Improving the luminous efficiency of red nanocolumn μ-LEDs by reducing electrode size to ϕ2.2 μm

A red InGaN-based nanocolumn micro μLED with an emission diameter of ϕ2.2 μm was demonstrated to achieve an on-wafer external quantum efficiency (EQE) of 2.1% at the peak wavelength of 615 nm. The LED was fabricated by repeating the electrode process on the same nanocolumn pattern area and reducing the emission diameter from ϕ80 to ϕ2.2 μm. The peak EQE, which was maximized at ∼25 A cm−2, increased by decreasing the emission diameter from 1.2% to 2.1%. This behavior, which differs from that of InGaN-film LEDs, is characterized as a unit of independent nano-LEDs with passivated sidewalls of nanocolumn LEDs.


M
icro LEDs (μLEDs) 1,2) have attracted attention owing to their potential applications in next generation displays.The two-dimensional arrangement of independently driven three-primary-color (RGB) μLEDs paves the way toward μLED displays.][5] Recently, multicolor (red, green, blue, and yellow) InGaN/GaN nanocolumn μLED pixels with 5 × 5 μm 2 emission windows have been monolithically integrated, 5) wherein the pixels were arranged two-dimensionally with a period of 10 μm.However, the red InGaN μLED pixel exhibited a low emission efficiency. 5)An improvement in the efficiency of red μLEDs poses a challenge for μLED displays. 16)GaN nanocolumns exhibit dislocation-filtering 17) and lattice strain relaxation. 18,19)Fine nanocolumns, with a diameter less than ∼200 nm, exhibited dislocation-free nanocrystals 17) and increased critical thickness 19) for InGaN in GaN nanocolumns, thereby suppressing the generation of misfit dislocations. 19)A 600 nm (amber) InGaN/GaN nanocolumn array exhibited a high internal quantum efficiency (IQE) of 22%. 20)The surface plasmon polaritons improve the IQE when the plasmonic metals (Au or Ag) are placed within tens of nanometers from InGaN. 21)This configuration can be realized for InGaN/GaN nanocolumn LEDs by depositing metals on the nanocolumn sidewalls.By depositing an Au film on SiO 2 -covered nanocolumn arrays, 5.2-and 4.8-fold enhancements in the photoluminescence intensity were observed at 600 nm (amber) and 680 nm (red) for triangle- 22) and honeycomblattice nanocolumns, 23) respectively.
Recently, the luminous efficiency of red-emission InGaN-film LEDs has been substantially improved.The external quantum efficiency (EQE) values of 2.9% (629 nm) 24) and 4.3% (621 nm) have been achieved. 25)To increase their light extraction efficiency (LEE), the LED chips were packaged with molded silicon resins, and the light output was experimentally measured in a calibrated integrating sphere.The EQEs of these LEDs were maximized at 9.5 and 10.1 A cm −2 , respectively, introducing the efficiency droops of 26% (at 50 A cm −2 ) 24) and 38% (at 100 A cm −2 ). 25)The emission area of these LEDs was approximately 2 × 10 5 μm 2 (∼450 × 450 μm 2 ).Recent investigation of μLEDs clarified that a decreased diameter of the LED chip raised the substantial deterioration in EQE via surface recombination and sidewall current leakage, 26,27) and sidewall passivation has been necessary to improve the EQE of μLED chips. 28)A large jump in EQE via sidewall passivation was recently obtained for 10 × 10 μm 2 amber and 25 μm size red μLEDs, reporting values of 5.5% at 15 A cm −2 and 5.95% at ∼80 A cm −2 , respectively. 29,30)At the same time, the on-wafer EQE of 6 × 6 μm 2 red (632 nm) μLEDs was low (0.2%) owing to an insufficient LEE. 31)his study investigated the electrical and optical characteristics of 2.2 μm diameter (f2.2 μm) red nanocolumn μLEDs and explored the size dependency of LED properties.The emission-window diameter (i.e. the size of the transparent p-type electrode) was reduced from f80 to f2.2 μm by repeating the electrode process on the same nanocolumn LED pattern area.The on-wafer EQE maximized at ∼25 A cm −2 and the peak EQE increased from 1.2% to 2.1% with decreasing LED size.
