InGaN-based red light-emitting diodes: from traditional to micro-LEDs

Our InGaN-based red LED epitaxial wafers were grown by MOVPE in a single-wafer horizontal reactor at 100 kPa. Trimethylgallium (TMGa), trimethylaluminum (TMAl), trimethylindium (TMIn), and ammonia (NH₃) are used as Ga, Al, In, and N sources, respectively. The n- and p-type doping precursors were N₂-diluted silane and bis(cyclopentadienyl)magnesium (Cp₂Mg), respectively. The carrier gases during MOVPE growth were high-purity H₂, N₂, or their mixtures.

The epitaxial growth structure of InGaN-based red LEDs is shown in Fig. 1(a). A c-plane GaN film is utilized to serve as a template for InGaN visible LED growth. However, GaN and high-In-content InGaN have a large lattice mismatch, which would cause many defects and degrade the quality of InGaN layers. Because the lattice mismatch was related to the in-plane strain of the epitaxial layers, understanding the influence of the GaN templates' in-plane strain helped to improve the crystal quality of high-In-content InGaN layers.

Zoom In Zoom Out Reset image size

There are two different ways to adjust the in-plane strain of GaN templates. One is to adjust the thickness of GaN templates, and the other is to choose GaN templates on different substrates. In our study, we investigated different GaN templates on c-plane patterned sapphire substrates (PSS) and Sn-doped β-Ga₂O₃ substrates. ^53,54) The GaN templates on PSS were prepared by our MOVPE system, and we grew different thicknesses from 4 to 10 μm for the GaN templates. The GaN growth rate was around 2.4 μm h⁻¹. The n-type GaN template on an Sn-doped β-Ga₂O₃ substrate was purchased from Tamura Co. and Novel Crystal Technology, Inc. SiN_x arrays were used at the interface between the GaN template and Ga₂O₃ substrate, which could enhance the light extraction efficiency and enable the carrier conductivity at the interface. ^55,56) Both of the FWHMs of (0002) and (10–12) planes X-ray rocking curve for GaN templates on PSS and Ga₂O₃ were below 300 arcsec, revealing good crystal quality of the GaN templates.

We regrew 1 μm thick n-Al_0.03Ga_0.97N or n-GaN on the GaN templates on PSS and Ga₂O₃ substrate, respectively. The n-Al_0.03Ga_0.97N layer was able to realize a smooth surface and a lower series resistance by high Si doping. Then, 15 pairs of In_0.08Ga_0.92N (2–2.5 nm)/GaN (5.5–6 nm) superlattices (SLs) were grown on the n-type layer, followed by a 15 nm thick n-GaN. The active region consisted of two components: a single blue In_0.2Ga_0.8N QW and two red InGaN QWs. This hybrid multiple QWs in our previous work have been shown to help improve light output power. ⁵²⁾ The In content in red QWs was adjusted by the growth temperature.

Also, we proposed complicated barrier layers [Fig. 1(a)] for strain compensation. ^51,53) The red QWs grown on the Al_0.13Ga_0.87N barrier layer could obtain fewer trench defects and realize a better crystal quality, which contributed to an enhanced device performance. The capping layer was GaN and AlN for the blue and red QWs, respectively, which were grown at the same temperature as QWs. The AlN interlayer could suppress the decomposition of the high-In-content red QWs during the barrier layer growth at higher temperatures. ⁴⁸⁾ Finally, a 100 nm thick Mg-doped p-GaN layer and a 10 nm thick heavily Mg-doped p⁺-GaN layer were grown on top of the active region.

Figure 1(b) shows a typical cross-sectional HRTEM image of our InGaN red LED structure. The LED structure could be seen clearly in this HRTEM image, demonstrating the same structure that we designed. The sharp contrast between InGaN and barrier layers in the active region and SLs indicated an abrupt interface of the structures.

