1. Introduction

With the development of multimedia technology, the quality of display devices is required to be higher and higher. Micro-scale light-emitting diode (micro-LED) is considered to be the most competitive next generation display technology [1–4], as it not only inherits the advantages of inorganic LED in brightness, stability and response speed, but also breaks through the resolution limit of traditional LED display. However, among red, green, and blue micro-LED chips, the surface recombination of conventional AlGaInP-based red one is most serious, and its external quantum efficiency (EQE) is not more than 1% in small size [5–7]. In addition, GaAs substrates for AlGaInP micro-LED are also brittle, bringing challenges to massive transfer. On the other hand, the bandgap of InGaN, which is primarily used for manufacturing blue and green micro-LEDs, can also cover red wavelength, but its surface recombination is much weaker [8–11]. Therefore, developing InGaN-based red micro-LEDs is expected by academia and industry to solve the above technical bottleneck problems of AlGaInP-based ones [12,13].

Red micro-LEDs with chip size below 20 µm require high-indium-composition InGaN active region. The high-indium-composition InGaN/GaN quantum wells (QWs) suffer from the large lattice mismatch and strong piezoelectric polarization effect [14–17], leading to a low EQE under a high current density for current InGaN-based red micro-LED [18–20]. Pasayat et al. reported 632-nm red micro-LEDs grown on a strained relaxed porous GaN template, with size of 10×10 µm² and EQE of 0.2% [18]. Similar results were also shown in the study of Zhuang et al. [19], wherein an EQE of 0.18% at 50 A/cm² was demonstrated with chip size of 17×17 µm². Up to now, the efficiency record comes from Li et al. [20]. They have achieved 607-nm red micro-LEDs with chip size of 5×5 µm² and EQE of 2.6% by using an AlGaN capping layer above the InGaN QW layer. Meanwhile, the current density at the peak EQE is about 3 A/cm², which barely meet the needs of low current density operation of micro-LED for display [21].

Compared with InGaN QW, InGaN quantum dot (QD) has the weaker strain, the easier indium composition incorporation, and the stronger carrier confinement [22–26]. Its carrier localization characteristic is also beneficial for suppressing the influence of sidewall etching damage. All these advantages make high-indium-composition InGaN QDs are promising to realize high efficiency red micro-LEDs, especially under a relatively-low current density, which is critical for display and has been demonstrated recently in a green micro-LED [23]. Frost et al. reported InGaN red QDs and laser diodes grown by molecular beam epitaxy (MBE) [24,25], wherein high QD density of 7×10⁷ cm^-2 is obtained and low threshold current density of 1.6 kA/cm² is achieved [25]. Compared to MBE, growth by metal organic vapor phase epitaxy (MOVPE) is more challengeable as there is no in situ monitor tool to reveal the morphology of QDs in real time. In 2013, Lv et al. reported red LEDs based on InGaN QDs grown with the growth-interruption method [26]. However, this method has to reduce the growth temperature to increase indium composition, leading to a low aspect ratio of QDs and poor material quality. Thus, the red LEDs with this method show serious leakage and low emission intensity under forward bias [26]. Till now, red micro-LED based on InGaN QDs has not been reported.

In this work, we report growth of high-indium-composition InGaN QDs based on a precursor-alternate-admittance method by MOVPE, wherein the growth temperature is relatively high and a typical QDs density as high as 3 × 10¹⁰ cm^-2 is achieved. The growth mechanism is intensively analyzed to optimize the growth parameters for improving optical properties. The red InGaN QDs with IQE of 12% is obtained for micro-LED fabrication. The typical performance of a 20-µm device is 570 nm emission with 4.92% EQE at 0.3 A/cm², or 630 nm emission with 0.03% EQE at 10 A/cm², exhibiting the potential for micro-LED display under low current density.