Nanocolumn LED crystals composed of ordered InGaN/ AlGaN nanocolumn arrays were grown using Ti-mask SAG, [12][13][14] wherein the nanocolumn period (L) was 350 nm.The patterned substrates for the SAG were prepared via electron beam lithography.The LED crystal comprises numerous p-n junction InGaN/AlGaN nanocolumns.The Ga-polar n-GaN nanocolumn arrays were grown on a GaN template/sapphire, followed by the growth of a 60-pair AlGaN/GaN superlattice (SL) underlying layer, 18-pair InGaN/GaN SL buffer layer, 5-pair InGaN/AlN/ AlGaN multiple-quantum-well, undoped GaN, Mg-doped p-type AlGaN electron-blocking layer (EBL), and Mg-doped p-type GaN cladding layers.Finally, an Mg-doped p + -GaN contact layer was grown at the top.The nanocolumn diameter (D) was ∼330 nm at the middle of the nanocolumn and decreased toward the top.The detailed growth process and layered structure are described in Ref. 32.During the growth of the InGaN/GaN SL, acute (10-11) semi-polar crystal facets were formed atop the nanocolumns on which the InGaN/AlGaN MQWs were grown.32) Figures 1(a) and 1(b) show the schematics of the two types of fabricated nanocolumn μLEDs (A and B).The electrode diameter decreased from f80 to f31 μm for Type A and f12 to f2.2 μm for Type B. The electrodes were reproduced on the same nanocolumn array area of 150 × 150 μm 2 .For Type A, the Ni (5 nm)/ITO (300 nm) electrode was deposited on the nanocolumn array LED crystal using evaporation coating and RF sputtering, respectively.Owing to the tight spacing of the nanocolumns, the electrode material was deposited on the shallow place of sidewalls within 100 nm from the nanocolumn tops (on p-GaN).However, the sputtering ITO electrode did not bridge the spacing between the nanocolumns, leaving air gaps, as the diameter decreased toward the c-plane top.The additional process was performed to fill the air gaps with an SiO 2 passivation layer via atomic layer deposition (ALD).Following the removal of SiO 2 from the nanocolumn tops covered with Ni/ITO, a 50 nm thin ITO layer was deposited, thus bridging the nanocolumns.Subsequently, to fabricate a Type A circular electrode mesa, inductively coupled plasma dry etching was performed.The mesa sidewall was exposed to air without passivation, and a row of nanocolumns along the mesa-side edge was partially chipped and damaged by etching, as shown in Fig. 2(a).
In the InGaN-film μLED chips of several μm in size, the carriers are diffused into the mesa sidewall and annihilated via non-radiative surface recombination. 26)In the nanocolumn μLEDs, most of the nanocolumns were standing inside the mesa [see Fig. 2(a)].The carriers were confined to independent nanocolumns, 5) preventing carrier diffusion to the mesa sidewall.As the mesa diameter decreases, however, the percentage of damaged nanocolumns along the mesa-side edge, which induces the leakage current, increases significantly.For instance, when L = 350 nm, the percentage is ∼4.5%, ∼11% and ∼48% for f31, f12 and f2.2 μm, respectively.Thus, when the mesa-side edge nanocolumns are included in the current injection area, the EQE can decrease for the mesa-diameter less than f31 μm.Therefore, for LEDs with small electrode diameters, the emission window was opened inside of the mesa, as shown in Fig. 1(b) (Type B).Notably, the mesa size can decrease such that it is close to the window size without decreasing the EQE.However, the mesa of the f31 μm diameter Type A was used for the electrode fabrication process of the Type B system, and the emission window was reduced from f12 to f2.2 μm, which was achieved by repeating the following process.After removing the SiO 2 passivation layer and electrode metals from the mesa, the spaces between the nanocolumns were embedded using ALD-Al 2 O 3 , although SiO 2 can also be used.Subsequently, a circular emission window was opened on these spaces, followed by the deposition of a Ni (5 nm)/ITO (300 nm) electrode.