After epitaxial growth, device fabrication is a further important step to realize high-performance InGaN red LEDs. Due to the low doping level and high work function of the p-GaN layer, high-quality p-GaN contacts are always the key issue for device performance. ITO is one of the most popular materials for the transparent p-electrode in InGaN-based blue and green LEDs. ^57,58) It is usually deposited by e-beam evaporation or magnetron sputtering. ^59,60) However, the deposition process is generally carried out in an oxygen atmosphere at high temperatures, which are complicated and require additional configurations for facilities. ^59,61) In addition, little attention has been paid to the optimization of ITO layers for InGaN-based LEDs in the red region. Therefore, the ITO layer for InGaN red LEDs needs further optimization.

We proposed a two-step annealing process under O₂-containing (Ar/O₂ mixture) and O₂-poor (Ar) atmospheres in sequence. ⁶²⁾ Figure 2(a) shows the sheet resistivity of ITO layers on double-polished sapphire substrates. The samples named S1–S3 were sputtered ITO while the sample named E1 was e-beam evaporated ITO. All the deposition processes were at room temperature (RT) without oxygen gas. As shown in Fig. 2(a), the two-step annealing could reduce the sheet resistivity of sputtered ITO to around 2 × 10⁻⁴ Ω cm. These results could be explained by the conductivity mechanism of ITO layers, which were related to oxygen vacancies. ⁶⁰⁾ Also, the transmittance of ITO layers could be improved after the two-step annealing, reaching over 98% for sputtered ITO in the 590–780 nm wavelength range [Fig. 2(b)]. As a result, we demonstrated that the two-step annealing indicates the excellent performance of ITO in the aspect of conductivity and transmittance.

Zoom In Zoom Out Reset image size

Meanwhile, we found that sputtered ITO exhibited better conductivity and higher transmittance than e-beam evaporated ITO, as shown in Figs. 2(a) and 2(b). This comparison result was attributed mainly to the crystal quality of ITO layers, which was characterized by the XRD diagram in our previous work. Finally, we utilized sputtered and e-beam evaporated ITO to fabricate red LEDs. Figure 2(c) shows the light output power of the red LEDs at different currents. The red LEDs with sputtered ITO were demonstrated to have a higher light output power as compared to those LEDs using e-beam evaporated ITO. This result was reasonable based on the performance comparison of sputtered and e-beam evaporated ITO in Figs. 2(a) and 2(b).

We measured the in-plane residual stress of the different GaN templates, as shown in Fig. 3(a). The in-plane residual compressive stress of the GaN on PSS was gradually reduced as the GaN thickness was increased from 4 to 10 μm. Also, in the case of GaN on Ga₂O₃ substrate, the 5 μm thick GaN template exhibited a few in-plane tensile or compressive stresses. The error bar was due to the use of different Raman shift values of the free-standing GaN from different Refs. 63, 64. Therefore, we could conclude that Ga₂O₃ substrates help relax the residual stress for the GaN template compared to PSS.

Zoom In Zoom Out Reset image size

Figure 3(b) shows the EL peak wavelength of InGaN red LEDs on the GaN templates on PSS with different residual in-plane stresses. The MOVPE growth conditions were the same for all samples. Clearly, the peak wavelength gradually red-shifted as the in-plane residual stress decreased. This phenomenon indicated that In incorporation was enhanced due to the relaxed in-plane residual stress in the underlying GaN layers. As a result, we can make a higher growth temperature of InGaN red QWs by using a thicker GaN template for the same emission wavelengths. Based on our experiments, ⁵³⁾ the increment of the growth temperatures for the case of GaN thicknesses of 4 and 10 μm was around 5 °C.

We realized the same peak wavelength of 635 ± 2 nm for red LEDs on GaN templates with different in-plane residual stresses by adjusting the growth temperature. As shown in Fig. 3(c), the EL intensity could be enhanced 1.3 fold because of the higher growth temperature for InGaN red QWs on more relaxed GaN templates. Another reason for the enhancement was that the possibility of the defect formation might also be reduced in the active region due to the lower lattice mismatch between the InGaN red QWs and the GaN templates.

