Significant reduction in sidewall damage related external quantum efficiency (EQE) drop in red InGaN microLEDs (∼625 nm at 1 A cm−2) with device sizes down to 3 μm

Ultra-small (10 μm) InGaN-based red microLEDs (625 nm at 1 A cm−2) are necessary for modern displays. However, an increase in surface-area-to-volume ratio with a decrease in the micro-LED size resulting in higher surface recombination causes a drop in efficiency with device size. In this letter, we demonstrate microLEDs from 60 μm down to 3 μm with significantly reduced sidewall-related efficiency reduction using a two-step passivation technique using Al2O3 and Si3N4. The peak on-wafer EQE changes from 0.21% to 0.35% as the device size reduces from 60 to 3 μm, possibly due to improved light extraction efficiency for smaller mesa-widths.


I
II-Nitride LEDs have been used extensively in commercial lighting applications for over 20 years. 1,2)icroLEDs which generally have a size under 100 μm offer several advantages over other conventional display technologies such as LCDs and OLEDs including lower power consumption, ultra-fast response times, higher brightness and contrast owing to higher pixel densities, and longer lifetimes. 3)][5][6] AR/VR displays require very high pixel density due to their proximity to the user's eye.Consequently, microLEDs smaller than 5 μm are needed. 7)[10] This is due to an increased surface recombination on the sidewalls of the devices.During the fabrication process, the individual device mesas are created using inductively coupled plasma or reactive ion plasma etching, which damages the sidewalls forming defects that act as nonradiative recombination centers.As the device size is decreased, the ratio of surface area to volume increases and thus the effects of damage caused by the etching becomes more prominent. 11,12)Currently, micro-LED displays use InGaN based blue and green LEDs, whereas AlInGaP based LEDs are used to cover the red region of the spectrum (>615 nm).AlInGaP based phosphide red LEDs demonstrate high EQE for larger device sizes, however, the efficiency decreases steeply as the size of the device is shrunk. 13,14)Fan et al. 15) reports a 3× reduction in EQE when the device size is shrunk from 160 to 20 μm.On the other hand III-Nitrides LEDs suffer less severely from surface recombination. 16,17)his is due to the Phosphide material system having an order of magnitude higher surface recombination velocities when compared to their III-Nitride counterparts and AlInGaP only shows limited improvements from sidewall passivation techniques. 11)owever, in InGaN based red microLEDs, the absolute values of EQE remain very low.][20][21] The root causes for this drop in efficiency are the quantum confined stark effect, 22) high dislocation density due to lattice mismatch, 23) low thermal stability of high composition InGaN quantum wells 24) and Auger-Meitner recombination. 25)In 2020, 26) Pasayat et al. demonstrated 0.2% on-wafer EQE for a 6 μm × 6 μm device at 10 A cm −2 emitting at 632 nm using porous GaN technology. 27)Other efforts in red InGaN microLEDs include use of a relaxed InGaN pseudosubstrates from SOITEC, 28,29) a thermal decomposition layer [30][31][32][33] and by using hybrid multiple quantum wells (MQW) stack.[34][35][36] Size independent EQE has been previously demonstrated 37) for InGaN based red microLEDs from 80 μm × 80 μm devices down to 20 μm × 20 μm devices.However, we are not aware of any reports of size independent EQE for InGaN based red microLED devices with sizes less than 10 μm × 10 μm. Recenty Panpan Li et al. fabricated 80 μm × 80 μm and 5 μm × 5 μm red microLEDs. 38)A peak EQE of 6% was achieved for the 80 μm device, while the EQE still reduced to 4.5% for the 5 μm device, showing a reduction of efficiency for smaller devices.
In this letter, we demonstrate significant reduction or elimination of sidewall damage related efficiency drop for InGaN red microLEDs from 3 to 60 μm.Our 60 μm device shows an on-wafer EQE of 0.21% at 5 A cm −2 with a peak wavelength of 614 nm, whereas our 3 μm device shows an on-wafer EQE of 0.35% at the same current density with a peak wavelength of 608 nm.We witness an improvement of EQE with lowering of the device size down to 3 μm due to increased light extraction in smaller devices.