The electrical and optical characteristics of the nanocolumn μLEDs were evaluated using on-wafer testing equipment.Figures 2(b) and 2(c) show the near-field emission images of the f55and f2.2 μm diameter nanocolumn μLEDs, wherein red emissions were observed.The electroluminescence spectra were measured using a spectrometer with a microscope that photographed the near-field emission images.The current injection was controlled using a parameter analyzer, and the radiated light was detected by calibrated photodetectors mounted on the front and back of the LED wafer.The intensity from both sides was almost identical, and the light output was evaluated as the sum of these intensities.
Figure 3(a) shows the current density versus the voltage characteristics of the μLEDs with emission diameters from f80 to f2.2 μm.As the emission diameter decreases, the turn-on voltage of Type B increases.Notably, Type A exhibited current leakage in the reverse bias direction.The leakage current density did not vary monotonically with the emission diameter.Therefore, the damaged nanocolumns lining the mesa sidewalls and singular growth of nanocolumns in the mesa area form the current leakage path.For Type A, clipped nanocolumns were observed at the mesa sidewall, as shown in Fig. 2(a).For Type B, the current leakage was suppressed owing to the missing leakage path at the mesa sidewall and decreasing event probability of singular growth points in the emission area.Note that the injection leakage occurred for the f31 μm diameter LED but was suppressed for the f12 μm diameter LED.
Figure 3(b) illustrates the EQEs of the nanocolumn μLEDs with diameters in the range of f80-f2.2μm as a function of the injected current density.The EQE values increased with increasing current density and were maximized at ∼25 A cm −2 to be 2.1% for the f2.2 μm diameter LED (electrode area: 3.8 μm 2 ).The nanocolumn μ-LEDs exhibited small efficiency droops.Among them, the highest efficiency droop of 10% was observed for the f2.2 μm diameter LED, for which the EQE decreased from 2.1% at 25 A cm −2 to 1.88% at 100 A cm −2 .
The EQE is expressed as EQE = IQE × LEE × CIE, where CIE denotes the current injection efficiency.By combining the three-dimensional finite-difference time domain with the ray tracing method, a theoretical LEE (from both sides of wafer) of ∼25% is provided for the same configuration (L and D) of the fabricated nanocolumn LED crystal; details will be presented elsewhere.As mounting LEDs in a package improves the LEE to 60%−70%, 24,25) the aforementioned on-wafer EQE is limited by a low LEE.The peak EQE at 25 A cm −2 was plotted as a function of the emission diameter [Fig.3(c)], and the EQE values increased monotonically with decreasing the emission diameter.The behavior of nanocolumn LEDs differs from InGaN-film LEDs because they are made up of numerous Al 2 O 3 -passivated nanocolumns.Each nanocolumn typically has a InGaN active layer (core) covered by a wide bandgap shell layer at the sidewall, 5,19,32) and therefore, the carriers are confined to the InGaN core.
The output power density increased monotonically as a function of the injection current density.The f2.2 μm diameter LED exhibited an output power density of 3.8 W cm −2 (0.14 μW) at 100 A cm −2 (3.8 μA) and a peak wavelength of 615 nm. Figure 3(d) shows the central wavelength of the electroluminescence spectra obtained via Gaussian fitting as a function of the current density.The central wavelength is believed to correspond to the bandgap wavelength of InGaN.However, the peak wavelength is observed at a different value because it is influenced by the nanocolumn photonic crystal (PC) effect, as discussed in subsequent paragraphs.For the f2.2 μm diameter LED, the central wavelength shifted from 631 to 606 nm as the current density increased from 0.7 to 103 A cm −2 .The observed blueshift of the 25 nm wavelength for the Ga-polar semipolar InGaN active layer is smaller compared to that of the c-plane InGaN-film red-emission μ-LEDs, which is typically 35-59 nm. 24,25,33)The central wavelength was 616 nm at ∼25 A cm −2 for the peak EQE.