Meanwhile, we compared the EL emission for red LEDs on PSS and Ga₂O₃ in Fig. 3(d). Although the QW growth temperature for the Ga₂O₃ substrate was 15 °C higher, the peak wavelength of red LEDs on Ga₂O₃ (665 nm) exhibited a large red-shift of 32 nm at 20 mA compared to LEDs on PSS (633 nm). This result further demonstrated that less in-plane residual compressive stress helped enhance the In incorporation. Figures 4(a) and 4(b) show the light output power and EQE of red LEDs, respectively. All the LEDs in our measurements were bare chips without resin encapsulation. The red LEDs on PSS exhibited a better performance, which originated from the lower In content in red QWs compared to LEDs on Ga₂O₃. The light output power of red LEDs on PSS could achieve 1 mW at a current of 30 mA. The EQE of red LEDs on PSS and Ga₂O₃ was 1.6% and 0.19% at 20 mA, respectively. In addition, the forward voltage of red LEDs on PSS and Ga₂O₃ at 20 mA was 3.3 and 2.45 V, respectively. So the WPE of red LEDs on PSS and Ga₂O₃ at 20 mA could be calculated as 1.0% and 0.14%, respectively.

Zoom In Zoom Out Reset image size

Furthermore, we investigated the effect of chip size on the optical performance of red LEDs on PSS. ⁶⁵⁾ We fabricated rectangular LED chips with different mesa areas, as shown in Fig. 5(a). The mesa width kept constant as 250 μm, while the mesa length reduced from 650 to 350 μm. So the active areas were calculated to be 0.1445, 0.1215, 0.0985, and 0.0755 mm², respectively. The EL images in Fig. 5(a) demonstrated that all chips exhibited homogeneous emission and had no significant current crowding. However, we also observed many dark spots in the emission area, which were regarded as threading dislocations or trench defects.

Zoom In Zoom Out Reset image size

The peak wavelengths of red LEDs at 5–100 mA are shown in Fig. 5(b). An approximately 50 nm blue-shift was observed for all chips with different chip sizes. The blue-shift amount for red LEDs was much higher than that for the blue and green LEDs, illustrating a very strong QCSE in InGaN red QWs. Also, smaller chips had a blue-shift compared to larger chips at the same current injection. This phenomenon was reasonable because of the screening of QCSE by a higher current density in smaller chips. Generally, LEDs would be operated at a certain current for a designed peak wavelength. Because the emission wavelength also depended on the chip sizes, researchers should consider the In content in red QWs and design the chip size together.

The FWHM dropped rapidly to the minimum values (around 60 nm) in the low current range and then increased slightly with currents in Fig. 5(c). The current dependent behavior of the FWHM was similar for all chip sizes. However, the smaller chips exhibited narrower FWHM at low currents but became broader at high currents when compared to those larger chips. This phenomenon was related to the two different mechanisms for the variation of FWHM. The narrowing of the FWHM was attributed to the filling of deep localized states, while the broadening of the FWHM originated from heat generation. ^37,66) As shown in Fig. 5(d), the duty cycle of the pulsed currents would determine the FWHM at high currents, which demonstrated that the heat generation was the main reason for the broadening of FHWM. Therefore, the higher current density for smaller chips would accelerate the filling of deep localized states to make the FWHM narrower, but generated more heat within chips to make the FWHM broader.

Interestingly, the peak wavelength and FWHM for different chips followed the identical current density dependence, as shown in Figs. 5(e) and 5(f). These results indicated that the residual strain and localized states in red QWs remained constant after fabrication. Also, the peak wavelength and FWHM of red LEDs on Ga₂O₃ were plotted in Figs. 5(e)–5(f) for comparison. The red LEDs on Ga₂O₃ exhibited a total blue-shift of 60 nm, which was 10 nm larger than that of red LEDs on PSS. The FWHM of red LEDs on Ga₂O₃ also increased by 5–7 nm at the same current density compared to red LEDs on PSS. These results illustrated that the QCSE and In fluctuation would become more severe upon a further increase of the In content in red QWs, ⁶⁷⁾ which explained the performance degradation of red LEDs on Ga₂O₃ in Fig. 4.

