Nature Communications | (2023) 14:7569 6
Content courtesy of Springer Nature, terms of use apply. Rights reserved

8 kV, giving rise to a penetration depth of approximately 450 nm in
GaN, was used. The beam current was set at 0.5 nA. An epitaxial wafer
with the same layer structure as that used for micro-LED fabrication
grown on a planar sapphire substrate was used because the patterned
substrate will cause scattering of light and lower the spatial resolution
of the mapping. The emission wavelength was measured to be
approximately 451 nm by photoluminescence excited by a 375-nm
laser (excitation power density: 7.5 W/cm2). An SiO
2
passivation layer
was not deposited on the sample surface to avoid the charge-up effect.
Data availability
The source data underlying all figures presented in the main manu-
script and Supplementary Information are provided in the Figshare
repository at https://doi.org/10.6084/m9.figshare.23961723.
References
1. Lin, J. Y. & Jiang, H. X. Development of microLED. Appl. Phys. Lett.
116, 100502 (2020).
2. Wierer, J. J. & Tansu, N. III-nitride micro-LEDs for efficient emissive
displays. Laser Photonics Rev. 13,1900141(2019).
3. Huang, Y., Hsiang, E.-L., Deng, M.-Y. & Wu, S.-T. Mini-LED, micro-LED
and OLED displays: present status and future perspectives. Light
Sci. Appl. 9, 105 (2020).
4. Park, J. et al. Electrically driven mid-submicrometre pixelation of
InGaN micro-light-emitting diode displays for augmented-reality
glasses. Nat. Photon. 15, 449 (2021).
5. Karpov, S. Y. Light-emitting diodes for solid-state lighting: search-
ing room for improvements. Proc. SPIE 9768,97680C(2016).
6. U.S. Department of Energy. Thermal Management of White LEDs.
https://www1.eere.energy.gov/buildings/publications/pdfs/ssl/
thermal_led_feb07_2.pdf (2007).
7. Shul, R. J. et al. Inductively coupled plasma induced etch damage
of GaN p-njunctions. J. Vac. Sci. Tech. A 18,1139–1143 (2000).
8. Minami, M. et al. Analysis of GaN damage induced by Cl
2
/SiCl
4
/Ar
plasma. Jpn. J. Appl. Phys. 50, 08JE03 (2011).
9. Smith, J. M. et al. Comparison of size-dependent characteristics of
blue and green InGaN microLEDs down to 1 μmindiameter.Appl.
Phys. Lett. 116, 071102 (2020).
10. Olivier, F. et al. Influence of size-reduction on the performances of
GaN-based micro-LEDs for display application. J. Lumin. 191,
112–116 (2017).
11. Karpov, S. Y. & Makarov, Y. N. Dislocation effect on light
emission efficiency in gallium nitride. Appl. Phys. Lett. 81,
4721–4723 (2002).
12. Finot, S. et al. Surface recombination in III-nitride micro-LEDs pro-
bed by photon-correlation cathodoluminescence. ACS Photon 9,
173–178 (2022).
13. David, A., Young, N. G., Lund, C. & Craven, M. D. Review –The
physics of recombination in III-nitride emitters. ECS J. Solid State
Sci. Technol. 9, 016021 (2020).
14. Himmerlich, M. et al. Confirmation of intrinsic electron gap states at
nonpolar GaN(1-100) surfaces combining photoelectron and sur-
face optical spectroscopy. Appl. Phys. Lett. 104,171602(2014).
15. Kim, P. et al. Impact of Ga and N vacancies at the GaN m-plane on
the carrier dynamics of micro-light-emitting diodes. Phys. Rev.
Appl. 19,014018(2023).
16. Jiang, F., Hyun, B.-R., Zhang, Y. & Liu, Z. Role of intrinsic surface
states in efficiency attenuation of GaN-based micro-light-emitting-
diodes. Phys. Status Solidi RRL 15, 200487 (2021).
17. Samukawa, S. A neutral beam process for controlling surface
defect generation and chemical reaction at the atomic layer. ECS J.
Solid Sci. Tech. 4, N5089 (2015).
18. Zhu, J. et al. Near-complete elimination of size-dependent effi-
ciency decrease in GaN micro-light-emitting-diodes. Phys. Status
Solidi A 216, 1900380 (2019).
19. Wang, X. L. & Samukawa, S. Damage-free neutral beam etching for
GaN micro-LEDs processing. in Semiconductors and Semimetals
(eds. Jiang H. & Lin J.) 106,203–221 (Academic Press, Cam-
bridge, 2021).