2. Result

2.1 Growth of InGaN QDs

The InGaN QDs are grown by a two-step method [27]. In the first step, triethylgallium (TEGa), trimethylindium (TMIn) and ammonia (NH₃) are alternately passed into the reaction chamber to form the wetting-layer-free low-indium-composition InGaN QDs [28]. In the second step, TEGa, TMIn and NH₃ are injected simultaneously to conformally grow high-indium-composition InGaN on top of the QDs formed in the first step. The final morphology maintains the same, while the emission wavelength is determined by the second step.

The underlying QDs formed in the first step has a decisive influence on the final QDs morphology. Therefore, the way how the precursors are alternately supplied in the first step should have a significant impact on the growth of QDs. Three ways for the precursors alternately on-off process are designed while keeping the growth temperature of 665 °C and cycle times of 120 constant, as shown in Table 1. After the first step growth, all the samples are quickly cooled to room temperature in an NH₃ atmosphere to maintain the morphology.

Table 1. List of InGaN QDs samples and their corresponding main growth parameters

View Table

Atomic force microscopy (AFM) images of the three samples are shown in Fig. 1(a)-(c), respectively. The morphology of the three samples has many characteristics in common. First, QDs with a high density of ∼3 × 10¹⁰ cm^-2 are all achieved on the surface. The diameter and height of QDs are ∼65 nm and 1.5-2 nm, respectively. In addition, the QDs of the three samples are all distributed along the atomic steps of the GaN surface, as indicated by green dashed lines in Fig. 1(a)-(c), respectively. However, there also exist some obvious differences among their morphologies. Sample A has almost no defects, while there are some obvious ones in samples B and C, such as the regions indicated by the yellow ellipses in Fig. 1 (b)-(c). In fact, the differences in their morphologies stem from the timing and duration of TMIn access. When the group-III precursors and NH₃ are separately supplied to the GaN surface, the free-migration capability of the indium and gallium atoms on GaN surface is strong, so metal atoms can quickly accumulate and form QDs to relax strain, and ultimately there are few defect areas generated (sample A). However, when TMIn and NH₃ are supplied at the same time (sample C), more nitrogen atoms can quickly capture indium atoms, reducing the ability of indium atoms to freely migrate on the surface. With the incorporation of more indium atoms, the migration ability of indium atoms decreases, so the more significant strain in sample C is released by generating defects. Sample B is in an intermediate state between sample A and sample C.

 

Fig. 1. AFM morphology and PL spectra of epitaxial samples based on Table 1. a-f The AFM images of QDs samples A-F, respectively. g The PL spectra of samples A, B, and C at room temperature. h The PL spectra of samples G, H, and J at room temperature. The insets show the PL photographs of the three samples.

Download Full Size | PDF

Figure 1(g) shows the photoluminescence (PL) spectra of the three samples at room temperature. It can be seen that the emission wavelength of samples A-C is 365, 374, and 384 nm, respectively. This indicates that sample C possesses the highest indium composition. Although sample C has the most defects, it has strongest PL intensity. This can be ascribed to a stronger confinement effect on carriers, as the indium composition increases. The PL results reveal that the indium composition in the first layer of QDs can be increased by introducing TMIn and NH₃ into the reaction chamber at the same time, during the alternately on-off process of precursor. But at the same time, the defect areas on the sample surface will also increase. In addition, the emission peaks of the three QDs samples are all in near-ultraviolet band, which means that the indium compositions in QDs formed in the first step are all as low as 5% or less.

In order to verify that the InGaN grown in the second step can maintain the morphology formed in the first step, three samples D-F are grown, wherein growth processes in the first step correspond to samples A-C, respectively. After the first step growth, an upper layer of InGaN with a higher indium composition and a nominal thickness of 1.5 nm is grown directly at 665 °C. Then they are quickly cooled after the second step of InGaN growth. In Fig. 1(d)-(f), their AFM images of the surface are shown. It can be seen that samples D-F all maintain the surface morphology of samples A-C very well. QDs with an average diameter of ∼75 nm (a little bit larger than those formed in the first step), a height fluctuation of ∼1.5-2 nm, and a density of ∼3 × 10¹⁰ cm^-2 are formed. In addition, their surface quality is also consistent with that of samples A-C. Interestingly, as shown by the red ellipse in Fig. 1(d), there are several QDs surrounding a hole. This phenomenon can be attributed to the threading dislocations in GaN penetrating to the InGaN QDs layer. Similar findings have also been made in our previous study [29].