The LED wafer was grown using metal-organic chemical vapor deposition in a commercial close-couple showerhead AIXTRON reactor.Ammonia (NH 3 ), TriEthylGallium (TEGa), TriMethylIndium (TMIn), TriMethylAluminum (TMAl), Silane (SiH 4 ) and Bis(cyclopentadienyl)magnesium (Cp 2 Mg) were used as precursors and dopants for the growth.The epitaxial structure is shown in Fig. 1(a).The growth started with the deposition of 400 nm of Si-doped In 0.07 Ga 0.93 N at 920 °C in the similar manner as described Pantzas et al. 39) The active region was composed of 3 MQW comprising of 2.7 nm InGaN QW, 1.5 nm AlGaN cap and 10 nm GaN quantum barrier (QB).The AlGaN cap was deposited in order to prevent Indium desorption during the QB growth as well as to balance the built-in strain due to the lattice mismatch between the high Indium containing QW and the underlying layers. 22)The quantum well and cap layers were grown at 800 °C, while the barriers were grown at 900 °C in order to improve material quality.Finally, the deposition process was completed after growing 120 nm of Mg-doped In 0.045 Ga 0.955 N layer at 940 °C and a 20 nm of heavily p-doped In 0.045 Ga 0.955 N as the contact layer.The NH 3 flow was kept at 178.5 mmol min −1 during the growth of all InGaN layers.The pressure was kept at 375 torr throughout the deposition.
After growth, the LED wafer was activated at 600 °C for 20 min in air ambient.The mesa of LED devices ranging from 3 μm × 3 μm to 60 μm × 60 μm were formed through lithography patterning followed by reactive ion etch (RIE) with chlorine gas chemistry.A low RIE power of 15 W was chosen to reduce the sidewall damage from dry etch without additional wet treatment. 40)The p-contact was formed by 70 nm Ni using E-beam metal evaporator.A double layer passivation of 30 nm Al 2 O 3 and 250 nm of Si 3 N 4 was deposited using atomic layer deposition (ALD) and plasma enhanced CVD (PECVD) respectively.This was previously demonstrated in Ref. 41.We need to deposit thick dielectric in order to passivate the damages caused by the dry etch and to provide electrical insulation between p-metal and n-InGaN.However, the time taken to deposit thick Al 2 O 3 using ALD will increase the overall processing time.Consequently, we used a thinner Al 2 O 3 layer and used PECVD to deposit 250 nm of Si 3 N 4 which was effective in suppressing the sidewall related damages in our micro-LEDs.After via opening, the metal stack of 30 nm Ti and 250 nm Ni was deposited using E-beam evaporator serving as both metal pads overlayer (on top of p-metal) and n-contact.Figure 1(b) shows the SEM image of a 4 μm device.
In this letter, current density (J)-voltage (V ) measurements were performed by using Keysight B1505A form factor probe station.The fabricated LED devices were measured onwafer and the light was collected from the backside using a cosine corrector (Ocean Insight CC-3-UV-S) positioned underneath the wafer.Then the light was guided through an optical fiber before it entered the Horiba iHR320 spectrometer using which we obtained the electroluminescence spectra in conjunction with a thermoelectrically cooled CCD detector (Horiba Synapse CCD).The optical setup was calibrated using a radiometrically calibrated light source (Ocean Insight DH-3P-CAL).Collecting light from the backside is not optimal as there can be scattering from all the layers underneath the active region including the sapphire substrate.Due to the limited light collection angle, we are not able to collect all the light that is being emitted.Based on estimates from Pasayat et al. 26) on their red LED work, the absolute EQE number may possibly increase by 3-5 times (compared to on-wafer EQE as presented in this work) with packaging and integrating sphere.
Figures 2(a) and 2(b) shows the EL spectra and the variation of peak emission wavelength and FWHM w.r.t.current densities for a 3 μm device.The peak emission wavelength shifts from 622 nm at 1 A cm −2 to 581 nm at 100 A cm −2 , resulting in a 41 nm blueshift which remains nearly identical across all our devices (Table I).This occurs due to the screening of the built-in piezoelectric field by charge carriers injected inside the quantum wells as the current density increases.Gradually, as the states which are generated by localised Indium fluctuation get filled, the band filling effect results in further blueshift of the emission spectra. 42)The blueshift is similar to previous reports on InGaN red microLEDs. 26,34)The FWHM initially starts at 68 nm and reduces down to 52 nm at 100 A cm −2 .The decrease of FWHM values at lower current densities has been attributed to the localised states getting filled. 43)igure 2(c) shows the variation of peak emission wavelength versus current densities for different device sizes.As can be seen from the plot, the peak emission wavelength of our devices stays nearly identical for greater than 5 A cm −2 across various sizes with a delta of ±6 nm, indicating that the different size devices can be utilized for EQE comparison.