The temperature dependence of the EL intensity for InGaN-based red LEDs on PSS and Ga₂O₃ was measured at a probe station. The stage temperature was controlled by a heater from 295 K (RT) to 373 K. Figure 6(a) shows the normalized EL intensity at different stage temperatures at 100 mA. The EL intensity dropped as the temperature increased, which is usually called the thermal droop. ⁶⁸⁾ The thermal droop could be described by the following formula ^48,65)

$$\begin{eqnarray}&&I/{I}_{T=295{\rm{K}}}=\exp \left(-\displaystyle \frac{T-295{\rm{K}}}{{T}_{0}}\right),\end{eqnarray} \tag{ 1 }$$

where I/I_T=295K is the normalized EL intensity by the EL intensity at 295 K, T [K] is a stage temperature, and T₀ [K] is the characteristic temperature.

Zoom In Zoom Out Reset image size

The characteristic temperature is usually regarded as a standard quantitative parameter to evaluate the temperature stability of LEDs. A larger characteristic temperature implies better temperature stability in the LED light output. By using Eq. (1) to fit the data points in Fig. 6(a), we obtained the characteristic temperatures at various current densities in Fig. 6(b). The characteristic temperature of AlGaInP red LEDs ⁶⁹⁾ was also plotted in Fig. 6(b) for comparison. Obviously, InGaN red LEDs on PSS exhibited higher characteristic temperatures at the current densities from 69 to 132 A cm⁻² compared to that of AlGaInP red LEDs. That indicates the benefit of InGaN red LEDs in temperature stability.

The characteristic temperatures of InGaN-based LEDs can be explained by the Shockley–Read–Hall (SRH) recombination. ⁶⁸⁾ Fewer SRH recombination centers in InGaN QWs led to a higher characteristic temperature. Therefore, the InGaN red LEDs on PSS exhibited better temperature stability than those on Ga₂O₃. Also, higher current density injection should benefit the level of saturation of the SRH recombination, which may increase the characteristic temperature of InGaN red LEDs at high current densities [Fig. 6(b)].

We also analyzed the temperature dependence of the peak wavelength for InGaN red LEDs on PSS and Ga₂O₃ [Fig. 6(c)]. The peak wavelength had a red-shift with an increasing temperature because the bandgap of InGaN shrank with the temperature. By using the linear fitting, we could estimate the red-shift coefficient of 0.066–0.091 nm K⁻¹ for red LEDs on PSS at different current densities, as shown in Fig. 6(d). The InGaN red LEDs on Ga₂O₃ had a smaller red-shift coefficient of 0.085 nm K⁻¹ compared to those on PSS at 69 A cm⁻². This difference would originate from the lower amount of bandgap shrinking for higher-In-content QWs with the temperature.

In addition, carriers in QWs could be excited to higher energy levels with increasing temperature, which conversely resulted in a blue-shift of the peak wavelength. ⁷⁰⁾ This blue-shift phenomenon would partly compensate for the red-shift of the peak wavelength. Because a percentage of carriers could be excited to higher energy levels at high current densities, the compensation by the blue-shift would become larger and lead to a reduced red-shift of the peak wavelength. Therefore, we could observe that the red-shift coefficient decreased as the current density increased.

The red-shift coefficient of AlGaInP LEDs ⁷¹⁾ was also presented in Fig. 6(d) for comparison. Although the operation mode was different, we could clearly observe that InGaN red LEDs exhibited smaller red-shift coefficients at different current densities, revealing the better temperature stability of the peak wavelength. Therefore, based on the above discussion, we could conclude that InGaN red LEDs were more stable at high temperatures, demonstrating their potential for temperature tolerant applications.