In order to further analyze the crystalline quality and luminescence characteristics of the QDs, three samples of G, H, and J are grown. Their QDs growth methods are the same as samples D-F, respectively. The only difference is that each sample consists of two layers of QDs. A 15-nm GaN barrier between the two QDs layers and a 15-nm GaN capping layer on the second QDs layer are grown at 740 °C.

As shown in Fig. 1(h), at room temperature, the PL peak wavelengths of samples G-J are 494, 530, and 545 nm, respectively. It is worth noting that the PL intensity of sample G is weaker than that of sample J. The difference of the PL wavelength and intensity among the three samples can also be perceived by the insets in Fig. 1(h).

2.2 Electron microscope analysis of InGaN QDs

In order to further explore the micro-morphology and growth kinetics of the InGaN QDs grown by the precursor-alternate-admittance method, and to investigate the essential reasons for the differences in the emission wavelength and intensity of the three samples G-J grown by the three kinds of the on-off process described above, transmission electron microscope (TEM) measurement and analysis are performed.

Figure 2(a)-(b) demonstrate the TEM EDS results of samples G and J, respectively. It can be seen that both samples have two active layers, wherein QDs with a diameter of 15-50 nm and a height of 1.5-3 nm are embedded. Compared with the AFM measurement results of the samples D and F without a capping layer, the diameter size of the QDs in TEM results has decreased, which can be ascribed to the further crystallization of the InGaN QDs during the growth process of the GaN barrier and capping layer. According to the indium atom EDS mapping and line scan results of samples G and J, it can be concluded that the InGaN QDs of sample J have a higher indium composition, and the aggregation of indium atoms in sample J is also higher, which takes responsibility for its longer PL wavelength.

 

Fig. 2. EDS and TEM results of samples G and J. The TEM energy dispersive spectrometer (EDS) results of (a) sample G and (b) sample J, respectively. A high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) image indicating the EDS region, three EDS mapping images for gallium atoms, nitrogen atoms, and indium atoms, respectively, and an EDS line scan data image for the gallium composition and indium composition along the red arrow in the HAADF STEM image are both contained, for samples G and J, respectively. (c) The dark field scanning transmission electron microscopy (BF-STEM) image for one single QD in sample G. The results of one-dimensional Fourier filtering of the QD BF-STEM image along (d) the y direction (growth direction) and (e) the x direction (the direction parallel to the interface of GaN and InGaN), respectively. (f) The BF-STEM image for one single QD in sample J. The results of one-dimensional Fourier filtering of the QD BF-STEM image along (g) the y direction and (h) the x direction, respectively.

Download Full Size | PDF

The difference in the emission wavelengths of samples G-J can be explained by the different indium composition of the grown QDs, then the remaining doubt is their difference in emission intensity. In fact, the high-resolution STEM images reflect the lattice arrangement of the epitaxial atoms. Therefore, one-dimensional Fourier filtering of the STEM images along a specific direction can usually be executed to reflect the lattice characteristics of the epilayer in this direction. Figure 2(c) and Fig. 2(f) are the high-resolution BF-STEM images of single InGaN QD of samples G and J, respectively, and they are subjected to one-dimensional Fourier filtering along the y-direction (growth direction) and the x-direction (the direction parallel to the interface between GaN and InGaN), respectively, resulting in Fig. 2(d)-(e) and Fig. 2(g)-(h). Generally, the arrangement distortion appearing along the y-direction stripes represents the edge dislocation in the lattice, and the arrangement distortion appearing along the x-direction stripes represents the screw dislocation or stacking fault in the lattice. The results show that the lattice arrangements inside the InGaN QD of both samples G and J are basically intact. Besides, no edge dislocations are found in the GaN material around the QD of the two samples. However, Fourier filtering results along the x direction indicate that there are a lot of screw dislocations and stacking faults existing in the GaN around the QD of sample G, while no such defects exist around the InGaN QD in sample J.