20. Ley, R. T. et al. Revealing the importance of light extraction effi-
ciency in InGaN/GaN microLEDs via chemical treatment and
dielectric passivation. Appl. Phys. Lett. 116,251104(2020).
21. Konoplev, S. S., Bulashevich, K. A. & Karpov, S. Y. From large-size to
micro-LEDs: scaling trends revealed by modeling. Phys. Status
Solidi A 215, 1700508 (2018).
22. Olivier, F., Daami, A., Licitra, C. & Templier, F. Shockley-Read-Hall
and Auger non-radiative recombination in GaN based LEDs: A size
effect study. Appl. Phys. Lett. 111, 022104 (2017).
23. Wolter, S. et al. Size-dependent electroluminescence and current-
voltage measurements of blue InGaN/GaN μLEDs down to the
submicron scale. Nanomaterials 11, 836 (2021).
24. Lee, T.-Y. et al. Increase in the efficiency of III-nitride micro LEDs by
atomic layer deposition. Opt. Exp. 30, 18552 (2022).
25. Park, J.-H. et al. Impact of sidewall conditions on internal quantum
efficiency and light extraction efficiency of micro-LEDs. Adv.Opti-
cal Mater. 11, 2203128 (2023).
26. Schubert, E. F. Light-Emitting Diodes Ch. 8 (Cambridge Univ. Press,
Cambridge, 2007).
27. Baret, S. et al. Surface potential of n-andp-type GaN measured by
Kelvin force microscopy. Appl. Phys. Lett. 93, 212107 (2008).
28. Henning, A. et al. Measurement of the electrostatic edge effect in
wurtzite GaN nanowires. Appl. Phys. Lett. 105, 213107 (2014).
29. Ohori, D. et al. Etching characteristics for InGaN/GaN hydrogen
iodide (HI) neutral beam etching for micro-LED fabrication. Nano-
technology 34, 365302 (2023).
30. Wang, S.-W. et al. Wavelength tunable InGaN/GaN nano-ring LEDs
via nano-sphere lithography. Sci. Rep. 7, 42962 (2017).
31. Ley, R. et al. Strain relaxation of InGaN/GaN multi-quantum well
light emitters via nanopatterning. Opt. Exp. 27,30081(2019).
32. Zhang, K. X. et al. High-efficiency nanodisk of InGaN/GaN MQWs
fabricated by neutral-beam etching and GaN regrowth: towards
directional micro-LED in top down structure. Semicond. Sci. Tech-
nol. 35,075001(2020).
33. Xie, E. Y. et al. Strain relaxation in InGaN/GaN micro-pillars evi-
denced by high resolution cathodoluminescence hyperspectral
imaging. J. Appl. Phys. 112, 013107 (2012).
34. Zhan, J.L. et al. Investigation on strain relaxation distribution in GaN-
based μLEDs by Kelvin probe force microscopy and micro-
photoluminescence. Opt. Exp. 26, 5265 (2018).
Acknowledgements
We would like to express our gratitude to Dr. Naoto Kumagai for his
valuable discussions on KFM measurements, Dr. Hisashi Yamada for his
advice on MOVPE growth on free-standing GaNsubstrates, and Dr. Reiko
Azumi for her continuous support of this work. Dr. Toshikazu Yamada
and Ms. Akiko Murai also provided significant help in wire bonding the
micro-LED chips.
Author contributions
S.S. invented the neutral beam technique. X.W. and S.S. jointly pro-
posed applications of the neutral beam technique to micro-LED pro-
cessing. X.W. performed most of the device fabrication and
characterization experiments and prepared the manuscript. X.Z. con-
ducted the majority of the SEM observations, while T.T. performed the
MOVPE growth for the KFM measurements. D.O. carried out part of the
NBE etching experiments. X.W., S.S. and D.O. participated in extensive
discussions on the experimental results.
Competing interests
The authors declare no competing interests.
Article https://doi.org/10.1038/s41467-023-43472-z
Nature Communications | (2023) 14:7569 7
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s41467-023-43472-z.
Correspondence and requests for materials should be addressed to
Xuelun Wang or Seiji Samukawa.
Peer review information Nature Communications thanks Byung-Ryool
Hyun, and the other, anonymous, reviewer(s) for their contribution to the
peer review of this work. A peer review file is available.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/.
© The Author(s) 2023
Article https://doi.org/10.1038/s41467-023-43472-z
Nature Communications | (2023) 14:7569 8
Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com