The EL characteristics were measured under direct current operation. The picture and cross-sectional schematic of the measurement configuration are shown in Figs. 8(a) and 8(b), respectively. An integrating sphere was located above the sample to collect the light output power and connected through a fiber optic cable to a spectrometer to record the EL spectra. The measurement configuration was called on-wafer testing. ⁸⁵⁾

Zoom In Zoom Out Reset image size

Figure 9(a) shows the EL spectra of 47 × 47 μm² L-InGaN and H-InGaN μLEDs at 4 and 30 A cm⁻², respectively. The peak wavelengths for L-InGaN and H-InGaN μLEDs were located at 626 and 631 nm, respectively. These peak wavelengths were very close to the primary red emission wavelength (630 nm) defined by Rec. 2020. The FWHMs of L-InGaN and H-InGaN μLEDs were 53 and 60 nm, respectively. The broader FWHM of H-InGaN μLEDs was reasonable because higher-In-content would cause more In fluctuation.

Zoom In Zoom Out Reset image size

We then investigated the current density dependences of the peak wavelength and FWHM for InGaN red μLEDs at RT, as shown in Figs. 9(b) and 9(c), respectively. Regardless of the In content in red QWs, the 98 × 98 and 47 × 47 μm² InGaN μLEDs exhibited the identical behaviors for the peak wavelength and FWHM at different current densities. This phenomenon was similar to the large-size InGaN red LEDs, implying the constant QCSE in red QWs with the chip size shrinking down to 47 × 47 μm².

Also, the L-InGaN and H-InGaN red μLEDs exhibited a large peak wavelength blue-shift of approximately 35 and 60 nm at 2 to 50 A cm⁻², respectively. The blue-shift of H-InGaN μLEDs was higher than that of L-InGaN μLEDs, especially at the low current injection region. This behavior originated from the strong QCSE and band filling effect in high-In-content InGaN red QWs. ^28,37) However, AlGaInP red μLEDs usually show a peak wavelength shift within 2 nm at different current densities, implying a stable peak emission compared to InGaN red μLEDs. ⁸⁶⁾

In the case of FWHM, the L-InGaN and H-InGaN μLEDs behaved with different current density dependences. At a low current density below 10 A cm⁻², the FWHM of H-InGaN μLEDs decreased rapidly with the current density. The total reductions of the FWHMs were in the range of 17–19 nm for 98 × 98 and 47 × 47 μm² H-InGaN μLEDs from 2 to 10 A cm⁻². However, the reduction of the FWHM for L-InGaN μLEDs was only 2 nm from 2 to 10 A cm⁻². The much higher reduction amount of the FWHM for H-InGaN red μLEDs originated from many deep localized states in H-InGaN QWs.

At the high current density above 10 A cm⁻², the FWHM of H-InGaN μLEDs continued to decrease but within 2 nm up to 50 A cm⁻². This implied that the localized states in the H-InGaN QWs still existed but became saturated with increasing current density. In contrast, the FWHMs of L-InGaN μLEDs started to increase slightly. This opposite movement was due to the heat generation in red QWs at the high current densities, which was identical to the FWHM behavior of large-size red LEDs discussed in Sect. 2.2.2.

We measured the average light output power of L-InGaN and H-InGaN μLEDs as a function of the current density by on-wafer testing in Fig. 9(d). Also, we were able to calculate the on-wafer EQEs at different current densities from the measured light output power, as shown in Fig. 9(e). The EQEs of InGaN μLEDs in the dimension from 98 × 98 and 47 × 47 μm² were quite similar regardless of low- and high-In-contents in red QWs. The EQEs of L-InGaN μLEDs corresponding to red emission at the peak wavelength of 626 nm were within 0.34%–0.36% at 4 A cm⁻². Interestingly, the EQEs of H-InGaN μLEDs corresponding to red emission at the peak wavelength of 631 nm were also within 0.34%–0.36% at 30 A cm⁻². This meant that the L-InGaN and H-InGaN red μLEDs could achieve similar EQEs regardless of operation at low and high current densities, respectively.

For L-InGaN μLEDs, lower-In-content in red QWs should exhibit a higher EQE. However, the SRH recombination was dominant at the low current density operation, which would lower the EQE of the devices. For H-InGaN μLEDs, the high driven current density was able to saturate most SRH recombination centers, resulting in the improvement of the devices' EQE. Therefore, the different operation current density caused the L-InGaN and H-InGaN μLEDs to exhibit similar EQE in Fig. 9(e).