It is believed that during the growth of InGaN QDs, owing to the formation of defects shown in the yellow ellipse in Fig. 1(c) (sample J), the residual strain can be completely released with no additional dislocations generated. However, a large number of screw dislocations or stacking faults are eventually generated in GaN around the QD in sample G, due to the lack of such a strain relaxation mechanism. The screw dislocations or stacking faults usually play a crucial role as the carrier non-radiative recombination centers, which will dramatically reduce the IQE of the InGaN QDs and leads to the lowest PL intensity of sample G.

2.3 Increasing QDs emission wavelength

Based on the AFM, PL, and TEM measurements, we have known that in the first step, the precursor alternate way which sample J used, is more conducive to the incorporation of indium atoms, and higher IQE. Therefore, the subsequent growth of InGaN QDs in the next study all adopts this alternately on-off method. On this basis, we still need to explore how to expand the emission wavelength to red light by controlling the growth parameters. While increasing the flowrate of TMIn, it is also necessary to study the growth temperature.

The growth temperature has a significant effect on the morphology of InGaN QDs. As shown in Fig. 3, six QDs samples were grown at different temperatures of 640 °C, 650 °C, 665 °C, 685 °C, 700 °C, and 720 °C, respectively, and the growth process all stopped after the first growth step, and then quickly cooled down in a NH₃ atmosphere, to observe the morphology of QDs. It can be seen that as the temperature increases from 640 °C to 720 °C, the average height of the QDs remains ∼2 nm, but the diameter of QDs increases from ∼50 nm to ∼200 nm. As the growth temperature is as low as 640 °C, Fig. 3(a) shows that not only dense InGaN QDs but also many large particles are generated on the surface of the sample, as circled in the red square. These particles are recognized as metal droplets as a result of the low decomposition capability of NH₃ and poor metal migration ability at a low growth temperature. The formation of metal droplets will deteriorate the crystalline quality and thus reduce the optical performance. Therefore, the temperature for growing InGaN QDs should be higher than 640 °C. As the growth temperature is lower than 685 °C, defect areas indicated by the yellow ellipses in Fig. 3(b)-(d) are formed on the sample surface, which is similar to the defects in Fig. 1(c). As the growth temperature rises, such defects will gradually decrease and finally disappear, but another type of defects begins to appear. As shown in Fig. 3(e)-(f), there are many holes distributed on the sample surface, as indicated by the purple triangles. These holes originate from the penetration into the QDs layer of the threading dislocations in the underlying GaN bulk material, as the growth temperature gradually increases. Counting the holes in Fig. 3(f), a density of 3 × 10⁸ cm^-2 can be calculated, which is consistent with the threading dislocation density of bulk GaN grown on c- patterned sapphire substrates by MOVPE [30].

 

Fig. 3. The AFM images of six QDs samples grown at different temperatures. a-f The AFM images of six QDs samples grown at different temperatures of 640 °C, 650 °C, 665 °C, 685 °C, 700 °C, and 720 °C, respectively.