Figure 3(a) shows the J-V Spectra of our devices.The turn-on voltage is around 2.4 V.Under reverse bias, the smaller devices do not show higher leakage current than larger devices.The smallest 3 μm device shows the lowest reverse leakage current while the other sizes lie between 38 and 54 A cm −2 owing to a device to device variation.030904-2 © 2024 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd Traditionally, smaller devices have larger leakage current due to increased sidewall damage.However, our double sidewall passivation method helps in reducing the sidewall damage for smaller devices (<10 μm), rendering almost no difference in leakage current to larger ones (>20 μm).The damage reduction is further evidenced by similar forward leakage current density for all micro-LEDs.The EQE of the device is defined as the ratio of the number of photons extracted from the device to the number of electrons injected into the device per second.Figure 4 shows the variation of EQE versus current density for devices with sizes from 3 to 60 μm.We have carefully calibrated our setup before measurement to prevent any setup related errors.Additionally, before every measurement, a reference device was measured to ensure that that there was no significant change in intensity compared to the previous measurement.The microLED is placed at the center of the cosine corrector in order to maximize light extraction.After collecting the spectrum data, we calculate the EQE using MATLAB using    030904-3 © 2024 The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd calibration files that were provided from HORIBA as well as our own calibration of the cosine corrector.For our largest devices with a size of 60 μm, we achieved a peak EQE of 0.21% at 5 A cm −2 , which gradually increased with reduction in device size, reaching 0.35% at the same current density for a 3 μm device.If sidewall damage was significant, the J Peak will move to larger current densities due to an increase in the Shockley-Read-Hall (SRH) non-radiative recombination coefficient. 41)J Peak refers to the current density where the peak value of EQE occurs.SRH recombination takes place through traps present within the energy band gap due to defect states in the bulk or at the surface.The J peak for our devices lie between 5 and 7.5 A cm −2 which shows very little influence of surface recombination related non-radiative recombination in our devices.At a low current density of 1 A cm −2 below J peak , we still witness significantly high EQE for all of our devices with size <10 μm.We believe that the increase in EQE with reduction in device size is related to the improved light extraction for smaller device size.This has been observed before in Ref. 41.In larger devices, most of the light rays directly strike the semiconductor-air interface and those with incidence angle greater than the critical angle undergo total internal reflection (TIR) and thus are not collected by the cosine corrector.However, as the device sizes are reduced, more and more rays strike the naturally inclined sidewalls which generate during the mesa etch process.Rays undergoing TIR at the sidewalls reflect to the top surface and much closer to the normal and thus have a higher chance of escaping, thus improving the light extraction efficiency.As our device size is much smaller than the diameter of the cosine corrector, we believe that there should not be any difference in collection efficiency across different device sizes.
In conclusion, we demonstrated the significant reduction or elimination of sidewall damage related efficiency reduction for InGaN based red LEDs.The on-wafer EQE for our smallest 3 μm device was 0.35% at 5 A cm −2 while that of our largest 60 μm device was 0.21% at the same current density.The higher EQE of our smaller devices may possibly be related to an improved light extraction.An optimized two step passivation process led to the suppression of sidewall related defects for small sized microLEDs.To the best of our knowledge, this is the first report of improvement of EQE with reduction in device size below 5 μm for InGaN red LEDs with top down process flow.We believe that this has great potential for the future of ultra-small red InGaN microLEDs.

Fig. 2 .
Fig. 2. (a) Electroluminescence Spectra of a 3 μm LED (the peak wavelength at each current density has been labelled), (b) peak emission wavelength and FWHM versus current density for a 3 μm LED, (c) variation of peak emission versus current density wavelength across device sizes from 3 μm × 3 μm to 60 μm × 60 μm.

Fig. 3 .
Fig. 3. JV Characteristics of the LED devices showing both forward and reverse leakage currents (Inset shows the JV plot but zoomed in at −10 V).

Table I .
Comparison of peak emission wavelength at 1 and 5 A cm −2 , blueshift from 1-100 A cm −2 and FWHM at peak EQE across all device sizes.