To compare the on-wafer testing and the measurement in the integrating sphere, we measured the absolute light output power of 47 × 47 μm² green and amber μLEDs. We defined the calibration factor as the following formula:

$$\begin{eqnarray}&&\begin{array}{l}{\rm{C}}{\rm{a}}{\rm{l}}{\rm{i}}{\rm{b}}{\rm{r}}{\rm{a}}{\rm{t}}{\rm{i}}{\rm{o}}{\rm{n}}\,{\rm{f}}{\rm{a}}{\rm{c}}{\rm{t}}{\rm{o}}{\rm{r}}\,=\\ \,\displaystyle \frac{{\rm{l}}{\rm{i}}{\rm{g}}{\rm{h}}{\rm{t}}\,{\rm{o}}{\rm{u}}{\rm{t}}{\rm{p}}{\rm{u}}{\rm{t}}\,{\rm{p}}{\rm{o}}{\rm{w}}{\rm{e}}{\rm{r}}\,{\rm{m}}{\rm{e}}{\rm{a}}{\rm{s}}{\rm{u}}{\rm{r}}{\rm{e}}{\rm{d}}\,{\rm{i}}{\rm{n}}\,{\rm{t}}{\rm{h}}{\rm{e}}\,{\rm{i}}{\rm{n}}{\rm{t}}{\rm{e}}{\rm{g}}{\rm{r}}{\rm{a}}{\rm{t}}{\rm{i}}{\rm{n}}{\rm{g}}\,{\rm{s}}{\rm{p}}{\rm{h}}{\rm{e}}{\rm{r}}{\rm{e}}}{{\rm{l}}{\rm{i}}{\rm{g}}{\rm{h}}{\rm{t}}\,{\rm{o}}{\rm{u}}{\rm{t}}{\rm{p}}{\rm{u}}{\rm{t}}\,{\rm{p}}{\rm{o}}{\rm{w}}{\rm{e}}{\rm{r}}\,{\rm{m}}{\rm{e}}{\rm{a}}{\rm{s}}{\rm{u}}{\rm{r}}{\rm{e}}{\rm{d}}\,{\rm{b}}{\rm{y}}\,{\rm{o}}{\rm{n}}-{\rm{w}}{\rm{a}}{\rm{f}}{\rm{e}}{\rm{r}}\,{\rm{t}}{\rm{e}}{\rm{s}}{\rm{t}}{\rm{i}}{\rm{n}}{\rm{g}}}.\end{array}\end{eqnarray} \tag{ 2 }$$

This calibration method was used in other works. ^83,87) All the μLEDs in our works were not fabricated on die chips and not encapsulated with resin. We used wire bonding for current injection and put the sample inside the integrating sphere for measurements. As a result, we found that the calibration factor was obtained between 1.7 and 2.6 based on Eq. (2). Therefore, the absolute EQE for our red μLEDs was estimated from 0.58 to 0.94%. The estimated results were comparable to 25 × 25 μm² AlGaInP red μLEDs operated below 30 A cm⁻², ⁸⁶⁾ illustrating the great potential of InGaN material for red μLEDs.

Generally, different applications require different luminance, which finally determines the operated current densities of μLEDs. Luminance was related to the output power density, so we calculated the output power density of L-InGaN and H-InGaN μLEDs at different emission wavelengths. We also presented the other reported data for InGaN large-size LEDs and μLEDs and AlGaInP μLEDs in Fig. 9(f) for comparison. All the data for μLEDs were estimated by on-wafer testing, while the data for large-size LEDs were obtained by measurements in the integrating sphere. The H-InGaN μLEDs that were operated at high current density could achieve a higher output power density than that of the L-InGaN μLEDs with the same emission wavelength. Also, the output power density of 20 × 20 μm² AlGaInP red μLEDs was approximately 0.5 mW mm⁻² at 30 A cm⁻² (measured by on-wafer testing with substrates). At a similar emission wavelength, the output power density of H-InGaN μLEDs was 2 mW cm⁻² at 30 A cm⁻², which was almost 4 fold that of AlGaInP μLEDs. ⁸⁸⁾ Although the output power density of AlGaInP μLEDs was obtained with the substrates, this comparison illustrated that the InGaN materials had great potential for high-brightness red μLED applications.