Download Full Size | PDF

Therefore, the growth temperature has a significant impact on the morphology of InGaN QDs. When the temperature is too low, there will remain a lot of metal residue. When the temperature is too high, the size of QDs will increase dramatically, and a large number of dislocation holes will be generated on the surface. Thus, the most suitable temperature for QD growth by the precursor-alternate-admittance method is ∼665 °C, and the approach adjusting the emission wavelength of QDs by changing the growth temperature is not applicable. Meanwhile, due to the shape retaining property of the upper layer, the second-step growth still maintains the unique morphology and stress release characteristics of QDs. Especially considering the indium composition of the first step is low, the final emission wavelength of QDs can be determined by the second step if further increasing its indium composition [26]. Thus, a red InGaN QDs sample with an emission wavelength of 613 nm was obtained by simply increasing the TMIn flowrate in the second step while keeping the growth temperature unchanged. The estimation from EL indicates that the In component in QDs is ∼35%. Figure 4(a) shows the temperature-dependent PL (TDPL) spectra of the red QDs. The excitation laser wavelength was 405 nm, the laser power is 50 mW and the diameter of laser spot is ∼50 µm. The IQE of the sample is 12%, which is calculated as the ratio of the integrated PL intensity measured at 300 K to that at 5 K (IQE = I_300K / I_5K) considering that the non-radiative recombination can be ignored at 5 K [31]. Figure 4(b) plots the temperature-dependent PL wavelength and spectral full width at half maximum (FWHM) of the red QDs sample. The FWHM of the sample is ∼64 nm, which is partially due to the nonuniformity of InGaN QDs. From 5 K to 300 K, the fluctuations of both the wavelength and the FWHM do not exceed 5 nm, reflecting the stable temperature characteristics of QDs.

 

Fig. 4. The TDPL results of the red InGaN QDs sample. a The TDPL spectra of the red InGaN QDs sample. b The evolution of FWHM and peak wavelength of the red QDs sample with temperature.

Download Full Size | PDF

2.4 Device fabrication and characterization

A red LED based on InGaN QDs active region, denoted as sample M is grown and its structure is shown in Fig. 5(a). The epitaxial structure mainly includes a 3.5-µm n-GaN bulk layer, 10-period In_0.03Ga_0.97N (3 nm) / GaN (3 nm) superlattices (SL) for pre-strain and nonradiative recombination centers suppressing, 5-layer of InGaN/GaN red QDs, a 20-nm p-AlGaN electron blocking layer (EBL) and a 180-nm p-GaN contact layer.

 

Fig. 5. Red QD micro-LED device structure. a-b Epitaxial structure of (a) sample M and (b) sample N, respectively. c-d The BF STEM image including SL, QD active region and EBL of (c) sample M and (d) sample N, respectively.

Download Full Size | PDF

To further improve the indium incorporation in QDs, another sample, labeled as sample N, is grown, wherein indium-composition-graded InGaN/GaN superlattices and 4-pair of blue In_0.2Ga_0.8N (3 nm) / GaN (10 nm) QW are inserted beneath the QD layers, as shown in Fig. 5(b). The indium-composition-graded InGaN/GaN superlattices are composed of 6-period In_0.03Ga_0.97N (3 nm) / GaN (3 nm), 6-period In_0.06Ga_0.94N (3 nm) / GaN (3 nm), 6-period In_0.09Ga_0.91N (3 nm) / GaN (3 nm), 6-period In_0.12Ga_0.88N (3 nm) / GaN (3 nm).

The BF STEM image of two samples mentioned above are shown in Fig. 5(c)-(d), respectively. It can be seen that in the complete LED structure, the QDs still maintains a good morphology during multilayer growth. In two samples, most of the dislocations in STEM images start from the third layer of QDs, which indicates that our growth conditions, especially the number of QD layers, still need to be optimized. Due to the increase of epitaxial thickness, the defect density in sample N is significantly higher than that of sample M.

Quick on-wafer tests were carried out and the EL photos can be found in Supplement 1. Red emission can be observed under low injection in both samples, but as injection increasing, the wavelength blue-shift of sample N is smaller than that of sample M. Micro-LED chips with 1∼20 µm size were processed by using direct laser writing lithography. Other steps were not optimized intentionally for micro-LEDs, including dry etching and passivation. For detailed epitaxial growth and device fabrication, please refer to Supplement 1.