Red nanowire (nanocolumn) μLEDs with small electrode sizes (0.75 × 0.75 μm 2 ), wherein the InGaN active region was prepared on underlying N-polar GaN nanowires with c-face flat tops, have recently been reported. 34)The EQE was maximized to ∼8.3% at 1 A cm −2 (5.6 nA) and rapidly decreased to ∼0.6% at 100 A cm −2 (efficiency droop: ∼93%).The wavelength shifted from ∼650 nm at 0.5 A cm −2 to ∼595 nm at 14 A cm −2 .The large efficiency droop and blueshift limit high-power emissions  014004-3 © 2023 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd in the red region. 34)The power density at the efficiency peak was 15 mW cm −2 .The near-field emission image of the f55 μm diameter LED at 2.1 A cm −2 is presented in Fig. 2(b).The intensity distribution contained dark areas even in locations under the p-electrode probe.This can be attributed to the in-plane nonuniformity in the lengthwise current injection along the nanocolumn axis; for example, the p-contact resistance is slightly high in this area.In contrast, a bright and uniform emission was observed at the center of the mesa, located away from the electrode probe.This suggests that the current spreads effectively through the electrode, bridging between the nanocolumns.Considering the microscopic distribution of nanocolumns, the top-view scanning electron microscopy (SEM) images indicated that the c-face p-GaN appeared at the nanocolumn tops [for example, Fig. 7(b) in Ref. 32].The c-plane nanocolumn tops have been considered to contribute to the uniform formation of a p-contact. 32)The c-plane tops, on which Ni/ITO are uniformly deposited, were preferentially exposed via oxide etching.However, the c-plane dimensions in this nanocolumn array varied among the neighboring nanocolumns in the range of 130-190 nm, resulting in a microscopic distribution of the p-contact resistance.Such nonuniformity of the nanocolumn tops in this device could be caused by slight randomness of the underlying Ti-mask nanohole-patterned substrate and crystal nucleation in the Ti nanoholes during the SAG of nanocolumns.An abnormal growth with small c-plane tops occurs over a wide area may account for the dark area shown in Fig. 2(b).Additionally, bright emission spots of several μm in size were scattered in this dark area, which may be understood as indicating the existence of a group of nanocolumns with high luminous efficiency.Thus, the improved uniformity of the nanocolumn array including the homogenization of the p-contact improves the EQE of nanocolumn μLEDs.Note that the randomness in crystal nucleation can be substantially suppressed using the nano-template SAG. 35)he distributions of the contact resistance and luminous efficiency varied the CIE and IQE, respectively, resulting in the scattered in-plane brightness of the nanocolumns.As the applied voltage increases, the dark emission nanocolumns gradually turn on in sequence, leading to increased intensities with small efficiency droops.The EQE did not decrease at a higher injection current [see Fig. 3(b)], although a slight efficiency droop was observed for the f2.2 μm diameter LED.The EQEs of the f2.2and f2.5 μm diameter LEDs were 2.1% and 1.7%, respectively, although the emission size was nearly the same.The latter EQE is comparable to that of the f12 μm diameter LED, which exhibits an in-plane intensity distribution containing bright emission micro-spots [see Fig. 6(e) in Ref .32].
Therefore, for the f2.2 μm diameter LED, which displayed a uniform near-field emission image [refer to Fig. 2(c)], it is believed that the electrode was positioned over the bright emission spot, characterized by the uniformly configured and efficient nanocolumns.With the removal of the in-plane inhomogeneity effect, the f2.2 μm diameter LED exhibits the normal characteristics of nanocolumn LEDs with the semi-polar InGaN active layers, having a small blueshift and efficiency droop.Generally, c-face InGaN-film LEDs exhibit larger efficiency droops and blueshifts compared with those with semi-polar 36) and non-polar 37) InGaN active layers.The efficiency droop has typically been explained using two mechanisms, namely electron leakage 37,38) and Auger recombination. 39)Following the former, a large polarization field induces the sloped triangular band bending in c-face InGaN/GaN MQWs and EBL, which causes an electron leakage toward the p-GaN cladding layer. 37,38)The current leakage is enhanced with the increased forward current (i.e. with the increased carrier density), inducing the efficiency droop.The experimental LEDs are considered to be composed of semi-polar InGaN active layers with a lower polarization field, observing the lower efficiency droop.Meanwhile, lowering the carrier density in the active layer could contribute to suppressing the efficiency droop. 