We also measured the temperature dependences of the L-InGaN and H-InGaN μLEDs by adjusting the stage temperature from 295 to 373 K. The characteristic temperatures were obtained by using Eq. (1), as shown in Fig. 10(a). As we discussed large-size LEDs in Sect. 2.2.3, the characteristic temperature was related mainly to the SRH recombination. Because the SRH recombination would be saturated at high current densities, the characteristic temperatures of both L-InGaN and H-InGaN μLEDs were increased by increasing the current density. Meanwhile, the sidewall treatment by TMAH could effectively remove some of the SRH recombination centers at the sidewalls induced by dry etching. ⁷⁷⁾ Therefore, the characteristic temperature of the H-InGaN μLED was higher than that of the L-InGaN μLED. However, the InGaN μLEDs still exhibited lower characteristic temperatures compared to the large-size LEDs, implying that the SRH recombination centers still existed at the surface of the sidewalls of μLEDs.

Zoom In Zoom Out Reset image size

The peak wavelengths of the InGaN μLEDs would also have red-shift with increasing temperature. Thus, we could calculate the red-shift coefficients at different current densities in Fig. 10(b). The red-shift coefficients decreased as the current density increased. This behavior is similar to those of the large-size InGaN LEDs, and we explained the reason in the above discussion. The red-shift coefficients of the H-InGaN μLEDs were larger than those of the L-InGaN μLEDs. The reason for this phenomenon was not clear yet. We believe that the higher defect densities in the H-InGaN red μLEDs will cause inefficient carrier injection into QWs because of more leakage currents through the defects. Thus, the inefficient current injection into the H-InGaN μLEDs would cause higher red-shift coefficients.

The color coordinates in the color space are important parameters for μLED displays because they determine the physiologically perceived colors in human color vision. Here, we plotted the color coordinates of the L-InGaN and H-InGaN μLEDs in both CIE 1931 and 1976 diagrams [Figs. 11(a) and 11(b)]. The operated current density was 2, 10, 20, and 50 A cm⁻² for L-InGaN μLEDs and 10, 20, 30, and 50 A cm⁻² for H-InGaN μLEDs. The primary red color according to Rec. 2020 is located at the coordinates of (0.708, 0.292) in CIE 1931 and (0.557, 0.517) in CIE 1976. These coordinates were marked as a red star in the CIE 1931 and 1976 diagrams, respectively.

Zoom In Zoom Out Reset image size

Clearly, at the beginning, the color coordinates of both L-InGaN and H-InGaN μLEDs were very close to the reference point of red color defined by Rec. 2020. However, the locations would move farther away from the reference point as the current densities increased in the CIE 1931 or 1976 diagrams. This movement caused the color variation and also reduced the color gamut area coverage, which are unfavorable features for micro-displays. The shifted distance for the H-InGaN μLEDs was much smaller than that of the L-InGaN μLEDs, demonstrating better color stability for the H-InGaN μLEDs than for the L-InGaN μLEDs. However, the color positions of H-InGaN μLEDs were not like those of the L-InGaN μLEDs at the edge of the color space in the CIE 1931 or 1976 diagrams. This gap illustrated that the color saturation for the H-InGaN μLEDs was insufficient, which was due to the broad FWHM of the H-InGaN μLEDs.

Because the peak wavelength has a very small shift at different current densities, ⁸⁶⁾ the position in the CIE diagram was expected to remain almost constant when the AlGaInP μLEDs were driven at different current densities, demonstrating good color stability for the AlGaInP red μLEDs. Therefore, to replace AlGaInP red μLEDs in μLED displays, InGaN red μLEDs required further improvements in the aspect of color stability.