The electrical properties of micro-LED chips are characterized by probe test. The light is collected from the bottom of the chip with an integrating sphere and analyzed by a spectrometer (see Supplement 1). Therefore, the measured EQE value will be only a part of the real value [18]. The EL results of 20-µm sample M and 10-µm sample N are shown in Fig. 6. The periodically arranged bright spots are the reflection of patterned sapphire substrate. The peak wavelength of the former could reach around 590 nm at 0.01 A/cm², while it is 630 nm at 10 A/cm² for the latter, indicating the indium composition in the latter has been improved effectively since the modified pre-strained layers were introduced. Due to the small volume, the increase of current density will lead to a more significant increase in carrier concentration in QDs. Thus, the band filling effect of QDs is very strong, leading to significant wavelength blueshift as the current density increasing. Interestingly, we can clearly observe the random distribution of red and green spots in sample M under 0.1 A/cm², as shown in Fig. 6(a), reflecting the non-uniformity of self-assembled InGaN QDs on indium composition and size. This non-uniformity also results in a large full width at half maximum (FWHM) of the EL spectra (∼300 meV).

 

Fig. 6. The EL characterization of the red InGaN QD micro-LED. a-b EL photographs of (a) 20-µm sample M and (b) 10-µm sample N at different current densities. c-d EL spectra of (a) 20-µm sample M and (b) 10-µm sample N at different current densities.

Download Full Size | PDF

Figure 7 shows the EQE and peak wavelength depending on the current density of samples M and N. When the chip size is downscaled, the peak wavelength of both samples M and N basically does not change, reflecting the insensitive strain of InGaN QDs on chip size. From 20 µm to 1 µm, sample M exhibits a decrease of the peak EQE from 4.93% to 1.78%, and an increase of the corresponding current density from 0.3 A/cm² to 20 A/cm². This typical size-dependent effect shows that the surface recombination at sidewall induced by dry-etching gradually becomes significant in the competition with Shockley-Read-Hall (SRH) recombination in the active region, as chip size downscales. For sample N, the peak EQE of all size devices is about 0.15%, and the corresponding current density is around 500 A/cm². The peak wavelength can reach 630 nm under 10 A/cm² for 10-µm sample N, with a reduced EQE of 0.03%. This insensitive size-dependent effect implies that the SRH recombination in the active region of sample N is dominant and the surface recombination can be ignored.

 

Fig. 7. The EL characterization of the red InGaN QD micro-LED. a EQE (upper) and peak wavelength (lower)-current density diagram of sample M with different sizes. b EQE (upper) and peak wavelength (lower)-current density diagram of sample N with different sizes.

Download Full Size | PDF

Till now, red micro-LEDs based on QDs still face some problems, including low EQE and serious blueshift under high current density. In order to improve EQE, the period of InGaN/GaN red QD layers should be optimized to avoid excessive strain accumulation. In order to solve the blue shift problem, we may use the coupled InGaN/GaN quantum well and quantum dot structure [23,32]. Coupled quantum well will serve as a “carrier reservoir” before the QD layer, and suppress the band-filling effect under different current densities through energy band engineering.

3. Conclusion

In summary, a two-step growth method for high-indium-composition InGaN QDs by MOVPE is proposed. The growth mechanism is intensively studied to optimize the morphology and luminescence characteristics of QDs, and a 613-nm red QDs sample with IQE of 12% is obtained. On this basis, QDs micro-LEDs with different sizes of 1-20 µm are fabricated and characterized. The 20-µm and 1-µm micro-LEDs show 4.92% and 1.78% EQE at 0.3 and 20 A/cm², respectively. By introducing InGaN MQWs as pre-strained layers, a 10-µm red micro-LED is demonstrated, which shows 638 nm emission with EQE of 0.03% at 10 A/cm². These EQE values are limited by the light collection in measurement. The present results validate the advantages of InGaN QDs to realize the display-oriented red micro-LEDs.