40)The nanocolumn effect suppresses the generation of misfit dislocation even for red bulk-InGaN, because of the strain relaxation effect. 19)We utilized bulk-InGaN active layers for the two-dimensional integration of multicolor nanocolumn μLEDs. 5)igure 4(a) shows the electroluminescence spectra of the f2.2 μm diameter nanocolumn LED for a current density ranging from 0.7 to 53 A cm −2 ; the light intensity was detected using ×40 magnification objective lenses with an acceptance angle of 73.7°.The spectral full width at half maximum of the f2.2 μm diameter LEDs was 50 nm at ∼25 A cm −2 and sharp emission peaks appeared at 636 and 615 nm.Based on the periodic arrangement of nanocolumns, the PC effect [41][42][43][44][45] enhanced the light intensity at the photonic band wavelengths, which are determined by the 014004-4 © 2023 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd periodicity of the nanocolumns.At the low current density, the peak at 636 nm was dominant.However, the central wavelength was blue-shifted as the current density was increased in the low current density region [see Fig. 3(d)], and the spectral peak switched from 636 to 615 nm.In the range above 7 A cm −2 , the peak wavelength of 615 nm became dominant and remained constant with the injection current density, as shown in Fig. 4(b).The PC effect in nanocolumn LEDs was first discussed in Ref. 41 considering radiation beam directionality and wavelength stability, followed by demonstrations of yellow, 42) red, 43) and green 44) emissions and monolithic multicolor integration of PC nanocolumn LEDs. 45)By adjusting L and D, the photonic band wavelength (λ B ) can be set to 633 nm.By aligning the InGaN bandgap accordingly, the stable operation of the μ-LED at 633 nm may be achieved, as shown in Fig. 4(b).Note that the relationship between L/λ B and (D/L) 2 of Fig. 5(c) in Ref. 45 can be utilized in the control of λ B and the device in this experiment operated using a higher order photonic band edge.
In summary, the electrical and optical characteristics of red nanocolumn μLEDs with semi-polar InGaN active layer were investigated as the electrode (emission) diameter was reduced.Type A and B μLEDs were fabricated on the same nanocolumn array area with dimensions of 150 × 150 μm 2 by varying their emission diameters from f80 to f31 μm for Type A and from f12 to f2.2 μm for Type B. The Type A electrode process was repeated inside of the emission area of the f80 μm diameter Type A μLED.A mesa configuration was fabricated with an electrode on top and decreased diameter, wherein the etched sidewall was exposed to air.The damaged nanocolumns located along the mesa-side edge [Fig.2(a)] and the growth singularity inside of the mesa provided current leakage paths.To exclude the effects of damaged nanocolumns along the mesa-side edge on the EQE, the Type B emission windows were opened inside the Al 2 O 3 -passivated mesa.As the emission size was decreased, the EQE increased owing to the decreased growth singularity and increased bright emission spots in the mesa.For the f2.2 μm diameter μLEDs, in which the electrode could be positioned over the bright emission spot, an on-wafer EQE of 2.1% was observed, but the EQE was limited by a low LEE of ∼25%.Simultaneously, the nanocolumn PC effect contributed to the stable peak wavelength operation of the μLED at 615 nm.
The EQE increased with decreasing emission diameter from 1.2% to 2.1%, behaving differently from InGaN-film LEDs.Because nanocolumn LEDs comprise numerous independent nano-LEDs (p-n junction nanocolumns) of Al 2 O 3 -passivated nanocolumns, the EQE behavior in this experiment was determined through the characteristics of each nano-LED and its in-plane fluctuation.The InGaN active layers in the nanocolumns are covered at the sidewall by wide bandgap layers. 32)Therefore, when the homogeneity of nanocolumn arrays is completed, the LED properties can be determined using independent nano-LEDs with similar emission characteristics; hence, the EQE can be less sensitive to the electrode diameter, in principle, down to the size of a single nano-LED.

Fig. 2 .Fig. 3 .
Fig. 2. (a) Birds-eye SEM image of the mesa-side area of the Type A LED, in which the ITO electrode top, etched bottom surfaces, and mesa sidewall are shown and the nanocolumn period (L) was 350 nm.Near-field emission images of the (b) f55and (c) f2.2 μm diameter LEDs at 2.1 and 6.6 A cm −2 , respectively.